Innovation Studies and the History of Technology

Home , Blue Envoy, Blue Peacock, Ernest Titterton, Ferranti Argus, Fort Halstead, High Explosive Research

Trying to secure the past: innovation studies and the history of technology

A thesis submitted to the University of Manchester for the degree of Doctor of Philosophy (PhD) in the Faculty of Humanities

Jonathan Aylen

2018

Contents page

Listing of Publications 3

Abstract 4 Declaration 5 Copyright Statement 5 Jonathan Aylen, Statement of Eligibility 6

Introduction 1. Selection of a coherent set of papers 8 2. Historical methods in the study of technology 23 3. The nature of the innovation process 30 4. Lessons from innovation research 41 5. Bibliography 46 6. Corrections and updates 57 7. Impact of this research 59

Papers

Blue Danube - Britain’s post-war atomic bomb 61

Stretch - how innovation continues once investment is made 62

Bloodhound - building the Ferranti Argus process control computer 63

Open versus closed innovation - development of the wide strip mill for steel 64

Construction of the Shotton wide strip mill 65

Development of computer applications in the iron and steel industry 66

“Trying to secure the past: innovation studies and the history of technology”

"People work much in order to secure the future; I gave my mind much work and trouble, trying to secure the past"

Isak Dinesen/also known as Karen von Blixen-Finecke (1885-1962), Shadows on the Grass, Harmondsworth: Penguin, 1990, essay “Echoes from the Hills”, p.116

papers:

1. Jonathan Aylen, “First waltz: development and deployment of Blue Danube, Britain’s post-war atomic bomb”, The International Journal for the History of Engineering & Technology, vol. 85, no.1, January 2015, pp.31-59

2. Jonathan Aylen, “Stretch - how innovation continues once investment is made”, R&D Management, vol.43, no.3, June 2013, pp.271-287

3. Jonathan Aylen, “Bloodhound on my trail: building the Ferranti Argus process control computer”, The International Journal for the History of Engineering & Technology, vol. 82, no. 1, January 2012, pp.1–36

4. Jonathan Aylen, “Open versus closed innovation: development of the wide strip mill for steel in the USA during the 1920s”, R&D Management, vol.40, no.1, January 2010, pp.67-80

5. Jonathan Aylen, “Construction of the Shotton wide strip mill", Transactions of the Newcomen Society, vol.78, no.1, February 2008, pp. 57-85

6. Jonathan Aylen, "Megabytes for Metals – The development of computer applications in the iron and steel industry", Ironmaking and Steelmaking, vol. 31, no. 6, 2004, pp.465-478

80,000 words

Abstract

These papers report untold and unfashionable stories in the history of technology. They focus on practical developments rather than science. Two papers (Stretch, Blue Danube) concentrate on practical issues of manufacture, maintenance and use. They show how solutions to technical problems emerge over time, often in unforeseen ways. Along with the paper on missile guidance systems, they highlight the impetus given by problem solving in the face of bottlenecks and “reverse salients” that obstruct progress on a technology. Innovation can be seen as a process of creative engineering to resolve a set of “problem sequences.”

One paper on construction (Shotton) highlights the dilemmas faced when adopting technology from outside the firm: who should supply the new technology and what scale of plant should be built? Another paper (Open versus closed innovation) considers the best form of organisation suited to developing innovations in house. Here too, cooperation with partners turns out to be the best way forward. The paper on process control shows how mobility of technical labour and a strong product champion encouraged the spread of a rapidly evolving computer technology throughout South Wales.

These papers are almost entirely based on primary sources, including company archives, oral history interviews, technical publications and correspondence with the innovators themselves. The strengths and shortcomings of this approach are explored, particularly in the defence sector. This approach of “history from below” is born of a conviction that even headline stories (such as the UK atom bomb) overlook the mundane nature of conventional engineering. In the case of atomic weapons, science was the heroic story - a myth which is still current. These papers show how technology actually develops and explore hitherto undocumented fields of technology.

All of these papers show how knowledge coalesces from diverse sources to generate innovations in an evolutionary way. Even a top secret project like the atomic bomb draws know-how from a wide range of suppliers. The papers focus on the way in which this diverse knowledge comes to be incorporated in new technologies. Evolution is a broad analogy. Most technologies evolve through purposive selection rather than random chance exhibited by the natural world of living beings. History matters when technology is being developed. Firms accumulate expertise and learn within networks and generate a variety of solutions to a given problem. Reverse salients often produce radical changes or mutations in technology. A selection process operates at all levels to identify satisfactory solutions which work. This is not necessarily optimal and organisations may get locked in to a particular technology as it develops in a path dependent way. The dynamics of innovation may be unforeseen in advance, but they can be traced in retrospect.

These case studies also offer lessons in the way technical development might be organised. Loose “communities of practice” drawing upon a range of skills and experience are often ideal at the development stage. Formal management is more appropriate to the production phase. Guided missile control at Wythenshawe developed in just such a loose way. Secrecy rules inhibited similar cooperation over the atomic bomb, but individuals were given remarkable freedom. The case of computer control highlights the role of product champions driving innovation at a senior level. Surprisingly, careful use of resources does not seem to inhibit innovation. Shotton, the atomic bomb and the wide strip mill were all developed in an atmosphere of parsimony.

Declaration

All of the work submitted here is the author’s own contribution. The articles are all single authored. The help of very many respondents is warmly acknowledged, but the author remains responsible for all that is submitted here.

All of the work submitted here was completed while the candidate was an employee of the University of Manchester Institute of Science and Technology and its successor, the University of Manchester between 2001 and 2018.

None of the work presented has been submitted in support of any application for any other degree or qualification of this or any other University or of any professional body.

I confirm that this is a true statement and that, subject to any comments above, the submission is my own original work.

Signed: ………………………………………………..……….. Date: . . . 21st November 2018 . . .

Copyright Statement i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The University’s policy on Presentation of Theses

Jonathan Aylen

Jonathan has enjoyed a varied career, as an economist and as an engineer, an innovation researcher and in environmental management. He now specialises in the history of technology. He is currently senior visiting research fellow in the Manchester Institute of Innovation Research within Alliance Manchester Business School at the University of Manchester where he was previously senior lecturer and a founder director of MIoIR. Jonathan Aylen was an Associate Editor of R&D Management from 2014 to 2016. He was Conference Chair for the R&D Management Conference in 2013.

His recent research on history looks at Cold War technology and at how the world went digital in areas such as missile guidance, process plant control and railway management. A paper published in 2015 on the technology of nuclear weapon design contradicts received wisdom on the issue. The research continues with a recent paper in the Royal Airforce Historical Society Journal presented earlier at their AGM. His work on RAF nuclear weapons has been presented to meetings of the Charterhouse Group and to the Defence Electronics Society. He was the invited keynote speaker at the International LETTERPRESS nuclear disarmament exercise in October 2017 and spoke at the Royal Navy’s Conference on the Continuous at Sea Deterrent in 2018.

Jonathan is currently Vice-President and future President of the Newcomen Society, the International Society for the Study of the History of Engineering and Technology. He is Chair of Newcomen North-West, an elected member of their national Council and a member of the Management Committee that runs its day to day affairs, with responsibility for publications. He is a former President of Manchester Statistical Society and a former member of the Economic Affairs Committee of ESRC. His latest research with the National Railway Museum explores the unrecognised history of computing at British Railways.

His work on wildfires concentrates on the likelihood and consequences of fires. He completed a NERC funded project on Wildfire Threat led by Julia McMorrow, working with the Forestry Commission to analyse the risk, spread and impact of wildfire, and worked with the Royal Society in 2015 on policy towards UK wildfire. His recent research on wildfire has helped inform policy and practice towards fire risk and has appeared in two joint papers in the Philosophical Transactions of the Royal Society, Series B in 2016. He is currently a government appointed SAGE Scientist, played an influential role in the National Risk Assessment and has helped train fire-fighters. His impact case on wildfire policy was submitted by the University of Manchester to the RAE.

He has written extensively on wildfire forecasting and on climate change. His latest work on weather and climate with Gina Cavan and Kevin Albertson appears in Climatic Change, following earlier published work in Climate Research and the Journal of Environmental Management. He was a co-author of an influential report on Climate Change and the Visitor Economy in the North-West.

Jonathan has written on the global steel industry for some 40 years. He has given extensive advice to national and international bodies on policy relating to steel, including various branches of the UK Government at critical moments, Parliamentary Committees, the Monopolies Commission, OECD, UNECE and the ILO. This work has focussed on productivity, international trade, privatisation and on materials recycling among other areas. Jonathan Aylen has worked extensively with the media for the past 30 years, making 50 appearances on BBC TV and radio in 2017 and 2018 commenting on steel issues. He previously wrote a regular

6 column for a steel industry trade journal, Steel Times International, some 40 columns in nine years. He was a Visiting Fellow, Science Policy Research Unit, University of Sussex in 1982 and a Visiting Researcher, National Institute of Economic and Social Research between 1980 and 1981.

As a technologist, he has written on rolling technology and his latest book with Ruggero Ranieri, Ribbon of Fire, shows how US strip mill technology spread across Europe. He was a member of the Bulk Metal Forming Committee, a technical committee of the Technology and Materials Science Division of IOM3 from 2007 to 2015 and a keynote speaker at a range of international conferences on rolling, most recently at IJmuiden and Birmingham. He also has experience as a forecaster and spent two decades advising a leading European steel company on purchasing. Jonathan has won two best paper prizes for his work relating to steel, including the Williams Prize of the Institute of Materials, Minerals and Mining (with Dr K. Albertson) in 2007 for the best paper in metals and a national Partnership Award for teaching excellence in economics while at the University of Salford presented by the Secretary of State for Education in 1995.

Jonathan Aylen has a B.A.(Hons) in Economics, first class, awarded by the University of Sussex in 1972. He was successively lecturer and senior lecturer in economics at the University of Salford from 1974 to 2001; senior lecturer in the Department of Mechanical and Aerospece Engineering at UMIST from 2001 to 2004 and senior lecturer at PREST/Manchester Institute of Innovation Research at the University of Manchester to 2018.

Introduction

1. Selection of a coherent set of papers and their contribution to knowledge

The articles presented here relate to the history of 20th century technology. They use the evidence of history to gain insights into the management of innovation. They focus on unknown and unrecorded aspects of technical history. For the most part, they cover a narrow time period between 1920 and 1965. This window of time offers lessons close enough to be valid for current technologies, and distant enough to offer detached perspective. Two of the papers use oral history sources. They are consciously directed at a wide audience, not just innovation scholars, but also professional engineers, defence experts and historians.

Contribution to Knowledge The articles offer contributions to knowledge in three areas: the history of technology, innovation studies and research methodology. These are empirical papers. In terms of history, they uncover fresh material on the history of the wide strip mill, a radical innovation in steelmaking. They show how the technology was developed, transferred to Europe and later automated. The work on computers covers the hitherto neglected field of process control. The papers here also explore the secret world of technical development in the Cold War in guided missile control and nuclear weapons.

The papers offer new insights into innovation. Open innovation has become a fashionable topic, but there is very little evidence on whether it offers a better way of organising innovation. The article on development of the wide strip mill considers a “natural experiment” whereby two competing teams used radically different approaches to engineering development in their race to build a mill to make wide steel strip in large commercial quantities. In this particular case it shows that collaboration promoted by entrepreneurs proved far more effective in generating radical and profitable innovation.

The articles on defence innovation in the Cold War show similar innovation processes at work in the secretive military sector as are found in open commercial activity. The successful development of the Argus 200 computer for Bloodhound 2 owed much to the loose community of practice that formed within Ferranti Automation at Wythenshawe. This approach was denied to those working on the “top secret atomic” development of Blue Danube who were constrained by official secrecy. Nevertheless, the same evolutionary dynamics of development were at work here as patterns of ad hoc collaboration with suppliers supported the emergence of a viable weapon whose design was repeatedly modified even as it entered service. The developers of Blue Danube exploited existing knowledge and explored new knowledge, especially in the areas of electronics and conventional explosives. In direct contradiction to the “linear model” of technical development, the weapon was deployed before it was tested.

In terms of methodology, these papers explore less orthodox approaches to evidence gathering in order to reveal underlying stories of practical development by engineers, technicians, draughtsmen (and women) far removed from the conventional accounts of highly qualified scientists in formal R&D labs. The papers explore new sources, notably uncatalogued company archives, prolonged oral history interaction with participants, access to artefacts and the technical trade press of the time. The papers also bring an innovation studies framework to an area of history that has emphasised technical stories.

These single authored articles were written at the University of Manchester over a period of fifteen years. They were written as a personal initiative alongside funded research projects in unrelated fields. Two articles were developed from an EPSRC funded project on response to climate change. Two have been republished as part of a jointly authored book. One article took over seven years to reach fruition (Blue Danube). Another was written and presented within a year, but draws on forty years of research (Stretch) reflecting a growing personal disenchantment with vintage models in economics.

One of these articles is highly cited (Open Innovation). One seldom cited article is among the most downloaded article from the journal concerned with over 1,700 downloads by November 2018 (Blue Danube). A well-cited article has been excluded because it was not published in a peer-reviewed journal , (Aylen, 2001). These articles were written to be read. Downloads and prolific correspondence show they have been widely read outside the academic community, by engineers and defence personnel. The Bloodhound paper is said to be the most frequently stolen item from the MOD Library, (it would not be politic to reveal the source for this). They have been presented at seminars and conferences across Europe and North America and to professional bodies around England and Wales.

They were written out of a compulsion to understand what went on in unknown circumstances and to understand how the process of innovation worked in practice. They build on detail. As Nathan Rosenberg (1994, 2-3) says: “by getting down into the trenches examining the particular sequence of events and institutions within particular industries, one can extract insights into the process by which technological knowledge grows.” They can be criticised for focussing predominantly on “boy’s toys” as Martin (2016) terms it. Nor do we consider the use and performance of these innovations here.

The evolution of knowledge All of these papers show how knowledge coalesces from diverse sources to generate innovations in an evolutionary way. They focus on the way in which knowledge comes to be incorporated in new technologies. These might be new products (a rolling mill, an atom bomb, a control computer), or new processes (process modifications, computer control). For the most part, the papers relate to hardware development, rather than services or business models. Development of new routines is tackled in a paper not included here (Aylen, 2016) – a weapons system is a bundle of military routines. Much of the knowledge exploited by these innovations is old knowledge, often rearranged and used in new ways. Some is new knowledge – insights often gained on the workbench, or through practical engineering in the laboratory, by

10 chance or by casual conversation. We overlook the key role of tacit knowledge held by engineers, technicians and draughtsmen.

There is very little formal science evident in any of these technical developments. Indeed, text-books on motion control by microprocessor were published a decade after the development of Bloodhound 2 and the Fleetwood soda ash plant (e.g. Tal, 1984). Process control illustrates the way in which new engineering disciplines begin to emerge in tandem with R&D in industry, as Mowery and Rosenberg (1998) illustrate for the USA.

Evolution is a broad analogy. Most technologies evolve through purposive selection rather than random chance exhibited by the natural world of living beings. (Although theorists such as Aghion and Howitt, 1998, ch.2 find it convenient to treat innovations as random arrivals.) As Ziman (2000, p.5) reminds us “the most obvious difference is that novel artefacts are not generated randomly: they are almost always the products of conscious design." Vincenti (1994) prefers the terms “blind variation and selective retention” to describe the learning process around technological evolution. Here the variety is “blind” in the sense that it is new to knowledge – the engineer cannot foresee if a technology will work. They can use their experience, training and judgement to conjecture, but they cannot know if a proposed solution will work before testing. Selection comes in to play through trials and engineering judgements which highlight the most promising options and developmental paths to follow (Vincenti, 1994). To echo Popper (1963), engineering design is a process of conjecture and refutation. Unlike the natural environment, engineers can be creative, they do not have to take their selection environment as given (McKelvey, 1996, p.39.) We might add, technological evolution is also inherently wasteful as it necessarily means that many ideas are pursued but prove to be misguided, not fit for purpose, too late (or too early), too expensive or just not work. Wasteful too, as the new destroys the old, captured by Schumpeter’s (1943, ch.7) notion of “creative destruction.”

We use the term evolution in one biological sense. We are not claiming there is anything optimal about these outcomes. Most engineering problems can be solved in

11 several ways. There are diverse solutions. There is more than one way to develop a nuclear weapon – using uranium or plutonium as fissile material, for example. Roller bearings were a false start for wide strip mills which soon switched to oil-film bearings. The accuracy of guided missile control for Bloodhound 2 would have been irrelevant if nuclear warheads had been used, as in the US Bomarc system with an X-40 warhead (Hansen, 1988, p.187; Morgan and Berhow, 2010). New technology can embody “false starts”. The first transistor based computer used point contact transistors (Grimsdale, 1995), whereas junction transistors had more potential. Again germanium transistors were used in the late 1950’s since they were the first available in commercial quantities (Braun and Macdonald, 1982, ch.6). But, silicon proved more efficient. Rather, these innovations are engineering solutions that were feasible given the objectives, the knowledge, resources and budgets available at the time. They were elegant solutions to the problem in hand, but not necessarily a “best” solution. Smith (1979, p.204) points out Felix Wankel looked for an “aesthetic” solution to the shaking and pounding of a reciprocating piston engine. His Wankel rotary engine was an attempt to provide a pleasing and elegant vibration free solution to the problem of providing power by internal combustion. The sheer variety of goods is overwhelming.

In any case, the concept of an “optimal” technical solution is hard to define in a changing environment where knowledge is limited, and technical possibilities and relative prices are altering all the time. The concept of satisficing, or a satisfactory solution seems far more realistic here (Cyert & March, 1964). Technical change is often path dependent too (Dosi, 1982; Rosenberg 1994, ch.1). One satisfactory solution leads on to the next. Admittedly, shifts towards the baroque are often curbed by engineers seeking elegant solutions and reductions in complexity (e.g. Marshall, 2013). Nevertheless, the account of technology adapting to its environment in a satisfactory way seems compelling. It can be argued that firms use complementary skills among their networks of suppliers and users and government support to reduce the remarkable variety of technical choices open to them into some “locally optimal” set of production choices (Frenken, 2000). Sub-optimality may itself be driving evolution in complex systems – the inherent instability driving change (Allen, 2014.)

Eldridge (1997) emphasises there are no biological parallels with the way that one technology supplier learns, borrows and steals from others. Items that are designed have different phyletic trees to those which evolve naturally through a branching process (Walker, 2003). There is no evidence plant species indulge in industrial espionage, for instance. Natural species do not “copy” from one another, although they may adapt in similar ways to similar conditions.

While Lamarckian evolution has been disavowed in biological accounts of species evolution (Jablonka, 2000), the notion that individuals strive for improvement within their ecosystem and directly pass on valuable traits to their immediate descendants is compelling in the technology sphere. As McKelvey (1996, p.21) points out, evolution proceeds quickly in the cultural sphere as knowledge is transmitted through the Lamarckian mode of evolution. This is not as straightforward as it seems.

Accepting the idea that there is variety and a selection mechanism at work in either a Darwinian or Lamarckian system, there are three issues that mark out a Lamarckian type of evolution (Kronfeldner, 2007): 1) The concept of directed change. 2) The direct inheritance of acquired characteristics. 3) The transformational pattern of evolution.

For Lamarck (1914), directed change arose from attempts to adapt to the environment and from the general tendency towards complexity. For Darwin, change was stochastic. Darwinian evolution depends upon chance mutations, genetic drift and natural selection. For our purposes, change is initiated by conscious problem solving by engineers, leading to variations in solutions to problems. The fact that technology is directed by firms allocating resources does not mean it is without its cul-de-sacs, reverses and failures. The technology watchmaker is not blind, but they can be wrong.

Secondly, direct inheritance is also a plausible account of the way technology proceeds. Technology is not an issue of genes: it is an issue of knowledge transmission. Individuals and design teams learn with experience and time.

Successive technology incorporates some of the changes of the past. The central thesis of the Stretch paper is the way in which existing plants can be modified over time in the light of operational experience to make extra output, or different products or use new inputs say. As innovations diffuse, they improve and these improvements, in turn, encourage further use. Metcalfe (1994) speaks of the joint effect of "diffusion through learning" and “learning through diffusion" as part of the dynamics of innovation.

Finally, evolution also carries connotations of smooth transitions and gradual adjustment. Yet the innovations studied here are disorderly, chaotic and opportune to modern eyes used to text-book accounts of formal project management. They were guided by chance and opportunity rather than roadmaps and scenarios (Saritas and Aylen, 2010). Information was often limited and uncertain. The designer of Blue Danube, Reg Milne’s selection of a hot-water bottle manufacturer in Barnsley to make a key component of Britain’s first nuclear deterrent is a clear example of opportune thinking. Again, the testing of Blue Danube went on long after it was in service and there were modifications to the fissile loading system once in service. This is not to deny Darwinian evolution cannot be characterised by dramatic jumps. Evolution as a process is sometimes characterised by “jumps” followed by periods of comparative “stasis” (Eldredge and Gould 1972; Gould and Eldredge 1993). This is spoken of as a process of “punctuated equilibria”, whereby the fossil record shows long periods of stasis, followed by bursts of rapid evolution – often in response to ecological changes. Rather we infer that transformational human effort is likely to produce more jumps than a random process. “Natura non facit saltum” is the epigram on the frontispiece to Alfred Marshall’s Principles (1890, title page). But, if nature does not jump – technology does.

Perhaps the clearest example of an evolutionary process at work is the development of the wide strip mill. Here is a clear process of “variety creation, selection and retention” (McKelvey, 1996). There are two rival schemes creating variety, the market clearly selected the developments of Columbia Steel at Butler since this formed the dominant design for all subsequent hot strip mills worldwide. The incremental developments of Armco were abandoned and their mill rebuilt. McKelvey et al (2015)

14 also distinguish between "blind variation" which arises from non-conscious actions, i.e. serendipity and chance, and "intentional variation" which emerges from deliberate actions. This raises the central responsibility of the entrepreneur. In the case of the wide strip mill, the key role of the entrepreneurs Townsend and Naugle was to consciously select the design features for development: four high stands, roller bearings, synchronised continuous flow through the finishing train, the use of coilers, a single reheating stage and so on. At Shotton, Richard Summers and Neville Rollason selected the technology, their suppliers and their supporting team to build their rolling mill with sharp entrepreneurial edge. Both wide strip mill examples show commercial acumen brought to technical development from the outset, which was not a feature in atomic bomb development.

There is also a puzzle posed by Schumpeter to the effect that you can add stage- coaches together, but you do not get a railway train. You can improve high-explosive bombs, but you do not get an atom bomb. You can increase the number of amplifiers in a series to improve the accuracy of an analogue trajectory calculation, but you do not get a digital system. You can put one mill stand behind another – as Armco did – but still not realise a continuous wide hot strip mill. These are radical, disruptive or drastic innovations. How do you mutate from one technological state to another?

The example of the Bloodhound control system clearly shows that reverse salients often force change. Derek Whitehead said he was forced to propose a digital solution to guided missile control because: "with the errors you could get in analogue terms at each of these stages I reckoned that there was no way that I could produce an analogue computer that would go out into the services field that would be accurate enough." In the case of ICI it was costs that drove a radical transition from analogue to digital. ICI appreciated that one central process control computer would be markedly cheaper than, say, a hundred three-term controllers distributed around a chemical plant. The wide strip mill was prompted by shortage of wide sheet steel for the all- steel car body (Nieuwenhuis and Wells, 2007). In the case of the atom bomb radical change was, indeed, provoked by scientific potential and the imperatives of war.

Untold Stories and the Role of Science Another theme of these papers is they report untold and unfashionable stories. They are subjects where there are few, if any, modern sources and no reliable entries in Wikipedia. They focus on practical engineering developments rather than science. At least two papers (Stretch, Blue Danube) focus on practical issues of manufacture, maintenance and use. They are told on the basis of a range of primary sources. This has incidental advantages: literature surveys can be brief and there are few secondary sources to be refuted. They are meant to start debates, not finish them.

As argued elsewhere (Aylen, 2010) “history often overlooks the mundane”. The case in point was process control computers. Academic papers on computer history emphasise scientific computing, technical breakthroughs such as arithmetic accuracy and processing speed. Whereas focus on the prosaic use of computers for automation gives a different agenda, emphasising widespread use, reliability, continuous operation in a hostile industrial environment, robust equipment and a different set of technical needs including data sampling, strict timing control and the use of interrupts to give priority to immediate calculations. Agendas differ significantly if you slightly alter the context of study.

Again, “The neglect of technical development of the atom bomb in the UK distorts our view of how technology develops. Accounts of the bomb focus on a polite story of intellectual effort and formal science rather than engineering and practical technology.” (Aylen, 2015). This article was a conscious attempt to rectify the bias of official history which emphasises the role of scientists and thinkers over technologists and practitioners. (e.g. Gowing, 1964). The bias in the UK nuclear narrative arises for two reasons. Firstly, the narrative is filtered through official secrecy (Gowing, 1978) and shaped by the nuclear state. It is an “official narrative” (Hogg, 2016, p.7). Secondly, it is a polite account which emphasises science and government decision making, rather than details of practical engineering and the reality of military deployment. As Scranton (2008) recognises:

"real engineering at the edge is a gritty process laden with fixes, errors, cursing, and painfully-incremental steps towards something that works, much less works reliably and safely. There's no romance in that, so a more marketable story has long been routinely fashioned".

In the case of atomic weapons, science was the heroic story - a myth which is still current. For instance, Farmelo (2013) only mentions the actual bomb once in his account of the development of the British Weapon (a point due to Keith Eldred, e-mail correspondence, 17th August 2015.) Nuclear culture resides as much with the hot- water bottle manufacturers of Barnsley who built key components for the first British atomic bomb as it does with the scientists at the laboratory in Harwell. As Carlson (1991, pp.4-7) emphasises “craft knowledge” and scientific knowledge are often intertwined in technical development, and the story of craft knowledge is often neglected.

Why study military technology? The focus on atomic weapons and guided missile control raises the question of why study military technology (Edgerton, 1988)? There is a pragmatic argument that military spending is important. It still accounts for 2 per cent of GDP in the UK and during the Cold War defence spending was a constraint on investment, growth and development of public services elsewhere in the economy (Geiger, 2004). Arguably, defence innovation has been neglected by liberal students of innovation, although left- wing scholars have done something to remedy this neglect (Kaldor, 1980).

Defence technology also opens a wider debate about the role of the State as an enabler of technology. Mazzucato (2013) argues only the State itself can cope with the high costs, huge uncertainties, extended time frames and long delayed pay-offs inherent in break-through technologies, which would include atomic weapons. (In any case, the state has a monopoly on violence.) The idea that the State is required to give order to capitalism goes back to Andrew Shonfield (1965) among many others. Government support for defence highlights Metcalfe’s (1994) point that innovation activities of firms depend upon a wide range of other institutions which supply knowledge, skills and markets for the efforts of individual firms. Foremost amongst these institutions is Government which supports education and skill acquisition, basic

17 research and the framework of law. These conclusions are consistent with Cohen et al. (2002) who find that public research (from Universities and Government Labs) in the USA is a crucial support to industrial R&D in a few industries and affects industrial R&D across much of the manufacturing sector, especially large firms.

John Pickstone (2000) argues military devices were developed in locations of “technoscience” where pioneering science, engineering and state-sponsored innovation coalesced. This is self-evidently true of Farnborough (Ellam, 1999-2000; Nahum, 1999). Again, guided missile development at Ferranti was based at a purpose built site composed of laboratories and final assembly and testing facilities at Wythenshawe, Manchester opened officially in June 1954 (MOSI, 1996). Ferranti at Wythenshawe was a location where state support and commercial goals combined to develop military equipment for government. Technical development at Wythenshawe drew on scientific knowledge and emerging technologies: "When the project started many of the branches of engineering we needed to use were also unknown." (Rowley and Metcalfe, 1968-9, p.6). So military spending was, at the same time, both pioneer risk taker and research funder in centres where knowledge accumulated. Such knowledge later diffused economy wide, through products (the Argus computer), through personnel and through the development of concepts pioneered by the military (digital control). Among the compelling examples of such military sponsorship of innovation is Silicon Valley (Leslie, 1993; Lécuyer, 2006)

Defence innovation raises two questions for the researchers: does military technology exhibit the same fundamental processes as civilian research and development, such as variety, evolution and selection? Basalla (1988, chap.5) argues military imperatives have led to the selection of technological innovations that have subsequently found a place in the civilian world. He cites the prosaic motor truck spurred on by the First World War and the case of nuclear power. This begs a second question, does defence technology generally spin-off into the civilian sector, or are these just well-chosen examples? We suggest here that military R&D exhibits the same evolutionary processes as civilian R&D. But the Bloodhound paper shows the notion of spin-off is

18 too simplistic and we need a nuanced understanding of the interplay between the defence and civilian economy.

Process Innovation: the advent of the wide strip mill Study of the wide hot strip mill is easier to justify. The continuous wide strip mill was one of the great process innovations of the 20th century, ranking alongside float glass (Bricknell, 2009), or oxygen steelmaking (Aylen, 1980). It made possible the mass production of cars, cans and consumer durables in inter-war America and in post-war Europe (Aylen & Ranieri, 2012). But, unlike these more familiar processes, the story of the wide hot strip mill has been neglected, apart from our recent book on the topic (Aylen and Ranieri, 2012) and a formidable technical literature (Ess,1941; Ess, 1970). There is one hagiographic account of the development of the wide strip mill which is simply propagandist and wrong (Borth, 1941, ironically titled “True Story” which attributes the development to a man named “Verity”.)

To add to the interest, development of the wide strip mill was a natural experiment where two sets of institutional arrangements competed side by side. Armco adopted a secretive, closed model based on in-house R&D. The successful team – Columbia Steel – adopted an entrepreneurial open-innovation approach, relying heavily on suppliers for technical advice and participation.

Sources of Evidence and the Issue of Secrecy These articles are based on three key sources: interviews with participants, archive material and access to artefacts. The archive sources were often unorthodox. Among the key breakthroughs in the open innovation study were working files maintained by a roller bearing manufacturer in Philadelphia, USA and boardroom archives kept by a rolling mill builder at Buttrio, near Udine in Italy who had acquired the business of United Engineering in the USA and shipped their records to Italy. The Shotton paper relied on access to the Record Centre maintained by Corus (as the firm was then known) at Shotton which maintains current records for the business such as personnel files and test samples. (One file in the archive related to this author – he did not read

19 it.) They coincidentally kept material which turned out to be of great historic interest, but again they are not a conventional business archive.

The author’s preferred approach to defence material is prolonged and repeat interviews with key sources. There are two reasons for repeat interviews: verification and depth of information. After a first approach to a respondent it is often possible to verify the interview accounts using archival material and other interviews. You can then return to an interview source to re-engage them in further discussion in the light of documentary evidence. It becomes clear that those claiming a key role without justification cannot sustain polite but probing questions during repeat interviews. In contrast, key participants offer ever richer technical detail, further contacts, previously undisclosed documents, sketches of design solutions and physical artefacts in support of their argument. Once a rapport has been established over a sequence of meetings, the respondent feels more relaxed. A number of respondents have been interviewed on three or more occasions for periods totalling twelve hours or more and provided follow up material through e-mails and written correspondence. Many of the respondents were of mature years. New evidence suggests that older people are the most reliable witnesses: less likely to lie and are more likely to tell the truth (Debey et al., 2015)

There is a danger in offering telling examples and “anecdotes” – however compelling (Silverman, 2001). We triangulated sources with documents and physical artefacts, where possible. Many of these documents come from unorthodox sources. Much of the written evidence is “grey material”. That is to say, material not kept in conventional archives. Rather it comes from garages, domestic studies, uncatalogued record centres and company filing cabinets. A surprising amount of secret material was retained by authors. This was seldom the case with “top secret – atomic” material which was consigned to a safe or guarded when in use. (Although one respondent was a rich source of photographs). Here, respondents offered specially written accounts and sketches reflecting their experience and designs. A key designer of Blue Danube is one such source. Interviews were an opportunity for participants to think- through and explain what they once did as a day to day occupation. Follow-up e-mails

20 often added precision to the verbal account, with names of personnel and dates clarifying their initial response.

The two defence articles draw on explicitly secret material. Bloodhound 2 was classified as “Top Secret” and Blue Danube “Top Secret – Atomic”. Both accounts involve self-censorship on the part of this author. In some cases, simply being told information meant it could not be published. This required obfuscation of the written account to protect sources without distorting the truth.

Artefacts also survive. In most cases, working technologies are relatively easy to examine. Thanks to the hospitality of steelmakers, the author has seen some thirty working wide strip mills in Europe, North America and Asia, including Shotton discussed here. But, in the defence field, access to this type of research evidence was given on a privileged basis. Visits to the Atomic Weapons Establishment, Aldermaston in Berkshire in July 2014 and October 2015 are examples. Material on Bloodhound was supported by access to Swiss Air Force Bloodhound 2 technology at Menzingen in Switzerland. Some physical evidence on Blue Danube was confirmed on the understanding that sources would not be traceable. Ultimately, this is an issue of trust on the part of contributors and the author.

Privileged access raises four concerns. There is a problem of verification since other scholars do not have an opportunity to refute what is said by this author. Secondly, it may be necessary to report facts known to the author without referencing sources. A wider concern is the author may be a “chosen mouthpiece” to place deniable information into the public domain. The author may well have been captured to enhance the narrative of the nuclear state. Finally, as we have indicated, there is an element of self-censorship and dissembling because the author is aware of the possibility of nuclear proliferation and reluctant to write on sensitive issues. As a result only part of the story emerges. However, this partial account reveals a more complete picture of the evolution of sensitive technologies than other archive based accounts.

Two underlying themes of the research deserve mention. The first is the generosity of those involved. In principle, it should not have been possible to write the account of Blue Danube. In retrospect, it is an understatement that;

“Atomic bomb design is an unfashionable, even unwelcome topic. The technology is complicated. It is shrouded in secrecy. Records are fragmented across Government Departments, or incomplete. Some archive records have been withheld, or subsequently withdrawn.” (Aylen, 2015, p.32).

However, the cooperation of 26 respondents – including three key designers and many unacknowledged sources – allowed the story to be told. On an emotional note, the generous help of so many people is very moving.

Secondly, there is an element of luck in research. The existence of detailed material on the collaboration between United Engineering and the Hydraulic Tool Company was exposed by a chance e-mail to a successor firm on the outskirts of Philadelphia. An e- mail from a well-wisher pointed me towards the role of the Royal Aircraft Establishment in nuclear weapons design and suitable contacts. The Blue Danube research was encouraged in the first place by a throw-away remark in a pub. The story of the Ferranti Argus was built on chance remarks by two former employees of the company when I was unaware of their roles in the project.

2. Historical methods in the study of technology

According to Derry and Williams writing in 1960:

“Since technology comprises all that bewilderingly varied body of knowledge and devices by which man progressively masters his natural environment, its history is a subject with wide and ill-defined ramifications”.

While this is comprehensive definition – though excluding the contribution of women (Jaffé, 2003) – it is not helpful in clarifying the current state of the discipline.

Since then, the history of technology has developed as a contested field with many participants. As David Edgerton (2010) says, “The study of technology is clearly not confined to self-proclaimed historians of technology”. There are many approaches to the appropriate subject of study.

Engineers are often concerned with the origins of a technology, key inventors and early instances of adoption. The clearest statement is from a President of the Newcomen Society for the Study of the History of Engineering and Technology :

“. . the history of technology comprised the study of the way technical ideas developed, how each stage of development is related to what has gone before, how new ideas are generated, how understanding gradually unfolds of how things work: the accurate description of how machines, devices, materials, are or were designed and made: the engineers and innovators themselves, and perhaps also the way they built up industrial organisations.” (Tucker, 1979, p.197)

We might add engineers often have a normative concern to forecast future developments on the basis of past experience (Bugliarello, 1979). Foresight scholars often turn to the past as a guide to the future (e.g. Cowan and Hultén, 1996). This is not to say that technology is an impersonal force with its own logic and trajectory that humans are condemned to follow as some determinists suggest (Heilbroner, 1967). Although Ceruzzi, (2005) reminds us the exponential advance of computer power does seem impervious to social, economic or political constraints.

By way of a caricature, it can be argued historians of technology represented by, say, the Society for the History of Technology are more concerned with the social impact of technology and the way in which technology is shaped by society (Bijker et al., 2012) – sometimes to the neglect of the material objects themselves. But the Social Construction of Technology has played a key role in drawing attention to the role of users in building technology, often as co-constructors (e.g. Von Hippel and Von Krogh, 2003).

Innovation scholars traditionally placed a strong emphasis on diffusion (Metcalfe, 1970; Stoneman, 1976; Aylen, 1980) and the research and development process, often with a view to promoting technological innovation as a policy tool for governments (Godin, 2016). With a few prominent exceptions, scholars neglect key areas of the discipline, including: technological failure (apart from Petroski, 1985); the persistence of old technologies (apart from Edgerton, 2006); and the decline of existing technologies (Pursell, 1995). Innovation is also strongly identified with masculinity (Pursell, 1995). Both the Bloodhound and Blue Danube studies showed women in key roles.

Ironically, historians are seldom represented among historians of technology. Questions such as: “what has been the impact of computers on 20th century British Society?” have been largely neglected (with the exception of Agar, 2006).

Our central concern here is the innovation process rather than, diffusion, use, social impact or social shaping. We seek to understand how technology was initially developed and transferred into early use. We can be accused of focussing on novelty and on radical shifts in technology (analogue to digital; conventional to nuclear weapons; batch process to continuous production) rather than looking at a gradual process of adoption, acceptance and rising living standards. Arguably, historians are too fond of sharp transitions at the expense of continuity (Burstyn, 1979, p.68; Agar, 2006). In many respects, this research is at the heart of the history of technology, using historical examples to demonstrate the nature of technological change (e.g.

Rosenberg, 1994) and accords closely with innovation-centric views of rising living standards over time (Landes, 1969; Pavitt, 1980).

We do not accept the distinction made by Edgerton (1999) between innovation and the history of technology. He emphasises the importance of the use of technology as being fundamental to the history of technology. It is certainly arguable that the history of technology focuses too much on the invention and development phases of technology, to the exclusion of users, especially the exclusion of those who are not empowered by governments or firms to make changes. But, the paper on “Stretch” shows that innovation continues during use, even for capital intensive equipment. The narrative of use is also a story of improvisation, adaptation and improvement which is central to the innovation process. Edgerton’s (2006) own book Shock of the Old focuses on the continuing adaptation and renewal of old technologies in use.

Historical Method in the Study of Technology A second contested area within the history of technology is methodology. The conventional approach is to use archives for information (Agar,2003). Sources such as the National Archives provide definitive information with a thirty year time lag. However, there are key shortcomings to a reliance on archives. There is an undue focus on elites and Government decisions. The National Archives report the activities of Civil Servants. Company Archives often focus on Board level decisions. Everyday designers, manufacturers, buyers and suppliers, support technicians, operators, programmers and maintenance personnel are overlooked.

Our approach here might be described as “history from below” (Lundin, 2012, p.20). We use ordinary sources: oral history, informal archives, the trade press, engineering journals, technicians and accounts from hands-on engineers - rather than scientists. These techniques are often used to raise awareness of powerless groups: in discussions of labour history, gender and ethnicity. These oral stories have low status in history and are mostly used to vindicate minority groups.

The power of oral history is that it reveals unknown stories. They may also contradict the official account. For example, archives show the leading aircraft designer Barnes Wallis was present at an early design meeting on the Blue Danube. (He was ostensibly representing Vickers.) It might be surmised he played a role in the design of the first atomic bomb. Yet one interview respondent present at that meeting reported in detail how Barnes-Wallis was then excluded from the meeting by the all-powerful chairman – Dr William Penney, a piece of information which completely alters the interpretation of the archive.

There is a third contested issue of scope. Hendry and Harborne (2011) point out there is a need to look at the “entirety of events”, including inconvenient phenomena. A longer time scale brings a changing perspective. In their case they were arguing Garud and Karnoe’s (2003) account of the supremacy of the Danish wind-turbine industry was less compelling if the story was brought up to date beyond the year 2000. Edgerton (1999) points out that Bijker ends a study of the bicycle in 1890; of Bakelite in 1920; and of fluorescent lighting in 1945 (Bijker,1995).

Truncating the Blue Danube study weakened the strength of some of the observations. Subsequent weapons use the same implosion principle for the core in miniaturised, albeit redesigned form. This implosion principle has again been selected as the primary in the first stage of a modern warhead. The development of Blue Danube, through to the smaller Red Beard, and ultimately the Polaris and Trident warheads illustrates evolution at work. As Pyne (2016) implies, Blue Danube set UK warhead design on to a plutonium trajectory.

Difficulties of Oral History in the History of Technology A number of these papers rely on oral history interviews to obtain data. Indeed, in many cases they are built around repeat interviews with key personnel involved in the original project.

Oral history interviews face a range of issues: access, reliability, ethical considerations and sample selection. They place a burden on the interviewer to be mobile, well

26 prepared and supportive, but sceptical. These issues have been rehearsed in medical literature (e.g. Karnieli-Miller et al., 2009). The broad idea is to build rapport and promote disclosure. This was often fostered pragmatically through buying respondents lunch. Follow-up interviews usually took place in respondents’ homes, which is an ideal non-threatening environment.

Given the deep involvement of respondents, they become not just a source of information, but also a participant in the research process. This is particularly the case in repeat interviews where we re-engage the respondent to help promote accuracy and to discuss new findings from documentary archives and complementary interviews. One sign of participant involvement is the substantial follow-up correspondence which occurs. Respondents often locate key documents after interviews and provide follow-up contacts. In some cases, the respondents have published their own technical accounts (Gribble, 1998; Allen,1999). Here the process of oral history comes closer to action research where the respondents supply part of the thinking behind the results (Bradbury and Reason, 2003) and join a wider community of enquiry as they call upon friends for clarification, confirmation and further knowledge as they reflect upon what they have said. The distinction between interviewer and interviewee is blurred. This is reinforced by the formidable intelligence of most respondents in these fields – they are, after all, quite literally “rocket scientists”.

This process of respondent participation is the converse of usual interview relationships where the interviewer is often argued to have the power and knowledge. In three cases it was clear that the respondent had used their contacts to vet the researcher for security clearance before the interview. These respondents could have stopped the whole study (perhaps the researcher’s career). Here was an interview programme where respondents possessed the power.

The close rapport established between researcher and respondent raises issues of detachment and ethics. It is difficult to maintain professional distance when respondents become personal friends and facts can be checked with an impromptu

‘phone call. There is a high level of participation long after the initial interview. One measure of the success of the research is not just getting an initial story, but also the process of repeated verification to test the detailed accounts given. The First Waltz paper was given at the Defence Academy, Shrivenham; to senior personnel at the RAF Club in London; and the private Charterhouse Conference among other venues. This was high level verification.

There are ethical principles that need to be adhered to. But, we emphasise: these ethical principles are a guide. Otherwise, they would prohibit research on topics such as nuclear weapons. Fortunately, others have already trodden this path including Cocroft et al., (2003) and Bud and Gummett (1999). By chance, we interviewed one of Mackenzie’s (1990) respondents from his study on missile accuracy. The initial help of the Ministry of Defence under the Freedom of Information Act was influential.

The first principle of oral research is that the aim of the study is made clear to the participants and the ultimate goal of publication is conveyed from the outset. This is crucial where there could be a technical breach of the Official Secrets Act. Timely reminders during the interview and follow-up discussions on disclosure helped here. Ironically, transcripts sent to respondents led to deletions of colourful language rather than redactions of official secrets. A vivid description of the AWE Burghfield site was lost through a respondent’s self-censorship.

Secondly, there was no obligation to participate. Respondents were keen to “tell the story”. Thirdly, there is an element of self-censorship in research disclosure, even where participants may have given information. The Blue Danube account does not discuss the fissile core of the weapon for proliferation reasons. The Bloodhound paper does not mention any details of the continuous wave radar. Neither study mentions criticisms of fellow researchers since they might be attributed to a respondent. So some research findings are omitted, but the overall conclusions can be validated from other evidence which is disclosed. In this sense, giving participants a key role does modify and limit research outcomes.

In both the Blue Danube and Bloodhound studies the respondents were offered anonymity, in which case they chose to be a “private source”. All respondents saw the draft papers before publication and were allowed to retract their names. Instead, at this late stage, two well informed “private sources” explicitly asked to be named in the Blue Danube study.

There is a problem of sample-selection in oral history interviews due to mortality and gender. Fortunately, many of the leading participants had survived in the Bloodhound and Blue Danube cases, as far as we can ascertain. But anno domini precluded interviews with the head of the Bloodhound team, Denis Best, and the head of Blue Danube development at ARL, the enigmatic Hugh Francis. One difficult issue is gender because of name change on marriage. We traced female programmers of Argus – notably Scilla Bretscher - through social connections, but the female designer of Blue Danube who worked in the drawing office at ARL, Brenda Adams, remains elusive.

One further stumbling block to oral history is that it was incumbent upon the interviewer to learn the detailed techniques and language of the sector concerned, given the close technical nature of the interviews. There are evident sensitivities relating to knowledge of the construction of an atomic weapon. Arguably, wide strip mills are more complicated if not so sensitive.

3. The nature of the innovation process

Innovation, Collaboration and the Growth of Knowledge

Ultimately, innovation is a process by which knowledge – new and old – comes to be embodied in new technologies (Rosenberg, 1994). But this statement conceals a complicated set of interactions. As Malerba (1992) emphasises firms learn in the realm of production, design, engineering, organisation and marketing, as well as formal R&D. Secondly they not only rely on internal know-how, but also absorb knowledge from external sources.

A range of analytical approaches have attempted to pin down these relationships between knowledge and technology. Neoclassical economics lays emphasis on factor endowments, relative prices and potential profitability as drivers of innovation – scarcity as the mother of invention (e.g. Nelson & Winter, 1974). But the structure of neoclassical theory is silent about the actual mechanisms which translate profit incentives into usable artefacts and services. Indeed, growth theories treat technological advance as a deus ex machina: a theatrical device to sidestep our ignorance of the innovation process. Instead, we appeal to the contribution of evolutionary economics associated with Schumpeter (1943, ch.7) and Nelson and Winter (1982) and highlight features of evolution from our case studies.

The Emergent and Distributed Nature of Knowledge

In both the case of the wide strip mill and the atomic bomb, there is a pattern of emergent problem solving as each obstacle is tackled to produce a successful device. This process of improvisation has been called “bricolage” (Garud and Karnøe, 2003; Baker and Nelson, 2005), the idea that innovation builds on the efforts of many participants using resources and networks close to hand. It is also related to the notion of “proximal innovation” where users call upon local resources to modify designs to suit their local needs (Usenyuk et al. 2016) But we prefer “distributed innovation” (Howells et al, 2003).

Distributed innovation describes a process whereby sources of knowledge and skill and the development of necessary artefacts are spread around a network of suppliers, customers and regulators. In the Danish wind turbine case discussed by Garud and Karnoe (2003) the actors were the designers and producers of wind turbines, the owners and users of the turbines, the Government sponsored test and research centre and the Danish Government regulators themselves. In the early stages of development, the wind turbines developed in small steps. Multiple sales of small turbines gave widespread learning opportunities and feedback from users. By a process of cooperative shaping, the local owners, the regulator and the producers worked together to develop better designs, helped by a strong component sector familiar with agricultural machinery. Later commentators have also pointed to the role of strong state support and formal R&D programmes in the development of Danish wind turbines (Hendry and Harborne,2011). This process of distributed innovation shows many similarities to those seen in medical innovation where no one surgeon is responsible for a breakthrough. Rather, the knowledge accumulates across medical teams until an agreed procedure is developed (Ramlogan et.al. 2007)

There was also learning within networks in the case of the development of the wide hot strip mill at Columbia Steel. The network of knowledge included the plant supplier United Engineering who had learned about the use of four high stands through supplying a mill to roll copper to Rome Brass & Copper Company. The bearings know- how came from the Hydraulic Tool Company. The electrical machinery supplier Westinghouse knew how to synchronise stands using Ward-Leonard controls to supply DC current to the variable speed drives. Key individuals acted as knowledge hubs for these suppliers, for instance F.C. Biggert and Lane Johnson at United and William Messinger at the Hydraulic Tool Company, the bearings designer, who negotiated with Harry Naugle and Gene Townsend at Colombia Steel. Even at the time, the key role of these engineers was acknowledged. William Messinger’s grandson claims he was the second highest paid engineer in America when he was helping develop the Butler mill (personal communication, 15 July 2015). But this pattern of cooperation to mutual advantage is not unique to steel. A similar process of cooeperation underpinned the expansion of capacity at Heathrow airport (Tether and Metcalfe, 2003).

Entrepreneurship The paper on open innovation at Columbia Steel, Butler begs the question, how did Townsend and Naugle recognise the technical opportunities presented by roller bearings in combination with four-high stands? As Shane (2000) points out: how can there be entrepreneurial opportunities if such technical knowledge is readily apparent to any knowledgeable observer? Instead, these two enterprising collaborators spotted the potential of an engineering breakthrough and a lucrative opportunity supplying the burgeoning US car body market. Yet the same principles eluded Armco who were technically well informed and had a large R&D Department. Armco at Ashland persisted with conventional brass friction bearings and two and three high mill stands. Clearly the opportunities were not obvious.

This divergence between Townsend and Naugle at Butler and Armco at Ashland illustrates a point central to Hayek (1945) and Kirzner’s (1997) work that different individuals have different information. It also highlights Townsend and Naugle’s willingness to act on this information by combining resources drawn from United Engineering, the Hydraulic Tool Company, Westinghouse and finances from the Mellon Bank of Pittsburgh. It is easy to see United’s motivation as a plant supplier as they could foresee a completely new market for wide strip mills, even though it would destroy their existing market for individual hand mills.

We know that Townsend and Naugle were not emotionally attached to their breakthroughs at Columbia Steel as they sold out to Armco, signed a non-compete clause and went on a world cruise. Or, as Time Magazine (1933) reported:

“American Rolling Mill (whose experiments along the same line were well advanced) finally bought the plant and patents, and Messrs. Naugle & Townsend promised to stay out of the steel business for five years. With $2,000,000 as their share of the profits, they departed on a leisurely trip around the world.” (Monday, June 19, 1933)

The same pattern of supplier selection and cooperation is evident with the first British nuclear weapon, although suppliers were not always aware of the final use of their

32 product. It is apparent that the Royal Aeronautical Establishment had close contact with suppliers such as Hudswell-Clarke, Frazer Nash and William Freeman of Barnsley. Again, it is possible to name the individuals who acted as knowledge hubs - Reg Milne and Peter Barker. In an interview, Reg Milne described how he approached two firms for support on the butyl rubber airbags surrounding the implosion sphere of Blue Danube.

“Dai Williams suggested air bags. I chose the rubber. I went to two firms, India Rubber, Gutta Percha and Telegraph Company where London City Airport is now and to a firm up in Barnsley. Now they made Suba-Seal hot water bottles amongst other things. They were a lot more accommodating and competent we thought. We settled on them.” Reg Milne (interview 6th August, 2013 Farnborough, Hants)

The parallels with medical innovation are evident here too, the combination of science, technology and practice that drives innovation forward.

The Relationship between Science, Technology and Practice

Our understanding of the complex relationship between science, technology and practice has advanced, largely through the process of detailed case studies. A century ago, the President of the American Institute of Electrical Engineers, John J Carty, (1916, 486) who was also Vice President of AT&T and Chairman of the Board of Bell Telephone Laboratories exhorted:

“the engineering student should be taught to appreciate the ultimate practical importance of the results of pure scientific investigation and to realize that pure science furnishes to engineering the raw material, so to speak, which he must work into useful forms”

He went on to advocate that engineers should observe a “pure scientist” at work to further their training after graduation. The prevalent view that science was the engine of technology found its apotheosis in Vannevar Bush’s (1945, summary) advocacy of science: “Progress in the war against disease depends upon a flow of new scientific knowledge. New products, new industries, and more jobs require continuous additions to knowledge of the laws of nature, and the application of that

knowledge to practical purposes. Similarly, our defense against aggression demands new knowledge so that we can develop new and improved weapons. This essential, new knowledge can be obtained only through basic scientific research.”

This caricature of the science-technology relationship has been dubbed the “linear model” and has come under attack from historians of science and technology (e.g. Scranton, 2006; Godin, 2006), although there have been attempts to rescue the idea (Balconia et al.,2010) and it still guides policy (McKelvey, 2014, p.75). Krige (2006) just sees scientific knowledge as one resource in technological problem solving. He points out that the process of innovation is not self-contained but depends upon a social context “that can doom it to failure or bring it to fruition”. So, simple models of science-technology relationships removes human creativity and motivation from the process. The debate ignores the animal spirits of the Schumpeterian entrepreneur (e.g. Dodgson, 2011) evident in the Open Innovation paper here.

There are numerous examples of technology progressing with little input from science. Buchanan and Watkins (1976, 124) show the efficiency of the stationary steam engine rose from consumption of 32lbs of coal per horsepower in 1712 to a mere 1½lbs by 1885. As much as anything, this was due to fierce competition among suppliers who offered a wide variety of designs. Taking one instance, textile mill engines were a major sub-sector of mechanical engineering between, say 1850 to the mid-1920’s. Buchanan and Watkins (1976, 114-120) list 28 makers of large mill engines, each with their own design teams. There were perhaps 30 firms all told, with an employment of 20,000, each with their own drawing office and machine shops, many with foundries and forges and each had their own distinctive designs and sub-sectors of particular strength. The result was an extraordinarily wide variety of designs and detail (Watkins, 1970, 1971). The selection mechanism was the mill owners, their architects and engineers, each seeking the most efficient solution for their particular niche in the spinning and weaving of textiles. The discovery of thermodynamics was an irrelevance. Donald Cardwell (1994, 314) concludes his evaluation of the steam engine in the nineteenth century: “In short, advance was evolutionary and along much the same lines as had been the case with the Newcomen engine in the eighteenth century and the Cornish

pumping engine of the early decades of the nineteenth century. Advance, while rapid, was achieved empirically.”

Mike Gibbons (1984, 99) sums up the position neatly when he points out that Carnot’s pioneering theoretical work on heat engines was prompted by existing steam engines – “it could be said that in this case, technology presented a problem for science”.

Our studies found little evidence of scientific input. Ultimately, the atom bomb seems the creature of science (Gowing, 1964, 1974). Yet the Manhattan project was essentially large scale chemical engineering (Hughes,2002). Alex Wellerstein (2013) shows 80% of the money spent on the Manhattan project went on the plants necessary for producing fissile materials at Oak Ridge (enrichment) and Hanford (plutonium production). Only 4% went towards research and development at Los Alamos. The Manhattan project may have inspired Vannevar Bush to write “Science the Endless Frontier”, but ultimately it was a lop-sided example. Science was necessary, but nowhere near sufficient to make the atomic bomb. Ultimately, the atomic bomb was a triumph of chemical, mechanical, electrical and military engineering. The official Smyth report (1945) only gives sketchy details on the industrial processes as these were not only secret, but also potentially of great commercial value to the companies involved. As a result, the official account displays extreme bias towards science. It was a convenient narrative.

Apart from a secret memorandum from Sir William Penney (AVIA 65/1156, 1953) there is little evidence scientists had much practical influence on atomic bomb design in the UK. Again, Ferranti Automation at Wythenshawe had no contact with the University of Manchester when developing the Argus computer, unlike their estranged colleagues at Ferranti Computers in Gorton. Certainly the wide strip mill developed through practical engineering.

As Mike Gibbons (1984) argues, the significance of science for innovation depends upon your prior choice of model. Criticising Project TRACES (Illinois Institute of Technology, 1968) he points out that the study itself is predicated on the disciplinary structure of science and the contribution of each sphere of science. Whereas a wider

35 industrial innovation study would point to the interaction of technological opportunity and market need and focus on what is necessary to achieve the development of a marketable product.

The idea is explored further by Fiona Murray (2002) who subtly shows that while the social structures of “science” and “technology” are often independent, scientific and technological ideas do in truth co-evolve. She studies an emerging area of biomedicine—tissue engineering and finds evidence of distinctive scientific and technological networks. However, “co-mingling” exists not where you might expect to find it in areas such as papers and patents, but through practical avenues such as founding of companies, licensing, consulting and advising.

What is the Nature of Technology ? Science, Technology and Practice Metcalfe (1995) emphasises “Technology can be treated in terms of knowledge, skills and artefacts and in each case there are different variety-generating mechanisms, different selection processes and different institutional structures. For policy purposes, the degree of connection between these different dimensions of technology is at the core of technology policy.”

This distinction between knowledge, skills and artefact finds many echoes. Various authors argue that artefacts tell their own story (Hoke, 1990; Staubermann, 2010), that they embody the technology and skills of their time and can be read like documents. Pickstone(2000, p.37) reminds us that we read every day “messages” into events such as illness, and the argument can be extended to artefacts. An extreme version of this view is Latour (1992) who argues artefacts are often “inscripted” by their designers to direct users, exemplified by the world of caravanning (Southerton et al., 2001)

The evolution of technology in use While innovation studies have been strong on invention, innovation and diffusion of new technologies, the story tends to stop there. Little attention is paid to the use of technologies and their subsequent modification and adaptation. We have tried to look at use in two ways: the routines and procedures which govern sensitive military

36 technology and the adaptation of technology with experience – the concept of “stretch”.

The research here shows strong evidence of trajectories (Dosi, 1982). Development of the wide hot strip mill has been driven by a combination of economic incentive and technical opportunity (Aylen, 2001 and 2010) since the success of Butler in 1926. But the fundamental layout of reheat furnace, roughing stands, finishing stands, run-out table and coiler remain the same. Rapidly growing demand and the imperative of scale economies drove strip mills to higher and higher outputs across Generations I to III between 1920 and 1980. Recall, the dilemma at Shotton was how much capacity to install. The energy crises of 1974 and 1979 forced a rethink in mill design. The shock of higher energy prices, coupled with slower growth in steel consumption and constraints on finance led to pressures for cheaper, more energy efficient mills. This need to cut capital outlays and reduce specific energy consumption prompted the emergence of Generation IV hot strip mills in the Far East during the 1980’s. The need to save energy and cut capital costs also drove the industry towards thin slab casting and mills which obviated the conventional roughing train for thick slabs. Thin slab casting was further encouraged by the opportunity it gave new firms to enter the industry for small capital outlay, especially in the USA and Italy. Direct strip rolling from thin slabs was taken up by conventional integrated steel firms in Europe. These Generation V mills offered a route to new products, including thinner gauge hot strip and ferritic rolling – although ferritic rolling proved to be a technical blind-alley.

The term “generation” was coined for successive waves of technical development in strip mill design and operation, no doubt due to the classic volume by Ess (1971). In the same way, we speak of “3G” or “4G” mobile ‘phones. Common rail diesel injection systems have been consciously named Generation 1, Generation 2 etc. with each stepwise increase in pressure (Catania and Ferrari, 2012). The notion of generations of technical development captures the way technology evolves over time and then “jumps” in response to technical opportunities and economic incentives before settling down to another evolutionary phase. Economists speak of “trajectories” of technical development – the way in which technology pursues a given path. Generations I to III

37 were steps on a path to higher output and lower costs per tonne. Weiss (1978) points out that technical innovation in mill design can be viewed as breaking a sequence of bottlenecks. Each of these generations overcame a set of constraints on the path to heavier coil weights and faster mill speeds. Generation IV and V represented a shift in direction towards lower capital outlay per tonne and less energy use. But within each generation there is constant evolution. Similar patterns can be seen in the design of nuclear warheads where both yield and miniaturisation were pursued.

Trajectories and Lock-in It is argued that once a technology has been selected, society as a whole gets locked in to early technical choices and it is hard to escape “lock in”. Malerba et al. (2001) argue the early computer industry showed strong dynamic increasing returns both in terms of technological capabilities and customer lock-in to the brand. Technology evolves within a paradigm or trajectory and a dominant design is selected. Worse still, better technologies or subsequent technologies may be “locked out” by the “dominant design”. Examples include the selection of the petrol engine for motor vehicles to the exclusion of electric power during the twentieth century (Cowan and Hulten, 1996) and the choice of water moderated reactors for power generation as a result of their selection for early nuclear submarine design (Cowan, 1990).

The papers here provide various examples of lock-in, both widespread and subtle. The most striking is the lock-in to plutonium in UK nuclear weapons design highlighted by Pyne (2016). Blue Danube was designed around a fissile core of plutonium due to the impracticality of enriching sufficient uranium given the time pressures and resources available. The implosion design was then transferred to the successor weapon, Red Beard and formed the primary of subsequent fission-fusion weapons in modified form.

Hot wide strip mill design has replicated the Butler model across successive generations of technology. The basic principles of reheating, roughing stands and four high finishing stands through to coiler have prevailed worldwide for ninety years, despite shifts in trajectory induced by market expansion and then by the energy crises and the need to make more parsimonious use of capital. Direct strip casting, for

38 instance by twin roll casters, was essentially “locked-out” despite its initial promise (Bessemer, 1865) and the emergence of commercial scale prototypes in the late 20th Century.

Escaping Lock-in

Cowan and Hultén (1996, p.61) suggest there are six ways to escape lock in: "crisis in existing technology, regulation, technological breakthroughs, changes in taste, emergence of niche markets, and new scientific results". In truth combinations of factors are often at play. Hand rolling on single stand strip mills could not produce sufficiently wide strip for the rapidly growing US car industry (a crisis in technology) but it required the technical breakthrough of the roller bearing to resolve the reverse salient in development. In much the same way, problems of accuracy and drift caused a crisis in missile guidance systems, but it needed a switch to digital techniques to resolve the problem.

In essence, the Stretch paper shows that lock-in is not the constraint that it appears to be. A central message of this paper is there is considerable scope for modifying many apparently fixed industrial process. The focus of the article is the frequent modification of post-war rolling mills built to established American “Generation 1” designs. We have written elsewhere of an ethylene cracker, Olefins-6 at Wilton, shifting from naphtha to ethane, freeing the chemical plant from lock in to a costly feedstock to a new, lower cost source of supply (Aylen, 2017). Admittedly the Stretch paper does not discuss the limits of modification. It might be more appropriate to mechanical technologies than to process plant. It is not technically possible to convert an advanced gas cooled reactor to a pressurised water reactor, for instance.

Omissions from this research: the Ghost of failure

This research is subject to selection bias: it is largely about success. Successful development of the wide strip mill in Pennsylvania; successful transfer of US technology to the UK at Shotton; successful imitation of the US atomic bomb;

39 successful use of digital control in guided weapons and process plant. Admittedly, Armco failed to produce breakthrough innovation at Ashland. We have not studied the short-live Thunderbird surface to air guided weapon which was a rival to Bloodhound 2, but was dropped after 18 years service because its clever but limited analogue system could not be modified, updated or “stretched”. The short life span of Thunderbird was however stretched by adopting the Bloodhound’s radar system.

As Vincenti (1994, 32) reminds us, we know the outcome of events as historians of technology. This colours our interpretation. The outcome to technological problems often looks obvious in retrospect. When studying evolution we need to consider not just the emergent successes, not only the failures, but the “also-rans” that didn’t quite make it.

4. Lessons from Innovation Research

We have focussed on the evolutionary nature of much technical progress. Our research thread also offers three practical insights into the conduct of research to develop new technology: parsimony in the use of resources; the importance of communities of practice when developing new technology, (even though this is contradicted by development of Blue Danube); and the role of product champions.

Parsimony in the Use of Resources Parsimony in the use of resources is one key theme runs through these papers from the construction of the Shotton strip mill, through the development of atomic weapons via Stretch. This is summed up by the insight from Peter Barker (e-mail, 11th August 2014, cited in shorter form in Aylen, 2015, p.15)

“With no wider overall Project input, the view of several of us in ARL Drawing Office was that it was inconceivable that ARL's small scale activity could possibly be a major part of the development of the British Nuclear Deterrent. Whilst perhaps it may actually have encouraged our determination to be the best, we joked that somewhere there must be a large, brilliant, team - with boundless resources - doing 'the real job', and that we were just a decoy.”

It is also evident in Richard Summers’ statement on the transfer of US strip mill technology to the UK quoted at the beginning of the Shotton paper (Aylen, 2008):

“The mill was to be a credit to those who built it, and to those who were to operate it. It was to be the best man could devise, but all extravagance and waste were to be avoided.” Richard Summers (1940)

Apparently Richard Summers would say “why spend sixpence when five pence halfpenny will do?” (Interview with Bill Summers, his son, Swansea, October 31st, 2014).

It seems that parsimony can be a spur to innovation. Massive deployment of resources may not always be the way to achieve progress in technology. Evidence of declining productivity in the pharmaceutical sector suggests that huge programmes of research are subject to diminishing returns. Scannell et al., (2012), argue “the ‘throw

41 money at it’ tendency” is one key reason for the secular decline of efficiency of R&D in pharmaceuticals. They attribute the rise in R&D spending to a number of factors including good returns on investment in R&D for 60 years, a poor understanding of the innovation process, the intense pressure to be the first mover in a winner-takes-all race and a bias associating success in large companies with the size of one's budget.

In many respects, the Stretch paper is the ultimate expression of parsimony – making the most of expensive capital equipment you have already built by modifications to produce new types of output, use alternative inputs, make more and higher quality products and perform more efficiently. The paper highlights resolving bottlenecks as a spur to innovation. Stretch is a process whereby a factory evolves through solving a sequence of problems, very similar to progress in the medical sector where treatments evolve as “problem sequences” resolving one shortcoming in treatments after another (Ramlogan et al., 2007) until a satisfactory procedure is achieved. Only in the medical field the solutions involve the coordination of widely dispersed actors, institutions and suppliers which weed out less satisfactory solutions and practices over time (Mina et al., 2016).

In the case of the wide strip mills at Butler and Shotton, efficient resource use is driven by entrepreneurial leadership by those with key stakes in the business. In the case of Stretch, competitive pressures bear down on firms in cost driven process industries. The Bloodhound case is akin to medical innovation in the UK where commercially orientated suppliers meet public needs and public administrators impose cost pressures and incentives as part of the procurement process. But the process is moderated by the pressing need to resolve problems and shortcomings in existing technologies.

Communities of Practice as a Culture for Innovation

In early stage research projects, engineers often coalesce around a particular design issue, recruit fellow engineers with related skills to the team, solve the problems in hand and then disperse upon completion. These loose assemblies of key engineers

42 and technicians have been called communities of technological practitioners (Constant, 1980, ch.1) or communities of practice (Wenger, 1998; Wenger and Snyder, 2000). Within these communities there is heavy reliance on tacit knowledge, intuition and experience coupled with formal analytic procedures. This highlights the fact that engineering design is an innately social activity (Law & Callon, 1988; Bucciarelli, 1988). Design development is a contact sport. As Cardwell (1979, p.8) stresses, “verbal communication seems to have played a much larger part in the transmission of technological information than it did of science”.

There has long been a tension in R&D Management between those advocating tight control of R&D projects and those wanting freedom for creativity and invention. Pioneer authors Burns and Stalker (1961, ch.6) observed that a mechanistic management system with hierarchical control is appropriate to stable conditions. Whereas an organic form of management is appropriate to changing conditions characterised by fresh problems and unforeseen requirements. Duncan (1976) suggested organisations needed to be ambidextrous to cope with both exploration and exploitation of innovations.

Evidence suggests participative teams perform best in terms of R&D creativity (Thamhain, 2003). Leitner’s (2009) survey of 50 industrial innovations in Austria shows that emergent, self-organised processes were a feature of a significant proportion of successful developments. Their conclusion is that interaction of deliberate managerial processes and sometimes chaotic local environments are often central to the success of developments, reinforcing findings of Koch and Leitner from the semi-conductor industry (2008). As Rowley and Metcalfe (1968-9, p.6) say about Bloodhound Project Management: "lessons emerge from this. The first is that, when you need invention, you should not hang too much ‘management’ round the project. What is needed is a close knit inventive team with freedom to manoeuvre.” We emphasise three aspects of communities of practice: social learning, practice and cohesion.

These informal communities are the place of learning and understanding about a problem and its possible solutions, often characterised by high creativity (Delemarle and Larédo, 2008). They are a way of working that allows engineers to share skills and

43 knowledge. Communities of practice offer a more realistic account of how development engineering progresses. Organisation charts show formal teams assigned to work on sub-assemblies against formal deadlines. Reality is less tidy. Small communities form to pool knowledge and solve problems and, as a result, creative solutions emerge, sometimes in a haphazard ways. Finally, communities of practice are transitory but coherent groups that are joined by common purpose, shared gossip, and war stories rather than formal job descriptions. Humour, not hierarchy, bonds R&D workers together around a common goal. Shared goals, easy familiarity and common language also make for rapid communication over emerging problems. Groups often self-assemble to share knowledge. Once a problem is solved, members re-form to move on to the next obstacle to development. So these communities are temporary, merely persisting until a problem is solved.

This joint Ferranti/ICI community emerges as democratic and consensual with clear recognition of the respective talents of participants. It was said that Syd Evans was selected by his fellows to lead the ICI task force from Ferranti’s side. They were subject to little, if any managerial direction and conformed to group norms rather than, say, working hours laid down by the job: “It was an invigorating environment, not a 9 to 5 job. . . . The view was ‘we have interesting problems’.” [Interview Evans, 25 Nov 2010]

These communities coalesce around solving shared problems. What captures interest among a technical group is a situation where participants do not immediately understand a problem (Bucciarelli, 1988, 163; Brown and Duguid, 1991). Engineers find recognition among their peers for solving progress-blocking problems. The respect and status earned by Maurice Gribble and Derek Whitehead within Ferranti Automation derived from their ability to find solutions.

Once a project moves through successful demonstration of a system prototype, the research phase is over and there is a need to move towards production planning. The R&D process ends. Derek Whitehead wrote (18th November 2013) “The transfer was very formal” [There were] “very detailed drawings to hand over” at a “formal sign-off meeting - a way of saying: ‘we think we are there’ ”. After that you need a move from loose to tight control. “It arises because you are in a different phase of the work”.

Product Champions

Project SAPPHO led by Chris Freeman (1973) studied pairs of successful and failed innovations in the instrument and chemical engineering sectors. Non-parametric tests were used to identify factors which discriminated between success and failure at innovation. One clear lesson to emerge was the importance of product champions, senior figures with the power to support innovations and maintain their momentum during the difficult stages of development, adoption and commercial use.

There are strong echoes of Freeman’s study here. The central role of Fred Cartwright as a promotor of computer control (e.g. Cartwright, 1964) at the Steel Company of Wales is evident in the process control paper here. The pioneering Llanwern wide strip mill installation was replicated at Port Talbot using a GE 412A (Morgan and Kirkland, 1965), full computer control commencing in September 1966, apart from mill pacing (Scott-Maxwell, 1966).

Computerisation at Llanwern itself was championed by Henry Spencer, Managing Director of Richard Thomas and Baldwins. He was a close personal friend of Leon Bagrit, the visionary head of Elliott Automation, who gave the 1964 BBC Reith Lectures on “The Age of Automation” (Bagrit, 1965) where Llanwern was described as “probably the most notable example of automation in this country” (Control, 1965).

Summing Up: a word of caution

These lessons of innovation highlight the role of key individuals: the entrepreneur, the product champion, the team leader of a community of practice. But, we return to where we began: the use of knowledge. It is difficult to identify any one individual’s contribution to a successful innovation. It is hard to say any one individual’s understanding or idea is truly significant, without the whole surrounding gamut of complementary knowledge: technical and scientific; lowly and high-brow; established and exploratory; inspired and mundane; tacit and codified. This is not to deny the role of outstanding individuals. But in the complex world of innovation, they are all ensemble players.

5. Bibliography

John Agar (2003), The Government Machine: A Revolutionary History of the Computer, London: MIT Press

Jon Agar (2006), "What difference did computers make?" Social Studies of Science, vol. 36, no. 6, 869-907

Philippe Aghion and Peter Howitt (1998), Endogenous Growth Theory, Cambridge, Mass: MIT Press

John E. Allen (1999), “Blue Steel and Developments”, The History of the UK Strategic Deterrent, Proceedings; London, Conference held on Wednesday 17 March 1999, section 4.1-4.25, London; Royal Aeronautical Society

Peter M. Allen (2014), "Evolution: complexity, uncertainty and innovation", Journal of Evolutionary Economics, vol.24, issue 2, 265-289

W. Brian Arthur (2009), The Nature of Technology: What it is and how it Evolves, London: Allen Lane

AVIA 65/1156 (1953), 10000 lb HE MC bomb: core size, file covers the period 1953-54, London: The National Archives (marked “Lord Chancellor retained” on 16 December 2013)

Jonathan Aylen (1980), "Innovation in the British Steel Industry", in Keith Pavitt (ed.), Technical Innovation and British Economic Performance, London: Macmillan, ch.12, pp.200- 234

Jonathan Aylen (2001), “Where did Generation V strip mills come from?” Steel Times, vol.229, no.7/8, July/August, 232-237

Jonathan Aylen (2008), “Construction of the Shotton wide strip mill", Transactions of the Newcomen Society, vol.78, no.1, 57-85

Jonathan Aylen (2010), "Promoting the Prosaic: the case for process control computers", IEEE Annals of the History of Computing, vol.32, no.3, 94-96

Jonathan Aylen and Ruggero Ranieri (eds.)(2012), Ribbon of Fire: How Europe adopted and developed US strip mill technology (1920-2000), Bologna: Pendragon

Jonathan Aylen (2015), "First Waltz: Development and Deployment of Blue Danube, Britain’s Post-War Atomic Bomb", International Journal for the History of Engineering & Technology, vol. 85, no. 1, January, 31–59

Jonathan Aylen (2016), “Building and deploying Blue Danube – Britain’s first nuclear weapon”, Royal Air Force Historical Society Journal 63, 6-31

Jonathan Aylen (2017), “Stretch: A cracking recipe for plant performance, R&D Today Downloaded from: http://www.rndtoday.co.uk/case-study/stretch-a-cracking-recipe-for-plant- performance on 28th June 2017

T. Baker and R.E. Nelson (2005), “Creating something from nothing: resource construction through entrepreneurial bricolage”, Administrative Science Quarterly, vol. 50, no.3, 329-366

Margherita Balconi, Stefano Brusoni and Luigi Orsenigo (2010), "In defence of the linear model: an essay", Research Policy, vol.39, issue 1, 2010, 1-13

L. Bagrit (1965), The Age of Automation: The BBC Reith Lectures 1964, London: Weidenfeld and Nicolson, London,

George Basalla (1988), The Evolution of Technology, Cambridge University Press

H. Bessemer (1865), "Manufacture of iron and steel", US Patent No. 49,053, patented July 25, 1865

W. Bijker (1995), Of Bicycles, Bakelites and Bulbs: Toward a Theory of Sociotechnical Change, Cambridge, Mass: MIT Press

Wiebe E. Bijker, Thomas P. Hughes and Trevor Pinch (2012), The Social Construction of Technological Systems: New Directions in the Sociology and History of Technology, Cambridge, Mass: MIT Press, Anniversary Edition. Original edition 1987.

Christy Borth (1941), True Steel: the story of George Matthew Verity and His Associates, Indianapolis: The Bobbs-Merrill Company, 1941.

Hilary Bradbury and Peter Reason (2003), "Action Research: An Opportunity for Revitalizing Research Purpose and Practices", Qualitative Social Work, vol.2, no.2, 155-175

Ernest Braun and Stuart Macdonald (1982), Revolution in Miniature: The History and Impact of Semiconductor Electronics, Cambridge U.P., 2nd. Edition

David J. Bricknell (2009), Float: Pilkington’s Glass Revolution, Lancaster: Crucible

J.S. Brown and P. Duguid (1991), “Organizational learning and communities-of-practice: toward a unified view of working, learning, and innovation”, Organization Science , vol. 2, no.1, 40-57

L.L. Bucciarelli (1988), “An ethnographic perspective on engineering design”, Design Studies, vol.9, no.3, 159-168

R.A. Buchanan and George Watkins (1976), The Industrial Archaeology of the Stationary Steam Engine, London: Allen Lane

Robert Bud and Philip Gummett (eds.) (1999), Cold War Hot Science: Applied Research in Britain’s Defence Laboratories 1945-1990, London: Science Museum

George Bugliarello (1979), “The engineer and the historian”, in George Bugliarello and Dean B. Doner (eds.) The History and Philosophy of Technology, Urbana: University of Illinois Press, chapter 5, pp.50-56

Tom Burns and GM Stalker (1961), The Management of Innovation, London: Tavistock Publications

Harold L. Burstyn (1979), “What can the history of technology contribute to our understanding” in George Bugliarello and Dean B. Doner (eds.) The History and Philosophy of Technology, Urbana: University of Illinois Press, chapter 6, pp.57-80

Vannevar Bush (1945), Science the Endless Frontier: A Report to the President on a Program for Postwar Scientific Research, Washington, D.C.: U.S. Government Printing Office

Donald S.L. Cardwell (1979), “Problems of the data base”, in George Bugliarello and Dean B. Doner (eds.) The History and Philosophy of Technology, Urbana: University of Illinois Press

D.L.S. Cardwell (1994), The Fontana History of Technology, London: Fontana

W. Bernard Carlson (1991), Innovation as a Social Process: Elihu Thomson and the Rise of General Electric, 1870-1900, Cambridge U.P.

W. F. Cartwright (1964), “How iron and steel will be automated”, Fourth annual lecture of the UK Automation Council at the Institution of Electrical Engineers, London, 6th October, summarised in Control, vol.8, no. 77, 580-581 and no.78, 636-7

J.J. Carty (1916), “The relation of pure science to industrial research”, Transactions of the American Institution of Electrical Engineers, vol.35, issue 1, 479-488

A.E. Catania and A. Ferrari (2012), “Development and performance assessment of the new- generation CF fuel injection system for diesel passenger cars”, Applied Energy, vol. 91, no.1, 483-495

Paul E. Ceruzzi (2005), “Moore’s Law and technological determinism: reflections on the history of technology”, Technology and Culture, vol.46, no.3, 584-593

Wayne D. Cocroft and Roger J C Thomas, edited by P S Barnwell (2003), Cold War: Building for Nuclear Confrontation 1946-1989, Swindon: English Heritage

Wesley M. Cohen, Richard R. Nelson, John P. Walsh (2002), “Links and impacts: the influence of public research on industrial R&D”, Management Science , vol. 48, no.1, 1-23.

E.W. Constant (1980), The Origins of the Turbojet Revolution, Baltimore: The John Hopkins University Press

Control (1965), “Two years’ progress at Spencer Works”, Control, vol. 9, no.80, 83-84

Robin Cowan (1990), "Nuclear Power Reactors: a study in technological lock-in", Journal of Economic History, vol. 50, no. 3, 541-567

Robin Cowan and Staffan Hultén (1996), “Escaping lock-in: the case of the electric vehicle”, Technological Forecasting and Social Change, vol.53, no.1, pp.61-79

Richard Cyert and James G. March (1964), A Behavioral Theory of the Firm, Englewood-Cliffs: Prentice-Hall

A. Delemarle and P. Larédo (2008), “Breakthrough innovation and the shaping of new markets: The role of communities of practice”, in A. Amin and J.Roberts (eds), Community, Economic Creativity and Organization, Oxford University Press, pp.178-199

T.K. Derry and Trevor I. Williams (1960), A Short History of Technology from the Earliest Times to A.D. 1900, Oxford: Clarendon Press

Evelyne Debey, Maarten De Schryver, Gordon D. Logan, Kristina Suchotzki, Bruno Verschuere (2015), "From junior to senior Pinocchio: a cross-sectional lifespan investigation", Acta Psychologica, vol.160, September, 58-68

Mark Dodgson (2011), “Exploring new combinations in innovation and entrepreneurship: social networks, Schumpeter, and the case of Josiah Wedgwood (1730-1795)”, Industrial and Corporate Change, vol.20, no.4, 1119-1151

Giovanni Dosi (1982), Technological paradigms and technological trajectories: A suggested interpretation of the determinants and directions of technical change, Research Policy, vol.11, issue 3, 147-162

R. Duncan (1976), “The ambidextrous organization: designing dual structures for innovation” in R.H. Killman, L. R. Pondy, and D. Sleven (eds.), The Management of Organization, New York: North-Holland, pp.167-188.

David Edgerton (1988), “The relationship between military and civilian technology: a historical perspective” in Philip Gummett and Judith Reppy (eds.), The Relations Between Defence and Civil Technologies, Dordrecht: Kluwer Academic Publishers, pp.106-114

David Edgerton (1999), "From innovation to use: Ten eclectic theses on the historiography of technology", History and Technology, vol.16, no.2, 111-136,

David Edgerton (2006), The Shock of the Old: Technology and Global History Since 1900, London: Profile Books

David Edgerton (2010), "Innovation, technology, or history: what is the historiography of technology all about?” Technology and Culture, vol.51, no. 3, 680-697

Eldredge, N. and Gould, S.J. (1972), “Punctuated equilibria: an alternative to phyletic gradualism” in T.J.M. Schopf (ed.), Models in Paleobiology, San Fransisco: Freeman Cooper, chapter 5, pp.82-115

Eldredge, N. (1997), “Evolution in the marketplace”, Structural Change and Economic Dynamics, vol.8, October, 385-398

Clive Ellam (1999-2000), "When engineering was fun: an apprentice's eye view of the Royal Aircraft Establishment, 1948-1953”, Transactions of the Newcomen Society, vol.71, 165-182

T. Joseph Ess (1941), The Modern Strip Mill. A recording of the continuous wide strip mill installations and practices in the United States, Pittsburgh: Association of Iron and Steel Engineers.

T. Joseph Ess (1970), The Hot Strip Mill. Generation II, Pittsburgh: Association of Iron and Steel Engineers

Graham Farmelo (2013), Churchill’s Bomb: A Hidden History of Britain’s First Nuclear Weapons Programme, London: Faber & Faber

Christopher Freeman (1973), “A Study of Success and Failure in Industrial Innovation”, in B.R. Williams (ed.), Science and Technology in Economic Growth. London, Macmillan, pp. 227-245

Koen Frenken (2000), "A complexity approach to innovation networks. The case of the aircraft industry (1909-1997)", Research Policy, vol. 29, issue 2, 257-272

Raghu Garud and Peter Karnøe(2003), “Bricolage versus breakthrough: distributed and embedded agency in technology entrepreneurship”, Research Policy, vol. 32, issue 2, 277–300

T. Geiger (2004), Britain and the Economic Problem of the Cold War: the political economy and the economic impact of the British Defence effort, 1945-1955, Aldershot: Ashgate

M. Gibbons (1984), “Is science industrially relevant? The interaction between science and technology”, in Michael Gibbons and Philip Gummett (eds.), Science, Technology and Society Today, Manchester University Press, pp.96-116

Benoît Godin (2006), “The linear model of innovation: The historical construction of an analytical framework”, Science, Technology and Human Values, vol.31, no.6, 639-667

Benoît Godin (2016), “Technological innovation: on the origins and development of an inclusive concept”, Technology and Culture, vol.57, no.3, 527-556

S.J. Gould and N. Eldredge (1993), “Punctuated equilibrium comes of age”, Nature, vol.366, 18 November, 223-227

Margaret Gowing (1964), Britain and Atomic Energy 1939-1945, London: Macmillan

Margaret Gowing with Lorna Arnold (1974), Independence and Deterrence: Britain and Atomic Energy, 1945-1952, volume 2, Policy Execution, London: Macmillan, 1974

Margaret Gowing (1978), Reflections on Atomic Energy History, the Rede Lecture 1978, Cambridge: Cambridge University Press

M.W. Gribble (1998), “The Argus computer and process control”, Resurrection, no.20, Summer, 20-29

Grimsdale, R.L. (1995), “The transition from valves to transistors”, Computer Resurrection, The Bulletin of the Computer Conservation Society, no.13, Autumn, 4-9

Chuck Hansen (1988), U.S. Nuclear Weapons: The Secret History, Arlington, Aerofax

F. A. Hayek (1945), “The use of knowledge in society”, American Economic Review, vol.35, no 4, 519-530

Robert L. Heilbroner (1967), “Do Machines Make History?” Technology and Culture, vol. 8, no. 3, 335-345

Chris Hendry and Paul Harborne (2011), “Changing the view of wind power development: more than ‘bricolage’ ”, Research Policy, vol.40, issue 5, 778–789

Jonathan Hogg (2016), British Nuclear Culture: Official and Unofficial Narratives in the Long 20th Century, London: Bloomsbury

Donald R. Hoke (1990), Ingenious Yankees: The Rise of the American System of Manufactures in the Private Sector, New York, Columbia UP

Howells, J., James, A., Malik, K. (2003), “The sourcing of technological knowledge: distributed innovation processes and dynamic change”, R&D Management, vol.33, no. 4, 395-409

Jeff Hughes (2002), The Manhattan Project: big science and the atom bomb, Cambridge: Icon

Illinois Institute of Technology (1968), Technology in retrospect and critical events in science, Vol. 1., Chicago: Illinois Institute of Technology Research Institute

Eva Jablonka (2000), “Lamarckian inheritance systems in biology: a source of metaphors and models in technological evolution,” in in John Ziman (ed.) Technological Innovation as an Evolutionary Process, Cambridge U.P., pp.27-40

Deborah Jaffé (2003), Ingenious Women: From Tincture of Saffron to Flying Machines, Stroud, Gloucestershire: Sutton Publishing

Mary Kaldor (1980), “Technical change in the defence industry”, in Keith Pavitt (ed.), Technical Innovation and British Economic Performance, London: Macmillan, ch.6, pp.100-121

Orit Karnieli-Miller, Roni Strier and Liat Pessach (2009), "Power relations in qualitative research”, Qualitative Health Research, vol.19, no.2, 279-289

Israel M. Kirzner (1997), “Entrepreneurial discovery and the competitive market process: an Austrian approach”, Journal of Economic Literature, vol. 35, no. 1, 60–85.

R. Koch and K-H. Leitner (2008), “The dynamics and functions of self-organization in the fuzzy front end: empirical evidence from the Austrian semiconductor industry”, Creativity and Innovation Management, vol. 17, no. 3, 216-226

John Krige (2006), “Critical reflections on the science-technology relationship”, Transactions of the Newcomen Society, vol.76, no.2, 259-269

Maria E. Kronfeldner (2007), "Is cultural evolution Lamarckian?" Biology & Philosophy, vol.22, no.4, 493-512

David S. Landes (1969), The Unbound Prometheus, Technological Change and Industrial Development in Western Europe from 1750 to the Present, Cambridge: Cambridge University

K.H. Leitner (2009), “Innovation and strategy as emerging processes: Empirical evidence from major Austrian innovations”, Paper presented to ISPIM, Vienna, June 2009

Jean Baptiste Pierre Antoine de Monet de Lamarck (1914), Zoological philosophy - An exposition with regard to the natural history of animals, trans. Hugh Elliot, London: Macmillan

Bruno Latour (1992), “Where are the missing masses? The sociology of a few mundane artefacts” in W. Bijker and J. Law (eds.) Shaping Technology/Building Society : Studies in Sociotechnical Change, Cambridge, MIT Press, pp.225-258

J. Law and M. Callon (1988), “Engineering and sociology in a military aircraft project: a network analysis of technological change”, Social Problems, vol. 35, no.3, 284-297

Christophe Lécuyer (2006), Making Silicon Valley: Innovation and the Growth of High Tech, 1930-1970, Cambridge, Mass.: MIT Press, 2006

Stuart W. Leslie (1993), The Cold war and American Science: The Military-Industrial-Academic Complex at MIT and Stanford, New York: Columbia University Press

Per Lundin (2012), Computers in Swedish Society: Documenting Early Use and Trends, London: Springer

Donald MacKenzie (1990), Inventing Accuracy: A Historical Sociology of Nuclear Missile Guidance, Cambridge, Mass: MIT Press

F. Malerba (1992), “Learning by firms and incremental technical change”, Economic Journal, vol.102, no. 413, 845–859

F. Malerba, R. Nelson, L. Orsenigo and S. Winter (2001), “Competition and industrial policies in a ‘history friendly’ model of the evolution of the computer industry”, International Journal of Industrial Organization, vol.19, no.5, 635-664

Alfred Marshall (1890), Principles of economics. Vol.1., London: Macmillan

E L Marshall (2013), “A Lifelong Love Affair — Sir Harry Ricardo and the Sleeve Valve”, International Journal for the History of Engineering & Technology, vol. 83, no.1, 62-95

Ben R. Martin (2016), "Twenty challenges for innovation studies", Science and Public Policy, vol. 43, issue 3, 432–450

Mariana Mazzucato (2013), The Entrepreneurial State: Debunking Public vs Private Sector Myths, London: Anthem Press,

Maureen McKelvey (1996), Evolutionary Innovations: The Business of Biotechnology, Oxford University Press

Maureen McKelvey (2014), “Science, technology and business innovation” in: M. Dodgson, D. Gann and N. Phillips (eds.),The Oxford Handbook of Innovation Management. Oxford University Press, ch.4, pp.69-82

Maureen McKelvey, Olof Zaring, Daniel Ljungberg (2015), “Creating innovative opportunities through research collaboration: An evolutionary framework and empirical illustration in engineering”, Technovation, vols. 39–40, May–June, 26-36

J.S. Metcalfe (1970), “Diffusion of innovation in the Lancashire textile industry”, The Manchester School, vol.38, no.2, 145-162

J. S. Metcalfe (1994), "Evolutionary economics and technology policy", Economic Journal, vol.104, no.425, 931-944

J. S. Metcalfe (1995), "Technology systems and technology policy in an evolutionary framework", Cambridge Journal of Economics, vol.19, issue 1, pp.25-46

A. Mina, R. Ramlogan, S. Metcalfe and G. Tampubolon (2016), "Coronary artery disease", in Davide Consoli, Andrea Mina, Richard R. Nelson and Ronnie Ramlogan (eds.), Medical Innovation: Science, Technology and Practice, Abingdon: Routledge, pp.32-47

H. D. Morgan and R.W. Kirkland (1965), “The application of automatic gauge control and computer control to an existing hot strip mill”, in Instrument Society of America, Digital Computer Applications to Process Control, New York : Plenum Press, pp.167-206

Mark L. Morgan and Mark A. Berhow (2010), Rings of Supersonic Steel: Air Defenses of the United States Army 1950-1979, Bodega Bay, California: Hole in the Head Press, 3rd Edition

MOSI (Museum of Science and Industry) (1966), 1996.10/2/4/903 1947 - 1953 [the 'Wythenshawe Project, 49 Items], Manchester: Museum of Science and Industry Archives.

David C. Mowery and Nathan Rosenberg (1998), Paths of Innovation: Technological Change in 20th Century America, Cambridge U.P.

Fiona Murray (2002), “Innovation as co-evolution of scientific and technological networks: exploring tissue engineering”, Research Policy, vol. 31, issues 8–9, 1389–1403

Andrew Nahum (1999), “The Royal Aircraft Establishment from 1945 to Concorde” in Robert Bud and Philip Gummett (eds.), Cold War Hot Science: Applied Research in Britain’s Defence Laboratories 1945-1990, London: Science Museum, pp.29-58

Richard R. Nelson and Sidney G. Winter (1974), "Neoclassical vs. Evolutionary theories of economic growth: critique and prospectus", The Economic Journal, vol. 84, no. 336, 886-905

Richard R. Nelson and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, Mass: Belknap Press of Harvard University Press

Paul Nieuwenhuis and Peter Wells (2007), "The all-steel body as a cornerstone to the foundations of the mass production car industry", Industrial and Corporate Change, vol.16, no.2 , 183-211

Keith Pavitt (ed.) (1980), Technical Innovation and British Economic Performance, London: Macmillan

Henry Petroski (1985), To Engineer is Human: The Role of Failure in Successful Design, New York: St.Martin’s Press, 1985

John Pickstone (2000) Ways of Knowing: A New History of Science, Technology and Medicine, Manchester: Manchester University Press

Karl R. Popper (1963), Conjectures and Refutation: the Growth of Scientific Knowledge, London: Routledge & Kegan Paul

Carroll Pursell (1995), “Seeing the Invisible: New Perceptions in the History ofTechnology,” ICON 1: 9–15

Katherine Pyne (2016) “The nuclear dimension – the development of warheads for the Royal Air Force ‘special weapons’”, Royal Air Force Historical Society Journal 62, 9-30

R. Ramlogan, A. Mina, G. Tampubolon and J.S. Metcalfe (2007 ) “Networks of knowledge: the distributed nature of medical innovation”, Scientometrics, vol.70, no.2, 459-489

Nathan Rosenberg (1994), Exploring the Black Box: Technology, Economics and History, Cambridge U.P.

D. Rowley and H. Metcalfe (1968-9), “Project management of Bloodhound”, Proceedings of the Institution of Mechanical Engineers, vol. 183, no.3, 1-9

Ozcan Saritas and Jonathan Aylen (2010), “Using scenarios for roadmapping: The case of clean production”, Technological Forecasting and Social Change, vol. 77, issue 7, 1061-1075

Jack W. Scannell, Alex Blanckley, Helen Boldon and Brian Warrington (2012), “Diagnosing the decline in pharmaceutical R&D efficiency”, Nature Reviews, Drug Discovery, vol.11, March, 191-200

I.S. Scott-Maxwell (1965/6), Monthly reports to the Board of the Steel Company of Wales, Corus Colors Record Centre, department 378, consignment 3, box 23, location 041328, sequence of Board Papers 1965 and 1966.

Joseph Alois Schumpeter (1943), Capitalism, Socialism and Democracy, London: Allen & Unwin

Philip Scranton (2006), “Technology-led innovation: the non-linearity of US jet propulsion development”, History and Technology, vol.22, no.4, 337-367

Philip Scranton (2008), “Technology, science and American innovation”, Momigliano Lecture 2005, in Ricerca avanzata e alta divulgazione, Le Momigliano Lectures 1997-2008, Terni, Italy: ICSIM, pp. 193-220

Scott Shane(2000), "Prior Knowledge and the Discovery of Entrepreneurial Opportunities", Organization Science , vol.11, no.4, 448-469

Andrew Shonfield (1965), Modern capitalism: the changing balance of public and private power, London : Oxford University Press for the Royal Institute of International Affairs

David Silverman (2001), Doing Qualitative Research: A Practical Handbook, London: Sage, 2000

N.A. F. Smith (1979) “The study of the history of engineering” Transactions of the Newcomen Society, vol.51, 202-5

H.D. Smyth (1945), Atomic Energy: A General Account of the Development of Methods of Using Atomic Energy for Military Purposes under the Auspices of the United States Government, Washington: Government Printing Office and London: HMSO

Dale Southerton, Elizabeth Shove, Alan Warde and Rosemary Deem (2001), "The social worlds of caravanning: objects, scripts and practices", Sociological Research Online, vol.6, no.2,

Klaus Staubermann (2010), “What machine tools can tell us about historic skills and knowledge”, International Journal for the History of Engineering & Technology, vol.80, no. 1, 119-132

Stoneman, P. (1976), Technological Diffusion and the Computer Revolution: The UK Experience, Cambridge University Press

Richard Summers (1940), The New Mill, 1940, Jonathan Cape and circulated by John Summers and Sons, Hawarden Bridge, Chester

Jacob Tal (1984), Motion Control by Microprocessor, Mountain View, California: Galil Motion Control

B.S.Tether and J.S. Metcalfe (2003), “Horndal at Heathrow? Capacity creation through cooperation and system evolution”, Industrial and Corporate Change, vol. 12, no. 3, 437–476

H.J. Thamhain (2003), “Managing innovative R&D teams”, R&D Management, vol.33, no. 3, 297-311

D. G. Tucker (1979), “Interpretations of the history of technology and the Newcomen Society tradition”, Transactions of the Newcomen Society, vol.51, 197-201

Svetlana Usenyuk, Sampsa Hyysalo and Jack Whalen (2016), “Proximal Design: Users as designers of mobility in the Russian North”, Technology and Culture, vol. 57, no.4 , 866-908

Walter G. Vincenti (1990) What Engineers Know and How They Know It – Analytical Studies from Aeronautical History, Baltimore: John Hopkins U.P.

Walter G. Vincenti (1994), “The retractable airplane landing gear and the Northrop ‘anomoly’: variation-selection and the shaping of technology”, Technology and Culture, vol.35, no.1, 1-33

Eric Von Hippel and Georg Von Krogh (2003), “Open Source Software and the ‘Private- Collective’ Innovation Model: Issues for Organization Science”, Organization Science, vol. 14, no. 2, 209-223

Gabrielle Walker (2003), “The collector”, New Scientist, vol.26, no.2405, 26 July, 38-41

William Walker (2000), "Entrapment in large technology systems: institutional commitment and power relations", Research Policy, vol.29, issues 7-8, 833-846

George Watkins (1970,1), The Textile Mill Engine, Vols.1 & 2, Newton Abbott: David and Charles

H. Weiss (1978), “Die anlagentechnische Entwicklung der Hüttenindustrie in den letzen Jahrzehnten”, Stahl und Eisen, vol.98, no.24, 1267–1276. (drafted by Wolfgang Rohde)

Alex Wellerstein (2013), "The price of the Manhattan Project" at http://blog.nuclearsecrecy.com/page/6/, posted May 17th 2013, downloaded 14th November 2013

E. Wenger (1998), Communities of Practice: Learning, Meaning, and Identity, Cambridge U.P.

E.C. Wenger and W.M. Snyder (2000), “Communities of practice: the organizational frontier”, Harvard Business Review, January-February, 139-45

John Ziman (2000), “Evolutionary models for technological change", in John Ziman (ed.) Technological Innovation as an Evolutionary Process, Cambridge U.P., pp.3-12

6. Corrections and Updates

Authors write in haste and repent at leisure. This is an opportunity to correct errors and record a small sample of follow-up comments from critics.

Comment on “Megabytes for Metals”

Oxford Author, Georgina Ferry (personal e-mail communication , 8th August 2006), pointed out Alan Turing was not responsible for Colossus (“Megabytes for Metals”, p.465). Turing’s work on codebreaking at Bletchley Park focussed on the use of mechanised crypto-analysis and he helped design the electro-mechanical Bombes used to help crack the Enigma codes. Credit should have been given to Tommy Flowers for leading the development of the valve based Colossus from 1943 onwards (Copeland , 2006).

The paper on process control overlooks the considerable use of computers on the research and development side. This would have highlighted the continuing and unexplored role of analogue computing in the iron and steel industry (Small,2001; Care,2010).

Since the paper was published a reader has donated the handbook for the GE412 computer to the Archives at Shotton (General Electric, 1964)

Comment on Shotton Wide Strip Mill

Photo captions are often the last item to be crafted. Figure 6 of the Shotton paper clearly shows “It follows that if you view the mill from within the motor room, the motor nearest the viewer is definitely the last stand, but equally it quite clearly has a reduction gearbox, so cannot be a direct drive to the pinion stands!” not the other way round as the caption suggests. (letter from Ewan Hewitt, sometime Technical Director of Davy United, Sheffield, 4th April 2008)

Comment on Stretch

The paper on “Stretch” discusses development of quick roll change rigs at United Engineering at Pittsburgh based on interview evidence with Jim Adair, Vice-President Engineering in Pittsburgh on Friday October 5th 1979. Writing 35 years later, (e-mail 4th December 2014) Ewan Hewitt pointed out the device could only be:

“. . economically applied to a total new installation.” . . .” Adair's design was brilliant for new mills, but that it was impossible to retrofit to an existing mill because of the huge cost of scrapping so many work roll chocks and modifying all the backup roll chocks, replacing all the pinion stands and fitting new ones incorporating the pushers.” . . “So you can see why, on future presentations on "Stretch" to the steel industry, I think you should mention that Adair's system only applied to NEW mills and cheaper retrofitting systems did later exist.”

Comment on Blue Danube

In one of many telephone conversations, Professor John Allen (Thursday 11th September 2014) disagreed with the view that Blue Danube was “deployed first, tested later”. He emphasised that “Testing was well established at Orfordness. The site was very advanced for its day with SE48 radar and kinetheodelites. The bomb was launched from eight miles up with an eight mile throw.” They also worked on scale models at Larkhill. They were testing ballistic

57 characteristics, safety and the installation of the bomb on the aircraft. They “needed to know the ballistic characteristics worked.” The fact remains that a complete operational weapon was not fully tested - a so-called “service trial” - until 11th October 15.30 1956 (Arnold and Smith, 2006, ch.8) at the Buffalo Trials at Maralinga in Australia.

Credit in the Acknowledgements should be given to Peter Barker (as it is in the text) not “Peter Barlow”. He was one of the leading designers at ARL and provided valuable insights to the design. "Hudswell Castle" should read Hudswell Clarke, top page 40. Atkins for Watkins in a footnote 66

Dr Richard Moore kindly pointed out the missing reference at footnote 10 is: PRO AVIA 65/1153, “10000 lb HE MC bomb: AS requirements”.

Peter Barker pointed out There was one nit-pick which I had forgotten to mention. On p.39 (re the flip-out tail) you say: “...This was achieved using compressed nitrogen to drive a single axial piston and four connecting rods, activated by a lanyard....” One con-rod per fin would almost certainly have resulted in the whole arrangement jamming well before the fins were fully extended. However, Ref 37 states the design correctly: “.... A blast of gas drove a piston and four pairs of connecting rods which pushed the fins upwards and outward by parallel motion linkage..” I doubt if too many people will register the difference between the two statements. Regards, Peter Barker

Dr Richard Moore wrote correctly on 2nd February 2015: You say (p.34) that "in simple terms, a sub-critical mass of plutonium was slammed into another sub-critical mass of plutonium by implosion of a surrounding jacket of high explosive". Do you have a source for the idea there were in fact two separate masses of plutonium? My understanding has always been there was one plutonium sphere with space for the urchin inside." He was right. I confused construction methods using two hemispheres with the overall design layout. ______

Laura Arnold and Mark Smith (2006), Britain, Australia and the Bomb: The Nuclear Tests and their Aftermath, Houndmills, Basingstoke: Palgrave Macmillan, 2nd Edition

Charles Care (2010), Technology for Modelling: Electrical Analogies, Engineering Practice, and the Development of Analogue Computing, London: Springer

Jack Copeland (ed.), (2006), Colossus: the secrets of Bletchley Park’s Codebreaking Computers, Oxford

General Electric (1964), I.G.E. Primary Data Input etc, Book B2, Richard Thomas & Baldwins, Ltd., 68” Hot Strip Mill, Book B (Automation) – Specification. Phoenix, Arizona: General Electric, Process Computer Section, Industry Control Department, last update January 1964, Tata Steel Records Centre, Works: Shotton; Department: Secretariat; Section: Chief Engineer; Department Code: 425; Departmental Consignment No. 59; Location no: 01/0070

James S. Small (2001), The Analogue Alternative: The Electronic Analogue Computer in Britain and the USA, 1930-1975, London: Routledge

7. Impact of the Research: Bibliographical Information on Papers Submitted

The applicant is asked to “comment on the standard of any journals and the reception of the publications as indicated by citations and reviews.”

These papers were written to be read by the wider community of those interested in the history of technology and engineering. We indicate the impact of the papers as follows: a) Citations b) Prizes c) Other measures of impact, including presentations d) Downloads (where available)

Citations Citation data is drawn from Google Scholar in November 2018. The data highlights the low citation rates for historical material compared with other fields, e.g. environmental management. The author’s overall Google Scholar score at the time was 1,419. These history of technology papers represent less than a tenth of his citations.

Blue Danube 2

Bloodhound 6

Stretch 17

Open Innovation 83

Shotton 4

Megabytes for Metals 12

Prizes The paper on Stretch was awarded was awarded the Best Workshop Presentation prize worth €1,000 by the industrial delegates to a Workshop “Managing R&D, Technology and Innovation in the Process Industries” at Grenoble on 5th/6th May 2011. The workshop was hosted by the Ecole de Management in Grenoble and the prize was funded by Promote, the Centre for Management of Innovation in Process Industry in Luleå, Sweden. The workshop brought together R&D managers from a wide range of European, American and north African companies and academic delegates from Europe and China.

Other measures of impact, including presentations and reprints These papers have been presented in appropriate forums.

Blue Danube: Defence Electronics History Society, Shrivenham; Nuclear History Group, Charterhouse; Newcomen Society, London, Manchester and Bristol; Royal Air Force Historical Society, London; Barraclough Memorial Lecture, Sheffield; four country LETTERPRESS Nuclear Disarmament exercise, RAF Honington.

Bloodhound: Conference on “Technological Innovation and the Cold War”, Hagley Museum, Delaware; University of Padua, Italy; Computer Conservation Society, London and Manchester; IET, Bristol

Stretch: Grenoble, workshop on “Managing R&D, Technology and Innovation in the Process Industries”; Outokumpu Oy, Avesta and SSAB, Luleå, Sweden; Cleveland Institution of Engineers, Middlesbrough; iCHSTM Conference, Manchester; Rolling Guild, Grange over Sands

Open Innovation: Conference on “Automobility: A Conference on the 100th anniversary of the Model T”, Hagley Museum, Delaware; R&D Management Conference, Ottawa; Rolling Guild

Shotton: Newcomen Society, Manchester; Rolling Guild,

Megabytes for Metals: Sheffield Metallurgical and Engineering Association; Lincolnshire Iron and Steel Institute, Scunthorpe

Reprints: The papers on Megabytes for Metals and on Shotton have been reprinted in revised form in Aylen and Ranieri (2012) and the paper on Open Innovation has been published in shorter form by the Manchester Statistical Society.

Downloads (where available) Download information is usually confidential for commercial reasons. The author has access to the following information on downloads since articles have appeared on the web.

Bloodhound 269

Blue Danube 1,716

Shotton 23

Other Recognition In recognition of his contribution to the history of engineering and technology, the candidate has been elected Vice-President of the Newcomen Society and is President elect for the years 2019 to 2021.

1. Blue Danube - Britain’s post-war atomic bomb

Jonathan Aylen, “First waltz: development and deployment of Blue Danube, Britain’s post-war atomic bomb”, The International Journal for the History of Engineering & Technology, vol. 85, no.1, January 2015, pp.31-59

61 int. j. for the history of eng. & tech., Vol. 85 No. 1, January 2015, 31–59

First Waltz: Development and Deployment of Blue Danube, Britain’s Post-War Atomic Bomb Jonathan Aylen Manchester Institute of Innovation Research, University of Manchester, UK

Blue Danube was the first British atom bomb deployed during the Cold War. The article focuses on practical issues of design, production, deployment, maintenance and testing of the weapon during the 1950s. Emphasis on scientific aspects of nuclear weapons means efforts to develop a workable technology have been overlooked. Blue Danube did not follow the usual pattern of technical development: it was deployed first and tested later. The first weapons were hand-crafted prototypes delivered to the RAF for familiarization and active service, rather than a durable and reliable deterrent. As service life progressed, components were tested and the technology developed in terms of safety, fuzing and arming. Electronic circuits were modified to improve reliability within the same overall casing. So, Blue Danube is not one weapon — more a sequence of modifications in response to a succession of problems. Development of the weapon required innovations in electronics, explosives and logistics. Conventional high explosive components were as much a constraint on deployment and serviceability as the novel radioactive parts. Learning how to manufacture, use, maintain and upgrade the weapon proved as important as building the device in the first place. The bomb was at the frontiers of practical application in electronic circuitry. Protecting electronic components by potting remained an uncertain art. Like many technologies, Blue Danube was a combination of the familiar and the advanced. keywords atomic bomb, Cold War, fissile material, plutonium, high explosive, potting, innovation, evolution, problem sequences, routines

Understanding nuclear bomb theoretical modelling is an essential start but designing it takes a much greater effort; making it is the largest task of all. Peter Jones, former Director, AWE (1999)1

Read at the Science Museum, London on 8 January 2014

Blue Danube was the first British atomic bomb deployed during the Cold War. There are accounts of the invention of the first atom bombs as part of the Manhattan Project and their subsequent re-invention in the Soviet Union and in Britain but these stories focus on the science and politics of early weapons development in an atmosphere of secrecy and espionage.2 History is largely silent on the physical technology and practical engineering of the first nuclear weapon in the UK. The rapid technical development of Britain’s first atomic bomb is examined, focusing on practical issues of procurement, production, assembly, deployment, maintenance and testing. Neglect of the prosaic aspects of atomic weapons is under- standable.3 Atomic bomb design is an unfashionable, even unwelcome, topic. The technology is complicated. It is shrouded in secrecy. Records are fragmented across Government Departments or incomplete. Some archive records have been withheld and some available records have been subsequently withdrawn.4 The neglect of technical development of the atom bomb in the UK distorts our view of how technology develops. Accounts of the bomb focus on a polite story of intellectual effort and formal science rather than engineering and practical technology. Interviews with those involved with Blue Danube design, production and deployment show an undocumented technological effort which encompassed a manufacturer of hot water bottles and a tank engine builder.5 Technicians struggled to perfect the production of items such as airbags, firing circuits and timing fuzes, often unaware of the full scope of the project they were working on. This paper takes an evolutionary view and argues that Blue Danube was founded on local knowledge and pragmatic solutions which produced a deterrent that was modified as problems emerged.6 Development work was distributed widely. The weapon system was built on the back streets of northern towns such as Leeds, Barns- ley and Mansfield and in southern suburbs such as Ilford, Weybridge and Shoreham as much as the more familiar sites of Woolwich, Aldermaston and Burghfield.7 Once the weapon came into service, feedback from RAF rehearsals, trial drops and volume manufacturing caused the bomb to be modified. Those involved were often expert on one aspect of the bomb’s design, but kept in ignorance about other features of the weapon. Designs were given a ‘Top Secret Atomic’ classification. Information was provided on a ‘need to know basis’. As one respondent from Farnborough said, ‘other things were going on like Yellow Sun and Red Beard, but those were in the next office or the next office but one and I can’t tell you anything about them’.8 Or, as a former RAF officer said ‘You were encouraged not to be curious. You never talked about “top site” matters outside the wire. You never, ever talked about it in the Mess. It was just “verboten”’.9 Nevertheless, de facto cooperation emerged as key individuals shared knowledge up and down the supply chain, though not always fully aware of the precise significance of the information conveyed.

Development of Blue Danube Blue Danube was conceived in secrecy and developed in haste. Work began before the go-ahead was given. Operational Requirement OR 1001 was issued in 1946.10 This initial specification for Blue Danube — the ‘special weapon’, or the ‘Bomb, FIRST WALTZ 33

Aircraft, HE, 10,000lb MC Bomb’ (initially, LC bomb), or ‘the store’, or ‘Small Boy’, or ‘Ten Thousand Pound Elsie’, as it was variously known — set the overall dimensions. The large size, weight and shape of the bomb were crucial as they determined the four designs for the new generation of jet-propelled V-bombers: Sperrin, Valiant, Victor and Vulcan. At the same time, the British Industrial Group of the Atomic Energy Organization began work in February 1946 to produce plutonium for use in bombs. The formal decision to proceed with a UK atomic bomb was taken at a small cabinet committee GEN 163 on 8 January 1947.11 The first planning meeting to build the weapon took place in a curtained and stuffy Library at Woolwich Arsenal in June 1947.12 The first British-made nuclear explosion took place at the ‘Hurricane Trial’ on the Monte Bello islands, off Western Australia on 3 October 1952 — a decisive demonstration that the UK’s warhead concept worked. Delivery of the Blue Danube weapon commenced at the Bomber Command Armaments School at RAF Wittering on 7 November 1953.13 The first operational V-bomber, a Valiant, WP201, was delivered to the RAF on 15 June 1955. The first live drop from a Valiant at 30,000 feet took place at Maralinga on 11 October 1956, three years after the first bomb had been delivered to the RAF. The Air Ministry recognized ‘for political reasons, Blue Danube was introduced into the service at a fairly early development stage, which would normally be regarded as premature’.14

Pre-history of UK development It is an irony of history that production of atomic bombs ceased briefly worldwide after their initial deployment at Hiroshima and Nagasaki and the post-war Bikini Atoll trials. When David Lilienthal became Chairman of the US Atomic Energy Commission in January 1947 he found just one deployable bomb in existence, few skilled personnel able to assemble other parts that were available and no capacity to manufacture more. The knowledge of how to build a bomb had been dispersed and the existing design had many shortcomings.15 Despite helping develop the first two atomic bomb designs as part of the Manhat- tan project and playing a part in the post-war tests at Bikini Atoll in July 1946, the British were cut off from any further American help by the terms of the USA’s McMahon Act which became law in August 1946. The UK Government decided they had to develop their own ‘special weapon’. The aim was to build a version of ‘Fat Man’ used at Nagasaki on the basis of partial knowledge acquired by British scientists working on the war time project. An organization called Basic High Explosive Research was formed in June 1947 under the leadership of William Penney who had worked at Los Alamos from June 1944, witnessed the Nagasaki explosion and advised the Americans at the Crossroads’ trials at Bikini Atoll. William Penney had the official title Chief Superintendent, Armaments Research. Penney had both the formal knowledge and the tacit know- how to transfer the essentials of bomb production to his new team. Otherwise, members of the UK team involved in the Manhattan project had dispersed to other roles, although Klaus Fuchs, Egon Bretscher and the circuit designer Ernest Titterton worked at Harwell on atomic energy research.16 34 JONATHAN AYLEN

The project name was soon shortened to ‘High Explosives Research’ (HER) and the design team were given headquarters at Fort Halstead, Dunton Green, near Sevenoaks in Kent with production facilities at Woolwich and testing facilities initially at Foulness in Essex and later also at Orfordness in Suffolk.17 The RAF and the Royal Aircraft Establishment (RAE) were involved almost from the outset. In due course HER moved to a new site of its own at Aldermaston, a former airfield in Berkshire, which became part factory and part R&D laboratory. As a result of incomplete knowledge, domestic production imperatives and the desire of technologist to modify, the UK bomb emerged as a local design. This fundamental redesign had major benefits:

Compared with the American weapons at the end of 1946 when all collaboration ended, the first British bomb had several advantages: about 20 per cent greater nuclear efficiency; easy change of explosive power; incomparably better ballistic and aiming qualities; rapid assembly; much safer loading; much improved firing circuits and detonators; sealed electronics, permitting release from a high altitude; and better storage characteristics.18

Design selection Blue Danube worked by nuclear fission. In simple terms, a sub-critical mass of plutonium was slammed into another sub-critical mass of plutonium by implosion of a surrounding jacket of high explosive. Reducing the volume of two hemispheres of plutonium so rapidly means the nuclei find plenty of neutrons to fission, but little surface area to escape from. Fission releases huge quantities of energy and subsequent radioactive fall-out. Two designs of atomic bomb were developed to fruition during the Manhattan Project of World War II. The first design triggered a chain reaction by firing a part icle of uranium down a tube inside the bomb into a mass of uranium. This ‘Little Boy’ design was dropped over Hiroshima on 6 August 1945. While the design was straightforward, production of fissile material U235 proved a bottleneck. So much so, that the bomb exploded over Hiroshima had not been tested due to shortage of uranium.19 In the second design of atomic bomb, ‘Fat Man’, a chain reaction was triggered by reducing the volume of fissionable material — in this case plutonium — using high explosives to compress the material symmetrically and with great force. This was the design exploded during the Trinity Test north-west of Alamagordo, New Mexico on 16 July 1945 and above Nagasaki on 9 August 1945. Here the technical difficulty lay in the design of the complex implosion mechanism that triggered a critical mass. The US design was aerodynamically poor as its length was constrained by the short bomb bay of the B29 delivery aircraft.20 Supply of fissile material helped determine the choice of British design of atomic bomb. Plutonium was largely manufactured in what seems, in retrospect at least, a ‘cheap and cheerful’ way in an air-cooled graphite pile at Windscale in Cumberland.21 An adjacent chemical separation plant used solvent extraction to separate plutonium from uranium in the irradiated fuel rods. The uranium fuel rods for Windscale were made at Springfields, Salwick, west of Preston in Lancashire using chemical separation and vacuum casting. The hollow spherical tamper inside the high explosive shell FIRST WALTZ 35 of Blue Danube was natural uranium 238, also from Springfields. A tiny amount of polonium 210 known as an ‘urchin’ was needed as an initiator, providing an intense source of alpha particles at the moment of prompt criticality.22 This polonium was made from irradiated bismuth at Windscale, and later Risley. The scientific team may also have appreciated the ‘Fat Man’ design was less likely to go critical by accident than the ‘Little Boy’ alternative, while plutonium is fairly benign and easy to handle in small finished quantities compared to uranium 235. Choice of the ‘Fat Man’ design for the UK atomic bomb had two consequences. Firstly, Blue Danube was large (Figures 1 and 2). Use of an implosion triggered by conventional high explosives resulted in a weapon weighing some 10,000 lb, just over 24 feet long and almost 62 inches in diameter.23 The overall bomb weighed 4½ tonnes, but perhaps 2½ tonnes of this were high explosive.24 Secondly, a key feature of the design was the double layer of explosive lenses that triggered the implosion. So, paradoxically, design and maintenance of high explosives and their accurate detonation were the key focus of an atomic bomb delivery system. Non-destructive testing of the explosive lenses using X-ray techniques became a key maintenance activity.25 The Blue Danube design did not replicate the US ‘Fat Man’.26 The American case was heavily armoured. The British version was a light internally braced ballistic

figure 2 An early general arrangement drawing from ARL at Farnborough supplied to Vickers at Weybridge to help with design of the Valiant. The lifting lug and finger chopper plate for in-flight loading have been omitted for clarity. RAF Museum Hendon

casing. The explosive sphere was redesigned from first principles using a combination of twelve pentagonal and twenty hexagonal shapes to form a sphere — not unlike the outer pattern of a modern soccer ball. The first version of the British bomb was primed in flight through In Flight Loading (IFL) of the core. The UK bomb was designed to work at higher altitudes and lower temperatures than its US equivalent.

Engineering design and the prototype Penney’s approach to engineering design was the familiar method of breaking down the weapon into its key packages and entrusting design, development and testing of the sub-assemblies to small teams.27 HER at Fort Halstead set up four working parties to complement their in-house work on the core of the bomb ‘to ensure that all aspects of the project would receive adequate consideration’.28 These committees were named C, A, F and H which covered Ballistic Case and Supporting Structure (C); Aircraft Installations (A); Fuse (sic) (F) and Handling and Transportation (H). The overall weapon would then be assembled into a prototype from its components. In turn, each team was broken down into smaller groups and allotted specific tasks. This helped maintain secrecy. What appeared to be a highly centralized and directed project was, in truth, widely distributed in terms of delivery across both public FIRST WALTZ 37 bodies and private firms. The project involved a high degree of coordination behind a wall of secrecy. While this process of deconstruction and reconstruction is a standard approach to engineering, it is not clear that many conventional engineers were involved in the lead team at HER at Fort Halstead.29 The initial team briefed by Penney — a mathematician — at Woolwich in 1948 was composed of a mix of scientists including chemists and fellow mathematicians.30 The conceptual and detailed design work for Blue Danube was undertaken within a narrow circle of government departments. This meant scientists were entrusted with production designs. As a result, ‘the design of the weapon is sound enough in principle, but unpolished in the engineering sense, leading inevitably to an expensive and long-drawn-out process of improving the design of many of the component parts’.31

The technologies of the Blue Danube bomb The atom bomb was a set of technologies. It brought together pioneering electronics with novel nuclear materials and conventional, heavyweight explosives.32 The casing posed aerodynamic problems. Preparation of the fissile material required feats of chemical engineering. The bomb was built up from six packages of components, each with its own set of development problems and manufacturing challenges. They were the outer ballistic casing; the suspension system for the central physics package (or warhead); the fuzes; the firing mechanism and detonators; the physics package combining a high explosive outer shell and the tamper, core and urchin; and, finally, a new set of organizational routines and facilities surrounding handling, storage and deployment. This omits associated developments which have commanded more public attention, such as a new generation of medium range jet bombers, the difficulties of making fissile materia l for the core, and the awkward nature of the nuclear testing process.

The ballistic casing and support structure — the role of RAE Farnborough The outer ballistic casing and internal support structure illustrate the combination of public sector research and private sector manufacture required to build Blue Danube. What appears to be a simple, delegated task actually required considerable experiment, technical development, design work and engineering ingenuity. The ballistic casing for an atom bomb has four functions. The outer shape of the case determines the bomb’s free-fall characteristics and accuracy. The strength of the case allows the bomb to be suspended in the bomb bay. The internal structure of the centre section holds the physics package in place and maintains gentle pressure on the outer explosive sphere. Finally, the casing encloses the arming mechanism, fuzes, firing mechanisms and batteries necessary to trigger the bomb. The front and rear casings also played a structural role as they were winched up against the fore and aft crutches in the bomb bay. At the front, a fibreglass radome allowed the primitive radar fuzes to operate. The design, testing and supervision of the development of the overall bomb casing was delegated to the RAE whose central role in design work and managing contracts 38 JONATHAN AYLEN is underplayed in official accounts. The RAE was to become even more prominent in later weapon developments such as adaption of US hydrogen weapons, conceptual design of Blue Steel and design of WE177.33 The RAE was involved in Blue Danube from 1948, not just on casing design and aerodynamics, but also the physical and electrical integration of the bomb into the aircraft. Farnborough had set up a division called ‘ARL’ — the ‘Airfield Radio Laboratory’. This innocuous name was retained as work began on Blue Danube under the leadership of Sid Hunwicks.34 Such misnomers were typical of the dissembling surrounding the weapon.35 RAE used a secure compound for atomic bomb development work located at the remote southern end of Farnborough airfield. This ARL compound had aircraft access and was occupied around 1950. It was in use by 1952 for bomb loading trials on the Valiant prototype flown in from Vickers at Weybridge (Figure 3). Photographic evidence shows ARL had a staff of at least twenty-two engineers and ten support staff working on Blue Danube in 1953, protected by an armed special police unit (Figure 4). The RAE was responsible for the design and ballistics of the complete casing and for elements of the implosion sphere such as the support structure and IFL. The casing had a flush riveted stressed skin over an inner airframe. The ballistic shape of the bomb was designed by Dr G. J. Richards. A programme of wind-tunnel testing, trials of rocket-propelled models and tests of 1/3.4 scale models, followed by full-scale trials at Orfordness were used to confirm the shape. Wind-tunnel testing at Farnbor- ough revealed the singular predicament that the bomb was reluctant to leave the

figure 3 A prototype Vickers Valiant enters the ARL compound for bomb fitting. Photo: RAE FIRST WALTZ 39

figure 4 The Blue Danube development team at ARL Farnborough and ‘Russian Spy’ around 1953. The interloper - Ivan - was smuggled in to the picture by Pete Barker. The picture shows the police and the RAF officers seconded to the development team as well as all the development personnel. The picture includes Sid Hunwicks, Reg Milne, John Allen. At the time of this photo by E. P. ‘Gred’ Greddington (top right) this was one of the most secret places in the world. © RAE

aircraft due to flow of air in the bomb bay with local lift at the nose and downward flow at the tail driving the weapon back up into the bomb bay. This would cause considerable embarrassment to the crew if the time-delay fuze kicked in.36 Two measures were taken to make sure the bomb did fall away from the aircraft. Blue Danube had four fixed fin stubs which were limited in span by the diagonal size of the bomb bay. Extendable ‘flip out’ fins were fitted to lengthen the span of these stubs or ‘fin gloves’. The flip-out fins were captured inside the fin stubs while it was stowed in the aircraft. The fins were deployed as soon as the bomb left the aircraft. This was achieved using compressed nitrogen to drive a single axial piston and four connecting rods, activated by a lanyard.37 The Valiants themselves were also fitted with six retractable dragon’s-teeth ‘fingers’ at the leading edge of the bomb bay to disturb the airflow to avoid the ‘nose-up’ effect which generated lift.38 The main casing of the bomb was fabricated by Hudswell Clarke at Leeds. The firm is best known for design and manufacture of industrial steam locomotives. They made the transition to aircraft, bomb and — later — rocket production through sub- contract work from November 1938, initially to Blackburn Aircraft’s Olympia Works on Roundhay Road, Leeds. The firm took over the works when Blackburn moved to Brough.39 They also made the centre section and nose of the various trial models, for which the tails were made by Follands. 40 JONATHAN AYLEN

The nature of the work at Hudswell Castle was apparent to the workforce: Every man on the shop floor would know what they were working on, because inside the Blue Danube tail joint ring (and other similar items) was a small brass plate engraved with the customer’s name (Atomic Weapons Research Establishment) and a serial number. No one was so dumb as not to know who that was, even in 1956.40

Suspension system — Percival Aircraft The suspension system for Blue Danube also shows how practical aspects of the technology were devolved to the RAE at Farnborough and onward to individual private contractors. A process of distributed problem solving which allowed HER at Aldermaston to get rid of their responsibilities for the delivery system. Again it involved the most unlikely participants. The physics package had to be suspended inside the ballistic case and insulated from the severe vibration and buffeting the bomb would experience on a jet aircraft during take-off, climb and cruising in the cold upper atmosphere. Structural design of the centre section of the bomb was undertaken by Percival Aircraft of Luton.41 There were hefty aluminium front and back formers for the cylindrical centre section of the bomb casing. These were separated by a horizontal structure — a rigid but light ‘strong ring’ made of cast magnesium with a circular hole to provide multi-point support for the sphere The outside of the explosive sphere was cushioned on thirty-two airbags matching each segment of explosive (Figure 5). Each airbag had an opening to allow for insertion of the detonator. This patchwork of air bags was held in place by a hinged aluminium cover, again made up of the same combination of pentagonal and hexagonal plates. The plates were laced together with substantial piano hinges attached to each plate edge and threaded together by hinge pins of 20 gauge steel wire. This surface covering of tessellated plates spread the weight of the explosive package under gentle tension. The centre of each plate was also open to a detonator cap of about 6 inches diameter. Lugs were fixed on the cover plates all around the equator. The lugs were fixed to the strong ring by attachment bolts. In this fashion, the whole explosive sphere, weighing close on three tonnes, was gently and evenly suspended under slight pressure within a cosseting layer of air. The butyl rubber airbags surrounding the explosive sphere were made by William Freeman & Company Limited in Barnsley, a well-known hot-water-bottle manufacturer (Figure 5). Reg Milne recalls: Dai Williams suggested air bags. I chose the rubber. I went to two firms, India Rubber, Gutta Percha and Telegraph Company where London City Airport is now and to a firm up in Barnsley. Now they made Suba-Seal hot water bottles amongst other things. They were a lot more accommodating and competent we thought. We settled on them.42 Reg Milne’s choices show the flexibility inherent in the devolved structure of decision-making. However, in contrast to the openness at Hudswell Clarke: I am quite clear that Freeman and his salesman and staff had no idea of the purpose of the air bags — just that they were important and needed promptly. They were made in the same workshop with moulds made to fit in the presses that were being used for all the other products. FIRST WALTZ 41

figure 5 A mock-up of the physics package at ARL, Farnborough shows the plates covering the air bags surrounding the outer explosive shell, the detonators distributed evenly over the surface and a handling frame which supported the package. Reg Milne is the young man on the right. © RAE

Fuzes Technical development of the various fuzes used by Blue Danube was given over to Royal Ordnance at Woolwich, along with the related firing mechanism and detonators. The slow development of the whole firing circuit highlights the troubleshooting necessary to convert design ideas into workable technology. Their slow delivery became a bottleneck and prompted design reappraisals. The electronics of Blue Danube shows how design often evolves by solving a sequence of problems, only in this instance the technology was already in service.43 To ensure the bomb imploded, there were a variety of fuzes, all with duplicate circuits. The fail-safe approach made the circuitry complicated. Blue Danube was a free fall bomb built for an air burst in either a ‘high’ position at 2450 ft or ‘low’ position at 1082 ft. Detonation was triggered by a ground proximity fuze (or W/R fuze) — a simple form of radar located in the nose of the bomb (Figure 6).44 Since each fuze was duplicated there were four proximity fuzes in the service weapon, arranged as a quadrant. To add to the difficulty, contracts for manufacturing these devices were divided for security reasons. Two outside suppliers — Plessey at Ilford 42 JONATHAN AYLEN

figure 6 The ground proximity fuze (W/R fuze) supplied by two different manufacturers to ensure secrecy. They were criticized for being unduly complex, but radar was an heroic technology after World War II. Crown copyright 2013, with permission

and Peto-Scott at Weybridge — were responsible for supplying the proximity fuzes. In addition, Woolwich made the first fuzes used on the trial weapons. To add to the complexity, a clockwork time fuze was added, again duplicated (Figure 7).45 If all else failed, either a graze fuze or a contact fuze detonated the weapon when it hit the ground. Supply was insufficient. So a ‘crash programme’ had to be set up in January 1955 to supply the ground proximity fuzes. This initiative seems to have been highly effective and the whole programme went up a gear during 1955 as the RAF Valiants were being commissioned. A design freeze was imposed. No more modifications were allowed. Production was switched to the Royal Ordnance Factory (ROF) at Burghfield. Key staff were loaned to suppliers. It was a classic instance of troubleshooting.46 Ironically, a shortage of clockwork delayed matters. A government evaluation of Blue Danube ruefully remarked: ‘instead of the supply of radio-active bomb components being the limiting factor, this doubtful distinction belongs to the Firing Installation and the Time Fuse’.47 FIRST WALTZ 43

figure 7 The clockwork fuze. Delivery problems with the clockwork jeopardized Britain’s strategic deterrent. Shortcomings in the cocking mechanism that wound the fuze led to premature detonation, a problem resolved with extra washers to eliminate asymmetry in the winding plunger. Crown copyright 2013. With permission

Detonators and firing mechanism Detonators were one area where US experience helped UK development. US technicians Luis Alvarez and Lawrence Johnston developed exploding wire detonators for ‘Fat Man’.48 These bridge wire detonators had the advantage they would vaporize simultaneously in microseconds in a predictable and precisely timed way thereby avoiding an asymmetric explosion where plutonium leaked out of one side of the sphere without forming a critical mass. These detonators were developed and tested at Fort Halstead but their manufacture was given to AWRE production unit located on Ha-Ha Road, Woolwich (Figure 8). Here rows of women worked with microscopes soldering together the electrodes and the platinum bridge wires — just two thousands of an inch in diameter and a tenth of an inch long — inserting them into plastic moulds and filling the moulds with PETN explosive to exactly the right density. [. . .] The result at the end of the line was an object that came to be called an exploder, a little bigger than a tennis ball, rounded on one side to be set into the bomb with two large connectors sticking out on the other side to be joined to the cables.49 44 JONATHAN AYLEN

Development of the firing circuits, fuzes, secondary fail-safe fuzes and circuits was led at HER, Fort Halstead, by W. J. ‘John’ Challens, a former rocket scientist and future director of AWRE.50 This was frontier technology. Firing circuits provided a precisely timed high-voltage pulse to reach all thirty-two detonator sites at exactly the same moment. To this end, all of the firing cables were the same length. This meant detonators close to the firing unit had a length of excess cable coiled nearby. These firing circuits were to prove problematic, illustrating a familiar truth that it is often the subsidiary aspects of a technology that stand between success and failure, not the main technical principle itself. Multipoint detonation of the thirty-two detonators was achieved with a high- voltage pulse. The firing mechanism had to discharge a precisely timed pulse to all of the detonators simultaneously (Figure 9). This posed novel design problems.51 Details are partly classified, but it seems various designs were used. Early components shown to the author discharged a bank of eight condensers (now known as ‘capacitors’). These are devices which store an electric charge until required. They were prompted to discharge their pulse by a very large relay driven by the fuze current. Later photos show a pulse transformer which may be part of a more reliable design. (Pulse transformers are not dissimilar to the magneto once fitted to cars.) This too would need to be moderated by condensers to prevent arcing. But limited evidence and newly released documents suggest there were a variety of components used at various stages in the evolution of the weapon. In particular, there were problems with the unreliability of condensers.52 The electrical pulse to the detonators needed to be discharged in less than a milli- second, requiring a very fast switching device. This was achieved using a trigatron, a FIRST WALTZ 45

figure 9 The final version of the firing unit with 32 sockets for each of the leads to the detonators spread round the outside of the physics package. This unit was duplicated giving a total of 64 leads to detonators. Crown copyright 2013, with permission

glass valve filled with gas which ionizes instantly as a plasma, thereby allowing a current to pass through. It switches ionically, just like a flash of lightning. Trigatrons were widely used in radar as transmit/receive switches during the Second World War, but only had to perform at low altitudes, and even so were prone to explode. The design of trigatron for Blue Danube was altered at a crucial stage and maintenance records for RAF Barnham show they were temperamental in service use.53 A trigatron switch has another advantage. An electric pulse is usually a drawn-out distribution of current. (Consider the slow pace at which an old light bulb element illuminated when switched on.) A trigatron trims the slow build up and run down of current, giving a short, sharp burst instead. This current was then fed to the detonators. These circuits were novel in the laboratory. The problem was to produce a reliable circuit which gave simultaneous switching regardless of temperature or air pressure, even after long periods of storage and despite vibration in flight. Just to make sure it worked, the firing mechanism was duplicated. As was remarked ‘By the very nature of the blast every effort had to be made to get 100% reliability’.54 46 JONATHAN AYLEN

Potting — a potted history of problems High technology often stumbles on prosaic problems. Delivery of an atomic bomb posed new challenges for electronics. V-bombers were to fly at high altitudes and currents spark across gaps at the low air pressure existing at high altitudes. Flying high meant components cycled from warmth at ground level to extreme cold at high altitude — shrinking with drop in temperature. V-bombers were to be deployed in areas of high humidity, such as the tropics, bringing the problem of moisture ingress. Circuits would vibrate during take-off and flight creating mechanical stresses to break connections. The solution to these problems was encapsulation of electronic components in resin through the arcane art of potting. This pioneering technology evolved in the late 1940s, but was not without shortcomings.55 It is difficult to get resin to bond to the components. Cracks and tracks might emerge as the resin shrinks giving alternative routes for stray currents. The durability of resins was not clear. The very process of encapsulation set up forces in components which could break soldered circuits. To add to the embarrassment, preparation of resin could be exothermic and resins would melt through plastic containers and heat up components. Many circuits which performed perfectly under test failed once they were encapsulated for use. For instance, a report on the proximity fuzes concluded: ‘cracking of a percentage of valves in Araldite blocks is envisaged’.56 The teething problems associated with potting are illustrated by attempts to manufacture the firing system. Volume production of the firing mechanism was first entrusted to Whiteley Electric- al Radio Company of Mansfield, Nottinghamshire. Whiteley was a private company founded in 1926 which specialized in loudspeakers for domestic radios and public address systems.57 Whiteley was an unlikely candidate for a crucial military electronics contract. However, the firm did have a vertically integrated production system, including electromagnets which were a key component of Blue Danube’s circuitry which amounted to a ‘mass of relays’.58 They would have set aside separate productio n facilities for the secret work. Components of the firing unit included a power supply, a trigatron to switch the pulse, a pulse transformer, a trigger isolator and an arming unit. This was coupled with a trigger unit, a discriminator and a junction box. Everything was duplicated and all wired into a common frame behind the explosive sphere. Surviving units show components were built up on a frame and then substantially encased in transparent resin. Individual electronic components came from a range of UK manufacturers, including Norbury Watch Co. who made the trigatrons and BICC who supplied the eight specially designed EHT condensers for each unit.59 An initial order was issued to Whiteley in Mansfield for assembly of components in 1953 but there was insufficient detail for a complete contract; so the actual contract was not issued until 8 January 1954.60 Even then design details were not finalized. There were constant modifications up to the point where none of the twenty-five initial sets of firing units had been delivered by the end of 1954. By that time Blue Danube bombs had been delivered into RAF service. So the electronic component supply chain was a severe bottleneck. The contract with Whiteley began at an initial value of £255,000 but soon doubled as the modifications flowed in. Individual components proved unreliable and there were no agreed test procedures. In the meantime, FIRST WALTZ 47 firing units were being delivered by the research laboratory of ARE(C) at Woolwich Common as a stop-gap measure. Indeed, the first complete and fully tested production firing unit was not delivered from Mansfield until July 1955. All firing units had to be tested in Stratosphere Cabinets. Sub-units passed individual tests and were successfully built up into sub-assemblies only to fail on final assembly of the complete mechanism once ‘potted’ in Araldite. Arming Units, Trigatrons and EHT condensers were prone to failure at this stage.61 By the end of December 1955 Whiteley had still only delivered twenty-six firing installations out of the 120 planned. This was not helped by a change in Trigatron design in July 1955. The contract was finally rescinded in 1955 and ended in acri- mony with the Ministry of Supply going to court to recover two stratosphere test cabinets loaned to Whiteley and not returned.62 The firing system contract was taken up by Dennis Ferranti Meters, at Bangor.63 The firing system for an atomic weapon presented an opportunity for new business to an aggressive newcomer. The firm specialized in precision assembly of meters. They had no experience of electronic components. Nevertheless, in 1955 a contract was issued for fifty-six firing units for £200,000 which proved successful. The components were assembled and tested amid great secrecy in a department with its own dedicated workforce, management and security ‘minder’.

Physics package Conventional explosives were central to the workings of Blue Danube. A carefully structured globe of high explosives was needed to initiate the implosion. Casting of the segments of the fabricated high explosive ball was initially delegated to the Royal Arsenal, Woolwich and the resulting spheres were taken to Foulness in Essex for testing. But full-scale production moved to Royal Ordnance Factory (ROF) Chorley in June 1953. The ball was made up of pentagonal and hexagonal sections which tessellate together to form a sphere. The outer explosives were built up as ‘lenses’ so as to focus the blast on the centre of the sphere. During manufacture, the explosives were poured as liquids into moulds in two stages to form the lenses. When solidified, shaped lenses of RDX/TNT curved outside the inner cones of Baratol. The plutonium core of Blue Danube remains secret, but it is clear the design evolved from that used by the original ‘Fat Man’ device. The warhead assembly was altered in service once practical shortcomings in arming the weapon became apparent. This evolution is illustrated by the development and demise of IFL which resulted in a redesign of the casing, a change in arming procedures on the ground and the warhead assembly being turned upside-down.64 Originally, with IFL, Blue Danube was assembled and loaded into the aircraft without its plutonium core.65 The first operational bombs were deployed on Valiant aircraft. The bomb was loaded and fixed precisely and firmly into the bomb bay with a 2-inch clearance below to allow the bomb doors to close. The bomb was locked into place using the same device employed for Second World War Tallboy and Blockbuster bombs on the Lancaster. This had two curved S-shaped arms locked into place at the top by a gudgeon pin.66 There was an access hole in the top of the bomb casing to allow IFL to operate. 48 JONATHAN AYLEN

IFL was promoted as a safety device to avoid an accidental yield or contamination. This inherently safe concept owed much to John Rowlands who was seconded to HER, Fort Halstead, from the RAF.67 But there was a more pragmatic reason for wanting to see the core separated from the bomb in flight: if the bomb had to be jettisoned, the precious plutonium core was retained on the plane. The arming device was initially loaded into the plane from above through a hatch on the top of the fuselage.68 A vertical box section inside the fuselage above the bomb bay contained the mechanism for inserting the core into the bomb. At its lower end, the loading device rested on the edges of an opening in the outer shell of the explosive of the bomb. A bayonet connection locked the insertion mechanism into the sphere by a push and turn action. The loading device allowed for movement of the bomb relative to the aeroplane. The arming device — or cartridge — was a slightly tapered test-tube-shaped device like a geological core sample. It contained the spherical plutonium core of the bomb at its base (with its initiator) plus a complete vertical bore through the rest of the sphere including a section of the uranium tamper and segments of the two types of explosives. When the IFL was actuated, an electric motor pushed the arming device into the heart of the sphere from above, inserting this missing section of the sphere precisely between the detonators. Once this transect plug was inserted, the bomb was complete and symmetrical and ready to be dropped. As the bomb fell away, the hole in the top of the bomb was closed by a spring loaded ‘finger chopper’ plate stowed just behind the hole. The IFL was designed, and then redesigned at RAE insistence, by Frazer Nash.69 The electric actuator motors were made by Miles Aircraft at Shoreham, with a built- in brake. A pair of actuators was used, coupled by a differential, to drive the vertical movement using two helical screws. Considerable technical effort went into refining the design of the IFL system by the RAE even after it was installed on Valiant. IFL was not adopted for Victor and Vulcan aircraft and abandoned sometime after summer 1956 for reasons which are not clear, but standing on the slippery back of a Valiant on a snowy, cold night handling a heavy cartridge containing radioactive material and explosives would not have been an attractive job for two aircraftsmen.70 IFL was replaced by Last Minute Loading whereby a Navigator-Radar from the flight crew inserted the necessary cartridge into the bomb from below to make it live just prior to take-off.71 This system too was not without its shortcomings.

Deployment and storage New technologies require new operating practices. Quite apart from new bombers and new strategic doctrines, the Royal Air Force had to develop a new set of routines to handle nuclear weapons, including armed convoys to move bombs around and ‘special storage’ facilities at depots and airfields.72 A new set of procedures was developed for the loading, arming and release of the weapon. These new measures forcibly illustrate the radical nature of the atomic bomb as an innovation in warfare. These routines were developed under the leadership of the Herod Committee (‘High Explosives Research Operational Distribution’) and the related Salome Com- mittee.73 A special training school was set up at RAF Wittering called the ‘Bomber FIRST WALTZ 49

Command Armament School’ (BCAS), formed on 1 August 1953 under Wing Commander J. S. Rowlands to train aircrew and engineers in the use and servicing of nuclear weapons.74 Two ‘special storage sites’ were built at RAF Barnham in Suffolk and RAF Faldingworth in Lincolnshire to store and carry out simple maintenance on Blue Danube.75 The ‘top site’ store at Barnham, south of Thetford, was operational for just ten years between 1953 and 1963. Blue Danube was made up a small fissile core and a large high-explosive sphere. This combination of radioactive material and high explosives dictated the layout of the site at Barnham. Within the inner compound or ‘danger area’ there were three high explosive storage sheds for the casings and explosives known as the DD Buildings. There were also four sets of individual ‘igloos’ for storage of the plutonium cores distributed around the site, totalling fifty-seven in all. There was a central maintenance building and another much smaller building for maintaining the weapons. The site is built around a circular inner road. The secure area is surrounded by a concrete panel security fence and an outer chain link fence a kilometre long with watch towers at each corner. The ‘free fire’ zone between fences was patrolled by armed guards. In common with other Cold War installations, the bomb store is on a large scale. Igloos for core storage were simple concrete shelters about 2 metres high fitted with wooden doors, remote alarms, secure door locking and a circular stainless-steel locked safe fitted into the floor (Figure 10). The cores were light and compact and amenable to manual handling so these storage huts were spread around the site reached by narrow concrete paths. In contrast, the heavy bomb carcase required road access, travelling cranes and mechanical handling devices (although the overhead gantries at the DD sheds were seldom used). Since the bomb casings contained high explosive, lined with their uranium tampers, the three large concrete-framed sheds had side walls that were not bonded into the pillars or the roof. In turn, these sheds were surrounded by earth banks to help contain and deflect any explosion upwards. These sheds were air-conditioned. Typically the bombs were stored in a fishbone pattern on their trailers allowing swift removal in the event of fire. New routines were developed for handling the new atomic bomb.76 There were always two RAF personnel in charge of handling or releasing a nuclear core. To release a core from storage needed an ‘Equipper’ with the combination of the igloo and an ‘Engineer’ with the key to the padlock of the safe inside. They would then walk together with the core to the load carrier. Casings were always transported separately from cores. Again, the two-man principle was always used on the aircraft: the ‘NavRadar’ would release the bomb, but the Captain controlled the bomb release safety lock on the final pin preventing the bomb dropping.

Supply chain: manufacture, assembly and delivery It is clear the Blue Danube supply chain shifted over time as various manufacturers gained competence and a Royal Ordnance Factory site at Burghfield in Berkshire emerged as an assembly centre, taking over from Chorley.77 Sections of the weapons were transported to RAF Barnham or RAF Faldingworth for onward deployment to V-bomber airfields in the UK and Cyprus or retained for ‘second strike’ storage. 50 JONATHAN AYLEN

figure 10 Igloos at RAF Banham Special Storage Unit were used to store the plutonium cores. This is number 35. The large number of igloos may have been designed to intimidate and impress as it is unlikely they were all used either here or at Faldingworth. Author

There were many curious features to these covert logistics. For instance, the tail was assembled in Nissen huts at a semi-derelict ROF site at Elstow, south of Bedford. Entrance to this unguarded and unmarked site was concealed behind rusty gates in a hedge off a roundabout on a main road.78

Conclusion: evolution of a technology Blue Danube was the first British nuclear deterrent deployed during the Cold War carried by the V-bomber force. Here we consider how the bomb was designed, pro- cured, manufactured, assembled and delivered to airfields and bomb stores, focusing on the practical evolution of the technology. Production of components was widely distributed across the public and private sector in England and Wales. Focus on the fissile aspects of atomic weapons development means the difficulties of making the whole weapon have been overlooked. It was a struggle to develop and manufacture reliable electronic components, notably the fuzes and firing units. There was an acute shortage of electronics components three years after the weapon was first supplied for operational use. Ironically, deployment was constrained not by lack of fissile material, but by a shortage of simple, clockwork-based time fuzes. FIRST WALTZ 51

So, development of Blue Danube is a story of paradoxes. Problems with high explosives turned out to be as much a constraint on serviceability as the novel fissile material. Extraction of weapons-grade plutonium was described as a ‘brilliant success’.79 Yet, HER struggled to design the complicated explosive jacket for Blue Danube. In broader terms, the story here contradicts conventional accounts of technical development. The ‘linear model’ of development suggests science leads to technology, then to prototypes, on to testing and design and then closes to a reliable final product.80 In contrast, Blue Danube was an imperative of ‘build first — test later’. Many components were only tested and reliably manufactured after the weapon was delivered. The operational weapon was modified, as if it were a prototype. There were two different ‘priming’ systems and casings during the bomb’s brief history. Production decisions were often based on rules-of-thumb and judgement rather than science and experiment. Blue Danube was more a process of learning than a finished product. It was deployed first and developed afterwards. Our account shows aspects of atomic weapons technology were distributed across agencies such as the RAF, the RAE, the Ministry of Supply and a range of manufacturing firms. There was a strong interlock between the military needs and the civilian economy. The many innovations needed to bring the weapon into use were spread around the UK. Blue Danube’s practical development was a product of a wider ‘warfare state’ of Cold War Britain, not just the product of a few boffins. In truth, the first atomic bomb was designed in suburban centres and built in ordinary factories down prosaic back streets by regular workers — men and women — in towns across industrial Britain. The atomic bomb was the product of northern industrial towns and the suburbs of southern England as much as the secret citadels of Fort Halstead, Aldermaston and Burghfield. So, Blue Danube was a provisional deterrent — frequently modified as problems cropped up. Blue Danube continued to evolve after it was designed. In this respect it was also a learning deterrent as the RAF developed new routines and the supply chain learned how to manufacture and maintain nuclear weapons. It was also a half-way house deterrent as those who had been at Los Alamos knew of the possibility of a fission-fusion weapon, an H-bomb. Above all, Blue Danube — however provisional — was a British deterrent. The weapon was developed and deployed with awesome professionalism and with emphasis on the credibility of the threat it posed. Cheap but practical development of the atomic bomb saved deploying resources elsewhere. As Peter Barker said of ARL’s Drawing Office:

We joked that somewhere there must be a large brilliant team — with boundless resources — doing the ‘real job’, and that we were just a decoy [. . .] it was inconceivable that ARL’s small-scale activity could possibly be a major part of the development of the British Nuclear Deterrent 81 It kept the UK rather forcibly in the great power game and encouraged the Americans to renew cooperation over nuclear secrets. But the triumph of the technology was as much prosaic as scientific. 52 JONATHAN AYLEN

Acknowledgements Participants in the design, procurement, manufacture, storage and deployment are the key source of evidence. They are a remarkable group of people who have been generous with time, advice and discussion of artefacts while maintaining an appropriate line between disclosure and secrecy. Some ‘private’ sources preferred not to be acknowledged. Among others, we acknowledge the welcome help of John Allen, Mike Allisstone, Adrian Armishaw, Peter Barlow, Doug Bateman, Brian Burnell, Wayne Cocroft, Tim Crichton, Keith Eldred, Mike Igglesden, Brian Jackson, Keith Julian, ‘Jeff’ Jefford, Reg Milne, Bill Taylor, Bill Tivenan, Mike Watkins, John Weller, Derek Whitehead and at least five unnamed contributors. Key public sources are the National Archives at Kew and the RAF Museum Hendon. In addition, Barnsle y Archives, Liverpool Maritime Museum, Mansfield Museum, Sywell Aviation Museu m and the Tasker Trust patiently fielded enquiries. The Ministry of Defence and AWE supplied material very efficiently under the Freedom of Information process. Earlier versions of the paper have been presented to the Defence Electronics History Society, Shrivenham; the Newcomen Society, London; IET, Doncaster and the Nuclear History Conference, Charterhouse.

Notes 1 Peter G. E. F. Jones, ‘Overview of History of in association with AWE plc, Aldermaston, UK Strategic Weapons’, in ‘The History 2000) is a cautious introduction. of the UK Strategic Deterrent’, The Royal 4 The Ministry of Defence helped at arm’s Aeronautical Society Proceedings, Wednes- length through a freedom of information day 17 March 1999, p. 2.2. procedure, declassifying a number of docu- 2 Richard Rhodes, The Making of the Atomic ments and photographs for this project. Bomb (New York: Simon & Schuster, 1986); They bear no responsibility for the views Chuck Hansen, U.S. Nuclear Weapons: The expressed here. On the difficulties of Secret History (Arlington, Texas: Aerofax, researching nuclear weapons, see John R. 1988); Margaret Gowing, Britain and Atomi c Walker, British Nuclear Weapons and the Test Ban 1954–1973 (Farnham: Ashgate, Energy 1939–1945 (London: Macmillan, 2010), ch. 1. 1964); Brian Cathcart, Test of Greatness: 5 Interviews are relevant since it has been Britain’s Struggle for the Atom Bomb (Lon- argued that nuclear weapon design is charac- don: John Murray, 1994); David Holloway, terized by a high degree of tacit knowledge. Stalin and the Bomb: The Soviet Union and But this is true of other advanced technolo- Atomic Energy, 1939–1956 (New Haven: gies, for example aircraft production. See Yale University Press, 1994). Donald MacKenzie and Graham Spinardi, 3 An exception is John R. Walker, ‘Potential ‘Tacit Knowledge, Weapons Design, and the Proliferation Pointers from the Past: Lessons Uninvention of Nuclear Weapons’, Ameri- from the British Nuclear Weapons Program, can Journal of Sociology, 101.1 (July 1995), 1952–69’, The Nonproliferation Review, 44–99, and Walter G. Vincenti, ‘Technologi- 19.1 (2012), 109–23 emphasizing the diffi- cal Know-How Without Science: The culty of manufacture and deployment. Innovation of Flush-Riveting in American David J. Hawkings, Keeping the Peace: The Airplanes, circa 1930–circa 1950’, Technol- Aldermaston Story: A Brief Account of the ogy and Culture, 25.3 (July 1984), 540–76. First Fifty Years of the Home of Britain’s Absence of tacit knowledge has not stopped Nuclear Deterrent, the Atomic Weapons proliferation of nuclear weapons, which sug- Establishment (South Yorkshire: Leo Cooper gests practical know-how can be developed FIRST WALTZ 53

in the same way that Britain and the Soviet 14 AVIA 65/1114 Ministry of Defence, Defence Union did from 1946. Research Policy Committee — Atomic 6 On technology shaped by local practices, see Energy Sub- Committee, Supporting Docu- Mikael Hard, ‘Technology as Practice: Local ments 1956, File No. 407/090, 29 November and Global Closure Processes in Diesel- 1956, ‘Serviceability of Blue Danube Mark I Engine Design’, Social Studies of Science, and Mark II, Note by the Air Ministry’. 24.3 (August 1994), 549–85. For technology 15 Gregg Herken, The Winning Weapon: The as evolutionary problem solving, see Atomic Bomb in the Cold War, 1945–1950 Jonathan Aylen, ‘Stretch: How Innovation (Princeton, New Jersey: Princeton University Continues Once Investment is Made’, R&D Press, 1988 edition), fn. pp. 196–97. Management, 43.3 (June 2013), 271–87, and 16 Ferenc Szasz, British Scientists and the Man- R. Ramlogan, A. Mina, G. Tampubolon and hattan Project: The Los Alamos Years (New J. S. Metcalfe, ‘Networks of Knowledge: York: St Martin’s Press, 1992). Gowing The Distributed Nature of Medical Innova- (Britain and Atomic Energy) concludes ‘In so tion’, Scientometrics, 70.2 (2007), 459–89. far as “know-how” went the British were 7 David Edgerton says England was above well equipped for the postwar construction all a technological nation when it came to of atomic bombs’ (p. 267). prosecution of war. Blue Danube was also 17 Wayne D. Cocroft, Fort Halstead, Dunton sourced from Llanishen near Cardiff and Green, Sevenoaks, Kent: A Brief Assessment Bangor in North Wales. David Edgerton, of the Role of Fort Halstead In Britain’s England and the Aeroplane: Militarism, Early Rocket Programmes and The Atomic Modernity and Machines, 2nd edn (London: Bomb Project Research Department Report Penguin, 2013), p. 103. Series No. 49-2010 (Portsmouth: English 8 Reg Milne, who looked after the centre Heritage, 2010); Wayne D. Cocroft and section of Blue Danube, discussing ARL Roger J. C. Thomas, Cold War: Building Farnborough, 6 August 2013. This individu- for Nuclear Confrontation 1946–1989, ed. by alized problem solving on Blue Danube P. S. Barnwell (Swindon: English Heritage, was very different from the fluid, open and 2003); Paddy Heazell, Most Secret: The Hid- cooperative ‘communities of practice’ that den History of Orford Ness (Stroud: History emerged in the development of the Blood- Press, 2010). On the workings of Fort hound guidance system. See Jonathan Aylen, Halstead, see Lorna Arnold with Katherine ‘Bloodhound on my Trail: Building the Pyne, Britain and the H-Bomb (Basingstoke: Ferranti Argus Process Control Computer’, Palgrave, 2001), ch. 6. International Journal for the History of 18 Margaret Gowing with Lorna Arnold, Engineering & Technology, 82.1 (January Independence and Deterrence: Britain and 2012), 1–36. Atomic Energy, 1945–1952, 2, Policy Execu- 9 Air Commodore Mike Allisstone, former tion (London: Macmillan, 1974), 473. Equipment Officer, 26 February 2014, 19 Ibid., p. 369. discussing RAF Barnham. 20 Professor John Allen, former Senior Scien- 10 We have not been able to locate this tific Officers, ARL, interview 5 September document. 2013. The first ‘Fat Man’ bomb dropped at 11 See Peter Hennessy, ‘Cabinets and the Bomb’, Bikini Atoll in July 1946 missed its target by British Academy Occasional Paper 11 2130 feet. See Hansen, p. 31. On the origins (Oxford: Oxford University Press, 2007), of ‘Fat Man’ and why the ‘Thin Man’ gun- pp. 49–59. type plutonium bomb was abandoned, 12 Cathcart covers Woolwich and the first see Lillian Hoddeson, Paul W. Henriksen, Hurricane explosion at Monte Bello Islands. Roger A. Meade, and Catherine L. Westfall, 13 Humphrey Wynn, The RAF Strategic Nucle- Critical Assembly: A Technical History ar Deterrent Forces: Their Origins, Roles of Los Alamos During the Oppenheimer and Deployment 1946–1969, a Documentary Years, 1943–1945 (Cambridge: Cambridge History (London: HMSO, 1994) covers University Press, 1993). delivery of the weapon (p. 92) and the trial 21 Lord Hinton, ‘Atomic Energy’, in A History drop (pp. 97–98 and 170–73). of Technology, Volume VI, the Twentieth 54 JONATHAN AYLEN

Century c.1900 to c.1950, ed. by Trevor I. 25 While there is no documentary support, Williams (Oxford: Clarendon Press, 1978), there is convincing private evidence this form pt 1, ch. 10, pp. 223–67 is a frank account of of non-destructive testing took place, for plutonium preparation for the bomb includ- instance in building 58 at RAF Barnham ing the choice of a packed column design Special Storage Site. Informant, 17 October for the solvent extraction process. Gowing 2013. Norman Skentelbery, Arrows to Atom argues secrecy ‘led to serious undervaluation Bombs: A History of the Ordnance Board, of a very large industrial enterprise which 2nd edn (London: HMSO, 1975) argues ‘the was notable for the magnitude of its tasks conventional explosives associated with and achievements’. Margaret Gowing, atomic weapons are much more a source of Reflections on Atomic Energy History, the risk than the stable fission material’, p. 175. Rede Lecture 1978 (Cambridge: Cambridge 26 Shortcomings of the US ‘Fat Man’ Mk III University Press, 1978), p. 20. weapon are documented in Hansen, ch. 2, 22 Gowing and Arnold, pp. 469–70. The tiny ‘Postwar U.S. Fission Weapons Development’. initiator encapsulated in beryllium was called 27 File AVIA 65/1163 Weapon Principle: Implo- an urchin, because it resembled a sea urchin sion, running from 1 January 1947–31 in size and shape. Code-named ‘kitten’. The December 1953 is listed at the NRA as being case is said to be the size of a golf ball. The ‘retained by a government department under ‘urchin’, or initiator, was an intense neutron Section 3.4 of the Public Records Act, 1958’. source at the heart of the core to make sure Peter Barker, former draughtsman at ARL said ‘The BD project predated the more the chain reaction in the fissile plutonium formalized ‘Systems’ approach to project started at the moment of implosion. The management [. . .] Hence there was no gener- design involves a combination of beryllium ally understood overall Project Structure, no and a tiny amount of polonium. The known formal definition or specification of polonium for the initiator was produced by Sub-Systems and Interfaces, no generally Harwell and Risley (and subsequently Wind- known Project Reviews [. . .] and so it was scale) from irradiated bismuth. Invented by always bravely “feeling its way”, hampered James L. Tuck who went to Los Alamos by the demands of security which argued for from Oxford in May 1944. See Hansen, compartmentalization which in turn led to p. 35. Tested at the ‘Operation Totem’ an inefficient project development due to Kittens trials between 26 September and enforced absence of communication between 6 October 1953 at the EMU site in Australia, design teams’ (correspondence, 11 August see J. L. Symonds, A History of British 2014). Atomic Tests in Australia (Canberra: Aus- 28 Arm. 1775/P/SAH/130, HER Working Party tralian Government Publishing Service for ‘C’ Minutes of the first meeting at Fort Department of Resources and Energy, 1985), Halstead on Thursday 28 September 1950 pp. 157–59. (private collection, Hampshire). It is clear 23 Drawing No. Arm. 37437/A Issue C. Osten- from TS 2872 that work at Farnborough, sibly drafted by ‘Tim’ Sims, the Senior such as wind-tunnel trials on high-speed Draughtsman of ARL, it was traced (and bomb release, began before May 1948. probably drawn) by Brenda Adams. See RAF 29 Walter G. Vincenti, What Engineers Know Museum, Hendon, AC74/27/2/11/5 Papers and How They Know It — Analytical Stud- relating to Blue Danube, 1953–1955; H/332/ ies from Aeronautical History (Baltimore: 3103/00002. John Hopkins University Press, 1990). 24 A note scribbled on a copy of TS 2872 30 Cathcart, p. 48; Gowing and Arnold, ch. 13 suggests the overall physics package weighed ‘The Men’. 6200 lb — 2.8 tonnes. ‘Ballistic Casing’. Notes 31 AVIA 65/1160, A Technical Review of Bomb, of Meeting held 11 a.m. 11 May 1948 — Aircraft, H.E. 10,000 lb., M.C., M.S.1804 (?), C.S.A.R. in the Chair. A meeting led by Dr /56/D.G. Eng., p. 7, para. 46, 12 December Penney ‘to initiate preliminary investigation 1956. A view confirmed by Mike Watkins, into the design of the ballistic casing’ (Pri- who said: ‘From my RAF experience I have vate collection, Hampshire). always felt that the electronics at least could FIRST WALTZ 55

have been designed better [. . .] surprised that with Dave King in charge. Neither com- lead acid batteries were used for power’ (14 pound is mentioned in a guide to the site, August 2013, email exchange). The official Royal Aircraft Establishment Farnborough, judgement is complicated by the fact that the Open Days, Friday 9 June 1961 and Saturday ‘engineer’ grades were only just being intro- 10 June 1961, Guide to Exhibits (Farnbor- duced into the Civil Service at the time, so ough: Ministry of Aviation, April 1961). Sid some of the scientists at HER Fort Halstead Hunwicks is not listed among the Heads of and many at RAE Farnborough were trained Department on p. 47 of the Guide. as engineers. 36 R. I. Vaughan and J. E. Allen, The Effect of 32 Innovation often occurs at the junction of High Release Speed on the Design of Blue technologies, see John Becklake, ‘The V2 Danube, AVIA 6/19446 (Royal Aircraft Rocket — a Convergence of Technologies?’, Establishment, Farnborough, RAE, Techni- Transactions of the Newcomen Society, 67 cal Note No. Arm. 502, December 1952) and (1996), 109–23, and Jonathan Aylen, ‘Open interview with J. E. Allen (5 September Versus Closed Innovation: Development 2013). The central problem is the bomb had of the Wide Strip Mill for Steel in the USA low density and a large surface area, see G. during the 1920s’, R&D Management, 40.1 (January 2010), 67–80. J. Richards, The Ballistics of Blue Danube, 33 On the growing role of Farnborough, see AVIA 6/17799 (Royal Aircraft Establish- AVIA 65/1166 Weapons Programme, file ment, Farnborough, RAE, Technical Note 14 July 1955 to 21 March 1961, and George No. Arm. 413, by May 1949). Hicks, ‘History of RAE and Nuclear Weap- 37 The flip-out fins were designed by ‘Cottie’ ons’, ed. by Roy Dommett, Prospero, 2 Cotsworth, then Leading Draughtsman (Spring 2005), 155–78. For the orthodox role at ARL (Peter Barker, communication 11 of RAE, see Andrew Nahum, ‘The Royal August 2014). The duplicate nitrogen bottles Aircraft Establishment from 1945 to Con- were supplied by Walter Kidde. Gas release corde’, in Cold War Hot Science: Applied was initiated by a bomb bay lanyard as the Research in Britain’s Defence Laboratories weapon dropped. A blast of gas drove a pis- 1945–1990, ed. by Robert Bud and Philip ton and four pairs of connecting rods which Gummett (London: Science Museum, 1999), pushed the fins upwards and outward by pp. 29–58. Particular thanks are due to Doug parallel motion linkage. The fins were locked Bateman for his research on the early history by driving the connecting rods over dead- of ARL (interview 30 May 2013, Farnbor- centre. The fins were zirconium-aluminium ough). Doug Bateman argues one reason and ran in guide rails which provided the for RAE’s initial involvement: ‘there was a necessary rigidity. The short stub fins at shortage of scientists and technicians at this their base were sometimes referred to as time and RAE was long established and had ‘glove boxes’ (John Allen, 5 September 2013). the knowledge and skills’. The design was modified at least once, espe- 34 Sid Hunwicks was, perhaps, a Principal Sci- cially after a battery door fell off and dam- entific Officer at this stage in charge of ARL aged a fin glove on a trial drop in Australia at Farnborough. He had more influence on as the bomb went transonic (interview, Mike nuclear policy than his lowly title suggests. He sat on the Director of Air Armaments Igglesden, 1 October 2013). On the use of committee at the Ministry of Aviation which lanyards to activate the tail fins, see AIR controlled the ‘Nuclear Weapon Programme’. 10/5677, otherwise AP 1664A, Bomb Han- See AVIA 65/1166. dling Equipment, Vol. 1 (Amendments), 35 Naturally, a radio laboratory needed to be as coverage 1950–59, Bombs and Supply Carri- far away from interference as possible — the ers and Associated Equipment. A three-page story told by John Allen, 5 September 2013. amendment dated August 1952 discusses ‘the Hicks gives a different interpretation of the Mark 2 Lanyard Retracting Unit — a spring ARL name. A second compound was set up operated device for use in conjunction with for adoption of US nuclear weapons from tail units having retractable fins’. The nylon 1958 in E Hangar, a quarter of a mile away lanyards were colour coded by length. 56 JONATHAN AYLEN

38 The retractable fingers for Valiant were front which led into a single storey work- developed and tested by Vickers at Wey- shop with roof lights. The workshop must bridge. Vortex formation at the front of the have been two houses wide’. Each air bag open bomb bay was worse on the Victor was fitted with a schraeder valve. with its upwardly sloping underbelly. Inter- 43 Problems with the fuzes are discussed view with Prof. John Allen (5 September extensively in Ministry of Supply file AVIA 2013 and subsequent conversations). 65/1160, 10,000 lb HE MC bomb: acceptance 39 A history of the firm coyly reports after the trials and acceptance for service use, 1951– war ‘Many secret projects were coming the 60. way of the Aircraft Section of the Railway 44 These W/R units were controversial. John Foundry from the Ministry of Defence. Allen claims they were unduly complex with Many of these cannot be recorded due to 80 valves in each unit and so prone to failure their classified nature’. The firm’s role in the (conversation, 8 November 2013). Trials to military-industrial complex belies their allay concerns that the various fuzes might benign ‘Thomas the Tank Engine’ image. interfere with one another are reported in A. See Ronald Nelson Redman, The Railway B. Hillan, ‘The F5 Trial — Performance of Foundry Leeds, 1839–1969: E. B. Wilson- W/R Units in the Blue Danube Nose Assem- Hudswell Clarke and Co Ltd (Norwich: bly’, Atomic Weapons Research Establish- Goose and Son, 1972), p. 113. ment, Report No. B21/56 Development of 40 Interview with Brian Burnell, former keeper Blue Danube W/R Fuzing System, Part IV/3 of drawings, Hudswell Clarke Aircraft Sec- (dated February 1957, received 21 January tion, London, 9 January 2014. Not one item 1957, AWRE Aldermaston, Berks, National relevant to Blue Danube remains among the Archive ES 2/67). 3000–4000 drawings of supplier Whiteley 45 The time fuze was designed by Vernon Rut- Electrical Radio Company lodged in Mans- land at Farnborough. These were duplicated. field Museum (search 17 December 2012). The two clocks were meant to start together, 41 Gowing and Arnold, pp. 470–71 suggest the activated by a bomb bay lanyard. They firm assigned two of their best designers to were worked by a plunger which drove the the work. Reg Milne of RAE who was clockwork on both fuzes. It needed a push to responsible for the centre section of Blue wind up the clockwork. Unfortunately, the Danube recalls these were R. P. Pedley and dimensions of the ‘cocking mechanism’ were L. G. Frise (Milne, interview, Farnborough, wrong. There was tilt on this mechanism 6 August 2013). Mr Pedley, Research Engi- and the fork would move to one side mean- neer, Percival Aircraft Ltd, received an MBE ing one fuze was wound more than another. in 1954. (Also see Wynn interview with The less wound fuze would run down ‘George’ Pedley, p. 88.) The firm operated sooner, leading to premature detonation. under the name Hunting Percival Aircraft Reg Milne’s solution was to add a couple Ltd from 1954. The title Hunting Aircraft of washers to prevent this recurring during Ltd was adopted in 1957. Later work was the trials at Maralinga. (Reg Milne, 8 May done as ‘Special Weapons Division of 2014.) Hunting Aircraft Limited’, later ‘Hunting 46 The troubleshooting episode is documented Engineering Ltd’, abbreviated ‘HEL’ in in E. T. Jenkins, ‘W/R Unit Production’, government documents. A history of the firm Atomic Weapons Research Establishment, is silent on their atomic bomb work. See Report No. B14/56 Development of Blue R. Silvester, Percival and Hunting Aircraft Danube W/R Fuzing System, Part III/1 (date d (Luton: R. J. Silvester, c. 1987). July 1956, received 7 May, 1956, AWRE 42 Reg Milne, Farnborough (interviews, 6 Aldermaston, Berks, National Archive ES August 2013 and 8 May 2014 and correspon- 2/61). dence). William Freeman & Company Ltd 47 This evaluation is given in file AVIA 65/1160, operations were then based at Subaseal A Technical Review of Bomb, Aircraft, HE Works, Peel Street, Barnsley. Reg Milne says 10,000 lb MC, MS.1004/56/D.G.Eng. dated he entered through an unassuming ‘shop 21 February 1957. Notice the distrust of FIRST WALTZ 57

industry by senior civil servants — this 20 April 1956, National Archive ES 2/64). It distrust was mutual. See Richard Moore, deals with attempts to reduce the failure rate Nuclear Illusion, Nuclear Reality: Britain, of condensers built into a component of the the United States and Nuclear Weapons, W/R ground proximity radar fuze. 1958–64 (Houndmills, Basingstoke: Palgrave 53 AVIA 65/1792 History of Nuclear Weapon Macmillan, 2010), pp. 246–47. Production in the UK is an invaluable source. 48 This detonator was patented: US Patent 54 Ibid., p. 7, quote on reliability. 3,040,660, ‘Electric Initiator with Exploding 55 AVIA 26/2033; G. W. A. Dummer, Potted Bridge Wire’, Lawrence H. Johnston, Los Electronic Circuit Techniques, RRE Report Angeles, Calif., filed 8 November, 1944, no. 3002, Malvern: Ministry of Supply, published 26 June 1962 assigned to the Unit- Radar Research Establishment, May 1954. ed States of America as represented by the Also advice from Derek Whitehead (10 United States Atomic Energy Commission. It September 2013), Fred Axon (15 September reveals a half a joule provides enough energy 2013) and an amusing but anonymous source to initiate an explosion and this can be ide- (Shrivenham, 17 October 2013). ally provided by a condenser and suggests 56 Jenkins. synchronous operation is possible across 57 Their surviving drawings attest to a business detonators to within half a microsecond. devoted to loudspeakers, early FM radios, Also see Rhodes, p. 655. amplifiers, cabinets and Bakelite switches. 49 Cathcart, pp. 70–71. The key task of devel- The firm’s lead product was its range of oping the detonators was led by E. H. Mott ‘Stentorian’ loudspeakers. See Whiteley Elec- and C. M. Bean. On the AWRE production trical Radio Company Limited, Mansfield, site, see Sarah Newsome and Andrew Notts., spiral-bound publicity pamphlet Williams, Woolwich Common, Woolwich, and catalogue, 1958, Mansfield Museum Greater London: An Assessment of the His- collection. toric Environment of Woolwich Common 58 Heazell, pp. 162–65 discusses trial drops of and its Environs, Survey Report, Research Blue Danube at Orforness ‘testing [. . .] the Department Report Series no. 098-2009 trigger mechanisms by sensitive radio telem- (Swindon: English Heritage, 2009), pp. 14– etry to check that the very elaborate firing 17. On detonator operation, see R. J. Baker relays would work correctly’. and C. M. Bean, ‘The “Potting” of Unit 18’, 59 These specially designed Extra High Tension Atomic Weapons Research Establishment, condensers were probably made at their Weapons and Assembly Division, Report No. B7/53 (dated July 1953, AWRE Alder- Wood Lane Research Centre, London (inter- maston, Berks, National Archives ES 2/7). view with Dr Keith Julian, former Director 50 On John Challens, see Gowing and Arnold, of Technology, BICC Cables Ltd, 7 March p. 464; Cathcart, pp. 66–70; John Allen 2013). Confirmed by National Museums interview, 5 September 2013. on Merseyside, BICC Collection, Item VI 70 51 A key textbook is Ian Alexander D. Lewis 343, ‘The Products of BICC. September and Frank Herbert Wells, Millimicrosecond 1954’: ‘BICC are prepared, wherever circum- Pulse Techniques (London: Pergamon Press, stances warrant it, to manufacture special 1954). Both authors were at the Atomic items. For this purpose research into associ- Energy Research Establishment, Harwell. ated problems will be carried out in any of Generous help from confidential sources is its Research Laboratories’. Even up to the warmly acknowledged at this point (inter- late 1950s capacitors were made of paper views, 19 December 2012 and 16 January soaked in a salt bath and prone to leakage. 2013). They are subject to dielectric failure and the 52 For example, J. S. Wise, ‘Oscillator Anode metalized paper volatilizes. In short, not very Capacitor Assembly’, Atomic Weapons reliable. Research Establishment, Report No. B18/56 60 AVIA 65/1792, 1954 onwards. Development of Blue Danube W/R Fuzing 61 Ibid., pp. 7–9. System, Part III/5 (dated June 1956, received 62 Ibid., p. 14. 58 JONATHAN AYLEN

63 The firm had been founded as Pearson 67 John Rowlands played the lead role in the Electrical Company. A scion of the Ferranti RAF’s adoption of nuclear weapons. Then a family, Dennis de Ferranti took over the firm Wing-Commander, he was one of ten serving and pursued a policy of price competition to RAF officers seconded to Fort Halstead to enlarge sales at the expense of the existing ensure the bomb met service specifications. meter cartel. See John F. Wilson, Ferranti: A He went on to make a huge contribution to History, Building a Family Business, 1882– weapon design, servicing, handling, training 1975 (Lancaster: Carnegie Publishing, 2000), and safety (Wynn). His remarkable role pp. 296–97. The firm moved to Bangor in is overlooked in ‘Air Marshal Sir John North Wales, occupying an old shadow fac- Rowlands’, Daily Telegraph, 7 June 2006. tory. A private source indicated their secure 68 Form 555. Blue Bunny had a similar mecha- component manufacture for Blue Danube nism, see AVIA 65/2036. ‘was done in an old wooden building out the 69 Frazer-Nash was an idiosyncratic firm which back’ (interview, 19 December 2012). produced gun turrets during the Second 64 Cathcart, pp. 136–40, speaks of three com- World War. This may have recommended peting core designs. The broad configuration them to Air Vice-Marshall A. D. ‘Dizzy’ of the physics package was due to Klaus Davis at the Ministry of Supply. They made Fuchs and permitted In Flight Loading. The two designs for In Flight Loading of Blue detailed local design was developed by John Danube. Components were manufactured by Corner and Herbert Pike and chosen Febru- Palatine Tool and Engineering located on the ary 1951. This approach lasted for, perhaps, A3 Kingston By-Pass at Surbiton in Surrey. four years. Cathcart wrongly suggests In The drawing on display at the AWE Educa- Flight Loading involved horizontal insertion tional Collection is a later RAE design by of the cartridge. This was the layout used by Ben Sykes and Reg Milne. The Frazer-Nash the Blue Peacock atomic mine AVIA 65/2036 design was subsequently adapted for other Blue Peacock: atomic land mine. Also see use by Peter Barker. David Hawkings, ‘Blue Peacock: The British 70 The decision to abandon In Flight Loading Army’s Forgotten Weapon’, Discovery (no and alter the design of Blue Danube is date, AWE house magazine, privately circu- anticipated in AVIA 65/1166 Weapons lated), pp. 42–43. The version of Blue Dan- Programme, ‘Notes on a meeting to discuss ube deployed on the Valiant used vertical In D.A. Arm’s Nuclear Weapon Programme Flight Loading. held at St. Giles Court on 16 July, 1956’, 65 In Flight Loading on Valiant is covered by section 2.1 Blue Danube. Form 555, Ministry of Supply, Report on 71 A cartridge or ‘gauntlet’ known in RAF par- Examination of Aircraft, To: R.D.Q.B., from lance as a ‘golden goolie’. Among others, D.A.L. Robson, R.T.O. Date: 7 November former ‘Nav Radar’ John Weller, interview 3 1954. Aircraft Valiant B Mk.1, Firm Vickers February 2014. The design of this cartridge Armstrongs Ltd. Address: Weybridge, Surrey was developed by Peter Barker, responsible entitled Installation of 10,000 lb. M.C. Bomb for the appearance of a ‘Russian Spy’ in Mk. I in RAF Museum Hendon, file Figure 4. The plutonium hemi-spheres were AC74/27/2/11/5 papers relating to Blue gold-plated. Last Minute Loading involved Danube, 1953–55; H/332/3103/00002. little change to the cartridge assembly, 66 A trial inert bomb fell off into the bomb bay although it was now locked in place manu- of a prototype Valiant over Dorking during ally rather than mechanically. But the Blue ballistic trials on 30 July 1952. It was jetti- Danube casings had to be redesigned to soned over the Thames Estuary with a near allow access from below rather than from miss on a yacht. It was not recovered. Brian above, and the explosive sphere was inverted. Trubshaw was the test pilot at the time (Reg 72 The notion of ‘routines’ evolving over time Milne, 6 August 2013; email exchange, Mike in response to changed circumstances is due Atkins; Brian Trubshaw and Sally Edmon- to Richard R. Nelson and Sidney G. Winter, son, Brian Trubshaw: Test Pilot (Stroud: An Evolutionary Theory of Economic Alan Sutton, 1998), p. 35. Change (Cambridge, Mass: Belknap Press of FIRST WALTZ 59

Harvard University Press, 1982). These many were active at any one time. Consensu s routines included a new bombing strategy: suggests 57 were deployed and one tested, no bomber stream or pathfinders, one pre- see Moore, appendix 1, p. 256. There were a planned target per aircraft, prior recruitment large number of trial, ‘practice inerts’ and and training, war of limited duration and replacement casings made — well in excess blind bombing by radar according to former of 250 to judge by component orders. ‘Nav Radar’ Dave Lingard, ‘Bombing on the They were used for trials at Orfordness, a tin triangle’, Lecture Newark Air Museum, programme of drops at Maralinga and for 20 October 2012. training purposes. A dozen Blue Danube 73 See Wynn, ch. 6. casings were used for Violet Club, a short- 74 Interview with Squadron Leader Bill Tive- lived variant fitted to Vulcans between 1958 nan, Monday 17 February 2014; Wynn, and 1959. p. 92. Also Michael Hely, ‘Nuclear Weapons 78 Interview Allisstone, and Mike Allisstone, Training — Groundcrew’, Royal Air Force ‘Nuclear Weapons and no. 94 MU, RAF Historical Society Journal, 26 (2001), 74–80. Barnham’, Royal Air Force Historical Soci- 75 Keith Eldred, Wayne Cocroft and Mike ety Journal, 35 (2005), 54–61. A rare insight Allisstone gave considerable advice on into the shifting supply chains is offered by Barnham. The special storage site at RAF W/R Unit, AWRE, Report B14/56, op. cit., Barnham is documented in Cocroft and which documents four changes in production Thomas; Wayne D. Cocroft and D. Gregory, routes for ground proximity fuzes between ‘Barnham, St Edmundsbury, Suffolk: RAF November 1951 and January 1955 when Barnham, Special Storage Site. Documentary ‘PLO’ — probably Plessey Nucleonics at Analysis of Sources in the National Archives’, Stour Road in Northampton — was finally Research Department Report 48/2011, dropped from the supply chain in favour of Swindon: English Heritage, 2011; also see ‘LPD 4A’ (possibly Borehamwood, North AIR 29/4157, No. 94 Maintenance Unit London) and ROF Burghfield. Also email RAF Barnham, Air Ministry and Ministry of communication, Mike Watkins, 14 August Defence: Operations Record Books, Miscel- 2013. laneous Units, 1 October 1957–30 June 1959; 79 Hinton, p. 265. and AIR 29/2522, No. 92 MU, Faldingworth, 80 On the linear model, see B. Godin, ‘The January 1956–December 1960. Linear Model of Innovation: The Historical 76 Interviews with Tivenan, Weller and Alliss- Construction of an Analytical Framework’, tone. Science, Technology and Human Values, 77 Production trials began in 1950 and full-scale 31.6 (2006), 639–67; Philip Scranton, ‘Tech- manufacture took place between 1952 and nology-Led Innovation: The Non-Linearity 1958. Official records suggest a total of 100 of US Jet Propulsion Development’, History active weapons and 20 training weapons and Technology, 22.4 (December 2006), were ordered at a cost of £2.2 million by the 337–67. end of 1962 according to AVIA 65/1792, p. 9 81 Peter Barker, correspondence, 11 August and appendix 2. Estimates vary as to how 2014.

Notes on contributor Jonathan Aylen is Senior Lecturer in the Manchester Institute of Innovation Research at the University of Manchester and a member of the Newcomen Society Council. Correspondence to: Jonathan Aylen. Email: [email protected]

2. Stretch - how innovation continues once investment is made

Jonathan Aylen, “Stretch - how innovation continues once investment is made”, R&D Management, vol.43, no.3, June 2013, pp.271-287

62 bs_bs_banner

Stretch: how innovation continues once investment is made

Jonathan Aylen

Manchester Institute of Innovation Research, Manchester Business School, University of Manchester, Manchester M13 9PL, UK. [email protected]

Process industries are characterised by large ﬁxed items of capital equipment. This paper asks how innovation takes place once these plants are built. Stretch is the mechanism by which established plants incorporate improvements in process and product technology, and make higher output and new products as a result. A taxonomy of ‘stretch’ is proposed looking at ﬁve interrelated features: improved intensity of hardware use through experience and better maintenance; system-wide effects of improvements in feedstock and downstream processing; ‘bolt-on goodies’ and physical reconstruction of existing plants; and quality enhancement and new products. Intensity of use encompasses familiar learning effects, but also enhanced maintenance. Evidence for stretch is given for wide strip mills built under Marshall Aid, but the concept has wider application across process industries, manufacturing, transport and services. From a theoretical point of view, stretch is the expression of evolutionary problem solving. The practical implication is that research and development management should encourage cross-functional collaboration. Creative and unorthodox personnel need to work in routine areas such as maintenance. Results-orientated production managers need to accept risky interventions in production schedules to allow continuing innovation.

Stretch, v. fig. ‘to enlarge or amplify beyond tonnes per annum. The Sidmar wide strip mill at proper or natural limits; to extend unduly the Ghent is an example of ‘stretch’ – the ability of a scope, application or meaning of’ 1553. process plant to produce output beyond its initial design capacity and produce a wider range of prod- Shorter Oxford English Dictionary, vol.2, ucts than originally envisaged because of improve- Oxford UP, 1973 ments in operating practice, developments in ‘bolt-on’ equipment and changes to feedstock after it 1. Introduction – the reality of stretch has been commissioned. Here we attempt to answer the question: how does n 1966 a wide hot strip rolling mill was commis- innovation take place in process industries character- Isioned in Ghent, Belgium for an initial capacity of ised by large items of fixed capital equipment once 1.75 million tonnes of steel strip per year. During the those plants are in place? As Thomas Lager (2002) years 2007–2009, the same rolling mill produced 6.5 points out, incremental development and refinement million tonnes of steel per annum – approaching four of existing products and processes is the most times its rated capacity. Recently the plant has run for common form of innovation in manufacturing indus- a number of weeks at a rate equivalent to 7 million tries such as metals, building materials, glass, paper

R&D Management 43, 3, 2013. © 2013 The Author. R&D Management © 2013 John Wiley & Sons Ltd 271 Jonathan Aylen and chemicals. Yet how does innovation continue Existing technologies also evolve as successive once the commitment has been made to build a large new plant is built. These new techniques can be ret- capital item? How does existing kit get upgraded rofitted to earlier equipment. Continuous wide strip once it is in production? rolling technology for steel moved through five gen- We define stretch as the continual modification of erations of development from its inception in 1926 a plant, system or service beyond its initial design (Aylen, 2001; Aylen, 2010). Mill designs reflect specification, with the main aim of increasing capac- growth prospects and relative prices of their time. ity, but also to increase product range and quality and The story of hot strip mill development is one of use new inputs. There may be incidental advantages gradual evolution, with occasional step changes of from stretch such as better environmental perform- direction – evolution as the engineers pursued the ance or improved working conditions. But the focus next logical technical improvement, changes of here is on modifying and upgrading existing equip- direction as economic imperatives forced a ment or services to make more output. re-evaluation of the path of technical development. Stretch was traditionally seen as a way of increas- Each generation of mill design accumulated techni- ing throughput in chemical production (Coe, 1962). cal developments of the past, but added more in Maxwell and Teubal (1980) suggest that stretch is response to changed circumstance. The first three often pursued when capital investment is not avail- generations of wide strip mill pursued economies of able. Here we use ‘stretch’ in a wider sense as the scale in growing markets. High-capacity Generation potential for further expansion, development and III strip mills were a product of optimistic growth upgrading of existing plant in the process industries, prospects, abundant capital and cheap energy before including capital spending on reconstruction. Stretch the oil crisis of 1974. The trajectory shifted towards is the mechanism by which established plants incor- compact layouts with Generation IV and V. Genera- porate subsequent improvements in process and tion V mills cut capital outlays and energy usage product technology and organisational innovation. It further by casting thinner steel slabs, an approach is making the most of the material immediately to that is viable on a small scale. The latest small mill, hand – existing equipment (Miner et al., 2001). Cremona 2 (Arvedi et al., 2010) is migrating towards The paper considers post-Second World War-wide a continuous flow process for rolling, thereby strip mills supplied under the terms of Marshall Aid. expanding capacity by increasing throughput, This has the advantage of longevity, giving a clear notably by higher casting speeds to increase mass perspective on retrofitting of technology developed flow to the rolling train. One theme has been consist- over the subsequent 60 years. We show that obsoles- ent, the move towards heavier piece weights as coil cence is not an automatic process of decay and sizes have steadily increased over time. Innovations replacement. Instead, old technologies remain in such as heavier coil weights can be retrofitted to widespread use, upgraded and shaped to modern uses earlier generations of equipment. (Campbell, 2009). Stretch is important for competitiveness. Evidence Admittedly, there is occasional and severe discon- suggests that plants that increase output over time tinuous technical change in process industries, which and keep up with new technology are those that requires complete new plant items and premature survive. A study of 17,000 manufacturing plants in scrapping of old technology. Oxygen steelmaking, Portugal finds current size and past growth are key continuous casting of steel, tunnel kilns for brickmak- determinants of plant survival (Mata et al., 1995). ing, float glass and dry process cement plants are all Again, a survey of 6,090 US manufacturing plants examples of radical innovations that diffused rapidly, during the 1987–1991 period shows that capital- supplanting existing techniques (Lacci et al., 1974; intensive plants and plants employing the latest tech- Aylen, 1980; Utterback, 1994; Alsop et al., 2007). In nology are less likely to fail (Doms et al., 1995). these circumstances investment in new plant brings Baldwin and Rafiquzzaman (1995) look at survival substantial productivity gains (Sterner, 1990). rates of new plants over two decades in Canadian These breakthrough innovations themselves open censuses of production. They discover that new up possibilities for further development. For the most plants that were smaller and less efficient did not part, process plant technology displays great longev- survive, and those that did survive this selection had ity once built. The fluid catalytic cracking process for improved their performance relative to established oil refining is a case in point. Indeed Yin (1994) firms through a learning process. argues returns from follow-on incremental improve- Everyday experience in process industries stands ments to a petroleum-refining innovation exceed in marked contrast to economic theory on capital those from the initial radical process innovation vintages (Salter, 1966). Economics quite simply itself. assumes that new plants incorporate the latest tech-

272 R&D Management 43, 3, 2013 © 2013 The Author R&D Management © 2013 John Wiley & Sons Ltd How innovation continues once investment is made nology. Subsequently, equipment deteriorates with staged in line with future market growth. Unrealised age and use, and becomes relatively obsolete as new potential for expansion may be a credible threat vintages of technology emerge (Fisher, 1963; Gort deterring new entry by competitors (Dixit, 1980). A and Boddy, 1967; Shone, 1975; Gomulka, 1978). likely competitor would think twice before entering a Kenneth Arrow (1962, p. 157) caricatures this putty- sector where established firms can readily expand clay model: ‘at any moment of time, the new capital existing plants, raise output and stage a price war goods incorporate all the knowledge then available, backed by latent new capacity. but once built their productive efficiency cannot be Other sources of stretch are unforeseen by initial altered by subsequent learning’. Change is further designers, builders and operators. Some improve- precluded because one item of sunk investment is ments occur by plant level invention, such as the interdependent on another (Frankel, 1955). Evidence predictive software developed by Voest on their Linz from a range of steel-production processes (Wolf and wide strip mill (Luger and Hubmer, 2001). Others Van der Rijst, 1982) suggests that technical advance arise from generic developments outside the industry. has progressively raised the output and cut the capital costs per tonne of capacity of new equipment. But there is a clear assumption that all technical change 2. A taxonomy of stretch is embodied in new equipment (Ruth and Amato, 2002), quite simply overlooking the extent to which Our taxonomy of stretch is based on a five-way clas- existing equipment is both reorganised and modified sification (Table 1). This distinguishes between in use. improved management and organisation, whether Moreover, a new plant does not always adopt best capital spending is involved, and considers upstream practice technology. Gregory and James (1973) and downstream system-wide effects as factors influ- found a wide dispersion of productivity across new encing the central plant unit under study. Finally, we Australian factories. The productivity of new plants focus on quality improvements and production of was no different from existing factories. Productivity new products unforeseen when a plant was built. So growth reflected the whole past history of the capital this classification looks at five interrelated features: stock, not the performance of the latest plants. So, the firstly, improved intensity of hardware use through idea that successive vintages of capital equipment experience and better maintenance. Secondly, incorporate the best practice of their day and inevi- system-wide effects of improvements in material tably supplant existing equipment is not borne out by feedstock and downstream processing. Thirdly, we the evidence.Yet, as Edgerton (2006, p. 8) points out, consider ‘bolt-on goodies’, including improved ‘the assumption that the new is much superior to instrumentation or control systems. Fourthly, we older methods is widespread’. consider physical reconstruction of existing plants. Instead, this paper develops a taxonomy of Finally, these improvements often enhance quality ‘stretch’, unpicking the sources of performance and offer scope for making novel products. These improvement that enhance output and increase the features of stretch are connected. Capital spending product range of established items of process plant. may be needed to accommodate heavier piece Stretch is related to key innovations in the process weights on rolling mills, for instance. New roll tech- industries. Some of these innovations are quite nology, such as high-speed steel rolls, are needed to generic, such as replacement of mechanical actuators make high-yield strength low-alloy steels. by hydraulic devices, digital control systems or high- power microprocessor management of variable speed 2.1. Pure learning alternating current motors. Other innovations are highly specific to a particular installation or process, One aspect of stretch, the notion of ‘learning’, has such as the quick roll change rigs and coilbox used on been widely studied. Learning is the ability to reduce wide strip mills. The taxonomy of stretch covers unit costs or raise output with cumulative operating interlocking changes in hardware, software, experi- experience (Argote and Epple, 1990). But the ence, intensity of use (including scheduling and concept of pure learning does not capture the hetero- maintenance), improvements in preparation of raw geneous mix of control equipment, capital spending, material feedstock and improvements in specification upstream innovation, additional equipment, feed- of finished products. stock changes and alterations in product specification Future stretch is often anticipated at the design that explain the steady expansion in processing stage, before capital investment begins. Initial capacity and product range of existing plants. The designs can provide for future staged expansion literature on learning curves or experience curves (Earnshaw, 1978), allowing capital spending to be conflate a number of factors that enhance productiv-

Table 1. A taxonomy of stretch Dimension Characteristics Impact

Management and operator skill 1. Pure learning Operating experience Enhanced throughput when operating and wider product range 2. Changes in Mostly experience, better monitoring and Shorter downtime, enhanced plant availability maintenance scheduling and better product quality Embodied in extra capital 3. ‘Bolt-on goodies’ Additional modules of equipment (e.g. Higher output, better quality and reduced instruments and controls) downtime 4. Reconstruction of Replacement and upgrading of core facilities Often higher output through removing existing equipment (e.g. furnaces and motors) bottlenecks System-wide improvement – upstream 5. Better availability Better scheduling of materials supply, more More regular operation of feedstock consistent supply 6. New feedstocks New materials, different dimensions Higher yield and heavier piece weights System-wide improvement – downstream 7. Better ‘take’ of Better scheduling, warehousing and product Longer production runs materials logistics 8. New capabilities Larger capacity items for downstream Higher yield and heavier piece weights processing New products 9. New qualities Improved quality of existing products e.g. Better surface ﬁnish, gauge, shape and dimensional accuracy 10. New products Complete new products e.g. High-strength low-alloy steels, transformation-induced plasticity steels

ity over time (Hart, 1983). For example, Van der Rijst Learning how to operate is the most widely dis- et al. (1978) suggest that maximum productivity of cussed aspect of increasing the intensity of use, blast furnaces for ironmaking is directly related to although there is little discussion of how it actually production experience and claim that there is a con- happens (von Hippel and Tyre, 1995). At its heart is stant rate of learning across Europe. This overlooks the growing confidence in operation and assembly, frequent upgrading of blast furnaces that takes place which improves the pace of operation. As a rule of during every periodic re-line. New refractories, new thumb, doubling output seems to be associated with a stoves, better top equipment, higher power blowers, 20–30% unit cost reduction. Lieberman (1984) esti- more efficient water cooling, new clay guns, better mates learning curves for 37 different commodity cast house equipment, new instrumentation and com- chemicals and metals across six broad product cat- puter control are routinely adopted during these egories using price as a proxy for cost. He finds that regular and prolonged stoppages for reconstruction. cost reductions are related to cumulative output and For instance, blast furnace 3 at SSAB, Luleå, Sweden the capital intensity of the process concerned. He was drastically rebuilt in 75 days during 2000 using concludes that research and development (R&D) prefabricated segments to give a modern 11.4-m spending accelerates the learning process. Higher hearth furnace. Regular upgrading has been spurred R&D intensity interacts with cumulative output. The on by tighter environmental regulation. In short, cost advantage was shared by batch and continuous experience is just one part of overall ‘stretch’. Other processes alike. Evidence for 221 special chemicals technologies are upgraded at regular intervals during within one large company suggests that growth in rebuilds. Cement kilns or oil refineries typically run past output is a strong signal of future returns to R&D for 330 days between stoppages. Float glass tanks are expenditure, and so reductions in unit cost may actu- relined every 7 years or so. These regular outages for ally reflect incentives to invest in process research prolonged maintenance give an opportunity for ret- (Sinclair et al., 2000). rofitting equipment or judicious expansion of indi- Learning continues over the life of a plant. One vidual plant items. factor is better scheduling (e.g. mixed rolling of elec-

274 R&D Management 43, 3, 2013 © 2013 The Author R&D Management © 2013 John Wiley & Sons Ltd How innovation continues once investment is made trical steels and carbon steels on wide strip mills), 18-year period between commissioning in 1969 and which is enhanced by the development of new com- 1987 (IISI, 1989, 3.22–3.23), which show a continu- puter programmes to optimise plant utilisation ing reduction in unexpected stoppages from 74 hr a (Cowling, 2003). month in the first year of operation down to 13 hr in Attempts have been made to widen the notion of the 19th year of operation. Production over the same learning towards a progress function. A wide ranging period grew by 24%. Most of the changes were survey by Dutton and Thomas (1984) disaggregates attributed to training, changing approaches to work the performance of firms into pure learning, technical (e.g. quality circles), eliminating demarcations and progress, local industry and firm characteristics, and reorganising maintenance. A regular pace of rolling scale effects. But these broad categories do not mill operation brings a higher yield of finished steel capture the nitty-gritty of how plants actually get from raw material, for instance through lower slag modified over time. losses on reheated slabs (IISI, 1992).

2.2. Changes in maintenance practice 2.3. ‘Bolt-on goodies’ – improvement of existing plant Maintenance is a neglected topic. As Edgerton (2006) argues, ‘although central to our relationship Incremental process improvements are colloquially with things, maintenance and repair are matters we known as ‘bolt-on goodies’. These are modular items would rather not think about’. Yet, maintenance is that can be attached to existing equipment often the third largest item in a steelmaker’s budget, after developed in collaboration with a specialised equip- raw materials and manpower (Aylen, 1989). Annual ment supplier, electrical engineering company or maintenance costs for chemical plants are typically bespoke software house (Hutcheson et al., 1995; 5–15% of fixed capital costs, divided evenly between Ozman, 2011). They are usually trialled on existing labour and materials (Sinnott, 2005, p. 262). plants before being specified on new equipment. They The military were among the first to appreciate are often supplied as ‘vendor packages’ (Hutcheson the output gains brought by maintenance planning. et al., 1996) from an equipment supplier who is Switching Royal Air Force Coastal Command air- responsible for the integration of the new equipment craft to ‘planned maintenance’ overcame a shortage into the existing power supply, process control and in skilled labour, increased their flying hours per instrumentation. The development of ‘bolt-on maintenance worker by 43% and raised their operat- goodies’ highlights the importance of machinery sup- ing time from 140,383 flying hours in 1942 to pliers as sources of innovation (von Hippel, 1977; 200,558 flying hours in 1943 (Air Publication, 1954, van Rooij, 2005; Reichstein and Salter, 2006; Lager pp. 171–176). and Frishammar, 2010). A high level of retrofitting Reduction in scheduled maintenance time from explains why Salvanes and Tveteras (2004) find that once a week to once a month typically releases an there are marked differences between the age of a extra 12% potential capacity (assuming a cut from 49 works and the age of its equipment in a survey of days to 11 days with a 2-week annual shutdown). Norwegian manufacturing plants (Dunne, 1994). This is usually achieved by better planning and There are at least 30 readily identifiable bolt-on scheduling techniques. A rare comparative study on developments used to enhance existing wide hot strip maintenance practice from the International Iron and mills over the past 50 years (Table 2). These include Steel Institute (IISI, 1989, 3.23–3.24) argues that modifications at every stage of the production line ‘detailed planning is recognised as having more from reheat furnaces (notably large walking beam effect on maintenance productivity than any other furnaces that lift and carry slabs through heating single factor’, highlighting the way in which organi- zones) right through to coilers. Some rely on generic sational innovation and technological innovation technologies, such as the use of hydraulics to interact in process industries. manipulate roll screw-downs, edger rolls and coiler Online maintenance while a plant is still in opera- wrapper rolls. Some are highly specific to rolling tion reduces down time. In the cement industry, the wide strip such as gauge metres, shape metres and kiln is the critical feature. Hot alignment techniques width metres for hot material. Roll developments allow corrections to kiln alignment while it is oper- might add another 10 incremental innovations. There ating (Saxena, 2009, p. 19). have been numerous product developments, too, with Reductions in unscheduled delays are more impor- perhaps 80% of the product range rolled on a modern tant because unplanned stoppages destroy the wide strip mill having been developed in the last 20 ‘rhythm’ of a working plant. Detailed results are years. This helps explain the predominance of available for Kimitsu hot strip mill in Japan over an quality-enhancing investments among the list in

Table 2. ‘Bolt-on goodies’ used to modify wide hot strip Table 2. But an incidental side effect of these quality- mills since 1960 enhancing innovations has been an increase in mill 1. Walking beam reheat Output, energy saving and output through faster rolling, higher yield of prime furnaces quality gain quality material and lower downtime because of 2. Direct hot charging Energy saving maintenance. 3. Computer control of Energy saving We illustrate three of these modular improvements reheat furnaces to wide strip mill technology in more detail in Boxes 4. New burners (e.g. pulsed, Energy saving and yield gain 1–3. These innovations have been retrofitted to exist- atmosphere control) ing equipment to enhance capacity, improve product 5. Waste heat recuperation Energy saving quality, cut energy use or more generally to obviate 6. Lift and carry furnace Quality gain discharge the need to invest in a complete new installation. Two 7. Slab skid mark Quality gain for slabs from of these examples require substantial mechanical compensation pusher furnace engineering enhancements; the third is a software 8. Hydraulically controlled Yield gain improvement. vertical edgers for slab The examples include quick roll change rigs, shaping which were almost universally adopted as an output- 9. Hydraulic screw down Quality gain and yield gain at coil ends enhancing innovation between the mid-1960s and the 10. High pressure descalers Quality gain mid-1980s. Coilboxes require more substantial 11. Delay table covers (e.g. Energy saving reconstruction and changes in operating procedures, ENCOPANELS) but they offer the chance to make much heavier coils 12. Coilbox Output, quality and product on existing mills with short delay tables. Predictive range modelling of microstructure is one way in which 13. Strip edge heaters (not Roll wear, so output steelmakers supplying the car industry have been common) able to satisfy more demanding quality requirements 14. Quick roll change rigs Output and obviate mill damage in terms of yield strength and surface finish while maintaining the pace of mill operation. Predictive 15. Zoom-rolling Output-offsetting thermal rundown of coil modelling is a highly R&D-intensive retrofit. 16. Back-up roll bending and Quality Retrofitting of modular items is not unique to work roll bending for steelmaking. Cement plants have upgraded clinker crown control coolers with higher pressure static coolers, added 17. CVC mills for crown Quality heat exchangers to cooler dust collection to enhance control capacity and switched to new materials for dust col- 18. Interstand cooling Output and product range lection in bag houses (Alsop et al., 2007, pp. 190– 19. Hydraulic loopers Maintenance and quality 192). 20. Work roll side shifting to Roll wear, so output distribute edge wear 21. AC motors under high- Energy saving power digital control 2.4. Reconstruction of existing plant 22. Automatic gauge control Quality, but also output 23. Computer control Savings all round and output A more drastic way of improving plant is reconstruc- (successive generations) gain tion. A number of European hot wide strip mills have 24. Laminar flow cooling Quality been converted from continuous layout to more effi- 25. Width meters Quality cient three-fourths continuous or semi-continuous 26. Shape meters Quality layout, which brings greater utilisation of the finish- 27. Automatic crop control Optimising yield, so output ing train by better scheduling of the roughing train 28. Revolving mandrel coilers Quality and yield and use of heavier incoming slabs (Aylen, 2001). 29. Hydraulic wrapper rolls Quality and yield Reversing roughing stands improve heat distribution 30. Lap detection Quality in the breakdown bar. This modification in layout is 31. Close-coupled coilers (not Product range often associated with the construction of larger common) walking beam reheating furnaces and, sometimes, Omitting false-start technologies, local developments such as six the use of a Stelco Coilbox on the delay table. Recon- high hot mills in Japan, roll improvements, a wide range of new struction typically involves mechanical preparation products and major changes in mill configuration such as thin slab casting. and civil engineering underneath the roughing train while the plant is still in operation followed by an extended summer shutdown. This approach has seldom been used in Japan or the United States,

Box 1. The Stelco Coilbox – increasing piece weights

The Stelco Coilbox was prompted by the conceptual planning of a new greenfield works on the north shore of Lake Erie, near Dover, Ontario. Stelco wanted to build the mill as cheaply as possible. The mill was laid out for in-line rolling directly from the continuous caster to save energy. Overall mill length was reduced by the use of a novel Stelco Coilbox in place of a lengthy delay table between the roughing stand and finishing train. This improved energy efficiency and reduced the capital costs of the mill. Obviating the delay table cuts, perhaps 75 m off mill length. Lake Erie was a logical way to develop conventional and long-established North American hot strip mill technology in order to pare initial capital costs to the bare minimum (Carroll and MacNeil, 1985). The coilbox simply receives the hot breakdown strip in the final pass from the roughing train as a large, open coil. In effect, coiling is a temporary storage device. Coiling the strip obviates a long, horizontal delay table. Coiling also promotes heat transfer so that the bar is at a more uniform temperature. Coilbox operation starts with the transfer bar from the roughing train being captured and deflected through bending rolls so that the hot material forms an open, loose coil on a simple cradle of rolls. The hot coil is then paid-off into the finishing train in the reverse direction. The cooler tail end of the breakdown strip now becomes the leading end as it enters the finisher, making for a more even distribution of temperature along the length of the strip. Finally, the coil is moved sideways once it has started to feed into the finishing train so that the cradle is free to receive the next transfer bar from the roughing train. The coilbox was conceived by William Smith. It is a tribute to the creative pressure imposed on good engineers by shortage of resources. It was trialled as a sequence of prototypes on Stelco’s existing Hilton Works hot strip mill in Hamilton to the point where it was suitable for commercial use. Commissioning of Lake Erie works was delayed for financial reasons. Les Gore of steelmaker John Lysaghts Australia saw the prototype at work and adopted the idea for their new wide strip mill in Australia (Gore and Shegog, 1979). Thereafter, the Stelco Coilbox was widely adopted in Europe and Canada both as a technique for saving energy and improving strip quality through more uniform heating. It is an ideal way of rebuilding old roughing trains as the original delay table can be cut and the space used for two widely spaced roughing stands rolling a heavy slab. By the mid-1980s Stelco had made $20 million in royalty on an initial $1 million outlay on R&D. By 2000 some 43 coilboxes had been installed – half since 1990.

Coilbox US Patents 3803891, ﬁled 13 November 1972 and awarded 16 April 1974, and US Patent 3805570, ﬁled 13 November 1972 and awarded 23 April 1974, both invented by William Smith and assigned to the Steel Company of Canada Limited, Toronto, ‘Method for rolling hot metal workpieces’ and ‘method and appa- ratus for rolling hot metal workpieces and coiler for use in coiling hot metal workpieces’.

where hot strip mills remain as built, suggesting that processing both within the works and on customers’ stretch is more widely practiced in Europe. facilities. One variant is to roll thicker gauge material on the hot rolling mill and rely on heavier reduction 2.5–8. System-wide improvement – in subsequent downstream cold rolling facilities. upstream and downstream A key factor behind stretch in the case of the Marshall Aid mills turns out to be heavier piece One of the easiest ways to stretch a plant is to weights. The Linz wide strip mill was built in 1952 increase the unit size or ‘piece weight’ of the inputs for 10.7-tonne slabs. The mill now rolls coils up to used by the plant. The Sidmar mill discussed in the 32 tonnes maximum, a threefold increase. Here opening sentence receives thicker slabs than it was again stretch conceals a sequence of technical inno- designed for, resulting in longer and heavier coils vations retrofitted since the plant was built: adop- produced by greater reduction along the mill. Aylen tion of continuous casting, higher capacity reheat (1982) finds striking differences in capacity between furnaces, more powerful edging rolls on roughing UK and German wide hot strip mills, partly reflecting mills, extra finishing stands, increases in motor higher drive powers and heavier coil weights of power, higher capacity coilers and more robust coil German mills. Heavier coils also make for cheaper handling facilities.

Box 2. Quick roll change rigs – reducing mill down time and damage

The original turntable roll change rig patent ﬁled in 1960 was an output-enhancing innovation (US patent 3208260). As the preamble states:

Not only is it extremely desirable to provide a roll changing system that will not require the use of any cranes and one that will reduce the manual assistance necessary, but even of more importance is the need for keeping at a minimum the downtime of the mill incident to roll changing. In the case of a six stand tandem mill, it is not uncommon to experience more than a forty-ﬁve minute delay in changing work rolls thereof. Moreover, this operation usually is performed as frequently as every two to four operating hours so that the aggregate of the lost production time is most costly.

As with many technical developments, the quick roll change rig was prompted by the needs of a user, in this case Inland Steel of Chicago, who wanted to change rolls in 10 min. Inland approached the plant supplier United who made little progress on a solution for 2 years until they came up with the idea of using a turntable. (Characteristically, Jim Adair said the idea came to him sitting on a revolving bar stool.) The quick roll change device met its brief. Ess (1970, pp.82–86) reports that a roll change on one ﬁnishing stand of a hot mill could be accomplished in 2 min and a full set changed in 12–15 min. The advent of quick roll change rigs was said to increase mill output by 5–8% compared with a precarious conventional C hook or porter bar roll change.

Quick Roll Change Rigs US Patent 3208260, ﬁled 18 August 1960 and awarded 28 September 1965, inventors Maurice Paul Sieger and James R. Adair, assigned to United Engineering and Foundry Company, Pittsburgh, ‘Rolling mill’.

Box 3. Real time prediction of metallurgical transformation – more uniform quality and higher production rates

Direct digital control of a complete wide strip mill began at Llanwern in South Wales in 1963 (Aylen, 2004). This computer controlled the whole process in real time, tracked material through the mill and logged the results. By 1970, direct control of strip mills was widespread in the United Kingdom and Germany. Attention turned to simulation and control of product quality (e.g. Van Ditzhuijzen, 1993). During the 1990s, large-scale computer models were developed that allow the integration of scheduling and process control with quality assessment. This required a real-time model of metallurgical transformation during the rolling process. Developments by Voest-Alpine Stahl and VAI at Linz allow immediate predictions of quality for the whole length of a rolled coil (Luger and Hubmer, 2001). Linz makes a range of demanding products to high standards and sells profitably to sophisticated customers, especially for automotive applications (Marsh, 2011). The starting point is a physical-metallurgical model to predict strip quality in terms of tensile strength, yield strength and elongation. This modelling project began in November 1995. Once it was established that an off-line model gave accurate predictions of the mechanical properties of hot-rolled coil, the model was used to actually control set-up and cooling on the finishing train from January 2000 onwards. Strip varies throughout its length. So the model tracks each segment of the strip so that microstructure can be predicted and modified during final rolling and cooling. Optimum rolling and coiling temperatures are crucial for high-strength low-alloy steels, for example. There are commercial advantages from being able to predict the quality of each coil straightaway. The coil is passed on for further processing immediately. The whole length of the coil is ‘checked’ by inference, whereas conventional measurement is restricted to samples from the head and tail end, which may be untypical. There are raw material savings as it is possible to predict mechanical properties such as tensile strength, rather than relying on expensive alloy additions at the steelmaking stage.

Use of larger inputs to an existing process system et al., 2002). This required mixed rolling of alternate is not confined to mechanical equipment. A study of thick and thin slabs, but the interleaving did not work Heathrow airport shows that existing runways, taxi- and the new machine was dismantled. ways and terminals were stretched by airlines using larger planes to overcome capacity constraints 2.9–10. Quality and product range (Tether and Metcalfe, 2003). In this way, slots were utilised more intensively through system-wide modi- Stretch can also focus on changing product mix to fication of operating procedures. Mixed approach meet shifts in demand. The petrochemical industry paths were used by traffic control to overcome the responded to rising demand for propylene, usually turbulence problem brought by larger, more powerful made as a by-product of ethylene manufacture, by engines of bigger jets, in the same way that wide strip developing ‘on purpose propylene’ technologies, for mill operators mix rolling of silicon steels at ultra- instance using Olefins metathesis (the metal cata- high temperatures with tough stainless steels as alter- lysed redistribution of carbon–carbon double bonds). nate slabs. ABB Lummus’ Olefins Conversion Technology uses A rolling mill designer (Weiss, 1978) sees the this approach by adding an extra reactor to an exist- process of technical improvement as one of breaking ing cracker (Plotkin, 2005). The addition of an extra successive bottlenecks. Once one constraint on process stage is analogous to the near universal adop- throughput has been overcome, attention shifts to the tion of ladle steelmaking since 1980, which provides next. So the process of stretch is one of stepwise a holding buffer and refining vessel between steel continual improvement to release the potential of melting and continuous casting, thereby allowing latent capacity upstream or downstream from a more efficient scheduling and production of a much problem area. Each scheme to enhance output then wider range of steel products (e.g. Price, 2007, pp. reveals another constraint elsewhere. Constraints on 392–395). Ladle furnaces provide a consistent flow technical change themselves throw up new opportu- of batches of molten steel at the precise temperature nities. While the nature of improvement is piecemeal, and metallurgical composition required for continu- the overall effect over a long period of time is a ous sequence casting. substantial enhancement of the productive capacity Stretch also expands product range. Making ultra- of the overall plant. In the same vein, Alsop et al. low carbon electrical steels is a step towards making (2007) write: the more recent interstitial-free ultra-low carbon steels for car body sheet. The spread of product range De-bottlenecking: this repulsive but descriptive reflects a broader point made by Malerba (1992) that term is a mandatory focus for cement plant learning adds to the stock of knowledge and technical engineers. Every process has one or more capabilities of the firm, which then opens up a range capacity limitations. Identification of the limit- of opportunities for technical advance, not just cost ing equipment is the first step. reduction. The Linz wide strip mill considered here opened up the development of high-strength form- Stretch also involves shifts in inputs in response to able coated steels for Voestalpine, which puts the availability of materials, change in prices and down- firm at the forefront of steel strip supply for auto stream quality requirements. The British chemicals bodies (Marsh, 2011). firm Imperial Chemical Industries offers an extreme case of changing feedstock where ammonia production at the company shifted from water gas and pro- 3. The Marshall Aid mills – hot strip ducer gas made using hot coke, to steam reforming of and cold war by-product hydrocarbons, to steam reforming of cheap naptha, all within 30 years (Gard, 1966; van The continuous wide hot strip mill was an American Rooij, 2004, chapter 9). Papermakers have retrofitted invention, pioneered by Columbia Steel, Pennsylva- process lines to de-ink and pulp waste paper as a nia in 1926 (Aylen, 2010). By 1953 there were 55 supplement to conventional wood pulp (Engstrand continuous wide strip mills and Steckel mills, and and Johansson, 2009). Adoption of supplementary three continuous plate mills in existence or under processes in existing plants is not always successful. construction worldwide (BISF, 1953). All but three As part of an expansion plan begun in the year 2000, mills worldwide were of American construction. the Terni wide strip mill of ThyssenKrupp Stainless American process plant builders emerged from supplemented the rolling of traditional thick continu- World War II with a technical and commercial lead in ously cast stainless slabs with thin slabs cast on a new many fields, notably steel and aluminium, cement machine feeding a long tunnel furnace (Brascugli and petrochemical plant. Basic equipment was

Table 3. The Marshall Aid wide strip mills in Europe Mill Commissioned Status Rated capacity Capacity at Ratio Age (US builder) when built closure/now (years) (ktpa) (ktpa)

Usinor, Denain March 1951 (United) Closed March 1985 1,300 2,000 1.5 34 Port Talbot June 1951 (United) Operating, semi-half 1,800 3,000 (estimate) 1.7 61 continuous Linz, Austria July 1952 (Mesta) Operating, semi-half 600 5,000 (in 2010) 8.3 60 continuous Breedband, October 1952 (United) Closed November 600 1,100 1.8 33 IJmuiden 1985 Sollac, Sérémange January 1953 (United) Operating, M-stand 1,800 3,500 1.9 59 added 1983 Cornigliano, Genoa end 1953 (Mesta) Closed 1984 900 1,700 (peak output) 1.8 31 Ougree-Seraing 1954 (Mesta) Closed 1980s 800 1,400 1.8 30–35

needed to support post-war reconstruction and pusher reheat furnaces. A single, large, reversing growth. The European Recovery Programme – col- rougher was installed in place of the continuous loquially known as Marshall Aid – funded the acqui- roughing train, a Stelco Coilbox was installed ahead sition of US technology for reconstruction and of the finishing train and two new 34-tonne coilers defence. The programme had twin objectives of put in behind the existing coilers, more than doubling defending Europe from communism and tied-Aid the coil weight (Kidd and Dimblebee, 1987). The easing the transition of American heavy industry to finishing train was little altered apart from an extra peace time production. stand and new motors. The mill production manager The European Recovery Programme led to seven was also the project manager to ensure continued major hot strip mill orders for US plant suppliers mill operation during radical reconstruction. United and Mesta (Ranieri, 2012). These were Gen- The outstanding performance of the Linz mill is eration I mills, a design that remained unchanged striking (Box 4, Figure 1). Linz is the only wide hot to 1960 (Ess, 1941). Some of these mills were built strip mill in Austria. It has a commanding market with low capacity, such as IJmuiden, Linz and among the car manufacturers and white goods Cornigliano, while Sollac and Port Talbot were makers of southern Germany and northern Italy. The standard US style continuous wide hot strip mills mill continually developed through the gradual supplied by United. accretion of additional innovations, which permitted Three of these rolling mills are still operating 60 heavier piece weights brought by improvements in years later (Table 3). The remaining four closed after upstream slab supply and downstream cold rolling, an operating life of over 30 years. The median and finishing facilities and higher quality products. increase in capacity over their working lives was Box 4 lists over 30 significant performance- 80%. Typical stretch in capacity across all the Mar- enhancing modifications for the Linz mill. shall Aid mills is actually less than 2–a median of 1.8. One reason for the relatively low growth is the subsequent construction of high-output Generation II 4. Discussion mills by the same company with first claims on orders and steel supply. Another reason is the matu- Detailed examples are drawn here from rolling mills rity of the European steel market after the first oil for steel strip, but similar instances are found in a crisis and recession of the mid 1970s. It is easier to wide range of process industries such as cement, energise changes when there is pressure on capacity. paper or petrochemicals. The principles hold across a The Port Talbot mill operated in standard form wide range of sectors, such as transport systems, the from commissioning in June 1951 for over 30 years. service sector where hotels and shops are routinely Ingot casting and slab rolling were replaced by con- refurbished, and defence where weapons systems are tinuous casting of slabs. There was a major recon- regularly upgraded. struction during the mid-1980s to a semi-continuous There are other explanations for the phenomena layout (Cook, 1979; Bryant and Dimblebee, 1986). we observe here. Plant items may have been overen- Two walking beam reheat furnaces replaced the gineered in the first place with generous safety

Box 4. Marshall Aid mill – the wide hot strip mill at Voest, Linz

The semi-continuous wide strip mill at Linz was built as part of the European Recovery Programme. The mill was supplied by the Mesta Machine Company of Pittsburgh under the terms of Marshall Aid. The wide strip mill for Linz was a revival of a pre-war Mesta contract for a strip mill at the Hermann Goering Works in Linz. A wartime armour plate mill was available for roughing down feedstock and a temporary sheet mill was installed after the war. But Voest was still keen to acquire the American continuous wide strip mill planned before the war. The contract was controversial because of concerns that the site beside the Danube was close to the Russian sector in Austria and might fall into Soviet hands in the event of conflict (Tweraser, 2000, pp. 312–313). The works itself is famous because it was the first location worldwide to use the now dominant basic oxygen steelmaking process. Only Manfred Wirth, a salesman with Voest, could get a visa to visit the United States to progress the order for the wide hot strip mill in March 1949 (Wirth, 2001), as other nominated individuals were on the US ‘watch list’. Wirth took advice from Stelco in Canada and ordered a wider finishing train from Mesta than envisaged pre-war – 66 inches – and one fewer finishing stand, making five in all, but with more powerful motors than those specified in the pre-war order. He also bought a slabbing mill from Mesta, one 100-tonne per hour reheat furnace, a flying shear from Wean and a single downcoiler. Wirth played a key role in persuading the Organisation for European Economic Cooperation to give Marshall Aid support to the mill in August 1949, reversing an earlier decision to exclude the Austrian mill from funding. The 66-inch semi-continuous hot strip mill was completed in July 1952, with an initial output of 280,000 tonnes of plate and sheet per year (Lovay, 1957). The same rolling train now makes 5 million tonnes. Linz runs 60 years on, ostensibly little changed as a semi-continuous wide strip mill. It is a leading supplier of high-quality car body sheet to the German and Italian carmakers. Marked increase in output and quality over these intervening years are due to small changes in plant configuration, upstream product supply and downstream processing. Major gains in throughput have been realised through heavier piece weights – larger slabs and coils. Advances in software and small enhancements to equipment coupled to higher plant availability have helped increase output and product quality.

Modernisation stages Linz hot wide strip mill

1953 Basic semi-continuous mill with five finishers and one fixed mandrel coiler (10.7 tonnes) 1956 Second 100-tonne/hr pusher reheat furnace 1957 Sixth finishing stand added 1958 Edger ahead of roughing stand 1959 Third pusher furnace 100 tonne/hr 1968 Two more pusher furnaces 1974/6 Two large pusher furnaces 350 tonne/hr, leaving four in total Extra finishing stand at F0 Roughing mill with attached edger Down coilers 3 and 4, three wrapper roll on extended run-out table Process computer for the finishing train 1981 Down coiler number 5 1984 Hot storage boxes for slabs 1985 Roughing mill computer 1987 New edger Skid mark compensation Reheat furnace automation 1989 Profile control 1991 Water descaler 1992 Furnace riders Low-NOx Burners for 350-tonne pusher furnaces 1994/5 New cooling lines on run-out table 1995 Basic automation 1996 Two new thickness gauges New fume collection for stands F4 to F6 New crop shear – two blade rotary shear New entry guide in front of finishing train Cobble pusher on delay table Pinion stand for rougher and F1 1997 Water treatment plant 2002 Remotoring 2010 Coiler revamp

This is a partial list, omitting quick roll change rigs. Over time, roll use has changed and the process control software has been upgraded.

Figure 1. Stages of modernisation of the Linz wide strip mill.

margins to help ensure that buyer performance guar- the existing mill (Price, 2007, p. 400). This is quite a antees were easily met, thereby realising the prompt literal ‘stretch’, but it requires a generously propor- release of final payments for equipment. The extreme tioned building. durability of mechanical items such as wide strip So far we have ignored the costs and revenues mills and chemical processes such as oil refineries from systematic innovation. Mills built at the outset suggests that they were substantially built in the first for staged expansion can usually be extended quite place. The Linz mill had a low initial rolling capacity cheaply for a small outlay on extra mechanical and but was installed in a huge building, which allowed electrical equipment, coilers and reheat furnaces. for a significant upstream spread to incorporate more Doubling the capacity might cost an extra 20% of the and larger reheat furnaces and a downstream exten- initial outlay. This is a clear reflection of the fact that sion to the run-out tables. So the overall length of the mechanical and electrical equipment would only take rolling line within the building has increased by just up one third of the cost of the initial contract, with under 40% (Figure 1). The Scunthorpe rod mill in fixed items such as site preparation, civil engineering the United Kingdom has been physically extended and building costs, water supply, pipe and power upstream in similar fashion with six new horizontal/ runs, and cranes accounting for the rest. Once a plant vertical roughing stands to roll larger billets ahead of is built, the marginal cost of additional capacity from

282 R&D Management 43, 3, 2013 © 2013 The Author R&D Management © 2013 John Wiley & Sons Ltd How innovation continues once investment is made stretching is very low. Presumably there are dimin- This is a challenge for managerial organisation and ishing returns from stretch determined by the basic personnel selection. Contingency theory suggests a initial design, but Linz suggests that these limits have firm’s organisation adapts to the tasks it faces not been reached yet. (Donaldson, 1996). Manufacturing operations are We focus here on a batch process with production likely to be characterised by formal routines and hier- along a continuous flow line. It is possible that proc- archical structures, while R&D departments are likely esses made up of discrete, individual stages such as a to be structured more organically in a way that allows machine shop have different stretch potential, with for trial-and-error learning and fluid movement of scope for modifying individual items without dis- personnel from task to task as problems evolve. rupting the rest. Continuous flow processes such as There is also a subtle danger of managerial selec- paper mills, petrochemical plant and float glass lines tion. As Schneider (1987, p. 439) recognises, may be restricted to occasional opportunities for ‘humans, at least in Western societies, are not ran- modification. The example of blast furnaces for iron- domly assigned to settings. Humans select them- making shows that there is considerable opportunity selves into and out of settings’. People are attracted for carefully planned stretch during periodic relines. to like-minded individuals and leave organisations However, a dedicated single-train chemical plant where their face does not fit. As a result, organisa- may be harder to stretch than a cracker making mul- tions evolve as they attract recruits, perhaps along a tiple products. trajectory set by the founders who may select and Stretch also treats the factory or chemical plant as ‘sort’ the type of people who work for a firm (Witt, a closed system. But in truth customers and suppliers 1998). Over time a process of attraction, selection are often closely involved in process improvement, and attrition produces a group of like-minded people for instance the way in which iron ore supplier who shape the organisation for which they work. An LKAB develops pellets to enhance blast furnace pro- extreme case is Firestone Tire and Rubber where Sull ductivity. Customers too benefit from expansion and (1999) reports by the early 1970s that all of Fire- quality upgrading. stone’s management team had spent their entire career with the company, two thirds were born and 4.1. Implications for R&D – the paradox brought up in the company town Akron, Ohio, one of stretch third followed their fathers as Firestone executives and most of the leading executives lived within a Stretch may be considered as an evolutionary process five-block radius of one another. So, within a where a factory evolves through solving a sequence company we expect a science-focussed R&D depart- of problems. The idea of ‘problem sequences’ is ment to have different characteristics to practical, familiar in the health sector where medical treat- technology-focussed production management as a ments ‘evolve along trajectories of change shaped by result of the selection of different competencies for the search for solutions to interdependent problems’ different roles. (Ramlogan et al., 2007). Again, Helfat (1994) finds One feature of stretch is the way that change is that individual US oil companies accumulate knowl- often initiated by search for solutions to practical edge by persisting with individual lines of R&D, operating problems. Innovation is driven by problem which evolve over time in divergent ways. She sees solving. Localised electrical heating of slabs to firms as learning organisations that search for solu- obviate skid marks caused by the pusher furnaces at tions in the face of bounded rationality (Simon, Linz is a case in point. A customer requirement to 1979). They do not consider all possible outcomes obviate these regular imperfection marks in coils but focus on existing lines of research where they are drove an ad hoc technical solution that resolved the best informed. problem of not having modern walking beam fur- Problem solving through development may be naces – a process of ‘resourceful improvisation’ hard to implement in a process industry. Stable (Hendry and Harborne, 2011) operation and strict adherence to routine are key fea- So the central organisational problem facing R&D tures of process plant operation. In contrast, R&D management is how do you encourage creative and requires creativity, project management skills and unorthodox personnel to work in routine areas such entrepreneurial implementation in the face of incom- as maintenance and persuade results-orientated pro- plete knowledge and changing circumstance. This duction managers to accept risky interventions in poses a dilemma: how do you stretch a process plant their production schedules in the interest of continu- through application of novel equipment, untested ing innovation? operating procedures and unfamiliar products when One solution is to refocus engineering teams on steady operation is at a premium? continuing development with the support of R&D

© 2013 The Author R&D Management 43, 3, 2013 283 R&D Management © 2013 John Wiley & Sons Ltd Jonathan Aylen personnel. This approach to continuous improvement Failure to recognise stretch reflects overemphasis helps solve the frustration of gifted engineers on novelty in economics and innovation studies, and working in specialised roles on routine and repetitive a neglect of the old. There is a corresponding empha- tasks where they are unable to use their full profes- sis in engineering training on new plant design rather sional skills (Holt, 1974). A study of Scandinavian than refurbishment. Conscious exploitation of stretch manufacturing finds that cross-functional collabora- is one strategic option for competing in mature indus- tion between departments is one of the key factors tries. Empirical evidence suggests that it is vital to explaining innovation performance (Frishammar and plant survival. But promotion of stretch requires a Hörte, 2005). This finding is supported by Love and re-engineering of R&D organisation and planning of Roper (2009) who find that cross-functional teams future process innovation. enhance technical elements in the innovation process. There is systematic scope for exploiting potential Acknowledgments stretch. This resolves the central problem facing operators of process plant of flexing production in the This paper benefited from advice at ‘Managing face of shifting market requirements (Edler et al., R&D, Technology and Innovation in the Process 2002; Larsson and Bergfors, 2006). There is a need to Industries’, Ecole de Management, Grenoble, May systematically plan innovation that increases product 2011 and workshops at Outokumpu Oy, Avesta and flexibility in established manufacturing plants. SSAB, Luleå, Sweden. Warmest thanks to Thomas Process innovation needs to be planned in the same Lager for encouraging me to write the paper, to way as product innovation (Pisano, 1997; Lager Ruggero Ranieri for help on Marshall Aid mills and et al., 2010). In truth, process and product innova- to Emilia Brodén, Chris Foster, Johan Frishammar, tions are intertwined with new process steps gener- Lennart Gustavsson, Lotta Jakobsson, Phil Judkins, ating new products and vice versa (Reichstein and Stan Metcalfe, Peter Samuelsson, Mick Steeper and Salter, 2006). Arjan van Rooij for excellent discussions. Research The notion of stretch has implications for devel- was supported by the EPSRC grant EP/D032709/1, opment (Maxwell and Teubal, 1980). China and ‘Unlocking low carbon potential’. India have purchased second-hand steel plant, machine tools and plastics equipment from the References United States and Europe to operate with local, low- cost labour. The existence of stretch suggests that this Air Publication. (1954) Official History: The Second World machinery has a considerable future, providing wear War 1939–45 Royal Air Force: Maintenance. London: parts, deteriorating components and electronic cir- Air Ministry, Air Historical Branch. CD 1131; reissued cuits are replaced during reconstruction. Rebuilding AP 3397, March 1966. provides an opportune moment for upgrading and Alsop, P.A., Chen, H., and Tseng, H. (2007) The Cement retrofitting key plant items. Data are not available Plant Operations Handbook for Dry-Process Plants: on the performance of these plants subsequent to The Concise Guide to Cement Manufacture. 5th edn. recommissioning. Dorking: Tradeship Publications for David Hargreaves, International Cement Review. Argote, L. and Epple, D. (1990) Learning curves in manufacturing. Science, 247, 4945, 920–924. Arrow, K. (1962) The economic implications of learning 5. Conclusion by doing. Review of Economic Studies, 29, 3, 155–173. Arvedi, G., Mazzolari, F., Siegl, J., Hohenbichler, G., and We propose a taxonomy of stretch based on five Holleis, G. (2010) Arvedi ESP first thin slab endless aspects of plant enhancement: (1) learning to operate casting and rolling results. Ironmaking and Steelmaking, and maintain; (2) addition of novel equipment and 37, 4, 271–275. reconstruction; (3) system-wide improvements Aylen, J. (1980) Innovation in the British steel industry. In: upstream, notably heavier piece weights; 4) system- Pavitt, K. (ed.), Technical Innovation and British Eco- wide improvements downstream to improve the nomic Performance. London: Macmillan. pp. 200–234. Aylen, J. (1982) Plant size and efficiency in the steel indus- offtake of materials; and 5) quality improvements try: some international comparisons. National Institute and development of new products. In the case of wide Economic Review, 100, 2, 65–76. strip mills built under Marshal Aid in Europe, stretch Aylen, J. (1989) A classic of its type: IISI on maintenance. enhanced their capacity by a factor of around 1.8 Steel Times, 217, 10, 566–567. over 30 years, and in one case the increase is over 8 Aylen, J. (2001) Where did Generation V strip mills come times the initial rated capacity. from? Steel Times, 229, 7/8, 232–237.

Aylen, J. (2004) Megabytes for metals – the development Engstrand, P. and Johansson, B. (2009) Paper recyling. In: of computer applications in the iron and steel industry. Ek, M., Gellerstedt, G., and Henriksson, G. (eds), Ironmaking and Steelmaking, 31, 6, 465–478. Pulping Chemistry and Technology. Berlin: Walter de Aylen, J. (2010) Open versus closed innovation: develop- Gruyter. pp. 391–428. ment of the wide strip mill for steel in the USA during Ess, T.J. (1941) The Modern Strip Mill: A Recording of the the 1920’s. R&D Management, 40, 1, 67–80. Continuous Wide Strip Mill Installations and Practices Baldwin, J.R. and Rafiquzzaman, M. (1995) Selection in the United States. Pittsburgh, PA: Association of Iron versus evolutionary adaptation: learning and post-entry and Steel Engineers. performance. International Journal of Industrial Ess, T.J. (1970) The Hot Strip Mill Generation II. Pitts- Organization, 13, 4, 501–522. burgh, PA: Association of Iron and Steel Engineers. BISF (1953) The wide strip mill. Monthly Statistical Bul- Fisher, H.R. (1963) Obsolescence and optimum replace- letin, 28, 7, 1–20. ment timing. Chemical Engineer, 167, 86–98. Brascugli, G., Mascanzoni, A., Schwellenbach, J., Muller, Frankel, M. (1955) Obsolescence and technological J., Quix, F., and Dzierzawski, J. (2002) CSP-concepts change in a maturing economy. American Economic for stainless steel and the first CSP production plant at Review, 45, 3, 296–319. AST, Italy. 85th Steelmaking Conference, Nashville, Frishammar, J. and Hörte, S.Å. (2005) Managing external TN, USA, 10–13 March 2002, 225–243. information in manufacturing firms: the impact on inno- Bryant, J.M. and Dimblebee, R.A. (1986) Refurbishment vation performance. Journal of Product Innovation of hot strip mill at BSC Port Talbot Works. Steel Times Management, 22, 3, 251–266. International, 10, 3, 16–26. Gard, N.R. (1966) Thirty years of steam-reforming – a Campbell, S.M. (2009) On the absence of obsolescence. review of ICI developments and experience. Nitrogen, IEEE Annals of the History of Computing, 31, 4, 118– the Journal of World Nitrogen, 39, 25–29. January– 120. February. Carroll, W.P. and MacNeil, P.R. (1985) Design and start-up Gomulka, S. (1978) Growth and the import of technology: of the Stelco 2050-mm hot strip mill. Iron and Steel Poland 1971–1980. Cambridge Journal of Economics, Engineer, 62, 4, 39–45. 2, 1, 1–16. Coe, B.P. (1962) Capacity stretch. Industrial and Engineer- Gore, L.G. and Shegog, C. (1979) Lysaght’s 2050-mm ing Chemistry, 54, 3, 47–51. Generation III hot strip mill. Iron and Steel Engineer, 56, Cook, J.W. (1979) Hot strip mill roughing train arrange- 3, 25–34. ment and power selection. Iron and Steel Engineer, 56, Gort, M. and Boddy, R. (1967) Vintage effects and the time 12, 25–31. path of investment in production relations. In: Brown, Cowling, P. (2003) A flexible decision support system for M. (ed.), The Theory and Empirical Analysis of Produc- steel hot rolling mill scheduling. Computers and Indus- tion. NewYork: National Bureau of Economic Research. trial Engineering, 45, 2, 307–321. pp. 395–422. Dixit, A. (1980) The role of investment in entry-deterrence. Gregory, R.G. and James, D.W. (1973) Do new factories Economic Journal, 90, 357, 95–106. embody best practice technology? Economic Journal, Doms, M., Dunne, T., and Roberts, M.J. (1995) The role of 83, 332, 1133–1155. technology use in the survival and growth of manufac- Hart, P.E. (1983) Experience curves and industrial policy. turing plants. International Journal of Industrial International Journal of Industrial Organization, 1,1, Organization, 13, 4, 523–542. 95–106. Donaldson, L. (1996) The normal science of structural Helfat, C.E. (1994) Evolutionary trajectories in petroleum contingency theory. In: Nord, W.R., Clegg, S., and firm R&D. Management Science, 40, 12, 1720–1747. Hardy, C. (eds), Handbook of Organization Studies. Hendry, C. and Harborne, P. (2011) Changing the view of London: Sage. pp. 57–76. wind power development: more than ‘bricolage’. Dunne, T. (1994) Plant age and technology use in U.S. Research Policy, 40, 5, 778–789. manufacturing industries. RAND Journal of Economics, von Hippel, E. (1977) Transferring process equipment 25, 3, 488–499. innovations from user innovators to equipment manufac- Dutton, J.M. and Thomas, A. (1984) Treating progress turing firms. R&D Management, 8, 1, 13–22. functions as a managerial opportunity. Academy of Man- von Hippel, E. and Tyre, M.J. (1995) How learning by agement Review, 9, 2, 235–247. doing is done: problem identification in novel process Earnshaw, E. (1978) Notes on the layout of wide hot-strip equipment. Research Policy, 24, 1, 1–12. mills. Steel Times, 206, 4, 303–316. Holt, K. (1974) Work frustration in engineering depart- Edgerton, D. (2006) The Shock of the Old: Technology ments. R&D Management, 5, 1, 17–21. and Global History Since 1900. London: Profile Hutcheson, P., Pearson, A.W., and Ball, D.F. (1995) Books. Innovation in process plant: a case study of ethylene. Edler, J., Meyer-Krahmer, F., and Reger, G. (2002) Journal of Product Innovation Management, 12, 5, 415– Changes in the strategic management of technology: 430. results of a global benchmarking study. R&D Manage- Hutcheson, P., Pearson, A.W., and Ball, D.F. (1996) ment, 32, 2, 149–164. Sources of technical innovation in the network of com-

panies providing chemical process plant and equipment. aspects. Working Paper no.36, Research Programme on Research Policy, 25, 1, 25–41. Scientific and Technological Development in Latin IISI. (1989) Maintenance for the 1990’s. Brussels: Com- America, Buenos Aires, ECLA Office in Buenos Airies. mittee on Technology of International Iron and Steel Miner, A.S., Bassoff, P., and Moorman, C. (2001) Organi- Institute. zational improvization and learning: a field study. IISI. (1992) Towards a Better Steelworks’ Yield. Brussels: Administrative Science Quarterly, 46, 2, 304–337. Committee on Technology of International Iron and Ozman, M. (2011) Modularity, industry life cycle and open Steel Institute. innovation. Journal of Technology Management and Kidd, P.G. and Dimblebee, R.A. (1987) The rebuilding of Innovation, 6, 1, 27–37. Port Talbot Works hot strip mill. Proceedings of the Pisano, G. (1997) The Development Factory: Unlocking Institution of Mechanical Engineers, 201, part A, 4, the Potential of Process Innovation. Boston, MA: 223–240. Harvard Business School Press. Lacci, L.A., Davies, S.W., and Smith, R. (1974) Tunnel Plotkin, J.S. (2005) The changing dynamics of olefin kilns in brickmaking. In: Nabseth, L. and Ray, G.F. supply/demand. Catalysis Today, 106, 1–4, 10–14. (eds), The Diffusion of New Industrial Processes. Cam- Price, D.J. (2007) Scunthorpe Steelworks 1973–2006: a bridge: Cambridge University Press. pp. 105–145. history of continuous development part 2 – process and Lager, T. (2002) A structural analysis of process develop- product developments. Ironmaking and Steelmaking, 34, ment in process industry–anewclassification system 5, 389–402. for strategic project selection and portfolio balancing. Ramlogan, R., Mina, A., Tampubolon, G., and Metcalfe, R&D Management, 32, 1, 87–95. J.S. (2007) Networks of knowledge: the distributed Lager, T. and Frishammar, J. (2010) Equipment supplier/ nature of medical innovation. Scientometrics, 70,2, user collaboration in the process industries: in search of 459–489. enhanced operating performance. Journal of Manufac- Ranieri, R. (2012) Wide strip mills in Europe from the turing Technology Management, 21, 6, 698–720. 1930s to the 1960s: a comparative perspective. In: Lager, T., Hallberg, D., and Eriksson, P. (2010) Develop- Aylen, J. and Ranieri, R. (eds), Ribbon of Fire: How ing a process innovation work process: the LKAB Expe- Europe Adopted and Developed US Strip Mill Technol- rience. International Journal of Innovation Manage- ogy (1920–2000). Bologna: Pendragon. pp. 79–130. ment, 14, 2, 285–306. Reichstein, T. and Salter, A. (2006) Investigating the Larsson, A. and Bergfors, M. (2006) Heads or tails in sources of process innovation among UK manufacturing innovation strategy formulation? Porterian or Penrosian, firms. Industrial and Corporate Change, 15, 4, 653– let context determine. International Journal of Process 682. Management and Benchmarking, 1, 4, 297–313. van Rooij, A. (2004) Building Plants: Markets for Tech- Lieberman, M.B. (1984) The learning curve and pricing in nology and Internal Capabilities in DSM’s Fertiliser the chemical processing industries. Rand Journal of Business. 1925–1970. Eindhoven: Aksant. Economics, 15, 2, 213–228. van Rooij, A. (2005) Engineering contractors in the chemi- Lovay, A. (1957) Walzwerksanlagen der Vereinigten cal industry. The development of ammonia processes, Österreichishen Eisen- und Stahlwerke in Linz a.d. 1910–1940. History and Technology, 21, 4, 345–366. Donau. Berg und Hüttenmännische Monatshefte, 102,4, Ruth, M. and Amato, A. (2002) Vintage structure dynamics 135–146. and climate change policies: the case of US iron and Love, J.H. and Roper, S. (2009) Organizing innovation: steel. Energy Policy, 30, 7, 541–552. complementarities between cross-functional teams. Salter, W.E.G. (1966) Productivity and Technical Change. Technovation, 29, 3, 192–203. 2nd edn. Cambridge: Cambridge University Press. Luger, A. and Hubmer, G. (2001) In-line control of the Salvanes, K.G. and Tveteras, R. (2004) Plant exit, vintage mechanical properties of hot rolled strip based on com- capital and the business cycle. Journal of Industrial Eco- plete process modelling. In: Rolling Committee of the nomics, 52, 2, 255–276. Institute of Materials (ed.), Modelling of Metal Rolling Saxena, J.P. (2009) The Rotary Cement Kiln: Total Produc- Processes – Through Process Modelling. London: Insti- tive Maintenance, Techniques and Management.New tute of Materials; 6 December. Delhi: Tech Books International. Malerba, F. (1992) Learning by firms and incremental Schneider, B. (1987) The people make the place. Personnel technical change. Economic Journal, 102, 413, 845– Psychology, 40, 3, 437–453. 859. Shone, R. (1975) Price and Investment Relationships: A Marsh, P. (2011) ArcelorMittal concedes defeat in patents Study in Applied Economics. London: Elek. battle with Voestalpine. Financial Times, March 28. Simon, H.A. (1979) Rational decision making in business Mata, J., Portugal, P., and Guimarães, P. (1995) The sur- organizations. American Economic Review, 69, 4, 493– vival of new plants: start-up conditions and post-entry 513. evolution. International Journal of Industrial Organiza- Sinclair, G., Klepper, S., and Cohen, W. (2000) What’s tion, 13, 4, 459–481. experience got to do with it? Sources of cost reduction in Maxwell, P. and Teubal, M. (1980) Capacity-stretching a large specialty chemicals producer. Management technical change: some empirical and theoretical Science, 46, 1, 28–45.

Sinnott, R.K. (2005) Chemical Engineering Design. Weiss, H. (1978) Die anlagentechnische Entwicklung der Volume 6 (of ‘Coulson & Richardson’s Chemical Engi- Hüttenindustrie in den letzen Jahrzehnten. Stahl und neering’), 4th edn. Amsterdam: Elsevier. Eisen, 98, 24, 1267–1276. Sterner, T. (1990) Energy efficiency and capital embodied Wirth, M.F. (2001) Die Breitbandstrasse Für Linz. Linz: technological change: the case of Mexican cement Geschichte-Club Voest Archive, mimeo. manufacturing. Energy Journal, 11, 2, 155–167. Witt, U. (1998) Imagination and leadership – the neglected Sull, D.N. (1999) The dynamics of standing still: Firestone dimension of an evolutionary theory of the firm. Journal of Tire and Rubber and the radial revolution. Business Economic Behaviour and Organisation, 35, 2, 161–177. History Review, 73, 3, 430–464. Wolf, F. and Van der Rijst, A. (1982) Der Einfluss des Tether, B.S. and Metcalfe, J.S. (2003) Horndal at Heath- technischen Fortschritts auf die Investions- und Verar- row? Capacity creation through co-operation and system beitungskosten von Massenstahl. Stahl und Eisen, 102, evolution. Industrial and Corporate Change, 12, 3, 437– 5, 233–238. 476. Yin, J.Z. (1994) Managing process innovation through Tweraser, K. (2000) The Marshall Plan and the reconstruc- incremental improvements: empirical evidence in the tion of the Austrian steel industry, 1945–1953: the petroleum refining industry. Technological Forecasting bureaucratic politics of trusteeship, nationalization, and Social Change, 47, 3, 265–276. and planning as reflected in the rise of the United Steel Iron and Steel Works in Linz. In: Bischoft, Jonathan Aylen is Senior Lecturer and a Director of G., Pelinka, A., and Stiefel, D. (eds), The Marshall Plan in Austria. London: Transaction Publishers. pp. 290– the Manchester Institute of Innovation Research, 322. Manchester Business School, University of Man- Utterback, J.M. (1994) Mastering the Dynamics of Inno- chester, England. He is a member of the rolling vation. Boston, MA: Harvard Business School Press. committee of the Institute of Materials, Minerals Van der Rijst, A., Derkx, H.H.J., and Kamerman, A. (1978) and Mining and chair of the Newcomen Society in How experience and attitude affect steelplant productiv- north-west England. Recent publications focus on ity. Iron and Steel International, 51, 319–329. computer history, transfer of technology, forecasting Van Ditzhuijzen, G. (1993) The controlled cooling of hot wildfires and on foresight techniques. rolled strip: a combination of physical modeling, control problems and practical adaption. IEEE Transactions on Automatic Control, 38, 7, 1060–1065.

3. Bloodhound - building the Ferranti Argus process control computer

Jonathan Aylen, “Bloodhound on my trail: building the Ferranti Argus process control computer”, The International Journal for the History of Engineering & Technology, vol. 82, no. 1, January 2012, pp.1–36

63 int. j. for the history of eng. & tech., Vol. 82 No. 1, January, 2012, 1–36

Bloodhound on my Trail: Building the Ferranti Argus Process Control Computer Jonathan Aylen Manchester Institute of Innovation Research, University of Manchester, UK

Digital computers for process control were developed at the end of the 1950s. They had different design objectives from computers for scientific or commercial use. The Ferranti Argus was among the first computers worldwide used for direct digital control. The Argus was invented at Ferranti’s Wythenshawe Automation Division, Manchester, by Maurice Gribble. The starting point was a prototype digital computer developed for the Blue Envoy guided missile using low power hearing aid transistors. Announced by Ferranti as the ‘process control transistor computer’ in November 1958, Argus came into civilian and military use in 1962. It was used for process control at a soda ash plant for ICI and as part of a Cold War missile guidance system for the Bloodhound Mark 2 surface-to-air missile deployed by the Royal Air Force. While a small team of engineers within Ferranti used Argus to develop digital techniques for guided missile control, another technically powerful group of civilian users led development of the Argus for direct digital control of an ICI chemical plant at Fleetwood, Lancashire. The paper shows how the computer was invented, how it was developed in military and civilian contexts by small communities of practice and how these groups coalesced, grew and dispersed. As projects shifted towards software development, teams became smaller and women programmers were given considerable responsibility. These events highlight a key transition from analogue to digital control in manufacturing industry and defence during the early 1960s. Use of direct digital control by ICI followed commercial logic. The military were forced to switch to digital computation because technical advances in radar meant analogue calculations would not be accurate enough for Bloodhound Mark 2. keywords Cold War, automation, guided weapons, digital computer, process control, defence innovation, transistors, ferrite core memory, interrupts, communities of practice, analogue control

Read at the Museum of Science and Industry, Manchester, 26 October 2010

If credit for ‘invention’ must be assigned, it should go to the person or team that first had a clear vision of the principle, saw its potential, fought for its acceptance and brought it fully into satisfactory use. (W. Brian Arthur)1 This is the story of an early computer, the Ferranti Argus, from concept to commercial sale. The focus is on the people who made the technology work — those who shaped the way the computer was developed and brought into military and civilian use. The account stands in stark contrast to most histories of UK computer development of the 1950s and 1960s. The Ferranti Argus computer was built by an automation group, not a computer division. Argus had its origins in attempts to control guided missiles. It was designed for process control, rather than scientific analysis or commercial transactions. Contrary to the usual narratives of failure surrounding UK computers, Argus was a commercial success.2 Process control computers, such as Argus, are technically significant because they had different design objectives from mainframe computers developed for general purpose scientific and commercial calculations.3 Different priorities induced innovation. The need to sample data from a wide variety of sensors, calculate time-based derivatives and distribute signals required very strict timing control. Conventional mainframe computers processed material in intermittent batches. Process control computers worked continuously. The imperatives of real time operation brought development of interrupts (direct memory access) for both processors and memory of the sort pioneered by Maurice Gribble in the design of the Argus 200. The overlap with defence brought early adoption of transistors and ferrite core memories and, subsequently, visual display units.4 There was a need to convert analogue to digital signals. Reliable, continuous operation in a hostile industrial environment required robust equipment as these computers had to operate for months on end, often away from an air-conditioned environment. Argus was built to tough defence standards. The Argus computer was invented and funded in a Cold War context. Early work on digital control was intended for the Blue Envoy surface-to-air guided missile. The evolution of Argus at Wythenshawe was independent of the activities of Ferranti’s own computer Division at Gorton. Development of Argus depended upon two key communities of practice that grew up around military and civilian use. Civilian adoption of the Argus was directed and single-minded, with ICI as a lead user. The switch to digital control on the military side owes a lot to chance events, personal friendships and the shortcomings of arithmetic calculation on analogue computers. The first part of this account sets the context of Ferranti Automation Systems in Wythenshawe and development of guided missiles, including the widespread use of analogue computers and an early prototype digital computer made with transistors. Wythenshawe was the site for the invention of the digital ‘Ferranti process control transistor computer’ by Maurice Gribble and the resulting computer hardware, which became the Ferranti Argus computer. Adoption of Argus by ICI at Fleetwood is then discussed and its use for the Bloodhound Launch Control Post. We briefly consider the subsequent development of Argus and conclude with implications for the history of engineering and technology. BLOODHOUND ON MY TRAIL 3

Ferranti and the Bloodhound Guided Missile

The ﬁ rm of Ferranti Ferranti was a family owned and controlled engineering firm, begun in 1882 as a small-scale heavy electrical engineering business. After the Second World War, the firm diversified towards electronic components, computer systems, avionics and radar while retaining family ownership and control.5 The firm had a tradition of technological leadership in its chosen sectors, investing for the long term and developing innovative engineering for its own sake. By 1965 the firm employed over 19,000 and had a turnover of £40.9 million (about £620 million at 2010 prices). The Ferranti board gave considerable discretion to divisional management. The firm was a set of feudal kingdoms with little interchange between the divisions. There was rivalry between Ferranti Computers at Gorton and Ferranti Automation at Wythenshawe, just 9 miles apart in Manchester. Ferranti entered the defence market in a major way in 1943 when the Scottish Group was formed to make gyroscopic gun sights for aircraft with the establishment of a factory in Edinburgh paid for by the UK Government. After the war the firm traded upon its technical excellence and became accustomed to dealing with Government defence procurement. By the 1960s, Ferranti depended upon retained profits for its internal funds. The military divisions at Bracknell, Edinburgh and Wythenshawe in Manchester cross-subsidized civilian activity, both in growth areas such as semiconductors and traditional sectors such as transformers. As the historic advantage conferred by technical leadership in wartime and Cold War waned, so did Ferranti. Ferranti struggled to adapt to civilian life.

Guided missile development — Red Duster to Bloodhound Mark 1 The British government invested heavily in guided missile defence after the Second World War. Guided weapons development was the second largest component of UK government defence Research and Development expenditure — after aircraft — during the 1960s.6 The Bloodhound Surface-to-Air Guided Missile was developed as part of this post-war programme. It is described as ‘one of the most successful, widely deployed and long-lived surface-to-air missiles of the Cold War’.7 Bloodhound was operated by the Royal Air Force in the UK, in the former West Germany, in Cyprus and in the Far East in Malaysia and Singapore. Bloodhound missiles and guidance systems were sold to the Royal Australian Air Force and neutral Sweden and Switzerland. Export orders were refused for South Africa due to an arms embargo imposed by the Labour Government in the UK.8 The Bloodhound missile began life in 1949 as a research project to develop a surface-to-air anti-aircraft missile called ‘Red Duster’ with the Bristol Aeroplane Company as lead contractor.9 The aim was to intercept high-flying bombers carrying atomic warheads against Britain. By 1950, guided missiles were accorded the same priority as nuclear weapons development and there was considerable overlap between the two military R&D programmes.10 Ferranti was responsible for the forebody of the Bloodhound missile, including the gyroscopes and radar dish, the launch control post and a number of incidental parts including test equipment, data links and fuze safety.11 Power came from two Bristol 4 JONATHAN AYLEN

Siddeley kerosene-fired ramjets and there were four detachable booster rockets for launch. The missile was conceived as a twist and steer monoplane with fixed tail fins and moving wings.12 Development of Red Duster began at Ferranti, Moston in north-east Manchester but work moved to a purpose-built site composed of laboratories and final assembly facilities at Wythenshawe paid for by the Ministry of Supply, officially opened in June 1954. Everyone at the site signed the Official Secrets Act. The Red Duster development team at Ferranti was led by Dr Norman Searby with strong technical support from Denis Best. Development was organized in small teams. Typically, each team was assigned to one of the labs at Wythenshawe. Each team had the responsibility of designing and testing a component part of the missile. The Red Duster guided missile entered service with the RAF as the ‘Bloodhound’ Mark 1 in October 1958.13 In many respects the Mark 1 Bloodhound was a stop-gap. Initially up to 368 missiles were deployed at eleven sites along the east coast of England. The associated Type 83 Target Tracking and Illuminating Radar was supplied by British Thomson-Houston. This pulsed radar was susceptible to jamming and suffered interference from ground clutter, so the missile could not fire against low-flying targets. These shortcomings brought a switch to continuous wave tracking and illuminating radar (Type 86).14 Along with other improvements, this culminated in the longer and more powerful Bloodhound Mark 2 deployed from 1 October 1963. The system was finally stood down in 1991. Longevity of Bloodhound Mark 2 as a weapons system is partly attributable to adoption of digital control which gave scope for up-rating as technology progressed.

Analogue and digital computing at Ferranti Ferranti Automation made considerable use of analogue computing for missile design, testing and control. Analogue computers were developed by US and UK aerospace companies during the 1950s to carry out engineering design calculations and model the behaviour of aircraft and missiles in flight.15 These analogue computers used continuously variable direct current voltages to represent values in a problem. The voltages were made to follow mathematical relationships ‘analogous’ to the problem under investigation. The electronic components of the computer forced these relationships to simulate the actual problem. The resulting current outputs gave the solutions. Analogue machines were usually built up from direct current amplifiers with associated resistive and capacitive networks to shape or modify the signal. A simple analogue computer could be made from radio components. Greater accuracy required an increase in the number and size of components. The operations of addition and subtraction were easy to perform, but multiplication and division were more difficult and required special components or servomechanisms. The accuracy of analogue computers is limited to about 0.1 per cent and errors accumulate. Analogue computers were ideal for simulating a missile and its target manoeuvring in three dimensions in real time.16 Circuits were constructed to replicate the pitch, roll and yaw of a fast moving missile. During the 1950s Ferranti Automation built at least two analogue computers in their simulator section to solve aerodynamic problems relating to Red Duster.17 The initial machine begun in 1950 modelled BLOODHOUND ON MY TRAIL 5 the three-dimensional trajectory for Red Duster. After trials with an early analogue computer, a ‘Complex End Course Machine’ started work in 1956 to model detailed aerodynamics of the missile and the propulsion system. The simulator section was led by Tommy Thompson and employed ten or a dozen staff, initially in Moston and then at Wythenshawe.18 It was normal for users of analogue computers to build their own machines. As Dr Williams of rival missile designer English Electric said, ‘the design and construction of simulators was, in those days, necessarily a ‘do-it-yourself’ activity since no standard computing units were available on the British Market’.19 Digital techniques were developed to overcome shortcomings in analogue, notably accuracy and replicability. Digital computers were also ‘universal machines’ which could be put to work on any task, rather than purpose-built to replicate a specific problem. The apparently radical notion of building your own digital computer does not seem unusual when you have been self-assembling analogue computers for five or six years and have the relevant skills in electronic circuit design and knowledge of components. Many aerospace firms turned their analogue expertise into commercial products. Ferranti’s competitor English Electric Co. Guided Weapons Division exhibited their LACE (Luton Analogue Computing Engine) at the same Olympia Exhibition where the Ferranti Argus was first unveiled.20 Fairey Aviation Co. Weapons Division also developed a similar analogue system shown at the exhibition, alongside machines from two other aerospace companies Short Brothers and Harland and Saunders-Roe. In contrast, Ferranti Automation went digital. This is an example of Vincenti’s ‘variation-selection model’ for the growth of engineering knowledge: all British aircraft companies faced the same problems of development, analysis, evaluation and optimization during the 1950s, but just one firm — Ferranti Automation — came up with a different outcome, a digital computer at the 1958 exhibition while the rest showed analogue machines.21 Ferranti had already entered the digital computer business after the second world- war, drawing upon wartime experience of radar and electronics research in the UK.22 Many key radar personnel ended up in Manchester after the war, either working for Ferranti or at Manchester University, having worked together at the Telecommunica- tions Research Establishment (TRE), Malvern — the ‘TRE Mafia’ as Peter Hall describes them.23 Ferranti found Manchester University was already working on the creation of the world’s first stored programme computer, the Small Scale Experimen- tal Machine. Ferranti received their first order for a digital computer from the University on 26 October 1948 and the firm formed a separate Computer Group in 1949 as an outgrowth of the Ferranti Instrument Department at Moston specifically to develop and manufacture digital computers. The Ferranti Computer Group became one of the world leaders in commercial and scientific computing until the end of the 1950s from their base in Gorton. The main Board of Ferranti was evidently supportive of digital computer development. The main significance of Ferranti Computers to the quite separate development of Argus is the brand name and marketing organization in London. Ferranti Computers became a trusted digital mainframe supplier with a loyal customer base among steel, chemical and engineering companies. These same companies played a leading role in 6 JONATHAN AYLEN adopting process control computers built by Ferranti Automation. Pegasus was a general purpose digital computer introduced by Ferranti Computers in March 1956.24 Three Pegasus computers were sold to the steel industry for research calculations; two more to Shell for research and one each to ICI and Babcock and Wilcox.25 These Ferranti mainframe users were to become Argus lead customers. Customers associated the Ferranti name with computer design. Otherwise there was very little technical cooperation between adjacent technical teams at Gorton and Wythenshawe, or between Wythenshawe and the Military Division at Bracknell, at least up until the mid-1960s.26

Blue Envoy, hearing aid transistors and development of the Ferranti process control transistor computer

Maurice Gribble The Argus digital computer was invented and designed by Maurice Gribble at Ferranti’s Automation Division, Wythenshawe, although he was supported by a team of growing size as the computer developed from experiment to prototype and on to commercial production.27 Maurice was born in Ipswich and a radio ham at the age of fourteen while still at Ipswich School: ‘I was always drawing radio circuits in French [lessons]’ he confesses.28 During the Second World War, Maurice trained as a pupil in the local power station for 2½ years before leaving this reserved occupation early in March 1942 to volunteer for the RAF. Here he worked on ‘Gee’ radar beams for torpedo boat and aircraft guidance, initially in the UK and then mostly in France but also in Germany at the close of the war. He went on to take a University of London degree in physics at Woolwich Polytechnic after the war. He was recruited by Vivien Bowden to work at Ferranti’s in 1951, initially on their display for the Festival of Britain at the Science Museum. He moved to Moston to work on the Ferranti Mark 1 computer for Manchester University and developed equipment to drive a big line printer made by Machine Bull in France. He moved on to Ferranti’s Automation Division because he wanted to do something ‘more theoretical’.29

Blue Envoy and command guidance control By the mid-1950s development of the Red Duster guided missile, which was to become the Bloodhound Mark 1, was well under way at Wythenshawe. At that stage experimental test missiles were under trial, first at Aberporth in Wales and then in the clearer weather of the Long Range Weapons Establishment, Woomera, in Australia. Around the same time, 1954/55, Ferranti Automation was also working on a quite separate surface-to-air guided missile with a very much extended range of 320 km. This surface-to-air guided missile was code-named ‘Blue Envoy’. The concern was that Soviet bombers would be equipped with stand-off weapons and needed to be intercepted at greater distance. The missile was intended for both land-based and naval use. Blue Envoy has been described as ‘possibly the most enigmatic project in the field of 1950s United Kingdom weapons development’.30 It was also known as Bloodhound Stage 1¾. In some respects, such as engine testing, Blue Envoy was a prototype for the shorter-range Bloodhound Mark 2. Otherwise comparison with Bloodhound is BLOODHOUND ON MY TRAIL 7 misleading since Blue Envoy was under command guidance from the ground during the first part of its flight, whereas Bloodhound used a reflected radar signal and on-board guidance to home in to the target. Blue Envoy was developed by Bristol Guided Weapons Department and Ferranti. The missile was made of stainless steel, equipped with double-delta wings and was slightly longer and faster than Bloodhound Mark 1. One anti-aircraft option was to fit a low yield nuclear warhead called Blue Fox. The Blue Envoy missile was tested at the scale-model stage, but the full-size missile never flew before it was cancelled in April 1957 as part of that year’s Defence Review. In any event, it lacked short-range capability against low-flying attacks under the radar. It is said that the concept of Bloodhound Mark 2 was invented in a London taxi as an urgent response to cancellation of Blue Envoy.31 As the name implies, ‘guided’ missiles need to be carefully controlled, both in terms of direction onto the target and stability during flight. Firing a guided missile is a classic dynamic control problem, like a dog chasing an agile hare. But at least the dog only pursues the hare on the ground in two dimensions. A missile chasing a target aircraft has to deal with three dimensions, predicting the future position of the target from its current location and flight behaviour. Air density varies, so an optimum course may involve a near-vertical ascent to escape the drag imposed by the dense lower atmosphere, before heading off towards the target.32 To complicate matters, the earth is not strictly round, so allowance has to be made for this feature too.33 During the initial command-guidance phase of flight, Blue Envoy received control signals from the ground. The first task was to get the missile on course after launch. The missile needed to correct its course so that it aimed in the right direction towards target interception. (Final closure onto the target was by riding a narrow radar beam.) But, before the missile can manoeuvre onto the target using directions from ground control, it needs to establish its angular error from the likely direction of interception. Circuitry was also needed to actuate the servomechanisms used to control the Blue Envoy missile during its prolonged mid-course flight. Maurice Gribble developed digital logic circuits to code the angular position of the missile for onward transmission from the ground by a guidance beacon.34 The data would be decoded when it arrived in the missile and translated into control signals to steer the missile in the right direction. An obvious next step was to see if some of the actual controls on the missile itself could be undertaken digitally in response to these guidance commands from the ground. To this end, a special purpose digital computer was built to experiment with direct control of the servomechanisms.

The hearing aid transistor computer Maurice Gribble was given considerable freedom. He was a development engineer: ‘I didn’t really fit in with anything, no one quite understood what I was doing. No one ever questioned. [. . .] There wasn’t the tight financial control there is today’.35 At the time, a couple of teams in the UK were experimenting with computers built around early point contact transistors in place of thermionic valves. These transistors essentially consisted of a single crystal of germanium with two fine wires.36 A research student at the University of Manchester, Dick Grimsdale, ran a transistor computer from November 1953 onwards and similar experiments on small-scale transistor-based computers were successful at Harwell from February 1955 onwards. 8 JONATHAN AYLEN

Against this background, Maurice Gribble developed a prototype digital controller around 1956 using low-frequency junction transistors meant for hearing aid use (25 kHz) made by Mullard known as OC71.37 His view was that ‘a basic clock frequency of 25 kilocycles per second was slow by the standard of mainframe computers working on thermionic valves, but at the time high frequency transistors were not available’.38 This ‘hearing aid transistor computer’, as it was colloquially known, could add, subtract and multiply. Division was performed by software using the Newton-Raphson algorithm, while the trigonometric functions used Chebyshev polynomials. The approach of the hearing aid computer was simple: a single address and serial operation. But, Maurice Gribble appreciated:

Memory was always a problem on early computers, in fact it is true to say that the type of memory largely determined the design of the computer.39 The immediate-access working memory was a set of transistor flip-flops with a capacity of 32 words (Figure 1). These were expensive since the number of transistors increased almost in line with the number of bits of information being stored. Another 35 words were wired into constant locations. But an important distinction can be made between remembering data and storing a programme. The designer says:

I realized that a computer for process control needed a relatively small amount of working memory, since data was read in, processed and output to the plant in real time. Data only needed to be stored until it was used. The program, on the other hand, would require far more memory and that is why a diode memory, in the form of a plug board, was used. This was cheap and fast, although not quite as flexible as storing the program on the same memory that was used for data.40

figure 1 The ‘hearing aid transistor computer’ was developed for digital control of the Blue Envoy Command Guidance Missile. Flip-flops cannister on the left and NOR gates cannister on the right. BLOODHOUND ON MY TRAIL 9

So the programme was a plug-board with capacity of 64 words in the main block and 12 sub-routine blocks of 25 words, giving a total of 364 words. The hearing aid transistor computer was, in short, a simple digital computer that could be programmed to give accurate, replicable results using binary arithmetic. The twist-and-steer elevons at the outer end of the wings of Blue Envoy were operated hydraulically. Circuits were developed to operate hydraulic valves by switching current into them for a length of time proportional to the digital output. The difficulty here is converting angles (or their sines or cosines) into digital form. You cannot readily jump from a 360° angle to a 1° degree angle in digital! The breakthrough was to develop a digitizing disk using a cyclic progressive code which moved gradually and neatly through the angles in small steps.41 The prototype computer was used for testing of digital servomechanisms, in much the same way that real components were already inserted into circuits in analogue computers. Here again the analogue culture of physical simulation of components reinforced the move to digital control.

A Royal demonstration The ‘hearing aid transistor computer’, was a major breakthrough in digital control of processes.42 Although Blue Envoy was cancelled early in 1957, Maurice Gribble carried on with development work.43 The computer might have remained a laboratory curiosity if Maurice Gribble had not used it in a simple and appealing demonstration of digital parallax correction to Duke of Edinburgh during an official visit to Ferranti at Wythenshawe, on 22 November 1957.44 This was a shrewd move as preparations for the Royal visit commanded priority in the technical support workshops and the demonstration had the attention of senior management and colleagues. This royal demonstration of digital control was set up in a Laboratory in H wing at Wythenshawe using two hydraulically operated turntables supporting optical projectors. One projected a green spot and the other a red spot onto a screen. The trick was to get the red spot to converge on the green. The angular position of each of the turntables was measured by digitally coded disks and a similar disk was connected to a control handle for the royal visitor. When the knob was turned, the two spots of light converged as the computer performed the trigonometry for the parallax correction and operated the servomechanisms beneath the lights. The demonstration used typical short cuts employed by engineers. Photoelectric cells were created by scraping the black paint off the glass capsules of transistors. Car headlamp bulbs were used for illumination. But it was clear evidence to visitor and management alike that digital techniques could be used to control an elementary process. It was always Ferranti’s intention to move into process control.45 These new digital techniques provided the opportunity.

The prototype Argus Within a year of the Royal demonstration, the next stage in the evolution of digital control — the Ferranti ‘transistorised process control computer’ — the prototype of the Argus computer — was displayed for commercial sale at the ‘Electronic Computer Exhibition’ at London Olympia, in November 1958.46 The machine was described breathlessly as: 10 JONATHAN AYLEN

[. . .] a type suitable for development for process control work, for use at the centre of an automation process. Its special interest lies in it being made entirely with transistors in place of thermionic valves, and it represents a most important advance in developing these new techniques.47 In truth, most of the contents of the cabinets on display were dummies. So when a Russian ‘diplomat’ showed particular interest in the computer Maurice Gribble had double reason for not revealing more details. The Argus name was developed in collaboration with Bernard Swann of Ferranti’s Computer Centre at 21 Portland Place, London. The name was chosen to complement other mythological names used for Ferranti computers such as Pegasus, Orion and Sirius. It was an apt choice because Argus was a supervisory computer: ‘Argus, the all-seeing, had a hundred eyes which slept in turns, so that he was at all times awake’. Bernard Swann saw beyond local differences in Manchester. He understood how a process control computer from Wythenshawe complemented the product range of mainframes from Gorton. The name Argus was used from May 1959 onwards. Argus later became the Argus 200 when a smaller version using only core memory, was designed by Dave Butler, Dave Senior, Stan Redshaw and others, called the Argus 100. This was a serial machine which was programmed from a paper tape. The Argus 100 variant was installed at industrial plants such as the Steel, Peech and Tozer’s electric arc steelmaking shop in Rotherham, while ICI bought six to use around its works, including a novel Paraquat plant at Widnes.48 The Argus computer was marketed as a digital computer for civilian process control 4½ years before it went into service with the RAF, even though development had been funded by defence contracts. Ferranti recognized the tasks of initial set-up and then controlling a guided missile in flight are the same problems as active control of fast-moving industrial machinery, such as multi-stand rolling mills, where roll gaps and speed need to be set up and immediate response is required in real time once production starts.

Innovations in the Argus computer — interrupts and memory There were no precedents for the Ferranti Argus. There were no other digital process control devices to copy. A remarkable sequence of innovations in digital computer design emerged at Wythenshawe during the year between November 1957 and November 1958, resulting in the prototype process control computer shown at Olympia. The beautifully written computer manual for Argus shows clear design objectives.49 The resulting technical developments included the ideas of interrupts and direct memory access, use of both serial and parallel operation, adoption of higher powered transistors and a ferrite core memory, and development of a novel peg-board for permanent programming — all developed to prototype stage in twelve months.

Interrupts The specification called for real time control and continuous operation. However, it was also necessary to read in and send out data and deal with any other urgent interventions while the computer was still in control. So the notion of interrupts was developed. Precise timing control was essential as the computer would be working in real time. BLOODHOUND ON MY TRAIL 11

Switching from analogue to digital control was not straightforward. At the time it was customary to run industrial and military processes with analogue devices. So control of servos was achieved by a continuously varying signal. Digital is different since the data is discrete and holds the same value until altered. So instead, digital data is sampled frequently to be ‘pseudo-continuous’. This was a design challenge. As Maurice Gribble said:

In the hearing aid transistor computer, the sampling rate was significantly low for it to be a sampling servo and the program was adjusted so that the samples occurred as regularly as possible, bearing in mind that multiplication, which used a short-cut method, took a variable length of time. I realized that this made programming very complicated and invented ‘interrupt’ where the computer stored what was in its registers and changed to a short program to read in data, output it, or operate a servo. In attempting to patent the idea, we found that it had been patented in the USA a few months previously.50 Argus used interrupts at various intervals controlled by a timer, but it was not allowed to interrupt a multiplication or division, or a jump to a sub-routine.

Core memory and read-only memory The Argus computer had two memories: a core memory and a semi-permanent peg-board memory. The core memory was a ferrite core memory. Ferrite cores were developed in the USA from 1950 onwards. They rely on the principle of coincident current selection. The memory is built up on a grid of fine horizontal and vertical wires, resembling the template for a game of noughts and crosses.51 A small magnetic core — known as a toroid — surrounds each intersection of the grid, like a diagonal bracelet. A half-current applied to either the vertical or horizontal wires alone is not sufficient to alter these magnets. However, if half-currents do meet at an intersection, the coincidence of the currents is sufficiently powerful to switch the magnetic core at that particular point. A third wire is used to read the state of these magnets. Since binary operations only require ‘on’ or ‘off’ — zero or negative current say, the state of each of these uniquely positioned magnets is sufficient to provide a permanent memory store. The settings were 0 volts (binary 1)/ minus 6 volts (binary 0). A novel contribution of Argus was the second memory, a semi-permanent peg-board memory.52 In contrast to mainframes, process control computers only require one dedicated programme with a specific purpose. Recall that the hearing aid computer used a plug-board to store the process control programme. It was a logical next step to develop a versatile and durable plug-board to store a programme for Argus. This was done using ferrite pegs. These pegs were the size of pencil leads and made up of baked ferrite, a ceramic material, essentially an iron oxide Fe3O4, evocatively known to geologists as lodestone. Since process control computers only need to run a single programme, the peg- board was a read-only memory. It was programmed by inserting a peg into the appropriate hole in a printed circuit tray. Each peg represented a binary 1, analogous to a hole in an 80-column punch card. Each line of pegs represented a 24 bit word of code. Each tray held 64 twenty-four bit words (plus a parity bit). Each store box held 16 such trays, making 1024 words in all. The launch control post had four stores, 12 JONATHAN AYLEN making 4096 words in total. This feature was only used on the first design of Argus, the ‘200’. The decision to use the peg-board was both considered and pragmatic:

In the case of the Argus 200, the reasons for using a separate magnetic peg board for program and constants were reliability and speed. The Argus used one of the first, if not the first, transistor driven core memories. Until that time, core memories had been driven by valves. The problem with the germanium transistors available at the time was that fast switching and the ability to handle the high currents and high voltages needed were incompatible, so it was difficult to find suitable transistors. Reliability was not very high, but this did not matter very much for data in a process control computer, since a random error would be corrected at the next cycle of computation. However, if the program were corrupted, it would be chaotic. The peg board memory was also a lot faster than a core memory read/write cycle of the time. We also had a patent on it.53 The insertion of small pegs into a board required great care, but did not have to be done very often since control programmes were seldom modified. Indeed ICI complained ‘manual programming is rather tedious and is inconvenient for small changes’.54 The delicate nature of the peg-boards became part of the folk myth of the Argus 200 since a day’s work could be easily undone by an inopportune jolt loosening the pegs, just like sweeping a jigsaw off a table by accident. To minimize these frus- trating accidents, the peg-boards of the early machines were fitted with a transparent cover. Derek Whitehead recalled Ferranti days when ‘we polished the lid of the peg board so the whole lot flew out on Pete Smith due to the static on the celluloid lid’.55 Again, Mark Walker remembers times at RAF West Raynham when careful work could be wrecked by a gust of wind catching the unwieldy board during the short journey outdoors from the section hut to a nearby Bloodhound launch control post.56

Speed of operation and testing At 500 KHz, Argus was far faster than the hearing aid transistor computer.

Argus 200 was designed for speed [. . .] Parallel arithmetic was much too expensive before the invention of integrated circuits, but a series-parallel system, using two bits in parallel, almost doubled the speed. A modified short-cut multiplier was used which dealt with three bits at a time and equally well with positive and negative numbers. A non-restoring divider also improved speed. Finally, the word length was made as short as possible with the option of double length operation when high accuracy was required.57 Even so, ICI criticized the speed of Argus in comparison with their mainframes, not that their elderly soda ash plant at Fleetwood required anything like the response speed they were offered.58 Argus ran on a machine code which was a modified version of the order code used for the Ferranti Pegasus mainframe computer. This was one of the few areas where Ferranti Automation borrowed ideas from the firm’s Computer Division. Use of Pegasus order code had the advantage that both ICI and United Steel could test programmes for the later Argus 100 on their Ferranti mainframes.59 Programming in machine code is tedious as it is necessary to keep track of the precise location of BLOODHOUND ON MY TRAIL 13

figure 2 Hearing aid transistors OC71 form the memory of the first process control computer developed for Blue Envoy by Maurice Gribble at Wythenshawe. Failure of the V10/1S at the production stage delayed Argus. every element in the memory and keep a note of the time taken for each operation. You have to remember the difference between two types of binary numbers: the memory address itself and the numerical contents of that memory address.

Technical setbacks and commercial manufacture It is a myth of technical development that things flow smoothly from prototype to production.60 Development of Argus was not without its setbacks. Selection of suitable junction transistors for Argus was one cause of technical difficulties. Maurice Gribble sought advice from the Radar Research Establishment at Malvern since they had experience of working with transistors. They recommended a transistor, V6R8, which would stand a high current. This was made by Pye at Newmarket. A sample was secured and proved better than anything else. Maurice Gribble decided Ferranti should build the whole computer with this one transistor and an order was placed for about 10,000. Faced with this large order, Pye’s subsidiary, Newmarket Transistors, brought out a high power version V10/1S especially for Ferranti Automation (Figure 2). However, Pye decided to manufacture them in a different way. The early model had a soldered can. Pye found they could speed up the manufacturing process by dipping the cans into flux before they were soldered. The finished transistors were then covered in silicone grease to keep the moisture out. But a fatal flaw was that soldering flux is 14 JONATHAN AYLEN hygroscopic. Moisture took time to work through the silicone grease: it took six months. After final manufacture the Argus computers started dying due to moisture ingress and Maurice Gribble got the blame. In particular, these faulty components delayed the whole Launch Control Post programme for Bloodhound 2 as much time was spent repairing and replacing dud transistors. Fortunately, Ferranti did not have to alter the printed circuits. But there was extensive reworking using an alternative GEC transistor (GET875s) with a welded can, which was less able to switch such high currents. But the Argus 200 did get into commercial production, selling 63 machines plus 14 of the stripped-down Argus 100 variant. The machine was built at Moston where assembly was led by Bob Grove, ‘a marvellous organiser’ who honed his skills as a production manager on Bloodhound Mark 1.61

Marketing and early sales of Argus for process control Marketing material for the Argus 200 emphasized reliability and adherence to AID defence standards in construction.62 The computer had to be reliable and robust to work in severe environments. The technology was — quite literally — ‘gold plated’: with gold coated connectors for durability and conductivity.63 Since the first application of Argus was to help make washing soda there could hardly be a more apt illustration of Whitfield’s tart observation relating to US Cold War equipment suppliers: ‘The push buttons that were designed to make housework easier came from the same laboratories as the push buttons for guided missiles’.64 The Argus computer was technically advanced. Core storage of programmes was emphasized — a key selling point at a time when machines with drum memory were slow and difficult to programme. The use of ferrite core storage and random access memory circumvented the slow speed of drum machines with their greater reliance on sequential memory. Transistor circuits, ferrite cores and random access memory became the ‘dominant design’ for process control computers of the first half of the 1960s. The first sale of Argus was to engineers Babcock & Wilcox for automated control of a power station boiler at West Thurrock in Kent following discussions in 1958 between Geoff Griffiths of Babcock & Wilcox (a keen Pegasus user) and Maurice Gribble and Chris Wilson of Ferranti.65 Boiler automation had particular resonance for Maurice who, as already mentioned, had trained as a pupil at Ipswich Power Station in the early 1940s. A team was set up, led by Norman Leece, to build and install an Argus 200 computer for boiler control at West Thurrock power station, then under construction in Essex for the Central Electricity Generating Board. The project ran into severe delays due to the slow pace of power station construction and the computer was put into store for a while.

ICI as lead user In every sense of the word, ICI was the lead user for Argus. ICI Alkali Division ordered an Argus 200 machine for an elderly soda ash plant at Fleetwood, Lancashire.66 ICI treated the Fleetwood plant as a commercial scale pilot plant. Adoption of Argus was effectively a giant R&D project.67 Ferranti wrote the software and provided the hardware (Figure 3). But ICI had the technical competence to BLOODHOUND ON MY TRAIL 15 develop the project by adding instrumentation, developing actuators to control valves and convert analogue signals to digital and back again.

Civilian communities of practice: the Argus 200 at ICI at Fleetwood Adoption of the process control transistor computer into practical use required development work by two teams, one military and one civilian. Such teams have been called communities of technological practitioners (by Constant), or communities of practice (by Wenger).68 Here we use the term communities of practice to refer to small groups of engineers and technologists who get together to implement a novel technology, coalesce around a shared problem, recruit outsiders with relevant skills and then leave to join another small community once the task in hand is complete. The central idea is that learning and problem-solving in technology is a process of social participation. In the case of the Ferranti Argus, the focus of these efforts was digital control of processes. In the civilian case, this was automation of a complete soda ash chemical plant. In the military case, it was use of digital control to direct a radar dish in the nose of a guided missile. These communities brought together people with considerable prior experience and previous patterns of cooperation, who shared their tacit knowledge and experience, focussed on solving problems of implementing novel process control technologies for their designated project and left a legacy of experts whose experience would be used on related projects. These communities were democratic and consensual with clear recognition of the respective talents of participants and deference to intellectual leadership. It is said that Dave Evans was selected by his fellows to lead the ICI team for Ferranti.69 They were subject to little managerial direction and conformed to group norms rather than, say, working hours laid down by the job. They had informal and sometimes unorthodox ways of working. Group cohesion around a common goal is reinforced by long hours of technical effort, the inevitable reverses and struggle towards an agreed solution. A strong feeling of community is emphasized by signing the Official Secrets Act. Participants could not talk about their work outside of the plant, but could talk to others bound by the same Act and share the gossip and scandal attendant upon such work. Communities of practice are not the same as a formal development team. These communities transcend organizational boundaries and recruit members from outside the team. The Fleetwood community encompassed both ICI and Ferranti personnel. The Launch Control Post team included Derek Whitehead who was formally assigned to missile development, not ground control. So, a community of practice relates to those who actually deliver the solutions rather than any formal organization diagram. The idea of applying direct digital control to a chemical plant seems to have been mooted within ICI during a visit to the USA by members of the Alkali Division early in 1957.70 ICI Alkali Division had a reputation as ‘gentlemen inventors and were prepared to spend money on developments that were not immediately profitable’.71 At the time, US process plant users were discussing the relative merits of analogue versus digital control, although the only practical experiment that had been attempte d 16 JONATHAN AYLEN was use of a digital computer to advise operators at a Du Pont plant at Niagara Falls down a phone line from Philadelphia. Alan Thompson, Instrument Manager of ICI Alkali Division, approached Ferranti around the time of the Olympia exhibition in November 1958 with a view to using their ‘process control transistor computer’ for direct control of a chemical plant. Negotiations took place with Bob Morley of Ferranti who was working on guided missiles at the time, along with Peter Corrin who was involved with design issues. ‘Alan Thompson was the key mover’, Maurice Gribble said:

Alan Thompson appreciated you could replace 100 three term controllers with a digital computer. It would then become economical. It was able to control valves. Servos need stabilisation. Alan Thompson realised its potential.72 The chemical plant chosen for the experiment was an obsolete soda ash plant at ICI Fleetwood Works in Lancashire built in 1926 at a site originally developed by United Alkali Company in the 1890s to avoid the grip of the Cheshire Salt Union. The plant ran continuously with a single shut down day each year.73 The plant was controlled manually using hand-turned valves and sight glasses. ICI claimed existing instrumentation was reaching the end of its life, but this was public relations gloss since there was very little instrumentation to speak of. New pneumatic valves were installed, actuated by electrical signals. Soda ash manufacture by the Solvay process

figure 3 A mimic of the ammonia soda ash plant at Fleetwood designed to show the role of the Ferranti Argus 200 used for process control. BLOODHOUND ON MY TRAIL 17 is safe and has no explosive or toxic hazards. The plant had complete duplication of pipe work so it was possible to install computer control on one pipe circuit to make sure it worked before implementation of the complete scheme. ICI sanctioned the project in March 1960 using the Ferranti Argus computer. The Argus 200 for Fleetwood was tested at Wythenshawe (Figure 4). A tiny simulation of the plant was built in a lab H2 at Wythenshawe to test hardware and software with the odd operating loop as a preliminary test. The equipment was not actually delivered until April/May 1962. The computer was disassembled and taken up to Fleetwood on the back of a lorry in parts. The soda ash plant was a hostile environment due to heat and dust. The inside of the plant was covered in a build-up of fine white powder, inimical to delicate circuits. So the Argus 200 computer at Fleetwood was installed in a purpose-built, air-conditioned cabin, complete with interlock doors some 30 feet off the ground right at the heart of the works above the hot moving stoker grates of the bicarbonate bands. The computer control room was both physically and socially isolated. Access was strictly limited to the ‘inside’ and ‘outside’ controllers who manned the room. The ten or so men trained to operate the plant were something of a labour aristocracy. The computer had 2048 words of peg-board storage and 1024 words of core storage. The computer controlled 178 valves using information gathered from 102 thermometers, 92 transducers, 8 analysers and 89 valve positions. Input and output signals were distributed over conventional analogue phone lines installed by ICI electrical fitters.

figure 4 Argus 200 destined for Fleetwood installed in Lab H2 of H wing at Ferranti Automa- tion, Wythenshawe, in late 1961 or early 1962. This photo shows the computer used at Fleetwood, later transferred to Winnington until 1980 and now in the Science Museum. 18 JONATHAN AYLEN

The team installing the Argus 200 at Fleetwood was led by Syd Evans as overall project manager. Mr Evans came from Wythenshawe where he had earlier worked on mid-course guidance as part of the cancelled Blue Envoy project, a clear transition from military to civilian work. The project team for Fleetwood began to coalesce early in 1961 and was composed of four or five Ferranti people, Syd Evans, Frank Moss, Dave Rushton, Norman Leece and Tim Gossling, a mathematician who was the programmer. To this may be added Alan Thompson from ICI who the team met frequently at Winnington and Wythenshawe along with his right-hand man, making perhaps eight engineers in total at the heart of the development. They made frequent visits to ICI offices at Winnington: ‘we coordinated with them, spoke the same language’.74 All the ICI experience was with two and three term analogue controllers. In the event, there were problems to overcome, such as the ‘noise’ problems in the Fleetwood installation caused by spikes in the electric current in the mains. (The whole works was run on 440 volt direct current supplied by in-house generating plant and later from outside through mercury arc rectifiers which were doubtless the source of the spikes.) They learned the hard way, the usual way to learn on these sorts of projects. Results showed 99.6 per cent availability for the system in the first 20 months after commissioning.75 Or, as a local electrician put it, ‘When it first started there was a lot of hiccups and that, but it actually settled down very well’.76

The role of programmers As the physical hardware of Argus developed, communities of practice coalesced, grew and dispersed. These teams became smaller as projects became an issue of implementation and software development. Here women were given great responsibility. Instead of a community of practice working on teething problems, there was typically a lone gifted programmer and a Ferranti engineer assigned to each project. Within a year, novel engineering problems had been largely solved. Instead there were now novel software problems. The Ferranti Argus was something of a standard for ICI. They bought six Argus 100 computers in December 1964, neatly illustrating the way in which an experimental prototype makes the transition to commercial product.77 One of these Argus 100s was installed at a new chemical plant at Widnes, the No. 2 Paraquat plant at the Pilkington-Sullivan Works of ICI Mond Division Works — ‘Pilks’ as it was known locally. This was perhaps the first time in the world a chemical plant had been designed and built for direct digital control from the outset. The low temperature sodium process to make Paraquat was both secret and dangerous and had been developed using a pilot plant, the ‘Gaskell-Marsh’ pilot plant (with a nominal capacity equivalent to 23 tons/year paraquat ion) built at a cost of £100,000 which commenced operation in March 1961.78 The Argus 100 was to be applied to a new, commercial scale No. 2 Paraquat plant at Pilkington-Sullivan Works at Widnes to produce 3500 tonnes/year of paraquat ion at a capital cost of £3.5 million.79 This illustrates three things about ICI as a company: Their confidence with the science of a new processes, having made inferences on the basis of a pilot plant. They were willing to take risks introducing a novel route and using direct digital control by computer on a dangerous process. Finally, their sheer professionalism in terms of BLOODHOUND ON MY TRAIL 19 chemical engineering. So ICI’s role as a lead user of Ferranti Argus computers was just one part of a corporate culture of innovation at the chemical company. Cooperation between ICI and Ferranti was much more limited at the No. 2 Paraquat plant, no doubt for reasons of commercial secrecy. Ferranti were given a list of items to be controlled, but their purpose was blacked out to preserve confidentiality. Eric Johnson from ICI was their link. But otherwise there was little or no interaction with ICI personnel. Ferranti received £123,000 for the hardware and software while ICI spent a further £457,000 on instrumentation and research costs for the plant.80 The Ferranti team for Widnes was composed of two people, Peter Sagar the engineer and Scilla Bretscher the programmer and former maths teacher, one of a team of women programmers who joined Ferranti Automation in Wythenshawe in September 1963.81 Scilla Bretscher had previous vacation experience as a student at Harwell in Oxfordshire programming in autocode for a Ferranti Mercury computer. Programming at Widnes involved frequent trips to the uncompleted plant to test and install programmes in spartan surroundings. This reflects the culture of the two companies: Ferranti left their largely self-taught programmers to ‘sink or swim’ with little or no supervision. ICI had confidence in their commissioning engineers. Internal ICI reports show computer control at Widnes was a great success and ICI soon issued an invitation to tender for an expansion of computer hardware on the plant in May 1969.82

A presumptive anomaly: adoption of Argus for the Bloodhound Mark 2 Launch Control Post It is widely assumed Argus was intended for use on the Bloodhound Mark 2 Guided Missile.83 The hearing aid transistor computer was developed for command guidance of Blue Envoy. But the idea that Argus was developed for Bloodhound seems an interpolation backwards from its eventual use on the Launch Control Post for Bloodhound Mark 2. The turn of events that actually led to use of Argus on Blood- hound is both more complex and far more interesting. Use of digital computers for military purposes faced strong opposition and military deployment owes more to friendship, chance and technical genius rather than any high level decision to use the technology for defence purposes. Use of digital control on the Bloodhound Mark 2 Launch Control Post was due to a problem that could not be overcome using conventional analogue computing. As Derek Whitehead said ‘I did it because I didn’t think analogue would do what I wanted it to do’.84 Adoption of digital computing is a clear example of a ‘presumptive anomaly’, where assumptions of science suggest the conventional solution will fail, so a new technical solution is called for.85 A closely related idea is Hughes’s notion of a ‘reverse salient’ where concerted action in the form of invention and improvisation is needed to maintain a path of development.86 English Electric persisted with analogue and tried to reduce similar errors on a rival missile, Red Shoes, by scaling down the analogue calculation to reduce the absolute error and then scaling up the answer. But Derek Whitehead’s solution using a digital computer was one that worked. The guidance system of Bloodhound Mark 2 used a continuous wave radar system which illuminated the airborne target. A radar dish inside the head of the missile 20 JONATHAN AYLEN picked up the doppler return reflected back from the illuminated target. Continuous wave radar had compelling advantages, including the ability to discriminate between targets at the same range and could detect moving targets amid the clutter of stationary objects which had very different frequency returns. It was less susceptible to jamming. Continuous wave radar is exacting, since it requires a very stable signal so that the missile can pick up the slight doppler shifts returning from the target.87

The Bloodhound Mark 2 Launch Control Post Missile engagement, set up and launch was the task of the Launch Control Post (Figure 5). The Control Post received incoming data such as bearing, elevation and distance of hostile aircraft from the Early Warning Radar via an operations room. The engagement controller in the Launch Control Post would track the assigned target using Target Illuminating Radar (Figure 6). The missile launcher would be moved to face the target and the missile receiver dish on Bloodhound would be tuned on where to expect the target after launch. Once the incoming reflections from the target were sufficiently strong, the computer would indicate a ‘free to fire’ message and the controller was free to launch. Technical development at Wythenshawe followed the standard pattern of aerospace engineering. The overall missile was broken down into packages for individual design teams to develop. In turn, these packages were devolved into sub-assemblies. The Launch Control Post for Bloodhound was one such sub-assembly. Design of the

figure 5 Bloodhound Mark 2 was an archetypal cold war weapon. The air portable Launch Control Post is shown to the left of the missile. BLOODHOUND ON MY TRAIL 21

figure 6 Bloodhound Mark 2 relied upon the doppler return from continuous wave radar reflected off an incoming target aircraft (copyright English Heritage).

Launch Control Post for the Mark 2 Bloodhound was undertaken by a Development team led by Frank Fensome who had early experience in radar receivers at Ferranti Edinburgh. Frank Fensome joined the receiver section of the guided weapons department at its inception in Moston in 1949. He went on to lead the ground equipment section in one role or another from 1954. He conceived the overall design of both the original Red Duster launch control and the uprated mobile launch control post for Bloodhound Mark 2, ‘Frank Fensome conceived the whole thing [. . .] from scratch. He then parcelled it out to three or four of us to develop our bits of it’.88 The team for the Mark 2 Launch Control Post was a very small community of practice made up of, at most, five people. Keith Barker was given the related task of developing the simulator.89 This community transcended the boundaries of task groups as Derek Whitehead was really part of a group working on servo control for the missile itself rather than the Launch Control Post.

Resistance to digital in the defence sector Derek Whitehead was initially assigned a different task — making an analogue computer to solve fuel/range computations.

I did a design and made a model, but because there were a lot of sites it seemed to me that this was going to be a hell of a job to build all these things and very expensive. It seemed to me to be the wrong way to do it.90 22 JONATHAN AYLEN

At that time, Derek Whitehead and John Waterworth went on a course to learn how to programme a digital computer at 21 Portland Place, the Ferranti Computers’ office in London.

Neither of us had done any programming before. In two weeks from a cold start we wrote a programme — I must admit we put some bloody hours in — we wrote a programme which took into account a squadron of Russian aircraft approaching from the east with missile sites in various places, predicting where the impact point was, building in the fuel/range coverage pattern which varies as a function of height, and when the predicted impact point came within the fuel/range coverage it gave a ‘free to fire’. Much to our surprise and pleasure, and everybody else’s surprise [. . .] it did work and it was startlingly good.91 So Derek Whitehead proposed the national Orange Yeoman surveillance radar system could also act as a digital computer-based centre for guided missile control for the UK. Derek took the idea of a system wide digital solution to the person responsible for military radars in England, Norman Alder at RRE.

I was rather surprised I was able to get an interview with him because he was roughly equivalent to God [. . .] And I took all this stuff in and I must have spent all of five minutes with Norman Alder because he listened to what I had to say and then he said ‘There is no place for digital computers in military systems and there never will be’.92 Whatever the official position, the idea of direct digital control of missile systems was out of the box.

‘Maurice’s Gizmo’ and the need for accurate arithmetic Derek Whitehead also faced the problem of directing the radar dish on the new Bloodhound Mark 2 Guided Missile which was to use continuous wave radar for guidance. The radar dish in the forebody of the missile was locked for launch (the missile was subject to some 30 g on take off) but had to point in the right direction as soon as it was unlocked. Remember there are perhaps only 30 seconds between unlock and hitting the target.93 When the dish was unlocked in flight, two features were crucial: that the dish pointed in the right direction to find the target and the doppler gate was set to the right frequency. Recall that continuous wave radar meant the dish had to search for velocity, as well as the angle and range required for the first Bloodhound. Swift and accurate calculation was imperative to aim the dish after launch. The radar dish did not have the luxury of time to search for the incoming radar returns. The radar on the missile is looking for a reflection. The aerial on the missile spins and forms a cone in space. The beam width at X band, that is to say 3 cm or 10,000 MHz, is about 1°. This gives a good signal when pointed in the right direction. But this is also a very narrow arc of sky. This incoming signal from the on-board aerial is crucial because it resolves to give a steering signal for the missile itself, both to track the target and to calculate a point of interception by working out the rate of change of the sight line angle, ‘the sight line spin’. (The missile does not aim directly at the target, but at an interception point. The missile had proportional navigation which always steered towards a collision point.) BLOODHOUND ON MY TRAIL 23

So the dish had to be set up to anticipate both the direction of the target and set the doppler gate. When the missile is on the ground, the doppler shift in the frequency of the signals reflected back from the illuminated target aircraft is known. But, once launched the missile will be travelling above Mach 2. So, when the dish unlocks, you have to anticipate both the speed of the target and the speed of the missile. You have to anticipate the doppler frequency of signals returned from the target at these relative speeds.

We can calculate and pre-set [. . .] where the doppler gates will be at the end of the boost phase. That’s easy. What’s bloody difficult is pointing the dish in the right direction in space.94 This involves some fast and accurate trigonometry, to estimate the combined effect which could not be accomplished by analogue computation:

So you got about four or five trigonometrical stages to get what the azimuth angle was and what the up-down angle was. And with the errors you could get in analogue terms at each of these stages I reckoned that there was no way that I could produce an analogue computer that would go out into the services field that would be accurate enough. I couldn’t, I knew I couldn’t, I knew I was beaten. Now, with the wisdom of hindsight which is granted to us all, that was probably the most important and valuable and worthwhile decision that I have ever made in my life, knowing that I couldn’t do it.95 Derek Whitehead was a friend of Maurice Gribble. He followed what Maurice had been doing on digital control.

Maurice Gribble was developing a gizmo. He developed for a demonstration for the Duke of Edinburgh a scheme whereby you could effectively have a thing in space and something there and something here and you could turn one of the knobs and it did some trigonometry and it would track the thing that was moving, digitally. Wow!96 So Derek Whitehead and Peter Smith proposed to Bruce Calveley and then to Denis Best that the ‘process control transistor computer’ should be used for positioning the radar dish after launch on the Bloodhound Mark 2.97 It is not clear when the Argus was adopted for the Launch Control Post, but surviving minutes suggest it was a possibility before the autumn of 1959.98 Peter Smith and Stuart Thomas wrote the software. The Launch Control Post was also designed to be compact so that it could be air transportable. The size of the computer meant the Launch Control Post was a cramped space for the military operators. There was much relief when the Posts were later upgraded with more compact digital equipment. Once it had been decided to fit a digital computer into the Launch Control Post, other tasks were added. At some stage the firing sequence came to be initiated from the computer. There was a further problem of crossing targets. When the target aircraft is almost at right angles, the doppler drops towards zero and is undetectable as you cannot measure low frequency returns.99 The target disappears. So you need to pick up the target again once it has passed that point. The ground illuminating radar had a crude analogue servo for predicting a crossing target of this sort. But the Argus computer in the Launch Control Post was much better at predicting through this zero doppler point. The computer commanded scans if the zero doppler point was about 24 JONATHAN AYLEN to be reached in order to look for a reappearance of the signal. The longer the time without a signal, the bigger the scan. There was feedback from the computer to the illuminating radar. So using Argus on the Launch Control Post for Bloodhound Mark 2 not only solved a problem of accurate arithmetic but also left computation capacity for additional tasks.

Subsequent technical developments and learning to market Argus Ferranti’s Automation Division survived sale of the firm’s mainframe computer business at Gorton to International Computers and Tabulators (ICT) in September 1963.100 The Ferranti Argus series became a mainstay of process control computing, selling some 1263 units by May 1979 (Table 1), excluding many used on weapons systems. The computer evolved, switching from transistors to microcircuits. Over half the total sales were realized for variants of the final Argus 700 system. The UK was a centre of process control innovation during the 1960s alongside the USA, playing a pioneering role in the automation of rolling mills, chemical plant and power stations.101 By 1965, Ferranti’s domestic rival, Elliott Automation, had 7 per cent of world sales of process control computers and by 1968 had become the world’s second largest supplier with total sales almost three times those of Ferranti (Table 2). So there was a strong potential market for Argus. Growth in market share by Ferranti was the result of a learning process as the Automation Division appreciated it had to supply tailored systems to meet customer needs and become competitive on price. Ferranti was forced to learn how to market and support a product in cost-conscious civilian markets.102 The Argus increased its share of the fast growing global market for process control computers from 1.8% in 1963, to 2.3% by 1965 and up to 4.1% by 1968 (Table 2) — the last year for which we have figures. This was despite new entry in the USA, Japan and Germany. Three breakthroughs enlarged the market for Ferranti — powerful lead users, stronger technical support and software development, and more aggressive pricing. Evidence on the pricing policy for the Ferranti Argus is limited. Eli-Lily used direct digital control to run twenty fermenters at a new factory built by their subsidiary Dista Products, in Speke in Liverpool. Their engineer, John Thorley said:

When establishing the plant we were told it had to be digitally controlled. We went to see an ICI plant at Fleetwood where a Ferranti computer controlled the valve actuators, but this was far too expensive for Dista Products.103 Dista turned to Elliott Automation. The 200 and 100 model Argus computers were over-specified for a civilian use. They used military components and sockets designed to withstand extremes of temperature, shock and rough handling. Even the early 500 series had military specification power supply and gold plated connectors. There was a transition period in the late 1960s when, as one project engineer put it ‘they had to stop thinking in military’.104 Wiring diagrams went through the drawing office at Wythenshawe and emerged as neat blueprints. Yet it was his experience that American civilian process control computer maker GE used simple typed lists. Ferranti wiring was routed in neat trunking and cable runs. At GE in Phoenix wiring was wrapped ‘point to BLOODHOUND ON MY TRAIL 25

TABLE 1 SALES OF ARGUS COMPUTERS TO END 1979 BY TYPE AND CUSTOMER

Industry Type and Model Total

Transistor Microcircuit 200 100 300 400 500 600 700 ETSFG Chemicala 1 5 2 17 2 1 2 30 Oil (including 13 8 14 8 43 offshore) Process 6 5 1 12 Manufacturingb 3 122154747 21825216 Steel Industry 4 1 1 24 1 19 32 6 8 96 Public Utilities 2 1 4 27 21 34 17 23 13 142 Extractivec 121 4 Paper & Print 2 2 1 5 Distribution 3 4 7 Commerce 324247 260 Transport 2 62 4 6 53 9 4 5 145 Communication 16 34 2 9 61 Public Service 3 8 29 21 10 2 73 University & 2 217221311334885 Research Software Houses 4 6 4 14 28 Printing and 215816 Typesetting CS7 Civilian 3 13 8 108 166 83 211 167 85 75 104 1023 Military 60 1 14 1 29 22 8 6 141 (Bloodhound?) ‘Wythenshawe’ 3 14 41 12 8 12 3 93 Service 6 loan 6 Total 63 14 25 123 236 101 241 179 96 75 110 1263

Source: Ferranti Limited, Wythenshawe Division, List of Principal Argus Computer Installations, Excluding Military Sales, May 1979, p. 2 (John Rylands University Library, Manchester). Notes: a — includes BP chemicals which might otherwise be classiﬁ ed under oil. b — includes aluminium foil mills which might be classiﬁ ed as ‘metals’ alongside steel. c — includes cement manufacture. point’ — a ‘rats nest’, but cheap to assemble. The American competition provided more cost-effective design and assembly techniques for civilian applications. Around 1970, Ferranti seem to have revised their pricing policy, winning the project at the British Steel Corporation Llanwern Scheme ‘C’ blast furnace in 1971 on price against competition from IBM, Honeywell and GEC helped by a strong reference list in South Africa.105 Ferranti also benefitted from customer lock-in. Machine specific 26 JONATHAN AYLEN

TABLE 2 MARKET SHARE IN PROCESS CONTROL COMPUTERS, 1968

Total installed July 1968

Manufacturer/Supplier Number sold Market share 1. General Electric (GE/PAC) 326 11% 2. English Electric (Elliott) 307 11% 3. IBM 270 9% 4. Westinghouse (PRODAC) 240 8% 5. Scientific Data Systems (SDS/Sigma) 216 7% 6. Honeywell CCD 195 7% 7. GEC-AEI 145 5% 8 Ferranti 119 4% 9 Systems Engineering Labs 119 4% 10 Digital Equipment (DEC PDP) 117 4% 11 Siemens 115 4% 12 Bunker-Ramoa 91 3% 13 Bailey Meter 74 3% 14 Toshiba (TOSBAC) 48 2% 15 Hitachi (HITAC, HICOM, HIDAC) 45 2% 16 Control Data 43 1% 17 Leeds and Northrup 40 1% 18 Foxboro 38 1% Others 342 Total Installations Worldwide 2,890

Sources: ‘A staff survey: On-line computer scorecard updated’, Control Engineering, 15.7 (1968), 79–90, table 1. Note: a — Some Bunker-Ramo machines may have been reassigned to GE. programming skills reinforced their hold over customers such as the British Steel Corporation and the nuclear power industry. These organizations had large in-house R&D departments with staff who could adapt the hardware to their needs or had Automation Development Departments specifically tasked with computer development.106 Ferranti began to sell complete packages of technology, notably in airline reserva- tions for BOAC. In 1967 they integrated their knowledge of cathode ray tubes for electronic displays with their process control expertise to win the BOADICEA contract for a seat reservation system for the airline BOAC. This alone resulted in a first order for 51 Argus 400B computers delivered during 1968 and 1969. Here too a small, enthusiastic community of practitioners clinched the deal. Sales of four Argus 500 computers to Esso led to the development of a control software language called Consul (Control Subroutine Language) applicable to any process industry. By 1968, Ferranti were selling a cut-down system Consul B which ran in core store on an Argus computer. Core storage was far more reliable than a disc system. Software was BLOODHOUND ON MY TRAIL 27 the missing part of the process control package for Ferranti. Writing bespoke software every time was ‘fraught and time consuming’ for both Ferranti and their customer.107

Conclusions The Ferranti Argus computer was invented by Maurice Gribble working in the automation division of Ferranti at Wythenshawe. He built a digital computer using hearing aid transistors to control the Blue Envoy command guidance surface to air missile. While the missile was cancelled, rapid development of the computer led to a prototype for civilian use which was displayed at a trade exhibition in London in November 1958 under the name the ‘Ferranti process control transistor computer’. At the time, Ferranti’s rivals were persisting with analogue computers. The commercial product was developed with the name Argus 200 both as a computer for civilian process control and as part of a military cold-war missile guidance system. The chemical firm ICI was a crucial civilian lead user. The deployment of Argus for defence purposes was due to a presumptive anomaly — potential shortcomings in use of analogue computation for control of the radar dish on the new Bloodhound Mark 2 missile. Personal friendship led Derek Whitehead to propose accurate digital calculation for the complex trigonometry required, despite opposition from RRE to any digital schemes. The paper tracks the way in which the computer was conceived, how it was developed in military and civilian contexts by small communities of practice, how these groups coalesced, grew and dispersed, and then shrunk to a handful of people as projects became routine (Figure 7). One community of practice emerged around the Bloodhound Mark 2 Launch control post and another around software and instrumentation for an ICI soda ash plant. Ferranti Automation Division learned to market and support a novel product in cost-conscious civilian markets. This account of development at Ferranti draws on interviews with the Argus inventor, engineers, builders, programmers and users. This approach gives a different perspective on the history of engineering and technology. Here we look at individuals working as designers, developers, problem solvers and troubleshooters, in small communities of practice characterized by group cohesion, easy exchange of technical ideas and a certain amount of gossip and scandal. These communities were ideal ways to share experience and tacit knowledge about the problem in hand. This practical focus downplays the role of senior managers and government officials. Rather, the computer emerges among gifted designers and practitioners, helped along by a certain amount of bootlegging, a great deal of engineering verve and enthusiasm, and some gifted women programmers. The focus of this narrative stands in contrast to archive- based treatments of computer history which emphasize the role of senior managers and government R&D institutions. In a wider context, development of Argus is an example of Cold War technology development which defies simple notions such as ‘spin out’. The computer was, at the same time, part of a secret weapons system as well as the focus of a group of civilian users who led adoption of Argus for process control. Official secrecy allowed free exchange of ideas among those with security clearance. As personnel moved from military to civilian applications, wider overlapping communities of practice emerged 28 JONATHAN AYLEN

figure 7 Communities of practice surrounding the development of the Ferranti Argus computer. around software and instrumentation for the Argus. Simultaneously, Argus was both part of the closed world of Cold War weaponry and at the centre of an open network of innovators.

Acknowledgements and data sources This paper draws on interviews with twenty-three people involved in the development path towards Argus and Bloodhound, focussing on their tacit knowledge and links BLOODHOUND ON MY TRAIL 29 between those involved. These respondents cherish many unpublished reports and memoranda. The author is deeply indebted to participants in the development process at Ferranti, ICI and the RAF who gave their time and help. Particular thanks are due to two people: Maurice Gribble who was tireless and a tremendous help across three interviews. Derek Whitehead inspired the paper with the remark ‘you had better come and talk to me’ and gave extensive and colourful advice over four years. In addition, Fred Axon, Keith Barker, Humphrey Bowen, Bruce Calveley, W. F. Cartwright, Peter Collins, Albert Dodd, Audrey and Syd Evans, Alan Foss, Peter Hall of ICL, Roger Houghton, Derek Hughes, Emily Innes, Tom Lunt, Scilla and Dave Senior, Jean Shaoul, John Thorley, Mark Walker and Peter Wolstenholme all responded patiently to questions. The paper has been strengthened by advice and corrections from sometime rocket scientist Dr Frank Fitzgerald; control pioneer Crawford MacKeand; Professors Simon Lavington, Peter Hall and Philippe Laredo; author Chris Gibson and participants at conferences on ‘Technological Innovation and the Cold War’ at the Hagley Museum and Library, Wilmington, Delaware, March 2007; the British Society for the History of Science, Manchester, June 2007; the Department of International Studies, Padua, June 2008; a joint meeting of the Newcomen Society, Computer Conservation Society and IET, Manchester October 2010; and a workshop in 2011 with colleagues Mike Pryce and Etienne Wenger. Warm thanks are due to John Blewett at Catalyst, Widnes; Fleetwood Museum; Fleetwood Library; the Imperial War Museum, Duxford; the Public Record Office, Kew and the Ferranti Pension Fund.

Notes 1 W. Brian Arthur, The Nature of Technol- the Prosaic: The Case for Process-Control ogy: What It Is and How it Evolves Computers’, IEEE Annals of the History of (London: Allen Lane, 2009), p. 126. Computing, 32.3 (2010), pp. 94–96. 2 The key exception is an account of 4 Ernst Braun and Stuart Macdonald, Ferranti Automation’s leading competitor, Revolution in Miniature: The History and Simon Lavington, Moving Targets: Elliott- Impact of Semiconductor Electronics, 2nd Automation and the Dawn of the Com puter edn (Cambridge: Cambridge University Age in Britain, 1947–67 (London: Springer- Press, 1982); E. W. Pugh, Memories that Verlag, 2011). For the decline thesis see Shaped an Industry: Decisions Leading to IBM System/360 (Cambridge, Mass.: MIT John Hendry, Innovating for Failure: Press, 1984). Government Policy and the Early British 5 John F. Wilson, Ferranti: A History, Build- Computer Industry (Cambridge, Mass.: ing a Family Business, 1882–1975 (Lancaste r: MIT Press, 1989). Carnegie Publishing, 2000) and John F. 3 Thomas Stout and Theodore J. Williams, Wilson, Ferranti: A History, Volume 2, ‘Pioneering Work in the Field of Computer From Family Firm to Multinational Process Control’, IEEE Annals of the His- Company, 1975–1987 (Lancaster: Crucible tory of Computing, 17.1 (1995), pp. 6–18; Books, 2007). On broader issues such as the Jonathan Aylen, ‘Megabytes for Metals: dominance of engineers in aerospace and The Development of Computer Applica- dependence of the industry on the State tions in the Iron and Steel Industry’, see: David Edgerton, England and the Ironmaking and Steelmaking, 31.6 (2004), Aeroplane: An Essay on a Militant and pp. 465–78; Jonathan Aylen, ‘Promoting Technological Nation (Macmillan, 1991). 30 JONATHAN AYLEN

6 Stephen Twigge, ‘Ground-based air defence on the tail, like the rival English Electric and ABM systems’, ch. 4 in Cold War Hot Thunderbird. D. J. Farrar, ‘The Blood- Science: Applied Research in Britain’s hound’, Journal of the Royal Aeronautical Defence Laboratories 1945–1990, ed. by Society, 63 (January 1959), 35–50; also see Robert Bud and Philip Gummett (London: [downloaded 23 7 Wayne D. Cocroft and Roger J. C. November 2006]. Thomas, Cold War: Building for Nuclear 13 J. M. Hallett, ‘Biplanes to Bloodhounds’, Confrontation 1946–1989, ed. by P. S. The Ferranti Journal, 17.1 (1959), pp. 2–5. Barnwell (Swindon: English Heritage, 2003), 14 Ferranti Defence Systems, Type 86 Target p. 173. Illuminating Radar for Bloodhound MKII 8 Ferranti, Museum of Science and Industry Missile: The First 25 Years, transcript of Archives, Manchester, File 1996.10/2/3/311, presentation given in 1987 supplied by folder marked ‘Bloodhound’. Correspon- Imperial War Museum, Duxford, Blood- dence relating to South African purchase hound Object File. enquiries, 1969. 15 E. Lloyd Thomas, ‘Analogue Computa- 9 The ‘Red Heathen’ specification, Opera- tion’, British Communication and Electron- tional Requirement OR.1124 that led to ics, 5.5 (1958), pp. 348–58. ‘Red Shoes’ for the Army (Thunderbird 16 T. E. Ivall, Electronic Computers: Principles from English Electric) and ‘Red Duster’ and Applications, 2nd edn (London: Iliffe (Bloodhound Mark 1 from Bristol and and Sons, 1960), ch. 6 ‘applications of Ferranti.) Public Record Office, Kew, analogue computers’; James S. Small, The London. Also see AVIA 13/1285 Ministry Analogue Alternative: The Electronic of Supply, Guided Weapons Department, Analogue Computer in Britain and the ‘Red Duster Development and Proposals USA, 1930–1975 (London: Routledge, 2001), — Missile’, original file number GWS\1238\ ch. 6, ‘The Origins, Commercialisation and 1\1 Part 1, Feb 1958 to Dec 1959; AIR20/ Decline of Electronic Analogue and Hybrid 10554 Air Ministry, ‘Air Defence S.A.G.W. Computing in Britain, 1945–1975’. Future Policy (Development of Bloodhound 17 I. N. Cartmell and R. W. Williams, ‘Guided and MK2)’ Runs from 20 May 1958 to Weapon Simulators’, The Aeronautical 7 March 1963; DEFE7/1338 Ministry of Journal, 72.688 (April 1968), 356–60; inter- Defence, ‘RAF Production Programmes, view with Audrey Evans, former Labora- Guided Weapons, Bloodhound’. Original tory Technician, Simulation Department, File no. T.S.95/03/0311/59 Annex ‘D’; Ferranti, 25 November 2010, Cumbria. T225/2488 Treasury Part A, Guided Weap- 18 A. G. Biggs and A. R. Cawthorne, ‘Blood- ons Research and Development Contracts hound Missile Evaluation’, Journal of the ‘Bloodhound Mk II’ (S.A.G.W.) Runs from Royal Aeronautical Society, 66 (September 4 April 1960 to 14 January 1964. 1962), pp. 571–85. 10 Twigge, p. 91; Cocroft, ch. 7. 19 Professor R. W. Williams speaking about 11 The fuze (sic) was a small pulsed radar the Guided Weapons Group at English proximity device just behind the nose cone. Electric, Luton in Cartmell and Williams, (Interview with Mark Walker, former p. 357. On building an analogue computer Launch Control Post Technician, A Flight, see: Granino A. Korn and Theresa M. 85 Squadron, RAF West Raynham, 10 Korn, Electronic Analog Computers (D-c January 2008, London.) This so-called Analog Computers), 2nd edn (New York: ‘pranger fuze’ was developed by Brian McGraw-Hill, 1956). Jackson of EMI to replace a previous design 20 ‘Electronic Computer Exhibition’, Process based on continuous wave radar which Control and Automation, 5.11 (1958), proved temperamental (interview with pp. 487–96. Derek Whitehead, Cheadle, 15 February 21 Walter G. Vincenti, What Engineers Know 2011). and How They Know It: Analytical Studies 12 The alternative is a Cartesian monoplane from Aeronautical History (Baltimore: with fixed wings and elevators and ailerons The John Hopkins University Press, 1990). BLOODHOUND ON MY TRAIL 31

Engineering does not always work in this 29 Interview with Maurice Gribble, 14 March neat evolutionary way: there is a role for 2008, Gwent. intuition, informed technical judgement, 30 Chris Gibson, ‘Blue Envoy’s Peaceful aesthetic considerations and elegant Legacy’, Prospero, Proceedings from the solutions in deciding what works best. British Rocket Oral History Conferences at 22 Brian Johnson, The Secret War (London: Charterhouse, 2 (Spring 2005), 65–78. Also BBC Publications, 1978), ch. 2, ‘Radar’; S. R. Twigge, ‘The Early Development of Anthony Gandy, ‘The Entry of Established Guided Weapons in the United Kingdom’ Electronics Companies into the Early (PhD thesis, University of Manchester, Computer Industry in the UK and USA’ awarded 1990, Joule Library, University of (PhD thesis, London School of Economics Manchester Thesis TH16433), pp. 497 and and Political Science, awarded 1993). 500. The USAF surface to air Bomarc B 23 Peter Hall, ‘West Gorton and all that’, guided missiles carried nuclear warheads. lecture to the Computer Conservation Part of the thinking was that a high flux of Society, Manchester: Museum of Science neutrons from a low yield atomic explosion and Industry, 21 November 2006; see also would ‘poison’ an incoming hydrogen Simon Lavington, Early British Computers: The Story of Vintage Computers and the warhead through the R-1 effect, thereby People who Built Them (Manchester reducing the yield if the bomb was University Press, 1980), ch. 7, ‘The detonated. Manchester Mark 1’. 31 Cancellation of Blue Envoy is said to have 24 Simon Lavington, The Pegasus Story: prompted the design of Bloodhound Mark A History of a Vintage British Computer 2 in a London taxi outside Ferranti’s head (London: Science Museum, 2000). office by David Farrar, Taffy Higgenson 25 D. G. Owen, Computers and Steel and Don Rowley. See A. R. Adams, (London: British Iron and Steel Research Good Company, The Story of the Guided Association, 1957); Tata Steel UK, Shotton Weapons Division of British Aircraft Cor- Records Centre, department 378, consign- poration (Stevenage, Herts: British Aircraft ment 4, box 8, location 041425, file ‘Com- Corporation, 1976), p. 55. Adams credits puters 1957/59’; Lavington, The Pegasus Norman Searby of Ferranti with lobbying Story, table 4.3. for the acceptance of Bloodhound Mark 2. 26 A point explored during interviews to the 32 Fred P. Adler, ‘Information Theory and point of tedium. Missile Guidance’, ch. 6 in Allen E. Puckett 27 Tom Lunt proudly said they had a ‘first and Simon Ramo (eds), Guided Missile class team for computer control of the Engineering (New York: McGraw-Hill, Bloodhound missile’. Interview with 1959). This edited book is an essential Dr Tom Lunt, former personnel director, source for those wishing to understand Ferranti Ltd, 7 October 2003, Manchester. early surface to air missile design. 28 Interview with Maurice Gribble, designer 33 I owe this recondite truth to interviews with of Argus, 28 January 2011, Gwent. Quotes Derek Whitehead, 12 February 2008, from Maurice Gribble draw on phone inter- Cheadle and 14 May 2008, Manchester. views, two day-long conversations with the inventor in Gwent and e-mail exchanges. Ballistic missiles face a similar problem Radio Hams played a central role in early of varying gravitational pull, see Donald A. computer development, see for instance MacKenzie, Inventing Accuracy: An Christophe Lécuyer, Making Silicon Valley: Historical Sociology of Nuclear Missile Innovation and the Growth of High Tech, Guidance (Cambridge, Mass.: MIT Press 1930–1970 (Cambridge, Mass.: MIT Press, 1990). For solutions, see John M. Wuerth, 2006), ch. 1. Arnold Spielberg, designer of ‘Principles of Missile Navigation’, ch. 7 in the GE412 process control computer, was Puckett and Ramo. another radio ham, see Joseph McBride, 34 M. W. Gribble, ‘The Argus Computer and Steven Spielberg: A Biography (London: Process Control’, Resurrection, 20 (Summer Faber and Faber, 1997), ch. 1. 1998), pp. 20–29. 32 JONATHAN AYLEN

35 Gribble interview, 14 March 2008, Gwent. commercial process was fitted to a revers- On the prevalence of individual or covert ing roughing mill for steel slabs on the research work see Peter Augsdorfer, For- continuous hot strip mill at the Aliquippa bidden Fruit: An Analysis of Bootlegging, works of Jones & Laughlin in the USA. Uncertainty and Learning in Corporate This had three distinctive features: use R&D (Aldershot, Hants.: Avebury, 1996). of digital control; storage of the rolling 36 Braun and Macdonald, ch. 4 ‘The Bell schedule in a memory; and the first use of Laboratories’; Lavington, Early British transistors in steel industry operations. Computers, ch. 9 ‘Transistor Computers’. Westinghouse Electric were responsible 37 The hearing aid transistor computer was for both drive motors and the PRODAC belatedly written up in D. M. Butler, computer control system. See A. W. Smith: ‘Technical Memorandum: Command Guid- ‘Card Programmed Control System Applied ance Two Dimensional Digital Computer’, to Hot Strip Reversing Roughing Mill’, Ferranti, Wythenshawe, Secret, mimeo, 14 Iron and Steel Engineer, 33.10 (1956), October 1958, document CG/26 WY73/14/ pp. 164–65. The market for process control JL (author’s collection). The title ‘Com- computers was developed in the USA by mand Guidance Two Dimensional Digital another guide d weapons developer, Ramo- Computer’ seems to have been an oppor- Wooldridge (variously known as Thomp- tune renaming for budgetary purposes in son-Ramo-Wooldridge and then Bunker 1958. It was referred to as the ‘hearing aid Ramo.) Ramo-Wooldridge announced the transistor computer’ in all interviews. first computer specifically designed for 38 Maurice Gribble was right to bet on the industrial process control — the RW300 improvement potential of junction transis- — in September 1957, fourteen months tors. New transistors were devised to ahead of Ferranti (see ‘Bunker-Ramo handle higher and higher frequencies and Dropping Process Control Computers’, speeds around 500 MHz were to become Control Engineering, 12.6 (June 1965), commonplace by the early 1960s. Braun p. 22.) In January 1958, the Texas Oil and Macdonald, ch. 7. Company announced the purchase of a 39 Email exchange between Maurice Gribble RW 300 to help control its polymerization and the author, 20 January 2011. process at Port Arthur, Texas which went 40 Gribble interview, 14 March 2008, Gwent. on-line in April 1959 (ISA, The Computer A flip-flop is a simple circuit where an Control Pioneers: A History of the Innova- external electric current will cause it to tors and their Work (North Carolina, ‘flip-flop’ from one conducting state to Research Triangle Park: Instrument Society another. In that fashion the circuit can store of America, 1992), ch. 2). But this did not binary information. A series of these achieve complete digital control of the sort circuits can be used to count. Two changes pioneered by ICI Fleetwood. in the state of the first flip-flop induce a 43 The idea of command guidance did not go change in the state of the second flip-flop. away. There is extensive discussion in the In turn, two changes in the second induce a Public Record Office about development of change in the third and so on to the end of a Bloodhound Mark 3 command guidance the circuit. Addition is just counting and missile to be fitted with a nuclear warhead multiplication can be done by successive called ‘Tony’ which was cancelled in Janu- addition. So a flip-flop is a basic building ary 1960. This was intended for defensive block of computation. anti-aircraft use (see DEFE7/1338; AVIA 41 Cyclic progressive codes are discussed by 13/1285 and AIR 20/10554). At one stage Gribble, ‘The Argus Computer’. Also see Bloodhound 3 was also considered as a F. G. Heath and M. W. Gribble, Chain tactical surface-to-surface nuclear missile to Codes and their Electronic Applications, be developed in collaboration with the IEE Monograph no. 392M (July 1960). French (see T225/2488). 42 Maurice Gribble thought it was perhaps 44 Disruption to tea breaks and keeping up ‘the first computer to control something’. appearances were major concerns relating Arguably the first computer to control a to the Royal visit. Dr N. H. Searby requeste d BLOODHOUND ON MY TRAIL 33

‘all sections, whether on the Prince’s route 49 Ferranti Limited, Argus Programming or not, will remain (ostensibly at least) at Manual (February 1962), code number their work [. . .] to give the best possible WY54/E6 published at: Automatic Control impression’, ‘Canteen Arrangements, Division, Shadowmoss Road, Wythen- 22.11.57’, Ferranti Wythenshawe, type- shawe, Manchester, 22 and London script, 15 November 1957. Computer Centre, 68–71 Newman Street, w 45 Gribble, ‘The Argus Computer’; Lunt London 1 (Ferranti Pension Fund Collec- interview, 7 October 2003, Manchester. tion, Ringway, Manchester). 46 Ferranti Limited, Computer Department, 50 Gribble, email exchange, 2011. The Ferranti Process Control Transistor 51 Pugh. Computer (West Gorton, Manchester and 52 M. W. Gribble, ‘“Argus” Pegboard Stores Program in Ferrite Plugs’, Control Portland Place, London: Ferranti Ltd, Engineering, 8.1 (1961), p. 123. Computer Department, November 1958), 53 Gribble, email exchange, 2011. The patents Temporary list DC. 39 (St George’s Library, were Maurice Woolmer Gribble and David University of Sheffield). This temporary Rushton, ‘Improvements relating to Infor- sales pamphlet is the only place where mation Storage Devices’, GB Patent 868,775, a Gorton address appears on an Argus application 17 March 1959 (published publication. 25 May 1961); Maurice Woolmer Gribble 47 Process Control, p. 492. and David Rushton, ‘Information Storage 48 J. Tippett, A. Whitwell and L. H. Fielder, Devices’, US Patent 3,061,821, patented 30 ‘Controlling Megawatts in Steelmaking’, October 1962. Control Engineering, 12.6 (1965), pp. 68– 54 See A. Thompson, ‘Operating Experience 70; Roger Houghton, interview with former with Direct Digital Control’, in Instrument programmer at Steel, Peech and Tozer, Society of America, Digital Computer 18 August 2010, Tabley, Cheshire; Dr Applications to Process Control (New Humphry Bowen, Discussion with former York: Plenum Press, 1965), pp. 55–78 at ICI Engineer, Winnington, 10 January 2008, p. 58. Surrey; Ferranti, List of Principal Argus 55 Derek Whitehead blames another member Computer Installations, Excluding Military of the Launch Control Post team, Dave Sales (Manchester: Ferranti Limited, Shanks for the prank (interview 2 June Wythenshawe Division, catalogue C31, 2010.) The heat of a ship’s cargo hold in the May 1979) (National Archive for the Red Sea led to a switch to metal covers as History of Computing, John Rylands the plastic covers buckled at high tempera- Library, University of Manchester). tures (interview with Peter Wolstenholme, Numbering of Argus models is confusing. 4 February 2011, Prestbury, Cheshire). The initial Argus became the Argus 200 56 Walker interview, 10 January 2008, London. when the Argus 100 was launched. Techni- 57 Gribble, email exchange, 2011. cal documentation distinguishes between 58 Interview with Sydney Evans, former various sub-models of the 100 series: the Project Engineer, Ferranti Automation, Introduction to ch. 4 ‘Core Store’ of a 25 November 2010, Cumbria; Ferranti, Ferranti manual explains helpfully: Wythenshawe, ‘Minutes of Commercial ‘Computers of the Argus 100 series may Progress Meeting held on Monday, 13 have one, two or three core store units. November 1961’, circulated by Dr C. M. Such computers are designated Type 104, Cundall, item 1. 108 and 1012, respectively, since their 59 Bowen discussion, 10 January 2008, Surrey, combined storage for program and data is and Houghton interview, 18 August 2010, approximately 4000 words per core store Tabley. Maurice Gribble suggested the unit’ (p. 4.1); see Ferranti, Argus Mobile Pegasus code was used on Argus to please Process — Control Computer I.C.I. Ltd. Babcock and Wilcox as a lead customer. Argus 104 (Ferranti Limited, Wythenshawe, The original process control transistor com- 1964), file 1996.6/3/18, Museum of Science puter used a code that Maurice invented and Industry, Manchester. (interview, 14 March 2008, Gwent). 34 JONATHAN AYLEN

60 On setbacks, cussing and swearing associ- 68 Edward W. Constant, The Origins of the ated with technical development see Philip Turbojet Revolution (Baltimore: The John Scranton, ‘Technology-Led Innovation: Hopkins University Press, 1980); Etienne The Non-Linearity of US Jet Propulsion Wenger, Communities of Practice: Learn- Development’, History and Technology, ing, Meaning and Identity (Cambridge: 22.4 (2006), pp. 337–67. Cambridge University Press, 1998); A. 61 Wolstenholme interview, 4 February 2011, Delemarle and P. Larédo, ‘Breakthrough Prestbury. Innovation and the Shaping of New 62 The high level of military reliability was Markets: The Role of Communities of appreciated in an industrial setting. Writing Practice’, in Ash Amin and Joanne Roberts about the Ferranti Argus series, Corbett (eds), Community, Economic Creativity (p. 5) speaks of ‘the impeccable reliability and Organization (Oxford University Press, of core RAM based process computer 2008). systems in 1969’ (Brian Corbett, ‘The 69 Interview with Keith Barker, 4 February First Thirty Years of Digital Control at Port 2011, Prestbury, Cheshire. Speaking of Talbot Steelworks’, paper presented to The Fleetwood he added: ‘We were all scared to Institution of Engineering and Technology death it wouldn’t work’. 34th Annual History Weekend Meeting, 70 A. R. Legard, J. S. Hunter and A. Thomp- University of Swansea, 7–9 July 2006.) This son, ‘Imperial Chemical Industries, Alkali is borne out by 99.6% availability for the Division, Notes on a Visit to America’, Argus 200 at Fleetwood. Rivals GE achieved internal ICI Report, 8 February 1957 similar remarkably high levels of reliability (private collection, Surrey). too (Aylen, ‘Megabytes for Metals’, p. 471). 71 Interview with Dr Humphrey Bowen, for- High levels of availability were one feature merly of ICI Alkali Division, Winnington, of the defence inspired dominant design of Cheshire, 22 February 2008, Leatherhead, process control computers in the 1960s. Surrey. 63 Fred Axon, Discussion with former Blood- 72 Telephone interview with Maurice Gribble, hound circuit engineer at Hawker-Siddeley, 23 February 2008. Salford, 24 February 2007. 73 Interview with Derek Hughes, former Elec- 64 Stephen J. Whitfield, The Culture of the trician at Fleetwood Soda Ash Plant of ICI, Cold War, 2nd edn (Baltimore: The John 27 January 2011, Fleetwood; additional Hopkins University Press, 1966), p. 74. information from email correspondence 65 Maurice Gribble, Computer Notes, manu- with Crawford MacKeand, May 2010. script, Ferranti Pension Fund Collection, 74 Evans interview, 25 November 2010, Ringway, Manchester. Cumbria. 66 Evans interview, 25 November 2010, 75 Thompson, ‘Operating Experience’, p. 67. Cumbria; Thompson, ‘Operating Experi- 76 Hughes interview, 27 January 2011, ence’; C. S. Evans and T. H. Gossling, Fleetwood. Digital Control of a Chemical Plant, mimeo 77 Email exchange with Humphrey Bowen, manuscript (Wythenshawe: Ferranti Ltd, January 2011. c. 1965) (Cumbria); Ferranti Limited 78 W. H. Wilson, Paraquat — Notes on Automatic Control Division, Wythenshawe, Future Development Programme Following Argus Project Report No. 1, Digital Com- a Discussion at Fernhurst 9th April 1962, puter Control of an I.C.I. Plant at Fleet- Imperial Chemical Industries Limited, wood, mimeo (c. 1963). The Argus 200 was General Chemicals Division, Development later moved to another soda ash plant Department, Note No. 1939, 19 April 1962, at Winnington, Cheshire and subsequently archives Catalyst, Widnes. acquired by the Science Museum, London 79 J. S. Smith, Imperial Chemical Industries who keep it in store. Limited, Mond Division, Pilkington- 67 Discussion with Crawford MacKeand, Sullivan Works, No. 2 Paraquat Plant, former Project Engineer ICI, Wilmington, The Paraquat Plant Computer from the Delaware, USA, 9 March 2007. Viewpoint of the Plant Manager, internal BLOODHOUND ON MY TRAIL 35

ICI report, 26 February 1968; A. N. A. had fallen behind the broad line of prog- Dicken and J. S. Smith, The Development ress. See Thomas P. Hughes, Networks of of a Process Optimisation Procedure for a Power: Electrification in Western Society, New Paraquat Plant, internal ICI paper, 1880–1930 (Baltimore: John Hopkins Mond Division, Research Department/ University Press, 1983), pp. 79–80. The Mathematics Group, Symposium on point is that analogue had reached its limits On-Line Process Optimisation, Billingham, in terms of accuracy and expense. 23/24 September 1965, both private 87 The ‘Fuze’, which used conventional pulse collection, Leatherhead, Surrey. radar, would only detect the target itself 80 Smith, Imperial Chemical Industries and detonate the charge when it got Limited. into very close range. Walker interview, 81 Interview with Scilla Senior (née Bretscher), 10 January 2008, London. Herefordshire, Wednesday 5 January 2011. 88 Discussion with Derek Whitehead, 1 Febru- She modestly argued Ferranti recruited ary 2011, Heald Green; Correspondence female programmers because they were from Mrs Jean Fensome, 12 January 2011. cheaper, but there was a tradition of 89 Barker interview, 4 February 2011, recruiting female programmers. Personnel Prestbury. policies at Ferranti were hardly enlightened 90 Discussion with Derek Whitehead, 15 Feb- but there was early implementation of ruary 2008, Stockport. On the complexity equal pay (interviews with Jean Shaoul, of range equations see Allen E. Puckett, Manchester, 12 February 2008 and Emily ‘Aerodynamics of guided missiles’ in Innes, former personnel manager, Ferranti Puckett and Ramo, ch. 2. 91 Whitehead discussion, 15 February 2008, Automation, Manchester, 19 February Stockport. Also interview with Dave 2008). The No. 2 Paraquat plant at Widnes Senior, former project engineer Ferranti was not the only example where women Automation, Herefordshire, 5 January 2011 programmers assumed awesome responsi- who attended the same course. bility for development. The ICI Nylon plant 92 Whitehead discussion, 15 February 2008, at Wilton was programmed by Shirley Stockport — edited to spare Derek White- Mitchell working with Roger Wilde as head’s blushes. Derek Whitehead had not engineer. heard of the American SAGE air defence 82 Smith, Imperial Chemical Industries system being developed along similar lines Limited. The upgraded Argus survives in at the time. If Norman Alder was aware of the Catalyst Museum store. what was going on with allies in the USA, 83 For instance, in Wilson, Ferranti: A he clearly did not approve! See Paul N. History, Building a Family Business, p. 429, Edwards, The Closed World: Computers he claims ‘This move was one of the era’s and the Politics of Discourse in Cold War most celebrated adaptations of a military America (London: MIT Press, 1996); Kent technology to civilian use, because in G. Redmond and Thomas M. Smith, From converting the computer employed in the Whirlwind to Mitre: The R&D Story of the Bloodhound launch control post into a SAGE Air Defense Computer (Cambridge: machine capable of controlling real-time MIT Press, 2000). processes [Ferranti] established the basis of 93 The Mark 2 variant of Bloodhound was a business which expanded rapidly over the designed to also work against low-level following twenty years’ which simplifies targets and at ranges as short as 6.9 miles, the complex route by which a computer so speed of computation was a key issue designed for Blue Envoy found its way into (Gribble interview, 28 January 2011, civilian use and thence into Bloodhound Gwent). Mark 2. 94 Whitehead discussion, 15 February 2008, 84 Discussion with Derek Whitehead, 14 May Stockport. For a discussion of these princi- 2008, University of Manchester. ples with crucial details omitted — presum- 85 Constant, pp. 15–16. ably for security reasons — see Renne S. 86 Hughes notion of a ‘reverse salient’ refers Julian, ‘Radar’, in Puckett and Ramo, to an area where the advance of technology ch. 13. 36 JONATHAN AYLEN

95 Ibid. (Whitehead). 103 Discussion with Eur Ing John F. Thorley, 96 Ibid. (Whitehead). 6 October 2004, Old Trafford, Manchester 97 Interview with Bruce Calveley, former and follow-up correspondence. It is apocry- project manager Bloodhound ground instal- phal that the Elliott watchword when lations, Handforth, Cheshire, 11 August selling was, in topical allusion to the 2010. Goon Show, the acronym Eccles — ‘Elliott 98 Ferranti, Wythenshawe, Minutes of Ground computers cost less’. Equipment Progress Meeting Held on 104 Interview with Ted Higgins, former Project Monday 21st September, 1959, circulated by Engineer Ferranti Automation Systems Mr F. Fensome; however, Peter Wolsten- Division, Wythenshawe, Manchester, 5 holme in his interview (4 February 2011, February 2007. Ferranti were not unique Prestbury) recalls an early presentation to among defence contractors in facing these staff in the cinema at Wythenshawe which difficulties. Similar issues were raised about suggested the radar dish of Bloodhound 2 US chipmakers which had been successful would be controlled by ‘a huge mass of in military market then faced difficulties (analogue) synchro resolvers’. winning commercial orders, where techni- 99 Interview with Dr Bill Penley, former head cal requirements were more diverse and of UK Military Electronics Development, changed more quickly. Non-technical Dorset, 25 March 2008. factors such as price and delivery schedules 100 G. Tweedale, ‘A Manchester Computer were far more important to civilian custom- Pioneer: Ferranti in Retrospect’, IEEE ers than defence contractors, see John A. Annals of the History of Computing, 15.3 Alic, Lewis M. Branscomb, Harvey Brooks, (1993), 37–43; Martin Campbell-Kelly, ICL: Ashton B. Carter and Gerald L. Epstein, A Business and Technical History (Oxford: Beyond Spinoff: Military and Commercial Clarendon Press, 1989). Technologies in a Changing World (Boston: 101 Aylen. Harvard Business School Press, 1992), 102 Lunt in his interview (7 October 2003, p. 259. Manchester) said ‘Marketing and sales were 105 Corbett. not in their make-up’. Similarly, Keith 106 Email interchange with David Robinson Barker joined Honeywell from Ferranti re automation at Richard Thomas and Automation and felt the American firm was Baldwins, in South Wales in the early 1960s, 1½ years behind Ferranti technically, but 18/19 March 2006. they made money. 107 Corbett.

Notes on contributor Jonathan Aylen is Senior Lecturer at Manchester Institute of Innovation Research, Manchester Business School, University of Manchester, Manchester m13 9pl, UK. He won a Partnership Trust Award for innovative teaching in economics and was joint winner of the Williams Prize of the Institute of Materials, Minerals and Mining in 2007. Recent publications focus on history of technology, econometrics of seasonality, forecasting wildfires and on innovation management. Correspondence to: Jonathan Aylen. Email: [email protected] 4. Open versus closed innovation - development of the wide strip mill for steel

Jonathan Aylen, “Open versus closed innovation: development of the wide strip mill for steel in the USA during the 1920s”, R&D Management, vol.40, no.1, January 2010, pp.67-80

Open versus closed innovation: development of the wide strip mill for steel in the United States during the 1920s

Jonathan Aylen

Manchester Institute of Innovation Research, University of Manchester, MBS Harold Hankins Building, Manchester M13 9PL, UK. [email protected]

A paired comparison is made between rival attempts to develop the first continuous rolling mill for wide strip in the United States during the 1920s. One firm was secretive, and the other relied on collaboration. Development of the wide strip mill is a natural experiment comparing closed and open innovation as two firms were competing for the same target using different institutional arrangements for their R&D. Wide strip-rolling technology was developed by rival teams in the United States during the mid-1920s. The less successful team at Armco, Ashland, Ky was closed to outside influences. Breakthroughs came from Columbia Steel at Butler, PA, which pursued an open pattern of cooperation with equipment suppliers. Columbia Steel’s collaboration with machinery suppliers, use of independent advice on bearing technology and willingness to learn from precursors in copper rolling enabled them to build a successful wide strip mill complex, commissioned in 1926. Butler established the dominant design for the next 80 years. The leading equipment supplier at Butler, the United Engineering and Foundry Co., led global sales of the technology for four decades. It is not clear how far this example of successful open innovation in the US inter-war economy is typical. Historical studies of the management of R&D focus on formal, science-based research in large corporate labs rather than engineering development.

trip through the plant causes to be (Howells et al., 2003), or networks of innovation ‘‘A raised the question – ‘how was it (Powell et al., 1996), captured by the idea of possible to assemble such rolling mill equip- ‘Open Innovation’ (Chesbrough, 2003). Develop- ment, so drastically different from that of any ment of the wide strip mill for steel in the United previous installation and adapted to the rolling States during the 1920s shows that open ap- of high-finished flat or coiled steel without proaches to innovation are not a new idea. Ches- encountering no difficult operating problems brough and Crowther (2006) suggest that Open at the outset?’ ’’Comment on Columbia Steel, Innovation is a useful paradigm for innovation Butler written in the vernacular by John D. beyond high technology and is appropriate for Knox (1927c, pp. 1400 and 1433). traditional and mature industries. This paper suggests that it is also a long-established approach. This paper reports a paired comparison of rival 1. Introduction attempts to develop the first continuous rolling mill for a wide strip in the United States during the Modern approaches to innovation management 1920s. One firm was secretive, while the other relied stress the importance of distributed innovation on collaboration. Development of the wide strip

R&D Management 40, 1, 2010. r 2009 The Authors. Journal compilation r 2009 Blackwell Publishing Ltd, 67 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main St, Malden, MA, 02148, USA Jonathan Aylen mill was a natural experiment on the effectiveness rolled through all the finishing stands at the same of open innovation as two firms were competing time. A central difficulty is that steel elongates as for the same target using different institutional it is rolled thinner; thus, successive stands have to arrangements for their R&D. The example also roll progressively faster and the speed of stands poses a subsidiary question: was open innovation has to be synchronised to prevent either breaks widespread during the early 20th century? under tension or accumulation of strip between stands as ‘cobbles’. (Some mills are ‘semi-continuous’ because they are configured with reversing 2. What is a continuous wide strip mill and roughing stands to roll down the initial steel slab. why was it so important? But they follow the same principle of unbroken flow from the reheat furnace to the coiler through The continuous wide strip mill for steel was a a continuous finishing train.) revolutionary innovation in 20th-century manufac- The ultimate prize was very long, wide coils of turing that enabled the mass production of cars, hot rolled steel of uniform quality. Coils meant cans and consumer durables. In particular, the much heavier piece weights bringing economies of unavailability of wide steel sheets was a key bottle- scale. Manual handling was eliminated. Coils neck for the early US auto industry. ‘It is the large were also a key logistics innovation because they volume buyer, such as the automobile builder, who are fed continuously into subsequent processing has given encouragement to the new pioneering’, stages such as cold rolling and manufacturing the trade journal Iron Age asserted in its editorial steps downstream such as car-body presses or on ‘The revolution in sheet rolling’ (1927a). canning lines. Hence, system-wide gains were Poor quality, limited dimensions and restricted realised by the switch to steel sheet in coils. supply of steel sheet constrained early car body design, raised costs and inhibited output (Fan- ning, 1952). Quality and availability of wide steel 3. Open versus closed innovation sheet was a ‘reverse salient’, a point where technology had fallen behind in the broad develop- Continuous rolling of wide steel strip was pursued ment of vehicles (Hughes, 1983). Rolling has been independently by rival teams at the American used for the plastic deformation of iron and steel Rolling Mill Company (Armco), Ashland, Ky, since early experiments by Henry Cort from 1780 and at Columbia Steel, Butler, PA. onwards (Beynon, 1956). Rolling was largely a Armco relied exclusively on its own organisa- batch process with individual pieces of steel rolled tional resources as America’s leading sheet pro- in one mill stand at a time. Steel needed to be ducer. Armco’s development of the continuous ‘squeezed’ down while hot from a cast ingot sheet mill was the outcome of a systematic re- through repeated steps of rolling on a mill stand search and development programme conducted in and re-heating as soon as the material cooled. The great secrecy. It proved a false start. hot material was largely manipulated by hand. In contrast, the Columbia Steel mill at Butler, Eventually, the thinnest sheets could be rolled PA, relied heavily on reciprocal cooperation with cold, although they required subsequent anneal- machinery manufacturers and bearing suppliers ing to remove ‘strain ageing’. for its successful development. As a result of Steel sheet rolled on hand mills was not suffi- collaboration, a technical community of practice ciently wide, uniform or ductile and exhibited developed around the idea of continuous rolling poor surface quality. The output of hand mills on four-high mill stands using roller bearings and was individual pieces of thin steel of limited electric motor speed control. The exchange of dimensions and varied metallurgical characteris- know-how seems to have developed on the basis tics. Hand mills had to contend with labour of personal relationships and trust. shortages – hand rolling was hot, dirty and labour In evolutionary terms, these two competing intensive, even by the standards of the time. developments represent the generation of variety The key barrier for strip was rolling large towards solving a pressing technical problem volumes of wide, flat steel continuously, rather (Nelson and Winter, 1982). Rolling mill builders than as individual sheets. Continuous rolling and engineering teams in US steel companies soon means hot steel flows through a sequence of selected the Columbia Steel approach developed stands, without interruption and without any through open innovation. Paired comparisons intervening re-heating or manual handling. On a between success and failure have a long tradition continuous wide strip mill, one piece of steel is in innovation studies (Freeman, 1973.) Here, we

68 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd Open versus closed innovation are testing a hypothesis, rather than seeking to first six stands and then ‘packed’ two pieces thick generalise on the basis of case studies (Eisenhardt, for rolling in the last two stands. High scrap loss, 1989; Yin, 2003). While it is tempting to general- poor yield of prime material and roll breakage ise, a single, paired comparison is not a sufficient caused the mill to be abandoned. A similar mill basis for a generalisation that the open approach was then built by the same company at their is superior – although it might well have refuted Mercer Works in Pennsylvania, commissioned the hypothesis. Innovation is far more messy and in November 1905, to roll thicker gauges, rolling contingent. We judge that the Columbia Steel packs of three sheets in the final three stands. This team of Townsend and Naugle were better experiment failed for the same reasons and the designers than Tytus and Hook, their rivals at mill was dismantled in 1910. Armco. But Columbia Steel also had more tech- Evidently, the idea of rolling a wide sheet on nical support as a result of their open approach. It successive stands was a problem to be solved, is difficult to establish how far the radical innova- driven by demand from the rapidly growing tion at Columbia was due to open collaboration American car industry. But the new continuous and how far it reflected the flair and inventiveness sheet mills were limited in width to around 20 in. of Townsend and Naugle. For example, Morgan Construction Company The history of the wide strip mill is unre- built a 21-in. multi-stand mill for Sharon Steel searched. Early developments of the wide strip Hoop Co. to continuously roll sheet bars and mill were shrouded in secrecy, or as it was deli- slabs (Iron Trade Review, 1920 and 1929). cately expressed at the time, ‘the companies own- These narrow mills were a step on the road to ing these mills have been extremely reticent in continuous rolling, but a long way short of the regard to operating details’ (Shover, 1928). modern wide strip mill as they did not roll the Armco did not allow any visitors to their plant width required by customers nor have four-high until it had been operating for three and a half stands for accurate gauge or the synchronised years (Crout and Vorhis, 1967, pp. 142–143). Loss motors that are characteristic of continuous of archives relating to the initial development wide strip mills. hinders research. This attitude of secrecy faded as the technology diffused across the US steel industry. Extraordi- 5. A choice of development routes narily detailed technical publications (Ess, 1941) and a sequence of visit reports attest to the Successful development of the continuous wide growing openness of US strip mill operators and strip mill in the United States represented a pride in their achievements (Sheet Trade Board, combination of existing technologies. Badlam 1938; Hot Mill Team, 1959). Their candour and (1927) argued that the wide hot strip mill had hospitality accelerated the diffusion of the hot two distinct origins. The first line of development strip mill worldwide, helping US plantmakers was mechanisation of a sequence of mills rolling export by offering potential overseas customers individual sheets to make a continuous flow-line access to their domestic reference plants. process. This solution was pursued by Armco at Ashland in Kentucky. Essentially, this was an incremental approach to the problem. Armco 4. Wittgenstein’s father and other linked a sequence of batch processes to make precursors individual sheets of steel. The second route to continuous strip rolling The continuous wide strip mill had precursors, was to make existing narrow continuous strip notably at Teplitz in Austria between 1892 and mills wider. This was the successful solution 1907, where Karl Wittgenstein, the philosopher’s adopted at Columbia Steel, Butler, PA, in 1926. father, was heavily involved in research and Butler pursued a continuous process making coils development as mill engineer and owner. In the from the outset. Columbia Steel also developed United States, the American Tinplate Company the associated technology of coil handling at their experimented with a continuous wide strip mill Elyria works in Ohio, which ‘served as a proving between 1902 and 1905 at their Monongahela Works ground and laboratory for the company’s newest (Ess, 1941, pp. 3–4; Hogan, 1971, pp. 846–847). installation at Butler, Pa’ (Knox, 1927c, p. 1433). Designed by C. W. Bray, the mill aimed to roll Making a narrow strip mill wider is not as thin tinplate sheets. The mill was a sequence of obvious as it sounds: there are difficulties of eight stands. A tinplate bar was rolled down in the steering the wider strip through the mill, deflection r 2009 The Authors R&D Management 40, 1, 2010 69 Journal compilation r 2009 Blackwell Publishing Ltd Jonathan Aylen across longer mill rolls leading to gauge variation intermediate ‘rough plate mill’. The finishing train across the strip, handling much heavier piece or ‘sheet mill’ was characterised by a reheat weights, dissipating the heat generated by friction furnace ahead of the first two stands and then on the roll necks that bear the load and coping intermediate furnaces ahead of the three final with the high power required to make a wider stands, again a replication of the hand-rolling strip. practice. The mill rolled individual short pieces of sheet that were generally only in one stand at a time, and worked at relatively slow speeds and 6. Armco at Ashland – rolling sheets in low temperatures. The final sheet passed through sequence no less than five intermediate furnaces in addition to the initial slab reheat furnace. In effect, Ash- Armco pursued the idea of rolling of individual land mechanised the traditional hand-mill tech- sheets in sequence. Armco adopted tandem roll- nology by developing the rolling process into ing of individual hot sheets at its Ashland mill in flow-line production, drawing on earlier develop- Kentucky, which first rolled plate in December ments at Teplitz in Germany and Monongahela 1923 and sheet in January 1924 (Iron Age, 1927c). and Mercer works in the United States. The mill is reputed to have cost US$10 million Armco’s contribution was the roll-pass design (around US$65 million at 2007 prices). The mill (US patent 1,602,468). It is easy to steer narrow was designed and built by Armco amid great strip using side guides. This was not possible with secrecy using bought-in components from estab- the wide strip. Armco developed the notion that lished machinery suppliers that were delivered rolls should vary in cross section to ‘steer’ the to the gate, including mill stands from Mesta strip through the mill finishing train, against the Machine Co. and a blooming mill from United prevailing view that roll surfaces should form a Engineering and Foundry Co. (United). The Ash- true cylinder (Knox, 1927e, p. 1596; Fanning, land site was surrounded by a woven wire fence 1952, pp. 196–200). Tytus ground the early rolls and guarded by security staff. There is no evi- in the mill with a slightly concave profile, so that dence that the equipment suppliers participated in the breakdown strip – or the ‘rough plate’ as it any aspect of the development apart from man- was known – had a slightly convex crown. The ufacturing components from drawings supplied crown was reduced as the strip moved through by Armco. successive stands until the final stand, where the Armco’s development of the continuous sheet sheets emerged almost flat. mill was the outcome of a systematic research and Even so, Armco hedged their bets: the roughing development programme led by John Butler Ty- train at Ashland was not straight. The bar plate tus and Charles R. Hook (Crout and Vorhis, mill, as it was known, had skewed roller tables 1967). Initial trials used sheet mills left idle during between each stand and the alternate individual the first world war when Armco turned to forging stands were slightly offset from one another in a shells (Knox, 1927d, p. 1534). At the time, Armco zig-zag fashion (Knox, 1927f, p. 1658). As a was exporting shell steel to the British govern- result, the breakdown strip was alternately forced ment. The specifications called for cropping of the against one side guide or another as it moved ingots by 25%. These discarded crop ends were successively from stand to stand. Hence, in many readily converted to sheet bars for research pur- respects, Ashland was a false start – a set of poses. They systematically investigated the beha- individual mill stands linked by offset roller viour of the roll gap, roll temperature, roll tables. It was later realised that decreasing the composition and gap setting during hot rolling. concavity of the rolls was not crucial either. Tytus and Hook convinced the board of Armco Rather, the important feature for tracking is the to sanction a commercial-scale pilot plant at symmetry of the rolls around their centre line. Ashland using a team of engineers with no experience of rolling so that prior knowledge would not prejudice their work. 7. Columbia Steel, Butler – the first The layout of Armco’s Ashland sheet mill was continuous wide strip mill unique. It was a three-stage process. Essentially, it aligned the separate process stages of hand roll- Columbia Steel developed the first modern con- ing. Ashland used mainly two high mill stands tinuous wide strip mill rolling coils at Butler, PA, and an intermediate reheating furnace between in 1926 (Ess, 1941, p. 4). This was the real the initial ‘bar plate mill’ (roughing train) and the breakthrough: with Butler, the first continuous

70 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd Open versus closed innovation wide strip mill had arrived. Generation I had industry’ (Knox, 1927a, p. 1272). In many re- begun (Figure 1). spects, Butler overshadowed the earlier Armco Columbia Steel Company at Butler targeted the sheet mill at Ashland, which did not open its automotive market with their 36 in. mill, which doors to journalists until a couple of weeks later. started up on Thanksgiving Day, 26th November The Butler mill combined a number of novel 1926. The mill was developed by Arthur J. ‘Gene’ features crucial to hot strip mill development, Townsend and Harry M. Naugle on the basis of including straight through rolling from a roughing earlier trials at Masillon, OH. The two met in train to a finishing train without reheating; four- Canton, OH, where Harry Naugle was chief engi- high finishing stands; a coiler; and adjustable speed neer of the Berger Manufacturing Company and individual DC drives. The mill had a semi-contin- Gene Townsend was an engineer at their supplier, uous layout, with a reversing two high universal the Stark Rolling Company. With Kenneth Jenson, roughing stand supplied by two continuous reheat they formed the National Pressed Steel Company in furnaces. There was a closely spaced finishing train 1916 to make components for the building industry, built by United, composed of four, four-high but with the ultimate idea of making long wide stands with loopers to control strip tension. The sheets. They sold their idea to Central Steel, the back-up rolls were equipped with roller bearings. company that acquired 90% of the stock. This The mill rolled at high speeds and high tempera- financed the acquisition of Columbia Steel at Elyria tures with the strip in all four finishing stands at 1 and the later purchase of the works of the Forged the same time. There was one 2 2 ton up-coiler. The Steel Wheel Company of Butler, PA (Heald, 1955). wide hot strip mill had arrived. The wide hot strip mill at Butler was originally known as a ‘stripsheet mill’ on account of the long length and width of sheet that it produced. It was 8. Butler – collaboration with suppliers later said to have cost US$2 million (Blast Fur- nace, 1958). Successful development of the wide hot strip mill at The significance of the Butler mill was recog- Columbia Steel, Butler, was helped by collaboration nised at this time. Iron Age (1927b) enthused ‘as a with their plant supplier: United Engineering and means of producing higher finished material in Foundry Co. of Pittsburgh. United had begun to large production, the Columbia development develop four-high stands. These had two crucial spells a new era in sheet steel manufacture’. The advantages when rolling a wider strip. Firstly, the Iron Trade Review used similar superlatives, back-up roll acts as a reinforcement to prevent ‘Never, so far as can be learned, has an installa- deflection of the work rolls so that the strip can tion adapted to the rolling of steel been completed be rolled to a near-uniform thickness across its full which has had so many new features incorporated width. Secondly, back-up rolls transmit rolling in its design. It is a milestone in the rolling forces from the work rolls through to the mill

Figure 1. The continuous wide hot strip mill finishing train of Columbia Steel, Butler, PA, with United Engineering and Foundry Company four-high stands and Messinger roller bearings – Townsend and Naugle appear to be standing on the extreme left of the picture (courtesy of the Hagley Museum and Library, Delaware). r 2009 The Authors R&D Management 40, 1, 2010 71 Journal compilation r 2009 Blackwell Publishing Ltd Jonathan Aylen stands. This meant that the stresses could be borne industrial supremacy’ (Pfeiffer, 1951, p. 12). Rome by roller bearings on the much larger necks of the was the first modern, four-high mill with anti- back-up rolls. Placing the stresses on back-up roll friction bearings (Badlam, 1933, pp. 343–344). bearings solved the problem of transmission of United had earlier developed a 204-in.-wide extremely high loads in a confined space (Buhlman, four-high plate mill at Lukens, completed in 1927). Use of roller bearings solved the problem of 1916 for quite different reasons – to avoid having neck friction and the consequent heavy power use to cast huge work rolls. A surprise side effect was and bearing wear associated with rolling wider that the four-high configuration produced more material on a two-high mill. Less wear meant uniform gauge. United were keen to help Colum- greater accuracy in rolling. bia experiment with rolling wide steel strip on The equipment supplier, United, had recent ex- four-high mills as they already had an enquiry perience at Rome Brass & Copper Company, from Weirton for an even more ambitious wide Rome, NY, with the use of roller bearings on a strip mill, also inspired by the Rome Brass and four-high mill for a copper sheet (Biggert, 1927a, Copper Company mill development (Pfeiffer, pp. 3–4) (Figure 2). The Rome mill was strongly 1951, p. 14). The crucial roller bearings for all supported by F. C. Biggert, the president, general these early mills came from William Messinger of manager and former chief engineer of United, and Philadelphia (then known as the Hydraulic Tool by Lane Johnson, an MIT-educated engineer, for- Company) (Figure 3). Messinger had designed mer Armco employee and chief engineer of United roller bearings running in a bronze cage that were from 1922 (United Effort, 1926, pp. 7, 10). This ideal for carrying heavy loads in rock-crushing wide four-high copper mill was developed by Co- machinery and paper mills. ‘The question of roller lonel R. C. Jenkins in 1925 in the face of internal bearing design was discussed with Mr W. Mes- opposition at Rome. It has been said that Colonel singer of Philadelphia, one of the outstanding Jenkin’s contribution to rolling technology should authorities in this field and having determined ‘rank high among those who helped build American sizes which, upon rather meager data, appeared to

Figure 2. The original four-high wide strip mill stand for Rome Brass designed by United Engineering and Foundry Company of Pittsburgh with roller bearings from William Messinger’s Hydraulic Tool Company.

72 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd Open versus closed innovation

Figure 3. United Engineering and Foundry Company approached William Messinger of the Hydraulic Tool Company in Philadelphia for advice on roller bearings for the back-up rolls to the four-high mill at Rome Brass. This is drawing 2298-1 of a Journal Roller Bearing, marked ‘ROME’, one of a number of drawings of journal and thrust bearings for the Rome Brass project dated 3-25-25. The bearings are composed of an outer forged steel ring and a fixed bronze-bearing race holding steel rollers in position (courtesy of Messinger Bearings, Drummond Road, Philadelphia). meet the requirements, a mill was built. It has 1927b, c). All the stands on the cold mills were also proven satisfactory’ (Biggert, 1927a). of a four-high configuration with roller bearings Messinger’s solution was to design a separate on the work rolls and back-ups, which suggests radial and thrust bearings for the Rome Mill that United shared their ideas with their rival (Figure 3). The relationship with United pros- Bliss, as the firm’s machinery catalogue for 1926 pered and Messinger went on to supply the radial shows no four-high mills at all (Bliss, 1926). and thrust bearings for the wide strip mills for Columbia Steel installed both a continuous pick- steel at Butler and Weirton (Link, 1929) and ler and a continuous annealing line for strip at collaborate with United on experimental designs. Butler.Eachcoilwasjoinedtothenextautomati- The four-high mill with suitable bearings in cally by spot welding on a carriage that moved with heavy cast housings made the stands sufficiently the strip. The strip was held in place by magnetic rigid to allow accurate rolling of wider material to clamps while the welding occurred. These radical lighter gauges within acceptable limits. innovations, and many other handling devices were The Butler wide strip mill of Columbia Steel pioneered at their Elyria works in Ohio, which combined a number of radical innovations that ‘served as a proving ground and laboratory for remain widespread 80 years later. A key technical the company’s newest installation at Butler, PA’ and logistic innovation was the manufacture of (Knox 1927c, p. 1433) (Figure 4). Columbia Steel continuous coils in place of individual sheets. had to develop coil-handling equipment as hitherto Butler was remarkable in terms of novelty as it allcoilshadbeennarrow,andthereforelight also included a continuous pickle line for cleaning enough to be man-handled (Knox 1927a, p. 1400). 1 the hot strip, a four-stand tandem cold rolling The Butler mill made coils up to 2 2 tons. mill brought into operation in December 1926, As with many innovations, the key was not one built by E. W. Bliss and two temper mills (Knox, individual development, but a combination of r 2009 The Authors R&D Management 40, 1, 2010 73 Journal compilation r 2009 Blackwell Publishing Ltd Jonathan Aylen

Figure 4. United Engineering and Foundry Company earlier collaborated with Columbia Steel to develop continuous cold rolling of strip at Elyria, OH from 1923. Although only a set of narrow two high stands, it helped Columbia Steel pioneer continuous cold rolling and coil handling. The loose coils in the foreground are narrow hot-rolled strip awaiting cold rolling (courtesy of the Hagley Museum and Library, Delaware). technologies built into a single plant. The principle tory electrification did not reach full fruition in its of back-up rolls, for example, had been around technical development nor have an impact on since the three-high Lauth plate mill of 1864 productivity growth in manufacturing before the (Hoare and Hedges, 1945, p. 37). Indeed, Ashland early 1920’s . . . This was four decades after the used three-high stands on its finishing train – first central power station opened for business’. probably because they were available second- Rolling mills were a key part of this electrification hand. The distinctive features of Generation I process in the 1920s. In terms of innovation wide strip mills, according to Ess (1941), were theory, the wide strip mill was a ‘complex’ innova- the application of four-high mills and the Ward– tion in so far as it drew upon knowledge of more Leonard control system for the electric motors. than one fundamental technology – both electrical Variable speed drives are crucial to rolling steel and mechanical knowledge was required to make strip. Before this, mill stands were driven by belts it a success. In effect, the wide strip mill was or gears (Badlam, 1933, p. 339). The speed ratio located at the junction of two technologies. between successive stands was fixed. While this was The use of coils converted down-stream finish- suitable for rolling simple products, it does not ing operations from a batch process to a contin- work for strip, as Teplitz showed. The development uous flow basis. Continuous pickling, cold rolling of individual, adjustable speed DC motors on each and annealing became viable. Subsequently, con- stand allowed rolling speeds to be adjusted to suit tinuous tinplate lines, zinc coating and organic any gauge. The well-established Ward–Leonard galvanising lines emerged. Canning lines and car power control system for electric drives at Butler industry process lines began to use coils in place allowed successive stands to be synchronised, mak- of individual sheets. ing continuous rolling possible (Burr, 1927, p. 298), Comparing the cost of the two experiments, a technical problem that challenged paper making Columbia Steel’s mill cost one fifth of the price of machinery too (Menzies, 1926; Goldfarb, 2005). the Ashland project. It used a second-hand build- Electrical progress was crucial for the successful ing and re-used old reheat furnaces. The mill at development of the hot and strip mill. The tech- Columbia Steel was also a more compact, less nical journal Iron and Steel Engineer was founded elaborate and cheaper design. in 1924 and early issues were largely devoted to It is simplistic to tell a narrative of unbridled electrical engineering. David (1990) points out success on the part of Columbia Steel. There were that the transformation of American industry by doubtless problems. As Scranton cautions (2008, electric power technology was long delayed: ‘fac- p. 207) ‘real engineering at the edge is a gritty

74 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd Open versus closed innovation process laden with fixes, errors, cursing, and equivalent to 16 million metric tonnes of capacity painfully-incremental steps towards something (Ess, 1941, p. 5). United implemented the lessons that works, much less works reliably and safely. of Butler at Weirton Steel, where they had a There’s no romance in that, so a more marketable shareholding. Both Ashland and Butler were story has long been routinely fashioned’. made wider and rebuilt in the light of experience. Armco’s second mill at Middletown, OH, was 9. Armco’s competitive response modified with four-high stands from United (Longenecker, 1936). United went on to sell the equivalent of 57 wide continuous hot strip mills Armco’s competitive response to Butler’s superior for steel during its lifetime as an independent, US- development was simple: Armco rapidly took owned builder. This compares with 36 hot strip over the Columbia Steel Company and its patent mills sold by its nearest rival, Mesta. portfolio in July 1927. Armco paid a price ap- A standard or ‘dominant’ design of wide strip proaching US$20 million (Butler Eagle, 1927). mill soon emerged with the construction of the Columbia was largely owned by Mellon interests 76 in. mill at Inland Steel by the Mesta Machine and they secured a shareholding in Armco. Company, commissioned in 1932 (Davis, 1934; The official reason for the takeover was to avoid Badlam, 1939, p. 34). This fully continuous mill patent litigation (Iron Trade Review, 1927). But, of had three pusher reheating furnaces, a scale- course, the takeover gave Armco complete and breaker stand, four roughing stands and six lucrative control of much continuous wide hot finishing stands. Most of these subsequent hot strip-rolling technology. Or, as Armco stated: ‘A strip mills were built with fully continuous rough- very substantial patent structure has been created ing trains as well as continuous finishing trains covering the practical, technical and mechanical to solve the problem of heat loss, although problems involved in the entire development at the Butler design of a semi-continuous mill Ashland, Ky, and Butler, Pa, works. This entire re-emerged post-war and became the basis for conception represents a mechanical and economic Generation IV designs from the 1980s (Aylen, 2001). development of first magnitude which will undoubt- Not all new mills initially rolled coils. The edly have a marked influence on the whole sheet continuous 42 in. mill at Gary for Carnegie-Illi- metal industry of the future’ (Armco, 1928a, p. 6). nois Steel Corporation commissioned in 1927 All of Townsend and Naugle’s patents were followed Ashland and rolled sheets in long assigned to Armco when they were finally granted, lengths, which were handled at the end of the notably the key hot rolling patent 1,736,324 on mill via a huge, 180-ft-diameter locomotive-size ‘Strip-Sheet Manufacture’ filed 24 May 1927 before turntable (Hoare and Hedges, 1945, figure 18; Ess, the takeover of Columbia Steel. The combination of 1941, layout 5). The mill was subsequently retro- Armco’s patent on roll shape and Townsend and fitted with up-coilers (Ess, 1941, table 21, mill 5). Naugle’s patents on plant layout effectively sewed up Traditional mills continued to be built too: United hot strip mill design. Armco also received the Town- was the supplier of 24 electrically driven handmills send and Naugle patent on cold rolling 1,781,123 – for Youngstown Sheet and Tube Company, In- first filed in 23 September 1924 but not granted until diana Harbor Works, which were obsolete once 1930 – developing the idea of continuous cold rolling installed (Hogan, 1971, pp. 857–858). of strip under tension. As a result of the takeover, The technology of continuous wide strip rolling Armco sold worldwide licences for their hot strip diffused extraordinarily rapidly over the next mill technology and charged royalties on all strip decade, encouraged by the demand for autobody rolled on continuous wide strip mills. sheet from US car makers. Apart from the 28 Ironically, it can be argued that Armco embraced wide strip mills built in the United States, six mills open innovation when they bought Columbia, pur- were also built in the USSR, Germany, the Uni- chasing the technology they needed for strip rolling ted Kingdom and Japan before the second world through buying an entrepreneurial competitor and war. Five were direct exports from American their intellectual property and bundling it together plant suppliers while the German mill slavishly with their own know-how on roll-pass design. copied the US design. Marshall Aid brought American strip mill technology to Europe again 10. Diffusion of the innovation after the second world war, and American management methods too (Ranieri, 1998, 2000). No less than 28 wide strip mills were built in the The savings were huge: Fanning (1952, p. 203) United States between 1924 and 1939, then reports that ‘a careful computation based on r 2009 The Authors R&D Management 40, 1, 2010 75 Journal compilation r 2009 Blackwell Publishing Ltd Jonathan Aylen present-day labor and material costs, applied to handling, pickling, annealing and continuous cold old hand mill yields and processing operations rolling were attributable to Townsend and Nau- shows that 20 gage cold rolled would be selling gle. Close personal cooperation brought other today at US$275 a net ton instead of the present rewards: Harry Naugle was elected a director of price of US$104 a net ton for the new and very United in 1931. much superior product’. In contrast, Armco had a long tradition in hand rolling steel sheet and building its own rolling mills. It had acquired new works for 11. Why open innovation? Secrecy versus expansion at Ashland, had its own drawing of- collaboration? fices and a team of engineers (Crout and Vorhis, 1967). Armco also had a track record of success- Why did Armco prefer secrecy and Columbia ful R&D in the manufacture of pure iron and an Steel choose collaboration? There were mutual R&D Laboratory at Middletown, OH (Armco, gains from collaboration: Columbia Steel wanted 1928b). Their know-how on conventional hand access to new mill stand technology while United rolling was in demand: Armco managed part of wanted to develop a new product for a market the sheet plant at Shotton in North Wales, help- with huge potential and needed access to an ing John Summers & Sons supply the car indus- operating mill to develop their know-how. try, in return for royalties (Summers, 1940; Aylen, Columbia had already worked with United 2008). They pursued an in-house R&D pro- Engineering and Foundry Company, Pittsburgh, gramme on deep drawing sheet suitable for auto- on their narrow continuous cold rolling mill at bodies (Griffis et al., 1933). Elyria, OH, commissioned in 16 February 1923. Closed innovation meant that Armco used United had developed the four-high principle at traditional brass roll-neck bearings, despite the Lukens Steel and was working at Rome Brass. friction, heat and the wear (Ess, 1941, part-II, United’s response to Columbia’s proposal to p. 2). Wear on conventional bearings is a parti- develop a continuous wide strip mill was cular problem when trying to roll thin gauges accurately. Friction consumes power (Weckstein, Here was a chance to try out the 4-high 1935). Timken estimated at the time that roller principle in a big way, but only imperfect bearings halved energy consumption (Pruitt and data on which to base designs were available. Yost, 1998) on steel rolling mills. Moreover, brass We decided to make a clean breast of the bearings need intermittent lubrication so they matter and see if Columbia would go along cannot roll continuously under load. Conversa- with us in a large scale experiment. The allure- tion with United would have revealed their plans ments were great and we convinced them for using roller bearings in four-high stands. that the experiment was worth making (Biggert, 1927b, p. 1237). 12. Open innovation: how common in United not only built the hot strip mill, they also history? added monitoring devices for research to measure pressure and friction to improve future designs. The Armco/Columbia race raises the question: There must have been a high degree of coopera- how typical was their experience? It is not clear tion: notice that United’s rival Bliss also used the whether Columbia Steel was an isolated instance four-high concept and roller bearings on the of open innovation as there are few historical continuous wide cold mill at Butler, although studies of R&D management in the United States they turned to a different bearing supplier. and even fewer in the United Kingdom. Columbia was open innovation in so far as the A survey by Edgerton and Horrocks (1994) key features of four-high stands and roller bear- shows that an important feature of inter-war ings would not have been pursued without mutual R&D was that many of the largest R&D perfor- cooperation between steelmaker, plant supplier mers in the United Kingdom were either foreign and bearings manufacturer. All parties saw clear owned or participants in international or national gains from pooling their knowledge. Personalities cartels, patents pools and scientific agreements. played a key role, in terms of informal coopera- The example of Metropolitan-Vickers in Manche- tion, transfer of tacit know-how and individual ster suggests that there was considerable coopera- inventiveness. As the patents attest, the numerous tion between industry and academia, largely as a innovations on layout, manufacture of coils, coil result of the first world war (Cooper, 2007). In

76 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd Open versus closed innovation chemicals, Du Pont and ICI had a patents and exchange of technologies embodied in a major processes agreement that led to extensive sharing chemical plant (also Rooij 2005). of know-how between 1929 and 1948. A distinction can be drawn between ‘open’ There is a problem of sample selection bias: information flow between equipment suppliers historians study what is amenable to study and and users discussed here and collaboration be- the archives of large R&D laboratories are tempt- tween companies who might otherwise compete ing targets. Studies also focus on aspects of directly in the product market. In the United corporate research that are close to science (Is- Kingdom, Sanderson (1978) shows that some 40 rael, 1992, p. 185). It is not clear how far the Sheffield steel companies joined a semi-secret documented experience of Du Pont, GE, Bell, body called ‘Sheffield Steel Makers Ltd’ to share Kodak or Alcoa are typical of all R&D conducted technical know-how and advice during the early in the United States during the 20th century 20th century, followed by a national cooperative (Graham, 1985; Reich, 1985; Dennis, 1987; ‘Iron and Steel Industrial Research Council’ Hounshell and Smith, 1988; Graham and Pruitt, formed in 1929 (Carr and Taplin, 1962, ch. 46), 1990). The case of Alcoa shows a high degree of in turn succeeded by the post-war British Iron research collaboration during the second war, and Steel Research Association (Spenceley and although this tailed off when competitors were Scholes, 2001). A rare study by Divall (2006) finds strengthened by the sale of wartime plants. Alcoa that the biggest inter-war railway company in the collaborated with United Engineering and Foun- United Kingdom relied on a network of external dry Company, reflecting United’s considerable experts as well as its own in-house engineers for pre-war experience in the development of large its industrial research. aluminium rolling mills for Soviet industrialisa- The full history of R&D management remains tion. Before this, in 1938, United boasted about to be written. Studies of past development pro- their mills at Zaporozhe that ‘not even the Alu- jects offer potential lessons for current technology minum Company of America has machinery as management, with the added advantage that the modern’ (Sutton, 1971, pp. 58–60). outcome of past research is known. Wider im- The history of formal R&D has been neglected, pacts of the R&D process on users of new but there are even fewer studies of the important technology, suppliers and rivals can be assessed role of engineering departments. In the context of with hindsight. Study of past research highlights early electrical engineering, Hughes (1983, p. 161) the changing history of formal R&D management observes ‘There are countless examples of critical (Boersma, 2007). By its nature, tradition is over- problems being solved by manufacturers’ engi- looked in a forward-looking discipline. neers’. Fox and Guagnini (1999) point to Ferran- ti’s struggle to commission the Deptford Power Station. Collaboration between producers and 13. Conclusion their equipment suppliers has a long history, dating back at least to Henry Maudslay’s colla- This paper reports a paired comparison of rival boration with Marc Brunel over the pioneering attempts to develop the first continuous rolling Portsmouth block-making factory started up in mill for wide steel strip in the United States 1803 (Gilbert, 1965; Cooper, 1984). during the 1920s. Development of the wide strip No doubt, networks were as crucial to success- mill is a natural experiment comparing closed and ful innovation in inter-war America as they are open innovation: two firms were competing for now – we do not have the evidence. Rosenberg the same target using different institutional ar- (1982) emphasises that machinery and equipment rangements for their R&D. One firm was closed suppliers play a crucial role in ‘learning by using’. to outside influences; the other relied on a net- Feedback from operational experience improves work of support. the design of capital equipment. In a case analo- The less successful team at Armco, Ashland, gous to steel, Freeman’s (1968) study of chemical Ky, was highly secretive. Breakthroughs came plant shows how process plant contractors – a from Columbia Steel at Butler, Pa, which pursued group of specialised design, development and con- an open pattern of cooperation with equipment struction companies – came to dominate the trans- suppliers. Columbia Steel’s collaboration with fer of continuous process technology, even though machinery suppliers, use of independent advice oil and chemical companies remained the major on bearing technology and willingness to learn source of technical innovations. Licence flows and from precursors in copper rolling enabled them to know-how agreements lubricated the complex build a successful wide strip mill complex, com- r 2009 The Authors R&D Management 40, 1, 2010 77 Journal compilation r 2009 Blackwell Publishing Ltd Jonathan Aylen missioned in 1926. The Butler continuous wide Armco. (1928b) The Development and Service of strip mill established the dominant design for ARMCO Ingot Iron Hagley Museum and Library rolling steel strip for the next 80 years. Trade Catalogue A4955. 1928. Middletown, OH: Development of the wide strip mill offers a Armco. sharp contrast between technical success and fail- Aylen, J. (2001) Where did Generation V strip mills ure, but inference about open innovation as the come from? A brief history of the hot strip mill. Steel Times, 229, 7/8, 227–236. best way to manage R&D should be more cau- Aylen, J. (2008) Construction of the Shotton wide strip tious. Arguably, the winning team of Townsend mill. Transactions of the Newcomen Society, 78,1, and Naugle at Butler were also better inventors 57–85. than Tytus and Hook at Ashland. While the race Badlam, S. (1927) Strip, sheet mill practice overlaps to develop wide strip rolling of steel refutes a null after 50 years of development: recent design of hypothesis of no difference between open and continuous mills marks important advance in art of closed innovation, there is sample selection bias. rolling – evolution traced. Iron Trade Review, 81, Development of the wide strip mill is one undo- 1090–1091. cumented case among many in an era of innova- Badlam, S. (1933) Recent developments in the rolling of tion. It is not clear how far this example of sheets and strip. Iron and Steel Engineer, 10, 12, 333– successful open innovation in the US inter-war 361. Badlam, S. (1939) Developments in rolling flat steel economy can be generalised, although it is not products in 1937–38. Iron and Steel Engineer, 16,1, unique as other innovators have collaborated 28–51. with equipment and component suppliers. His- Beynon, R.E. (1956) Roll Design and Mill Layout. tories of R&D management focus on formal, Pittsburgh: Association of Iron and Steel Engineers. science-based research in large corporate labora- Biggert, F.C. Jr. (1927a) Developments in 4-high roll- tories rather than the process of engineering ing mills. Paper read before Iron and Steel Division development. The history of R&D management of American Society of Mechanical Engineers, remains an under-researched area. Youngstown, OH, November 10, 1927. United Ef- fort, 7, 11, 3–7. Biggert, F.C. Jr. (1927b) Power consumption not Acknowledgements affected by working roll diameter. Iron Trade Review, 81, 1237–1239. The paper benefitted from discussions at: The R&D Blast Furnace. (1958) The first continuous mill on Management Conference, Ottawa, June 2008; Auto- which strip was rolled successfully and today’s mod- mobility: A Conference on the 100th anniversary of ern mill that has replaced it, Blast Furnace and Steel the Model T, Hagley Museum and Library, Plant, November, p. 1194. Wilmington, DE, November 2008; and the Fifth Bliss (1926) ‘Bliss’ Rolling Mill Equipment – Cold Roll- ing. Catalogue 48, Section R-1. Brooklyn: E.W.Bliss. European Rolling Conference, London, June 2009. Boersma, K. (2007) Managing between science and Dr Frank Fitzgerald, Professors Richard Nel- industry: an historical analysis of the Philips Re- son and Phil Scranton gave valuable comments, search and Development department’s management. while Mick Steeper, Ewan Hewitt and Eric Earn- Journal of Management History, 13, 2, 122–134. shaw provided technical advice. Mr Roberto Buhlman, F.H. (1927) Factors governing the design of Borsi of Danieli at Buttrio, Udine, provided roller bearings for roll necks. Iron and Steel Engineer, access to the United Archives. Particular thanks 4, 6, 302–307. are due to the Hagley Museum and Library; Burr, W.H. (1927) Electrical developments in the iron Butler Public Library; Masillon Museum; Kurt and steel industry. Iron and Steel Engineer, 4, 6, 297– Garvey and Daniel P. Matthews of Messinger 302. Bearings; Randy Smrek of Butech Bliss; the Butler Eagle Columbia plant in Butler holds valuable patents. Acquisition by American Rolling Mill Com- Walsh Library at Fordham University; and Corus pany avoids litigation: stabilization for industry is Colors Record Centre, Shotton. The author is likely: method of rolling sheet iron developed by two indebted to two kind referees. companies same time. Butler Eagle, 30, 21 July 1927, 1, col. 8. Carr, J.C. and Taplin, W. (1962) A History of the References British Steel Industry. Cambridge, MA: Harvard University Press. Armco. (1928a) Twenty-Seventh Annual Report of The Chesbrough, H. (2003) Open Innovation: the New Im- American Rolling Mill Company for the Fiscal Year perative for Creating and Profiting from Technology. Ended December 31, 1927. Middletown, OH: Armco. Boston, MA: Harvard Business School Press.

78 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd Open versus closed innovation

Chesbrough, H. and Crowther, A.K. (2006) Beyond of US Manufacturing 1880–1930. Industrial and high tech: early adopters of open innovation in other Corporate Change, 14, 5, 745–773. industries. R&D Management, 36, 3, 229–236. Graham, M.B.W. (1985) Corporate research and devel- Cooper, C.C. (1984) The Portsmouth system of opment: the latest transformation. Technology in manufacture. Technology and Culture, 25, 2, 182– Society, 7, 2–3, 179–195. 225. Graham, M.B.W. and Pruitt, B.H. (1990) R&D for Cooper, T. (2007) The early development of scientific Industry: a Century of Technical Innovation at Alcoa. research in industry: the case of Metropolitan-Vick- Cambridge, Cambridge University Press. ers Ltd, 1901–1933. In: Pickstone, J.V. (ed.), The Griffis, R.O., Kenyon, R.L., Burns, R.S. and Hayes, A. History of Science and Technology in the North West. (1933) The aging of mild steel sheets. Year Book Manchester, Manchester Centre for Regional His- of the American Iron and Steel Institute, Vol. 23. New tory, pp. 84–105. York: American Iron and Steel Institute, Crout, G.C. and Vorhis, W.D. (1967) John Butler pp. 142–159. Tytus: inventor of the Continuous Steel Mill. Ohio Heald, E.T. (1955) The Stark County Story, Volume IV, History, 76, 132–145. part 1, Free People at Work in Stark County, Ohio, David, P. (1990) The dynamo and the computer: an 1917–1955. Canton, OH: Stark County Historical historical perspective on the modern productivity Society. paradox. American Economic Review, 80, (Confer- Hoare, W.E. and Hedges, E.S. (1945) Tinplate. Lon- ence issue), 355–361. don: Edward Arnold. Davis, R.W. (1934) 7600 Hot strip mill, Inland Steel Co., Hogan, W.T. (1971) Economic History of the Iron and Indiana Harbor, IN. Iron and Steel Engineer, 11, 12, Steel Industry in the United States, Vol. 3: Parts IV 486–490. and V. Lexington, MA: D.C. Heath & Co. Dennis, M.A. (1987) Accounting for research: new Hot Mill Team. (1959) Newport project visit to USA and histories of corporate laboratories and the social Canada April/May 1959 hot mill team report, RTB, history of American science. Social Studies of Hot Mill Team, Report no. R.120 12 June 1959, at Science, 17, 3, 479–518. Coed Parc, Bridgend, Corus Colors Record Centre, Divall, C. (2006) Technological networks and industrial Shotton, Code 359, consignment 5, box 1, location research in Britain: The London, Midland & Scottish 11946. Railway, 1926–47. Business History, 48, 1, 43–68. Hounshell, D.A. and Smith, J.K. (1988) Science and Edgerton, D. and Horrocks, S.M. (1994) British in- Corporate Strategy: Du Pont R&D, 1902–1980. dustrial research and development before 1945. Eco- Cambridge, Cambridge University Press. nomic History Review, 47, 2, 213–238. Howells, J., James, A. and Malik, K. (2003) The Eisenhardt, K.M. (1989) Building theories from case sourcing of technological knowledge: distributed in- study research. The Academy of Management Review, novation processes and dynamic change. R&D Man- 14, 4, 532–550. agement, 33, 4, 395–409. Ess, T.J. (1941) The Modern Strip Mill: A Recording of Hughes, T.P. (1983) Networks of Power: Electrification the Continuous Wide Strip Mill Installations and in Western Society, 1880–1930. Baltimore: John Practices in the United States. Pittsburgh: Associa- Hopkins University Press. tion of Iron and Steel Engineers. Iron Age. (1927a) Editorial: The revolution in sheet Fanning, F. (1952) Wide strip mills – evolution or rolling. The Iron Age, 119, 20, 1462. revolution? Yearbook of American Iron and Steel Iron Age. (1927b) Continuous rolling of sheets at Institute 1952, pp. 194–221. Butler, Pa: how the Columbia Steel Co. makes Fox, R. and Guagnini, A. (1999) Laboratories, work- stripsheets from slabs without reheating. The Iron shops and sites: concepts and Practices of Research Age, 119, 20, 1435–1439. in Industrial Europe, 1800–1914, the realities of Iron Age. (1927c) Exceeds ton of sheets per minute: industrial research. Historical Studies in the Physical Armco continuous mill at Ashland is epoch-making and Biological Sciences, 29, 2, 251–289. as to productivity and unique in incorporating roll- Freeman, C. (1968) Chemical process plant: innovation ing principles established by special research. The and the world market. National Institute Economic Iron Age, 119, 24, 1731–1737, and 1792. Review, 45, 29–51. Iron Trade Review. (1920) New sheet bar mill now Freeman, C. (1973) A study of success and failure in rolling. Iron Trade Review, 66, 161–164. industrial innovation. In: Williams, B.R. (ed.), Iron Trade Review. (1927) Purchase unites continuous Science and Technology in Economic Growth. Lon- sheet and strip rolling processes. Iron Trade Review, don: Macmillan, pp. 227–245. 81, 201–202. Gilbert, K.R. (1965) The Portsmouth Block-making Iron Trade Review. (1929) New continuous strip mill Machinery: A Pioneering Enterprise in Mass Produc- delivers stock at 2400 feet a minute. Iron Trade tion. London: Science Museum Monograph, HMSO. Review, 85, 262–263. Goldfarb, B. (2005) Diffusion of general-purpose tech- Israel, P. (1992) From Machine Shop to Industrial nologies: understanding patterns in the electrification Laboratory: Telegraphy and the Changing Context

r 2009 The Authors R&D Management 40, 1, 2010 79 Journal compilation r 2009 Blackwell Publishing Ltd Jonathan Aylen

of American Invention, 1830–1920. Baltimore: John Reich, L.S. (1985) The Making of American Industrial Hopkins University Press. Research: Science and Business at GE and Bell, 1876– Knox, J.D. (1927a) Rolling stripsheets on new contin- 1926. Cambridge, Cambridge University Press. uous mill: how latest commodity is processed. Iron Rooij, A. van (2005) Why do firms acquire technology? Trade Review, 80, 1271–1275. The example of DSM’s ammonia plants 1925–1970. Knox, J.D. (1927b) How strip sheets are rolled on Research Policy, 34, 6, 836–851. continuous mill. Iron Trade Review, 80, 1344–1346. Rosenberg, N. (1982) Inside the Black Box: Technology Knox, J.D. (1927c) Continuous annealing promotes and Economics. Cambridge, Cambridge University uniform structure in strip sheets. Iron Trade Review, Press. 80, 1398–1400, and 1433. Sanderson, M. (1978) The professor as industrial con- Knox, J.D. (1927d) Rolls sheet steel direct from ingot sultant: Oliver Arnold and the British steel industry, without reheating: continuous mill at Ashland, Ky, is 1900–1914. Economic History Review, 31, 4, 585–600. based on a rolling principle heretofore undiscovered. Scranton, P. (2008) Technology, science and American Iron Trade Review, 80, 1532–1535. innovation, Momigliano lecture 2005. In: Amatori, Knox, J.D. (1927e) Continuous sheet mill is based on F. and Amendola, M. (eds) Ricerca avanzata e alta proportional roll convexity. Iron Trade Review, 80, divulgazione, Le Momigliano Lectures 1997–2008. 1593–1596. Terni, Italy: ICSIM, pp. 193–220. Knox, J.D. (1927f) Sheets rolled to 16-Gage on three- Shover, B.R. (1928) The rolling mill in 1927. Iron and high backup mill – skew type feed table and side Steel Engineer, 5, 1, 4–6. guides enforce straight line movement to piece in Sheet Trade Board. (1938) Report of Delegation to the transit between stands. Iron Trade Review, 80, 1656– United States of America - 16 March to 11 April, 1659. 1938, Corus Colors Archive, Shotton, Group Secre- Link, E. (1929) Vier und Sechswalzengeru¨ste in tary Code 425, consignment 13, box 2, location 3527. amerikanischen Walzwerken. Stahl und Eisen, 49,2, Spenceley, G.D. and Scholes, P.H. (2001) BISRA - The 37–40. British Iron and Steel Research Association. In: Longenecker, C. (1936) Modern equipment for making Bodsworth, C. (ed.), British Iron and Steel AD1800- ‘Armco’ strip steel. Blast Furnace and Steel Plant, 2000 and Beyond. London: IOM Communications, December, pp. 1061–1066. pp. 221–300. Menzies, W.E. (1926) Regulator System, US Patent Summers, R.F. (1940) The New Mill 1940. Chester: 1,605,051, awarded 2 November 1926, filed 14 June Jonathan Cape for John Summers & Sons. 1923, assigned to Westinghouse Electric & Manufac- Sutton, A.C. (1971) Western Technology and Soviet turing Company, A Corporation of Pennsylvania. Economic Development 1930 to 1945. Stanford, CA: Nelson, R. and Winter, S. (1982) An Evolutionary Hoover Institution Press at Stanford University. Theory of Economic Change. Cambridge, MA: Bel- United Effort. (1926) 1901–1926 Silver Anniversary. knap Press of Harvard University Press. United Effort, 6, 7, July. Pfeiffer, J.E. United for fifty years: golden anniversary Weckstein, S.M. (1935) Precision Mills for Rolling number. United Effort, July 1951. Strip, Bars and rods, Year Book of the American Powell, W.W., Koput, K.W. and Smith-Doerr, L. Iron and Steel Institute 1935. New York: American (1996) Interorganizational collaboration and the Iron and Steel Institute, pp. 80–96. locus of innovation: networks of learning in bio- Yin, R.K. (2003) Case Study Research: Design and technology. Administrative Science Quarterly, 41,1, Methods, 3rd edn. Los Angeles: Sage. 116–145. Pruitt, Bettye and Yost, J.R. (1998) Timken: From Missouri to Mars — A Century of Leadership in Manufacturing. Boston, MA: Harvard Business School Press. Ranieri, R. (1998) Learning from America: the remo- Jonathan Aylen is senior lecturer and director of delling of Italy’s public sector steel industry in the external relations of the Manchester Institute of 1950s and 1960s. In: Kipping, M. and Bjarnar, O. Innovation Research, Manchester Business (eds), The Marshal Plan and the Transfer of US School, University of Manchester, UK. He won Management Models. London: Routledge Studies in a Partnership Trust Award for innovative teach- Business History, pp. 208–228. ing in economics and was joint winner of the Ranieri, R. (2000) Remodelling the Italian steel indus- Williams Prize of the Institute of Materials, try: americanization, modernization and mass production In: Zeitlin, J. and Herrigel, G. (eds). Minerals and Mining in 2007. He is a member Americanization and its limits: Reworking US of the rolling committee of the Institute of Mate- technology and management in post-war Europe and rials, Minerals and Mining. Recent publications Japan. Oxford, Oxford University Press, pp. 236– focus on the history of technology, econometrics 268. of seasonality and forecasting wildfires.

80 R&D Management 40, 1, 2010 r 2009 The Authors Journal compilation r 2009 Blackwell Publishing Ltd 5. Construction of the Shotton wide strip mill

Jonathan Aylen, “Construction of the Shotton wide strip mill", Transactions of the Newcomen Society, vol.78, no.1, February 2008, pp. 57-85

Published by & (c) The Newcomen Society place. the mostdramaticeventin sagawasthedecisiontoadoptnewtechnologyin thefi them breathingspacetodevelop newmanagementmethodsforstripproduction. Perhaps run astripmillforthemselves. Evenherefortunesmiled,aswartimesalescontrols gave vened, soJohnSummersand Sonshadtolearnthemanagementtechniquesrequired to engineering wasBritish. how’ andnobleequipmentwereAmerican.But,inspending terms,overtwo-thirdsofthe civil engineering,buildingsandalltheelectricalequipment totimeandcost.The‘know- clear. Soistheimportanceofasophisticatedlocalinfrastructure intheUKwhichprovided role oftheUSplantsupplier,Mesta,intransferring‘know-how’ aswellequipmentis adopting AmericanwidestripmilltechnologyinEurope, itsoutputandprofi rolling technologyinaBritishcontext.Shottongives a detailedinsightintothecostsof profi a deadlineofimpendingwar.Oncecommissioned,the millproducedaconsistentfl conceal thetruthofaprojectthatwasexecutedrapidlywithmeticulouscostcontrolagainst out andtheAmericancommissioningcrewreturnedtoStatesassoontheycould. across theAtlanticunderwartimeconditions.Themillwascommissionedafterwarbroke crossing fromBaltimoretoLiverpoolin1939andcrucialreplacementshadbebrought and Incidentals’. handwritten quotereads‘Fill,RailChangesandIncidentals’insteadof generous amountsofalcoholwereconsumedwiththelunchprecedingestimate:The and managingdirector,respectively,ofJohnSummersSon.Aslipthepensuggests Pittsburgh, inthepresenceofMrRichardSummersandNevilleRollason,chairman on hotelnotepaperbyLorenzIversen,PresidentoftheMestaMachineCompanyin shown inFigure 1, anestimatefor$10 020 a suiteattheSavoyHotelinStrand,London,onsummerafternoon,17May1937.As 000 (thenjustunder£2million)waswritten Trans. NewcomenSoc. The Shottonwidestripmill ts. be thebestmancoulddevise,butallextravaganceandwasteweretoavoided. The millwastobeacreditthosewhobuiltit,andwereoperateit.It It wasnotastraightforwardtransferofAmericanproduction techniques.Warinter- The focushereisontheprocessofpurchasing,building andcommissioningofUS The Shottonmillwasalsoaprojectofparsimonyandprofi During construction,machinerywaswashedoverboardinstormyweatheronthe Read attheMuseumofScienceandIndustry,Manchesteron26October2004 Construction oftheShottonWide , 78 (2008), 57–85 1 wasaprojectofromanceanddrama.Themillplannedin Strip Mill Jonathan AYLEN by Richard Summers, DOI: 10.1179/175035208X258284 t. Thesebravurastories The NewMill Roll Changes tability. The tability. , 1940 ow of ow rst Published by & (c) The Newcomen Society which rougheddownsmallingotsintolong,lightslabs—or‘sheetbars’forhandrolling. iron. Thefi made atShottonintwocold-chargeopen-hearthfurnaceshopsfedbyscrapandcoldpig single-stand sheetmillswherematerialwasmanipulatedmanually,Figure 2. Steelwas During the1930s,JohnSummersandSonstheirassociatedplantsoperatedsome70 58 nishing millsweresuppliedbyacoupleoftwostandcross-country‘bar’ rolling complexdrawnupattheSavoyHotel i.1 EstimatefortheShottonwidestrip Fig. 1. i.2 Anexampleofalabour-intensive Fig. 2. elsewhere inthepre-warsheet industry. hand millthatoperatedatShotton and SHOTTON WIDESTRIPMILL FIRST MOVES in London. Published by & (c) The Newcomen Society market asshowninFigure 3. RichardSummersrecalled: Vale representedaseverecompetitivethreattothepositionofJohnSummersinstrip action ontwocounts:Firstly,thecontinuouswidestripmillofRichardThomasatEbbw radical innovationbytheircompetitors. Rolling Millsin1946forrollingcopper.Finally,JohnSummersandSonswereforcedinto the ShottonSendzimirmillfromoutsetandunitwassoldsecond-handtoEnfi into coilsforcoldrollingwithsuccess. ogy wasnotascrazyitsounds:RenaultinFranceweldedindividualhotrolledplates which werethenweldedtogethertoformanendlessbeltofstrip.Suchintermediatetechnol- not availableforthe‘Z’mill,soitwasfedwith20feetlengthsrolledonathree-highmill adjacent continuousgalvanisinglineinstalledaroundthesametime.Continuousstripwas tandem millforcoldrolling. sheet requiredbytherapidlygrowingUKcarindustry,helpedinvestmentinatwo-stand other worksinStalybridge.JohnLysaghtproducedperhaps90percentofthefi galvanised material.Afterall,Summerswere—quiteliterallyclogironmakersattheir was 859 000 tons.However,JohnSummersspecialisedinrelativelylow-gradesheetand Thomas &Coand97 000 tonsatJohnLysaght&Co. output was241 000 tonsofsheetayearin1932,comparedwith131 000 fromRichard the outputoftheirnearestrival,RichardThomasandCo.EstimatessuggestSummers’ the UK,accountingforwelloveraquarterofUKsheetproductionandhadalmosttwice and tinplateproducersduringthe1930s.Secondly,ArmcoofAmericahadmanaged Richard Thomashadalreadygainedmarketsharethroughacquisitionofsmallersheet sheets upto40in.wide,commissionedinNovember1936. promise, whichwasinstallationofasmall,GermanbuiltSendzimirmillforcold-rolling Ashland, Kentucky.AsecondvisittotheUSAin1935wasfollowedbyashort-termcom- early as1927whenGeoffreySummersandanotherdirectorvisitedArmco’snewmillat installed inSouthWales,andwecouldnolongerdelay. Towards theendof1936,however,itappearedalmostcertainthatastripmillwastobe Construction oftheEbbwValewidestripmillinSouthWales1936wasaspurto At thebeginningof1930s,JohnSummersandSonswerelargestsheetmakerin The possibilityofinstallingacontinuouswidestripmillatShottonwasexploredas i.3 Comparativeperformance ofthe Fig. 3. Ebbw ValeandSummershotstrip mills. 3 SHOTTON WIDESTRIPMILL 5 Doubtswereexpressedabouttheperformanceof 6 2 TotalUKsheetoutputthatyear 4 Themillalsorolledcoilforthe eld ne 59 Published by & (c) The Newcomen Society 60 the UK,withoutSummers’knowledge. Sons byofferingRichardThomasanexclusivelicenceonwidestriprollingtechnologyin but Summersprevaricated.Finally,ArmcobroketheiragreementwithJohnand construction ofthesmallSendzimir coldmill. they confi Sendzimir continuousgalvanising lineatShotton,onlythesecondinworld.Otherwise, per centmoresteeloutput. Thelargestrecentinvestmentprojectwasinstallation ofa then around260 000 tonsa year. Yettheirnewmillcomplexwouldpotentiallyrequire60 based onmeltingcoldpigironandscrapinopenhearth furnaces.Theirsteeloutputwas 1937; workbeganimmediately,withtestpilingonthesite beginninginSeptember1937. tions. Sobureaucracymovedswiftly.Themillwasannounced in Iron andSteelFederations,isdated25June1937—just sixweeksaftertheSavoynegotia- Advisory Committee’rapidlygavethemillgo-ahead. AcrucialletterfromtheBritish tion fromSchleswig-HolsteinandOvesenwasafellow Dane.TheUK‘ImportDuties Copenhagen, HenrikOvesen,asprojectmanager.Iversen wasofDanish–Germanextrac- same time,IversenencouragedSummerstorecruitan experiencedprojectengineerfrom The dealwaslargelynegotiatedduringIversen’ssubsequent visittoLondon.Atthe • • Summers andSonsboard,bothacceptedunanimously: Pittsburgh for1 sandwiched betweentworailroadlinesandstretchingalongtheMonongahelaRiverin strip mills. President andpart-ownerofMestaMachineCompanyPittsburghvisitedAmerican consult MestaaboutanewrollingmillcomplexforShotton.TheymetLorenzIversen, semi-continuous narrowstripmilltoMachinoimportoftheSovietUnionin1937. for continuousstripmilltechnology,havingwonacontracttosupply32-in-wide 18 ordersplacedwithUnited.Bythen,Mestahadalreadysecuredtheirfi mill, Mestahadwonorderstodesignandbuild13continuoushotstripmillscomparedwith builders until1928.Theysoonbegantocatchup.By1938,whenSummersboughttheir strip millordersduring1926and1927,butMestadidnotenterthemarketashot burgh. MestahadbeenasteportwobehindUnitedduringthe1920s.wonfi United’s Americanrivalinstripmillconstruction,theMestaMachineCompanyofPitts- Steel atButler,Pennsylvania, continuous widestripmillthroughcollaborationonthefi Foundry Company. relationship inJune1938. construction ofahotstripmillatShottonastheonlyviableroutetomakingqualitysteel, Fine SheetDepartmentatShottonunderaprofi The fi JohnSummersandSonsmusthaveup-to-dateequipmentinthe formofhotandcold ‘We shouldbewelladvisedtoputourselvesunreservedlyinMr.Iversen’shands’. installation ofhotandcoldmills. continuous stripmillstomaintaintheircompetitivepositionintheUK; The millwasahugegambleforJohnSummersandSons. TheShottonworkswassmall, Summers andRollasonreturnedfromAmericawithtworecommendationstotheJohn In March1937,RichardSummersandMrNevilleRollasoncrossedtheAtlanticto The EbbwValeMillwassuppliedfromAmericabytheUnitedEngineeringand ned recentcapital spendingtosmallschemes,suchasre-motoringsheet mills,and rm shouldopennegotiationswithMrIversenofMestaforthesupplyand 12 MestahadahugeplantequippedwiththelargestfoundryinNorthAmerica, ⅝ miles. 9 UnitedEngineeringplayedapivotalroleinthedevelopmentof 10 SHOTTON WIDESTRIPMILL subsequentlytakenoverbyArmco.SoSummersturnedto 13 Or,asRichardSummersrecalledmorefulsomely, 8 ArmcoandJohnSummersSonsseveredtheir t-sharing agreement.Armcoarguedfor rst successfulmillatColombia The Times rst exportorder on6August 11 ve 7

Published by & (c) The Newcomen Society cold mill;batchannealingfurnaces andtempermills. by newsoakingpits;acontinuous widehotstripmill;twopicklelines;athreestand tandem The fi John SummersandSonsbought acompleterollingmillcomplex—notjusthotstrip mill. four oftheeightslabbingmillsoakingpitscamefromAmsler MortoninPittsburgh. generating sets.Thetwo-slabreheatfurnacesforthehot stripmillwereboughtlocally,but Summers turnedtoUKheavyelectricalequipmentmanufacturers forthemillmotorsand might otherwisehavebeenprocuredfromestablished andexperiencedUSsuppliers. blooming millbuiltbyMestaatGreatLakesSteelCorporation. mill standsuppliedbyMestawasessentiallythesame design asanearlier40in.reversing and rollsthanintheiroriginalbid,lessfor engineeringcharges.Theslabbing their trustinMestaandgratefullyboughtwhatwasonoffer. The accountofthemillprojectwrittenbyRichardSummerssuggeststheynaïvelyput consistently profi and afurther£2millionwasraisedthroughdebentures. subscribed £1.2millioninsharecapital.TheBankofEnglandpledgedanother£1 through adealwithUnitedSteelsandbackingfromtheBankofEngland. £2.8 million.Extraworkingcapitalwasalsoneeded.Financefortheschemenegotiated for theschemeanewhotstripmillandcold-rollingcomplexslabbingwas the stripmillschememeantSummersfamilylostcontroloffi later on1November1938for$531 475 —17percentlessthantheinitialbid. faced withadoptingdisruptivetechnologiesofthissort. steelmaking capacity.AsChristensenargues,eventhebestmanagedfi of theirexistinghandmills.Thepotentialoutputthenewhotstripmillexceeded in rollingpractice.AnewmillcomplexatJohnSummersandSonsimpliedtheclosure and bargaininghardwithMesta. reducing theweightofcomponents,turningtolocalsuppliersinUKforstandardparts on anAmerican-madereplacement.Theyusedtimehonouredmethodstocutthecost— already hadtwoslabbingmillsofsortsonsitesotheycouldaffordtodelaythedecision 1937 for$643 780 based onsupplyfromtheUSA. get. Mestahadtosubmitnotonebuttwoquotesfortheslabbingmill.Thefi slabbing milltobebuiltonexistingfoundations(OrderJ5).HereSummersplayedhard and associatedpicklelinecoldmills(OrderJ310).Thesecondcontractwasfora came tobothprocurementandcostcontrol. otherwise. Inreality,themanagersofJohnSummersandSonswerehardheadedwhenit John SummersandSonsreliedonUK-basedsuppliers forarangeofequipmentthat Summers boughttheirmillfi Adoption ofcontinuoushotandcoldstriprollingatShottonwasacompletechange There wereactuallytwomajorcontractswithMesta.Thefi nal scopeoftheprojectincludedanewslabbingmillbuilt onexistingfoundationsfed table fromtheoutset. SHOTTON WIDESTRIPMILL rst andonlythenwonderedhowtopayforit.Thesizeof 20 Asithappens,Mestagotslightlymorefortheirstand PROCUREMENT NEGOTIATION 18 Thesecondbidwassubmitted16months 15 16 Yetthenewmillwastoprove 17 rstwasforahotstripmill Businessrecordssuggest rm. rms runscaredwhen 14 Theinitialbudget rst wasinJune 19 Summers 61 Published by & (c) The Newcomen Society 62 gross tonsofstripayearwhereastheShottonmillmighthaverolled375 000 tonsayearof Maryland PlantofBethlehemSteelcompletedbyMestain1937wasdesignedtoroll600 000 of astandardMestacontinuouswidestripmill.Thefull-size56in.hotmillatthe third reheatfurnaceandaseconddown-coiler.Ineffect,Shottonwaslowoutputversion time. Summerscutcostsbyeliminatingafourthroughingstand,sixthfi instead oftheusualtenstandsfoundonAmericancontinuouswidehotstripmills followed byfi with attachededgers.Thisabbreviatedroughingtrainofjustthreehorizontalstandswas train hadafour-highreversingstand,followedbytwo‘one-way’roughingstands six four-highfi was aroughingscalebreaker,fourfour-highstands,fi Point. Sheet &TubeCo.millislittledifferentfromBethlehemSteel’s56in.atSparrows Harbor Works. capacity millorderedaroundthesametimeforYoungstownSheetandTubeCo.’sIndiana mills. TheShottonhotstripmillspecifi mill wasnotthewidestavailableattime:Mestaalonehadalreadybuiltsixmuchwider barrel lengthforShottoncomparedwith56in.thetinplatemillatEbbwVale.A60 mill commissionedin1932. US widestripmilldesignduringthe1930sfollowedastandardlayoutsetbyInland’s79in. The Shottonhotwidestripmillwasan‘economyversion’ofaprovenAmericandesign. John SummersandSonsadoptedUStechnologywiththrift.TheShottonroughing Richard SummersandNevilleRollasonboughtawidermillthanEbbwVale—60in. 22 SoMestasoldessentiallythesamedesignatleastthreetimesover. ve fi ve nishing stands—makingteninall. 21 YoungstownandShottonaresuccessiveorders.Inturn,the nishing stands,Figures 4, 5and6.SoShottonhadeight standsintotal, i.4 Thethree-stand continuousroughingtrainlooking Fig. 4. west towardsthefi 23 Thestandardlayoutforapre-warcontinuoushotstripmill SHOTTON WIDESTRIPMILL MILL SPECIFICATION probably late1939. ve-stand fi cation isnotthatmuchdifferentfroma54in.higher nishing trainatShotton, nishing scalebreakerand nishing stand,a Published by & (c) The Newcomen Society more power,afasterfi a nine-standcontinuousmill(plusroughingandfi compares thework-upcurvesofEbbwValeandShotton hotstripmills.EbbwValewas of fewerstandsandlimitedreheatingcoilercapacity canbeseeninFigure 3 which per weekwithtworeheatfurnaces,butonlyone3½ton up-coiler,Figure 7. The Shottonhotstripmillwaslaidoutforarelatively lowinitialcapacityof10 000 tons out duringconstructionbyindependentconsultingengineersconfi had two-thirdsofthecapacityatypicalAmericanhotstripmillday. average UShotstripmillin1938hadacapacityequivalentto568 000 tonnes.SoShotton fi UK. Powersupplyandmotors accountedfor15percentofthecosthotand cold Power equipmentwassupplied byheavyelectricalequipmentcontractorslocated inthe specifi Shotton. Thehotstripmillspecifi added atEbbwValesoonafterthewar.Consequently, Ebbw Valealwaysrolledmorethan nal productsgiventheoperating practicesofthetime.Again,datafromEss every reasonthatitshouldbecapableofmakingthesamekindproducts. we candefi Summers werebuyingtriedandtestedtechnology.Anauditfortheirbankerscarried cations in1940aregivenTables 1a, 1band1c,respectively. nitely statethatthisinstallation is typicalofsimilarplantsinU.S.A.,andthere nishingspeedandtwodown-coilers. i.5 ShottonRoughingStandNo.2with Fig. 5. attached edgersbeforecommissioningin SHOTTON WIDESTRIPMILL cation in1939andtheslabbingmillcoldstrip MILL POWER November 1939. nishing horizontalscalebreakers)with 27 Afurtherfi rms: nishingstandwas 25 26 24 Theimpact suggestthe 63 Published by & (c) The Newcomen Society were includedintheEnglish Electricorder,butseemtohavebeensub-contracted to GEC Generating Setswhichsupplied thedirectcurrentforfi 64 Ltd inStaffordforthepower supplyanddrivesforthehotstripmill(Table 2). mill complex.Thelargestcontract for£112 516 wasplaced withTheEnglishElectricCo. Shotton wereall3000hpdrivingthroughgearboxes.Thefi i.6 Fivedcfi Fig. 6. looking east.Thefi motor nearestthecamerais2000hpdrivingdirectly. either engagesthecoiler,orallowscoilerto i.7 Thesingleup-coiler,seenhereinnew Fig. 7. wrapper rolldrivemotors,butitwastoprove condition beforecommissioning,wascheaper be dummiedsothatcutsheetscanpiledat than themoreelaboratedown-coilerwith down-coilers. Theleverintheforeground unsatisfactory andlaterreplacedwith SHOTTON WIDESTRIPMILL the endofrun-outtable. nishing trainmotorsfromEnglishElectric, rst fourmotorsonthefi nishing trainat nishing traindrivemotors nal 28 TheMotor Published by & (c) The Newcomen Society Statistics alternating currentmotorsto drivedynamoswhich,inturn,makedirectcurrentto power direct currentsuppliedbytwo motorgenerator(MG)sets.AnMGsetusesconventional naturally chosevariablespeed dcmotorsforthefi power supplyforthehotand coldstripmills—doubtlessunderadvicefromMesta. They J17). (order £23 750 motors andgeneratorsforthecoldmillcamefromBritish ThomsonHoustonLtd.for equipment, Mestawasonlyresponsibleforthegearboxes andtheedgermotors.Themain at Wilton,Birminghamtojudgebyphotographicevidence. Amongthemotorroom fromMGset fromMGset fromMGset fromMGset fromMGset of maximumwidth53in. Sources: IronandCoalTradesReview(1953,p.80);Mesta MachineCompany, 3tonslimitedbyslabweight, 10 000 tonsperweek Building 1 Capacity Oneup-coiler– initially 4-high 500 Coilers 17¾milesperhour 4-high speed Finishing 22½ 4-high 4-high F8 22½ 22½ 4-high F7 22½ 22½ 4-high F6 22½ 22½ 4-high plus dc F5 stand roughing No.3 stands plus Finishing F4 motors stand length roughing No.2 Pinch rolls broadside edgers length synchronous edgers synchronous 40 2-wide rolling) (also tonsperhour ‘Spreader to4-tonslabs,150 induction stands Verticalscalebreaker Roughing 4-high stand’ furnaces pusher Reheat Two, furnaces, 38 3-zone 1½ continuous

Continuous andSemi-ContinuousHotStripMills John SummersandSonsscrupulously followedAmericanpracticeofthedayover , Code426,Consignment13,box12,Location11/2577. p long x 100 Shotton hotstripmillspecifi p wide SHOTTON WIDESTRIPMILL TABLE 1a ; AndrewReid(1948); nishing train ofthehotstripmill,with q q &49 diameter q q q q q q q cation in1939 &44 &44 &44 &44 &44 &44 &44 qx qx qx qx qx qx qx qx 96 x 60 60 60 60 60 60 60 q 17 2 500 hp ac hp 2500 q q q q q q q 00 pd 200/400rpm dc 2000 hp 200/400rpm dc 3000 hp 200/400rpm dc 3000 hp 200/400rpm dc 3000 hp 200/400rpm dc 3000 hp ac 2500 hp ac 2500 hp q 2 x John Summers&Sons. . . 300 hp Reference List 65 Published by & (c) The Newcomen Society 66 box 12,Location11/2577 aaiy 6 000 tonsperweek,upto52in.wide Sources: AndrewReid(1948,p.37); Threestand,4-high,20½ Capacity Finishingspeed,13½milesperhour Coilers One Two,sulphuricacidlines tandem coldmill Three stand Pickle lines up-coiler GuideBridge,Manchester 1, 1938toJohnSummers&Sons,Limited Watkins (1979),pp.87–90; 120tonsperhour;600 Sources: 000 ingottonsperyearworking7 200 hours Capacity slab comingfromthereheat furnace.Hereasophisticatedandexpensivewound rotor high reversingroughingstand hadtocopewithheavyloadsasittookthefi necessary, byadjustingthevoltage outputofthestandgenerator. all conditionsofload,principally byadjustingthefi able tosetandmaintaintherequiredspeedoneachstand withinveryclosetolerancesunder dc generators—oneforeachfi sets. EachMGsetwasdrivenbyasynchronousinduction motormechanicallycoupledto range of2or2½to1.InthecaseShotton,dccurrent wassuppliedbythetwoMG between two-to-oneuptoten-to-one.Thenaturalchoice isadcmotorwithspeedcontrol of reduction.Theratioingoingthicknesstofi the materialelongatesasitisrolled.Therelativespeed of eachstandvarieswiththedegree mill andtosupplydcpowerthefi mills —Weirtonwasapartialexceptionalthoughrectifi the mill.Mercuryarcrectifi Steampressure120 lbs /in 3Cylinderhorizontalsimplesteamenginewithcylinders40 Drive when stand wide from sheared 20 Four,SalemRoundTypesoakingpits—holdingcapacity,96tonsperpit Reversingstand,one2-high,rolls7 Slabbing mill Four,Amsler-MortonSquareTypesoakingpits—holdingcapacity,100tonsperpit. Soaking pits The roughingstandsofthe hotstripmillweredrivenbyacmotors.Thefi Successive standsofahotmillfi John Summers&Sons. . . Statistics Proposal fromMestaMachineCompany,Pittsburgh,PA,USA,November qx Shotton coldstripmillspecifi Shotton slabbingmillspecifi ers toconvertacdccurrent werenotusedinearlyUSstrip 40 q nishing standdrive.Thecontrolsystemsforthedriveswere SHOTTON WIDESTRIPMILL ingots;maximumslabsection36 John Summers&Sons. . . Statistics nishing trainatDinslakenwidestripmillinGermany. nishing trainoperateatfasterandspeedssince 2 ; maximumspeed180 rpm. Builder,Scott&Hodgson, , Code426,Consignment13,box12,Location11/2577; TABLE 1b TABLE 1c q &44 p 8 nal outgoingstripgaugecanvaryanywhere q qx body 55 elds ofthedrivemotorsbut also,when cation in1940 cation in1940 q , 1 500 hp perstand,dcworkrolldrives. x 38½ ers wereusedonaUSwirerod q diameter.Torollslabs36 qx 5 , Code426,Consignment13, q 29 ; maximumlength16 q bore x 54 rst biteofthe q stroke. rst four- rst q p Published by & (c) The Newcomen Society ) oprmtrgnrtn e £304 £208 £9 581 twocoolingandventilatingequipmentsforhotmillmotorroomamake-upair p) Motor room £1 463 setofcontrolgearforfi o) n) fi loopermotorgeneratingset fi m) volt 600 2 000hp one l) fi volt 600 3 000hp four £988 (presumablysub-contractedtoGEC) k) withafl two5 000 kW 600voltmotorgenerating(MG)sets j) Finishing train(directcurrentmotors) operatorcontrolpulpitforroughingmill £1 226 i) twocontactorcontrollersfor150hpedgermotorsonMestaHotMillMotorlist h) slipregulatorforstartingandcontrollingofspreadermilldrive(c) g) tworeversingcontactorcontrollersforedgerdrives f) e) two fi ) w 0 h eesn de rv oos £895 £22 298 (1 motors synchronous 2 500hp two d) fl 2500hp c) two300 hp reversingedgerdrivemotors b) Roughing train(alternatingcurrentmotors) 15panel 6.6 kV mainswitchgearwithcontrolboard a) Switchgear Derived from Ovesen, StripMillEngineering,Departmentalcode426;consignment4;boxno.3;location6758. Source: Total acceleration andthedivisionofloadbetweenfl induction motorwasused,controlledbyaslipregulator whichhandledthestartingpower, speed synchronousmotors. American practice,stands2and3weredrivenbythecheapest typeofacmotor—constant carrying thesteelforwardthroughrolls,evenif circuitstripunderload.Following kinetic energyandsohavetheadvantageoflowering peakloadsonthemillmotorsand and know-howUKbased civilengineering,constructionandelectricalengineering. In The millwasnotaturnkeyproject, butamixtureofUSmechanicalandfurnaceequipment fi and loopergeneratingsets edcnrlpnlfrsnhoosmtr fte5 0 k eeaigst j £328 eld controlpanelforsynchronousmotorsofthe5 000 kW generatingsets(j) lter andfan Requisition, TheEnglishElectricCo.Ltd.,Stafford,4November1937 edcnrlpnl o h ycrnu oos() £328 eld controlpanelsforthesynchronousmotors(d)

yn ha eeao n eaaeectrst £23 669 ying sheargeneratorandseparateexciterset

yhe nuto oo srae ildie £10 720 ywheel inductionmotor(spreadermilldrive) Order toTheEnglishElectricCo.Ltd.,Stafford(Orderno.B.21) Shotton hotstripmilldrivesandelectricalequipment nsigml rv oo o 0/5 rm £3 687 nishing milldrivemotorfor200/350 rpm nishing millgeneratingsets,mainmotors,ﬂ nsigml rv oosfr2040 p £20 007 nishing milldrivemotorsfor200/400 rpm 31 SHOTTON WIDESTRIPMILL SUPPLIERS —STRIPMILLS st and2 US suppliers TABLE 2 nd uieslml rvs £7 956 universalmilldrives) ywheel andthemotor. yn ha oos £8 858 ying shearmotors, , requestedby:MrH. . 30 Flywheelsstore £112 516

67 Published by & (c) The Newcomen Society 68 8 45 56687 5123 27192 16042 667500 41 582 283435 13 453 25617 20322 9802 onwards bringingtheultimate payment to$3,848,746byNovember1939. 176147 123421 *There werealsomiscellaneous followuporderstoMestaforsmalleritemsfromNovember 1938 16259 plus costsheet,group6‘Soaking Pits’insameFile. £ 135960 revised estimate28February1940 3350852 204 21358 80210 252 Contract’ circaNovember1939, allinFilelabelled‘StripMill’including November 1938plushandwrittenworksheets‘DetailsofPayments toMesta’and‘AmslerMorton 302,367 Sources includeuntitledtypescriptmemorandumfromMrH. OvesentoMrRichardSummers,22 67 265 equipment Lubricating 101610 Bowser, FortWayne,Indiana 49011 Soaking pitsfortheslabbingmill Order J113‘Soakingpits’ 880734 619577 AmslerMorton &Co.Pittsburgh,Pennsylvania(ordered3May 1938) 81292 106792 services Engineering 1511836 $ Spare partsandsparerolls Equipment Slabbing mill Order J5‘Slabbingmill&spares’ services Engineering rolls Spare parts Spare Equipment Cold mill rolls Spare parts Spare Equipment Hot mill Order J310‘Hot&ColdStripMillcomplete’ Mesta MachineCompany,Pittsburgh,Pennsylvania(asat21December1938)* engineering man-hours—thereisaveryheavyitemforthisincludedinthebudget and machineshopsfullyoccupied.ItisalsotruethatJohnSummerssuppliedconsiderable ferred totaketheirprofi for 18.5percentofMesta’stotalpricethiscontract.WecanspeculatethatMestapre- per centofthefi mill complex(Table 3). Thecontractincluded£123 421 forengineeringservices—just5 contract withMesta—J310accountedfor28percentofthefi shears, coiler,therollfi British. budget terms,thestripmillandslabbingprojectswere30percentAmerican,70percent Mesta wasresponsibleforthesupplyofall‘noblemachinery’:millstands,rollertables, nal projectcostfor‘know-how’.Putanotherway,engineeringaccounted Main UScontractsfortheShottonstripmillandslabbing ll anddrivegearboxesforboththehotcoldmills.Themain t ontheactualequipmentitself,andkeeptheirPittsburghfoundry , GroupSecretaryCode425,Consignment 36,box6,Location4747 SHOTTON WIDESTRIPMILL TABLE 3 nal costofthehotandcold StripMillConstruction, Published by & (c) The Newcomen Society design ofrecuperativesoakingpitpre-heatedtheairandwasenergyeffi Mesta, accountingfor10percentoftheoverallcostslabmillscheme.ThisAmerican Morton CompanyofPittsburgh.Thesecostmorethantheslabbingmillequipmentfrom supplied themillstanditselfandassociatedgearbox. ing giventhataseniordesignerfromMestawastiedupontheprojectandonly payment toMestaunderthecontract—some28percentoftheirprice,perhapsnotsurpris- overall costofthescheme.However,thissumrepresentsasignifi project. Heretheengineeringservicescomponentaccountedforlessthan4percentof built onexistingfoundationsonlycost14percentofthetotaloverallslabmill sold theirrevolutionaryknow-howataveryreasonableprice. are nocomparablefi duopoly atthetimeoverradicalinnovationsembodiedinstripmillequipment. acquiring suchpowerfultechnology.Afterall,MestaandUnitedhadavirtualworld the stripmill.Evenmakingtheseallowances,Mesta’schargesstillseemreasonablefor Shotton installationsweretheonlypitsoftheirtypeoutsideNorthAmerica. used inMarch1936atCarnegie-IllinoisSteelCorporationandspreadrapidly,althoughthe sterling. Atthetimecontractwassigned,exchangeratefi possibly couldfromdomesticsources. accounted foranother5percent.ItisclearthattheSummersteamboughtasmuchthey — over92percentaccountedforbypaymentstoMesta.Soakingpitstheslabbingmill four circularsoakingpitscamefromSalem. although theratedrifteddownwardsduring1938and1939topointwhereSummers Four oftheeightingotsoakingpitsforslabmillwereboughtfromAmsler- The supplyoftheslabbingmillitself(seeFigure9)fromMesta—contractJ5tobe The costofprocuringfromtheUnitedStateswashelpedbypurchasingpower The totaldollarelementoftheShottonprojectwas$4 165 508 bytheoutbreakofwar bay. ThismaybetheUScommissioning crewatworkin1940. i.8 Thethree-standtandem coldmillbuiltacrossthe Fig. 8. gures forotherprojects,butadetachedviewsuggeststheAmericans SHOTTON WIDESTRIPMILL cant partofthetotal cient. Theywerefi xed atUS$5 33 Theother 32 There = £1, rst 69 Published by & (c) The Newcomen Society 70 November 1938. placed ayearbeforethefi The fi value ofthisparticularcontract was£91 265. Theyalsosupplied50motorsfor £7 400. gear andmanipulatorsfrom thenewlyfoundedLoewyEngineeringCo.Ltd.The total were suppliedfromtheUK.Theseincludedmillapproach tables,aslabshear,discharging The 40-inchreversingslabbingmillcontractwasnegotiated separately.Basicitems structural steelforthebuildingsandrailssite. and Sons,theSheltonIronSteelCoalCompany, wasresponsibleformuchofthe tanks andannealingequipmentwereallreadilyavailable. AsubsidiaryofJohnSummers time. Itemssuchascranes,coilconveyers,rollertables, hydraulicequipment,pickling Photographs suggestitwasacarburettedwatergasplant. was essentiallyalocal,small-scalecoalgasifi provided bya‘producergasplant’,perhapsuniquearrangementamongstripmills.This needed fuel.Afterall,therewerenocokeovensorblastfurnacesonthesitethen.Gaswas Along withotherancillaryitemsthecontractwasworth£13 191. Thereheat furnaces Lamberton wonthecontractforfurnacepusherswhichphysicallyhandledslabs. sources, fromStein&Atkinson—acontractfor£47 650 includingthecontrolequipment. Hawarden andBroughton. At thetime,shadowfactoriesforarmamentsproductionwerebeingbuiltclosebyat Richard Summerswrote: not withoutitsdiffi The restofthemillcomplexwasprocureddomesticallyinUK.Localsourcing 1d. 15s £883 043 of thepoundhadcostSummersanextra£49 942 outofatotalexpenditureondollars only received$4.67¾perpoundbyAugust1939.Bytheoutbreakofwar,depreciation components refl came fromDarlingtonForge.Thediversityofsupplyforawiderangemechanical and Co.theedgergearsfromDavidBrownSons.Therollchangeliftingbeams the NationalSteelFoundry;pinionhousingsandreductiongearsfromDuncan,Stewart nearby WidnesFoundryandEngineeringCompany;thebearingboxesforhotmillfrom and gearingweresourcedlocally.Theseitemsincludethemillshoeplatesboltsfrom Most ofthiswenttotheInternationalConstructionCoLtd. short-lived installationwas£64 000 —slightlymorethanthereheatfurnacesthemselves. builders werefullyoccupied. and diffi We foundconsiderablediffi The twopusherslabreheatingfurnaceforthehotstripmillcamefromdomestic Standard itemsofmechanicalequipmentforthehotandcoldmillssuchascastings rst domestic ordersforslabmillequipment,suchasthe auxiliary motors,were cult strugglesingettingsatisfactorypricesanddatesofdelivery,asallmachinery ects thestrengthofheavyengineeringinfrastructureinUKatthat 34 culties assupplyfi nal contractforthestanditselfwasconcluded with Mestain 35 culty inobtainingdraughtsmen.Wealsohadmanylong SHOTTON WIDESTRIPMILL Suppliers —SlabbingMill Domestic suppliers rms wereatfullstretchin1937duetore-armament. cation plantbuiltrightnexttothestripmill. 36 Thefi nal budgetforthis Published by & (c) The Newcomen Society rebates andextrashadfl for aninitial£120 000 althoughthe fi buildings, millandmachinery (orderJ187on19August1937)wenttoA.Monk& Co. Ltd concrete caissonwithcellars below.Themaincontractforthefoundationsstrip mill £27 423. Steelworkwasthenerectedontop.Themilllineitselfareinforced of sand.PeterLindgotthecontractforpiledriving andfortheconcretepads,worth the siteandthencappedwithconcretebeamstoforma skeletonframeworkamidadesert and machineryrestedonconcretepiles.Thereinforced concretepileswerefabricatedat lagoons formedbetweenearthembankmentsandthewater lefttodrainaway.Thebuildings from theadjacentRiverDeetomakeatwenty-sevenacre platform.Sandwaspumpedinto level. Thenewsitewasraised17feetbypumpingthree-quarters ofamilliontonnessand in September. construction beganwithtestboringsonthemarshysite inearlyAugust1937andtestpiling Henrik Ovesen,assistedbyCyrilBottomleyofJohnSummers andSons.Asnotedearlier, Manager ofJohnSummersandSons.Day-to-dayprojectmanagementwasinthehands The wholeprojectwassupervisedbyReithGray,whowentontobecomeGeneral The millcomplexwasbuiltonDeeestuarymarshland thatwasoriginallyatsea emerging fromtherubbleofold‘barmill’foundations.Thismillcouldnot roll slabsuptothefullwidthrequiredbyhotmill.Itwasreplacedafter i.9 Aseparatecontractcoveredthesupplyofaslabbingmill,seenhere Fig. 9. war, butagainthefoundationsprovedunsatisfactoryandanotherslabmill owed backandforthbetweencontractor client. SHOTTON WIDESTRIPMILL was installedjustbeforeclosure. CONSTRUCTION nal outturnwas£1930more by March1940afterhefty 71 Published by & (c) The Newcomen Society 72 ‘Belship’ fl twenty daysatsea,Figure 10. Twofurther cargoesweredeliveredwithoutincidentbyother total of120tonswerelostoverboard. weather intheAtlanticandwasreporteddistress.Fourpiecesofmachineryweighinga machinery, includingthehousingsforhotmillstands.Butvesselstruckfoul heavy liftcargoship,theMVBelnor.ItsetoutfromBaltimorecarrying2 517 tonsof in March1939.Thefi under way. electrical equipment.Assemblyofthetwolocallysuppliedreheatfurnacescouldthenget strip millwerevirtuallycompletebymid-October1938,wellaheadofthemechanicaland May 1939,bywhichtimethefi towards themfromtheslabyardateastend.Theyweretomeetinmiddlearound building fromthewestendinwardson31August1938,whileLysaght’sstarted ends. Twostructuralsteelworkcontractorswereused.RuberyOwenbeganthetwo-bay furnaces. Themainhotstripmillbuildingwasstartedsimultaneouslyfromtheeastandwest fi building. Themillbuildingsthemselvestookayeartocompleteoncethesitewasready. £4 001. which ranbelowthehotmillrun-outtable—ajobforHolst&Cowerepaid Considerable effortwentintoprotectingthewaterpipeswithareinforcedconcretetunnel rst mechanicalitemtogetunderwaywastheproducergasplantsupplyreheat The projectwasnotwithoutdrama:deliveryofplantfromtheUSAbecameaconcern The hotandcoldmills,annealingfurnacescut-uplinesinitiallyoccupiedatwo-bay A watermainfromWalestoBirkenheadrandiagonallyacrossthesiteofhotmill. eet heavyliftshipson1Mayand 29May1939.Warhadbrokenoutbythetime i.1. Thefi Fig. 10. Docks fromPittsburghviaBaltimore aftertheheavyliftship MV Belnorranintodiffi rst cargoofcomponentsfromMestawasshippedaboardaNorwegian during stormyweatherintheNorth Atlantic. rst millstandswerealreadyinplace.Foundations forthehot SHOTTON WIDESTRIPMILL rst cargoofequipmentarrivingatBirkenhead 37 Finally,thedamagedshipreachedBirkenheadafter culties andlostpartofhercargo Published by & (c) The Newcomen Society iigmcie3 21 6 10 Source: Manuscript‘TheStrip Mill’,code426,consignment12,box7,locationF10765. 16 Oiling machine3 Oiling machine2 Oiling machine1 Inspection 10 Re-shears 9 Shipping department he-tn admcl il 8 Three-stand tandemcoldmill neln tcs 18 11 18 21 14 5 23 29 Annealing stacks Broadside coldmill Skin passmill Rotary shears3 Rotary shears2 Rotary shears1 Pickle line“A” Pickle line‘B’ iigmcie4 5 Oiling machine4 o ti il 15 Hot stripmill 1 Mesta wasrenownedformaintainingsparecapacitytomeetunforeseenorders. MV ManchesterDivision,whichberthedatLiverpoolon19September1939.Fortunately, with theslabbingmillstandsixdaysbeforeoutbreakofwaron28Augustaboard the fi mill standswereplacedinpositionfi of theBelnor,millhousingsbegantogointoplacealongrollingline.Thehot provided slabsforthenewmill.) judicious useofexplosives.(Presumably,theotherbarmillatNo. 1 open-hearthplant mill wasdismantledinearlyAugust1939andtheassociatedrun-outtabledemolishedwith kept inoperationaslongpossiblesupplyingtheexistinghandmills.TheNo. 2 bar ‘bar’ millandassociatedingotsoakingpitsoftheadjacentNo. 2 open hearthplantwere operation. Thefullchronologyofthestart-upisgiveninTable 4. Theexistingtwo-stand teething troubles,wentstraightintoproduction,showingaprofi Cutts, supervisedinstallationofalltheelectricalmachinery. fi and October1939,startingwiththeroughingtraindrivesmovingdownmillto in JuneandJuly.Theelectricalequipmentwasassembledveryquicklybetween of tandemcoldmillstandswereinposition.Thedrivebasesandgearboxespositioned By 18June1939,alleightpairsofhotmillstandswereinplace.4July,thethree following throughthefi nishing trainmotorsandtheMGsets.Summers’ ownChiefElectricalEngineer,Alfred Once themachineryarrived,projectmovedrapidly.Withinaweekofarrival The hotstripmillitselfstartedworkon9November1939at17.50and,withfew nal componentswereshipped.Replacementsforthelostpartsshippedalong nishing train,coldmillstandsandtempermill,workingwestward. Commissioning datesofstripmillunits SHOTTON WIDESTRIPMILL rst, startingwiththeroughingtraininearlyApril1939, th th st th th th Sit 2 Shift rd th th st th th th th th th st th Dc 3 5 20 Dec.’39 April’40 My’1 May’41 Oct.’40 Aug.’40 Fb 4 Feb.’41 30 May’40 Mrh’0 17 12 March’40 Nov.’39 Jl 4 July’40 July’40 18 9 July’40 June’40 17 Feb.’41 July’40 Feb.’40 Sp.’0 14 Sept.’40 TABLE 4 th th nd th th th th th th th Ot 4 24 24 Oct.’40 Dec.’39 Sit 3 Shift Fb 4 28 15 26 Feb.’41 27 May’40 19 Feb.’40 24 22 Oct.’40 April’40 May’40 Feb.’40 t initsfi rst fullyearof rd th th th th th th th th nd Shift 38 Feb.’41 June’41 Aug.’40 March’40 Jan.’40 Oct.’40 May’40 June’40 April’40 73 Published by & (c) The Newcomen Society 74 run-out tableontothefl taken offthedelaytable;slab7madeitthroughfi Richard Summersrecalled: However, wordgotroundthatthehotmillwasduetocommissionon9November.As seems, thecomprehensivephotographicrecordofsiteendingabruptlyinOctober1939. razamattazz usuallyassociatedwithnewplant.Notevenanycommissioningphotos,it mills andallthewerecommissionedunderwartimeconditions.Therewasnoneof experienced hands. of YoungstownSheetandTubeCompanyintheUSA.Hewashelpedbyjustfourother of commissioning,havinghadexperienceasasecond-in-commandatthehotstripmill Another American,GeorgePaul,supervisedtheerectionofmachineryandtookcharge otherwise allthedrawingoffi slabbing mill.MrKalberkampwasaccompaniedbyaChiefDraughtsmanfromMesta,but help withUKprocurementforthestripmillsandwentontosuperviseconstructionof project. AdesignerandChiefEngineerwithMesta,H.J.Kalberkamp,wassentacrossto little experienceofthenewtechnology.OnlyahandfulAmericanshelpedwith with ahighlevelofdomesticprocurement.Yet,theprojectwasmanagedbyteam These baldfactsconcealaprojectthatwascompletedatspeed,almostexactlyonbudget, the wholestripmillproject. cause ofahugecostover-run£58 000 —amongtheveryfewunanticipatedcostitemsof heavier slabs.Hastyreinforcementofthesefoundationsunderwartimeconditionswasa tions oftheoldbarmillwerenotreallyadequatetosupportnewslabwithits Morton soakingpitswerequicklybuiltduringSeptemberandOctober1939.Thefounda- powerful steamdrivefromthesecondstandofearliermillwasre-used.TheAmsler run-out tablewereemerging—quiteliterallyfromtherubbleofoldbarmills.The stand wasinplacebytheendofOctober1939and,then,foundationsnew mill standnexttoNo. 2 furnaceshop.Theslabbingmillprojectmovedswiftlyasthenew to hismateandsaid,‘Well,now Iguessitlooksjustlikehome’. by aremarkmadeoneofthe Americans,whopointedtothestriponfl fl out asfi the facesofGeorgePaulandBrown the slabsbyrunningupanddownbuilding,nordoesitmention thelookoftriumphon This report,however,doesnottelloftheexcitedcrowdwhich triedtofollowtheprogressof mill. appeared tomethatpracticallyeverybodyinouremployhad foundhiswaytothestrip and mengangafta-gley’,insteadofthemillbeingcompletely freefromspectators,it novices atrollingstrip.)But,asRobertBurnssoaptlysaid,‘The bestlaidschemeso’mice upset theoperators(allofwhom,withexceptionfour Americans,werecomplete We decidedtotrythefi The fi The SecondWorldWarbrokeoutduringcompletionofthefi The newMestaslabbingmillwasthenbuiltontheoldfoundationsofsecondbar oor withits threeorfourred-hotstripsrestingatfantasticangles. . . .I wasmuch amused 40 rst coupleofslabsstuckintheroughingtrain;3to6wererolledasplateand nished strip.These willlongremaininmymemory,andsothepictureof mill oor —andsoit continued, atypicalstartup: rst rollingafter5o’clocksothattherewouldnotbealargecrowdto ce workwasdonebySummers’ownorganisationatShotton. SHOTTON WIDESTRIPMILL COMMISSIONING 41 whenthefi rst piecestoenterthefi nishing trainbutcamberedoffthe nal stagesofthehotstrip nishing standscame oor, andturned 39

Published by & (c) The Newcomen Society project. SurvivingdocumentssuggestOvesensignedoffalltheorders. suggests theDanishengineerwasverywellinformedonallfi spending profi 4747. reinforcement oftheoffi pencils, anearlyHoovervacuumcleaner(£21),coathangersfromWoolworths(£1)and ink andthoseabovebudgetinred.Everyexpenditureitemislisted,includingtracingpaper, ment ofthesecostcodesagainstbudget.Itemsrunningbelowbudgetwererecordedinblue project aregiveninTable 5. individual orders,whileothergroupscoverarangeofsuppliers.Theout-turncoststhe 80 isdutyandfreight.Group85inventory.Sometimesthesecodescorrespondto mill, picklelines,annealers,coldmillandtempermills.Group79iscontingenciesgroup cost groupsrelatetothesoakingpitsandslabbingmill,1478coverhot The projectwasmeticulouslymanagedwithjustover80costcodesor‘groups’.fi Steamdrivere-used 416100 355143 Corporation,extracrane 5098 25000 18000 5000 11000 Holst&Co. 34000 9577 12000 revised estimate28February1940 Derived frommanuscriptaccounts inFilelabelled‘StripMill’including MostlyGECandLoewyEngineering. installation StandfromMesta,extragearand & Cables only insulators valves, Chimneys, Owen Smith Soaking pitsfromAmslerMortonand Wellman Miscellaneous supplier Main 120 000 40 002 Erection 3000 113 64 850 000 13 couplings Modifi 31 42000 716 11 68 Spares 000 12000 Tables&motors 10 Salem 33000 6000 Slabmillstand ordered 9, 12 Huge over-runshere,mostlylabourto 8000 8 Soakingpits system 60 Fuel 000 7 building 8 000 Cranes 6 22000 5 Buildings 20000 estimate 4 Foundations £ design Engineering £ 3 Out-turn Comment Original 2 1 number Group Item Construction costsoftheslabbingmill,Shotton,estimatedtotalexpendituretocompletionat Cost controlwasstrictandpaymentstosupplierslaggedwellbehindtheactual The archivesshowcostcontrolwasachievedthroughhighlydetailedmonthlyassess- le. InternalcorrespondencebetweenHenrySummersandHenrikOvesen ce windowsagainstpossiblebombblasts. SHOTTON WIDESTRIPMILL , GroupSecretaryCode425, Consignment 36,box6,Location COST CONTROL 28 th February1940 TABLE 5a cations toexistingbarmill nancial aspectsofthe Strip MillConstruction, rst 13 75 Published by & (c) The Newcomen Society ThomsonHouston,Igranic projectwhileunderway coiler,80%Mestasupply instruments,spares,offi Over-runpartlyduetoextrapickle when added capacity annealing Extra losses, exchange Contingencies, 82 000 81500 the to added 2,356,857 line pickle Second 30000 204729 Atkinson,Lambertonpushers British 95200 Ltd,SummersShelton GEC, EE, supply: UK 370000 48000 foundations,Birkenheadtunnel patent 360000 Miscellaneous uncoded 2,403,569 equipment + Electrical 79 toMestafordesign Annealing machinery 41 75–78, project lines Pickle payments 74 73, Stands,bases,gears,drives,tables, 729 813 Adamson Joseph pits, 726 Mostly & scale 71 Stein hydraulics, 178 70, from supply, furnaces Water Reheat 82858 202176 69000 Hotandcoldmill underway International from 43–69, 30,72 plant 76950 188000 gas Producer line 78000 &pushers 64000 handling and Cranes spares and Services 31–34, 40 WCJ 68500 Owen, Rubery Lysaghts, John furnaces Reheat 181000 29 21–26, plant, gas Producer 28 186500 18, payment and Shotton at work Design 42 20, Construction Co., Dredging 192000 Westminster 254022 175000 19 215000 Buildings design Engineering 63 39, 35, 17, estimate engineering Civil 36–38 15–16, £ 14 Original £ Out-turn numbers Group Item Comment 76 oa 2 0 00 99 6 19 6 20 0 2708969 200700 109 669 393100 theoriginal 2225 869 2909669 notinoriginal 2800000 23000 177700 2000 60957 46712 90000 416 100 Total 2403 569 Location 4747. 88000 2356857 355143 Construction, revisedestimate28 February1940 Tables 5band5cderivedfrom manuscriptaccountsinFilelabelled‘StripMill’including freight and Duty estimate mill Strip mill Slab 80 14–79 £ £ £ £ estimate 1–13 £ items Original Out-turn Over-run Main numbers Group Item items Final cost of Construction costsofthehotstripandcoldmillcomplex,Shotton,estimatedexpenditureto Total costofthestripmillschemeatShotton,estimatedexpenditure tocompletionat completion at28 SHOTTON WIDESTRIPMILL 28 th February1940 TABLE 5b TABLE 5c , GroupSecretaryCode425,Consignment 36,box6, th February1940 ce equipment. Strip Mill Published by & (c) The Newcomen Society executives. accustomed todoingbusinesssolelyonthebasisoforalagreementwithacustomer’stop optimistic abouteveryitem,exceptthemachinerytobesuppliedbyMesta.ButIversenwas 1938 and1939.TheoriginalIversenquoteattheSavoyequivalentto£2millionhadbeen fi complex itselfcamein£91,000 extra £200 700 ofspending;hence,excludingtheadditionalequipment,stripmill annealing basesandcraneswereaddedtotheschemeatanearlystageamounting just over£2.9millioncompareswithaninitialestimateof£2.8million.Extrapicklelines, fi The totalcostofthewholemillschemewas£2 909 669, oratleastthiswastheestimated earned £2 347 275 million in its fi its in million £2 347275 earned to 1942andthenmonthlyuntiltheendof1946. Summary departmentalaccountssurviveintypescriptorhandwrittenformfortheyearsup with theslabmillfoundationswereclearlyidentifi host ofsmallsurveyingitemssuchWellingtonboots,oilskinsandtheodolites.Diffi wildest dreams.Yet,intheevent,only£20 000 ofthecontingencywasused,mainlyfora of theestimatedcostwholeproject.Suchahugemarginexceedsprojectmanager’s • • • The reasonsforsuccessareclear: recovered itsfullcapitaloutlayby1947. point remains—themillprojectwasahighlyprofi Lysaght duringthewarandtheseorderswerenotalwaysprofi cross-subsidising obsoletefacilities.TheShottonstripmillalsoundertookhotrollingfor Sons, therebyinfl transferred atcosttotheadjacentMarshmillsandStaffordshireofJohnSummers weeks of1939(Table 6). Intruththeseprofi tently profi nal stagesofconstructionanddespitedepreciationthepoundagainst thedollarduring nal costwhenthedetailedaccountsclosedatendofFebruary1940.Thissum Summersweresupportedbyengineeringconstruction companiesandelectrical RichardSummershimselfpretendedtobenaïveandtrusting.In truth, themanagement Summers boughtatriedandtesteddesign.TheyacquiredUStechnology‘offtheshelf’. engineering fi specifi construction withcare.They appointedanexperiencedprojectengineerwith the were extremelygoodatcost control;paringthecostsofmillandmonitoring its of experiencedUScommissioningengineers; They reliedonanestablishedsupplierforthecriticalcomponents andhadthesupport Contingencies forthestripmillweresetatagenerous£204 729 —some8.7percent c taskofmanagingthewholescheme; tableapartfromabriefinitiallossof£3 600 duringtheworkupinlastfew 42 rms intheUKwithconsiderable experienceoflargeprojects.Despite ating profi ating REASONS FORTECHNICALSUCCESS ts atthesedownstreamoperations.Ineffect,thenewmillwas SHOTTON WIDESTRIPMILL rst sixyearsofwartimeoperation.Theschemewasconsis- OVERALL COSTS PROFITABILITY below ts are understatedassemi-fi budget,despitetheoutbreakofwarduring 43 ed undertheappropriatecostcode. Fromthese,weknowthewidestripmill table investment.Themillschemehad table. Butthecentral nished materialwas culties 77 Published by & (c) The Newcomen Society 78 US. of millconstructionbackhome. Theoutbreakofwarprecludedfurthertravelback tothe strip millatEbbwVale.Richard Summershimselfwastoopre-occupiedwiththe realities Summers andSonsorRichard ThomasandCo.Ltd,whowerethencompletingtheir wide more ofa‘jolly’andthatlittle newwasdiscoveredbyeithertherepresentativesfrom John to havetakenplace. led aUKdelegationtouringAmericanstripmillsin1938, wherefrankdiscussionsaresaid Otherwise therewaslittlecontactwiththeUS.SirRichard’s brother,GeoffreySummers, • • • • 90 4 89 273838 Hotandcoldmills 244879 Hotstrip 1941 1940 1939 (onemonth) tons Output £ Profi 96 7 46 418018 354160 496104 384746 375436 345964 361089 358357 9 consignment 36,box3,location4744;J.H.Pearce(1941), Sources: File‘SummariesofResultsOctober1942toDecember1944’,(actually1946),code425, Notes: fi 1950 489 1949 399 1948 363 1947 390 1946 1945 1944 1943 1942 000 000 000 000 File 10 th Therewasanopen‘trust’relationshipwiththeUSmillsupplier.Clearly USmill Summersunderstoodthepotentialmarketfornewmillinautobodysheet and ThehighvalueofsterlinghelpedSummersprocurethemillcheaplyfromUSA; WhilethereisnodirectdocumentaryevidenceintheJohnSummersandSonsarchive, August1941,GroupSecretaryCode426,Consignment12,Box7,LocationF10765,FileboxE, Ukraine andJapanforthenarrowmillMestabuilt inRussia. world. Nodoubtasimilarstorycouldbetoldformills thatUnitedsuppliedtothe builders playedacentralroleindiffusingthenewwide strip milltechnologyaroundthe helped themilltoworkupquickly; customers. Anundemandinginitialmarketforhotrolled stripforAndersonshelters defence equipment.Theywereessentiallysupplyingan upgradedproducttoexisting are striking; decision makingin1937bytheBankofEnglandandImportAdvisoryCommittee it iscleartheimminentoutbreakofwargaveimpetustoproject.Thespeed within cost; pressures onoutputcausedbyre-armament,thesesuppliersdeliveredscheduleand gures in italics Output andprofi partlyestimated 44 But,theblandnessofconfi SHOTTON WIDESTRIPMILL tability, hotandcoldstripmill,Shotton TABLE 6 282 700 320 500 0 0 –3600 10 000 Report onSalesOrganisation dential reportsuggeststhetripwas 397 775 444 252 ts . Typescript Published by & (c) The Newcomen Society fl were insuffi deadlines. Therewerelogisticproblemstoo.Parsimonyincapitalspendingmeantthere diffi bearings fromSKFinSwedenalsodriedup. replacement sparepartsfromGermanywereunavailableduringwartimeandthesupplyof the diffi worn rollsuntilalocalUKsupplywasfound.Thesewereminorirritantscomparedwith mill werelostintransit,causingproblemsofrollshortageandexcessivewearonexisting the diffi of thehomeguard.Thehoneymoonwasover. end. Inaddition,therewerewartimeproblemsofblack-outs,air-raidsnearbyandtheclaims shortcomings atthehotstripmill.TheeasyrunofordersforAndersonshelterscametoan ing furnacesfi practices. ThelasttwoexperiencedAmericanrollermenleftinJune1940.slabreheat- ments onhotmillfi sensitive tofi for subsequentrollingonhandmillselsewhereinthecompany.Noneoftheseproductswas light plateforthewareffort,14gaugesheetsAndersonair-raidsheltersandrough with, hotstripmilloperationwasfairlystraight-forwardasthekeptbusyrolling in aReportbytheGeneralManagerReithGraytoBoardAugust1941. Wartime operationhaditsadvantages,butalsoposedkeyproblemswhicharedocumented profi tion oftechnicalinnovation witheffectivemanagementstructuresthatbringssustained Again, experienceatNucor’s stripmillatCrawfordsvillesince1989showsitisacombina- adopted AmericantechnologyatCornigliano,but technology, butUSmanagementpracticestoo.Ranieri reportsthatFinsidernotonly plant andgettingitrunning.Thepost-warmillatCornigliano inItalyacquirednotonlyUS Evidently, successfuladoptionofstripmilltechnology involvesmorethanbuyingnew Despite theseproblems,milloutputandprofi oor hamperedmovementofwheeledvehicles. culties withorderhandling,millscheduling,progresschasingandmeetingdelivery tability. relations, businessstrategyand organisation. carefully masteredandcopiedAmericanmethodsinmanagement culture,industrial were affectedandmanyofourtroublescanbetracedtothismalignant source. Materials, equipmentlabour,transportandallthethingsthatmakeforsmoothproduction we werecompelledtodothiswithalltherestrictionsandrestraintsthatwarimposed. inexperienced operativesanewandcompletelyrevolutionarymethodofproduction,yet I thinkitwillbeagreedthatnobodywouldchooseaperiodofwarinwhichtostartupwith The salesreporttotheboardinAugust1941concluded: The samewartimereporttotheboardofJohnSummersandSonsdetailsconsiderable Access toUSadvicewastruncatedbythereturnofpersonnelStatesand However, startupofthetandemcoldmillinApril1940begantoimposestrictrequire- culties ofsendingmanagerstoMestaforadvice.Replacementrollsthecold culties facedbySummers’earlierGermanbuiltSendzimir coldmill.Naturally, cient cranestomovematerialaroundthemillsthemselves,whilecollapseof nishing temperature. 48 red byproducergasprovedproblematic.Therewereminormaintenance nishing temperatures,whichcalledfor newandunfamiliaroperating SHOTTON WIDESTRIPMILL MANAGERIAL LEARNING WARTIME OPERATION 47 ts grew. 46 45 Tobegin 79 Published by & (c) The Newcomen Society 80 Rolls-Royce atnearbyCrewe. ther careerinRotherhamafteritwasfi engine waseventuallysoldsecond-handbyJohnSummers Sonsandwentontohaveafur- powered slabbingmillsatShottonsince1917,ifnotearlier. Thethreecylinderhorizontal full widthforthemill. of over40in.wasrequired. be broadsiderolledontheNo. 1 roughingstand ofthehotstripmillifafi Mesta slabbingmillcompletedin1940. fi thrift withtheexistingdrivesre-usedwhereverpossible. the coldmillin1955,whilere-motoringofdisplayedcharacteristicSummers Major modifi were added.AutomaticGaugeControlfollowed. mented andthenwasreplacedbyconventionaldowncoilers.Extrastandsnewmotors An extrareheatfurnacewasaddedin1946andthetroublesomeup-coilerfi mill productswere346 000 tonsin1945;612 000 tonsin1955;and1 267 Output oftheShottonhotstripmillexpanded,doublingeverytenyears:deliveries 000 tonsin1965. Shotton certainlydrewcomparisonwiththemillatEbbwVale. problems associatedwithtemperaturecontrol.Tradeunionnegotiationsoverpayratesat Gray soughtadvicefromEbbwValebothovermanagementpracticesandthemetallurgical there werebetweenNorthWalesandSouthWales. across theAtlanticbetweenShottonandPittsburgh(viaLiverpoolNewYork)than 3-stand tandemcoldmillatShotton. a Rolleronthe5-standTandemMillatRichardThomas’tohelptrainworkersfornew construction andcommissioning,althoughSummersrecruited‘ayoungmanwhohadbeen competition gaveabreathingspacetolearnnewmanagerialskills. on output,pricesfi Shotton from1928to1938.ButSummerswerefortunate.Pressingwartimedemands leaders. TheyhadsupportfromArmcointherunningofFineSheetDepartmentat John SummersandSonswereusedtobuyingmanagerialexpertisefromtheUStechnical wartime NorthAtlanticcutcommunicationswithAmericaatthetimeofcommissioning. USA, theydidnotpickuptheassociatedmanagerialtechniques.Thehazardsof to rollforjust16months. 1975 andcommissionedon 13 August1978inanadjacentbay.Thefi second-hand millwasbrought fromtheBritishSteelCorporation’sScunthorpeWorks in and Sonswerecertainlyfrugal intheiruseofcapital. nish high-qualitywidesheetforthecarindustry—presumablytofi The replacementslabbingmillstandstillusedthethree cylindersteamdrivethathad A newslabbingmillstandwassuppliedbyDavy-United in1950toreplacethelimited Curiously, anew80in.,singlestandreversingcoldmillwasbuiltattheworksto A fourthstandwasaddedtothecoldrollingmill,infrontofexistingstands,1948. Wartime pressuresseemtohavebroughtcooperationwithSouthWales.Reith There seemstohavebeenlittleinterchangewiththeirrivalsatEbbwValeduring While Shottonacquiredtechnicalknow-howonwidestripmilloperationfromthe The Davy-Unitedslabbing mill wasitselfreplacedaftershiftingonitsfoundations. A cations weresubsequentlymadetothefeedandmechanicaltake-offendsof xed bycontrolsatalevelthatallowedprofi 55 ThenewDavy-Unitedstandrolledheavierslabsupto their 53 SHOTTON WIDESTRIPMILL SUBSEQUENT HISTORY 50 nally supplantedatShottonin1957. Beforethewar,itseemstherewerecloserlinksbysea 54 WiththeoriginalMestaslabbingmill,slabshadto 51 52 t overcostsandlimited 49 nish importedcoilsfor nal slabbing millwas nished stripwidth 56 JohnSummers rt supple- rst Published by & (c) The Newcomen Society of its40 The Shottonhotwidestripmillrolledforjustoverfortyyears,closingwithinsixmonths quality withnewerGenerationIImills. had neitheracompetitivefeedstocknorthemodernequipmenttocompeteoncostsor Summers andSons’worksatSheltonduringthe1960s.Ultimately,stripmillShotton new technologiesofoxygensteelmakingandcontinuouscastingwenttotheotherJohn post-war developmentscheme,Shottonbecameafullyintegratedworks.Butfundsforthe furnaces andnewopenhearthstosupplythegrowingstripmilloutput.Asaresultof some 15yearslater. seven. Thenovelthree-quarterlayoutanticipatedtheemergenceofGenerationIVmill which wererelocatedtothefi three-quarter continuouslayoutbyMestaduring1960.Thisfreed-uptworoughingstands struction ofthewidehotstripmill.Theroughingtrainwasrebuiltfromfullycontinuousto The projecthistorywasreconstructed fromfi Summers andSons’otheroperations. rapidly. Themillwashighlyprofi mill wascompletedandcommissionedindiffi manager fromabroad.Theyalsoreliedonlocalsupply for70percentoftheproject.The shrewdly andbymaintainingsuperbcostcontrol.They recruitedanexperiencedproject ment. Summersthemselvescontrolledcostsbyspecifying aloweroutputmill,purchasing nology. TheUSmillbuilderplayedakeyroleinthetransfer ofknow-howaswellequip- strip milltechnology.However,theyminimisedtherisks, buyingtriedandtestedUStech- John SummersandSonstookadramaticdecisionwhen theypurchasedcontinuouswide documents inthearchivesof JohnSummersandSonsBritishSteelStripProducts at the fi tonnes ofslabsleftinstockattheendnationalsteelstrikewhichlastedthroughout million tonnesofstrip.Theendwasananti-climaxasthelasttasktoprocess85 000 potential differenceincostsafterexpansionthatswayedthedecision. was £42.43pertonnein1972/3,comparedwith£42.95atPortTalbot. costs betweenShottonandPortTalbot.Thestandardcostofahotrolledcoilat continuous casting.Atthetimeofdecision,therewasactuallylittledifferencein port, hadalreadyadoptedlarge-scaleoxygensteelmakingandwasideallyplacedtoadopt offered greaterpotentialforlongrunexpansionthanShotton.PortTalbothadadeepwater rst quarterof1980. Initial postwarexpansionattheShottonsitewasdevotedtocokeovens,blast The Summerstraditionofparsimonysupportedalimited,innovativebutthriftyrecon- Hot rollingceasedatShottonon23May1980afterthehotmillhadrolledover35 th birthday.TheBritishSteelCorporationdecidedin1974thatPortTalbotWorks 58 SHOTTON WIDESTRIPMILL nishing train,bringingthetotalnumberoffi ACKNOWLEDGEMENTS table almostfromthestartandunderpinnedJohn SUMMARY CLOSURE cult wartimecircumstances, butworkedup nancial accounts,photographs,fi 57 Ratheritwasthe nishing standsto ls and lms 81 Published by & (c) The Newcomen Society 82 ) ‘HawardenBridgeSteelWorks’, a) ThehistoryofShottonWorksJohnSummersandSonsthesubsequentsite 1. remaining faultsandviewsexpressedhere. and especiallyEricGeorge.Givensuchextensivehelp,Iamemphaticallytoblameforany Summers employeesclarifi editor of Chief EngineerofDavyMcKeeinSheffi Shotton andRonEllwood,formerChiefEngineeratShotton.EricEarnshaw, written twohistoriesofShottonsteelworkshimself. Shotton, gaveadviceonsources,havingplayedacrucialroleinpreservingdocumentsand provided awarmwelcome.GordonSmith,formerInformationServicesManagerat manager, RolfHolthöfer,andhisteampatientlytrackeddownrequestsformaterial Information ServiceManagerRayWellsgrantedaccesstotheRecordsCentre.Thecentre Corus Colors,Shotton.AllthephotographsarecopyrightColours, ) JohnSummersandSonsLimited, e) ‘ProductionofSheetSteel—InterestingEnterprises:No.32,JohnSummersandSons,Ltd’, c) b) Andrew Reid, d) 2 Steven Tolliday, BrianRedheadandSheilaGoodie, 2. i) h) Gordon P.S.Richards,‘TheHawardenBridge,Shotton,Chester,IronandsteelworksofMessrs. g) f) Smith, Gordon Smith, . JohnL.Young,‘Finishingequipment forwidestripmills’, 3. . Thereisavisualrecord of thisstop-gapinvestmentinPhotographstheSendzimirmillproject, 4. The authorreceivedextensivecommentsfromJohnBryant,formerWorksManager, are outlinedin Mass.: HarvardUP,1987. Sendzimir, Stahl undEisen location 0109961,abriefaccount inA.Nöll,‘Present-dayproblemsoftherollingmillindustry’, on 20November1936)Departmental Code426,consignment17,aisle5,departmentalbox 1, Sendzimir millinGermanyaround 1935,installationatShottonandcoldrollingofthefi (an untitledleathercoveredalbumofSummersphotosrelating tothetrialassemblyof November 1938,p.18. Edinburgh on29April1953), Mass Production May 1936,supplement, Trades Review Hawarden BridgeSteelworks:ATechnicalSurvey,SpecialIssueoftheIronandCoal 1987. Products, 1996and Corporation, December1980, John SummersandCo.’, Hedgcoe, Bridge Steelworks,1948–1958 Steel TimesInternational British Patent478,361 , vol. 56, 17September1936,pp. 1104–1106 andadescriptioninTadeusz Business, bankingandpolitics:thecaseofBritishsteel,1918–1939 , London:20April,1953(accountpublishedbeforethevisitofDuke Continuous Venture—TheStoryofaSteelWorks Full Circle:TheStoryofSteelmakingonDeeside , vol.25(4),April1949,pp.40–47,72–74, A CenturyofShottonSteel(1896–1996) edmanypointsofdetail,includingIanCugley,RonaldHodges NOTES ANDREFERENCES SHOTTON WIDESTRIPMILL Flintshire HistoricalSocietyJournal , wasaconsistentsourceofadviceandsupport.Former , January1959withcolourphotosbyAdolfMorathandJohn , ‘Improvementsintensionrolling methodandapparatus eld, savedmefromglaringerrorsandDrTimSmith, Wrexham LeaderandMold,Deeside&Buckley The SummersofShotton The StoryofDevelopmentandExpansionatHawarden Iron andSteelEngineer , London:Hodder&Stoughton, , London:BritishSteelStrip , vol.25,1971,pp.103–123, , Shotton,Cheshire,1948, , Shotton:BritishSteel , Cambridge, , vol.15(11), rst strip , 8 Published by & (c) The Newcomen Society 20. Kenneth Warren, 19. 18. 17. Richard F. Summers 16. Clayton M. Christensen, . Summersaccountoftheseeventsisreportedin 8. Tolliday, 7. RichardF.Summers, 6. Young, 5. 5 Itishardtoestablish preciselywhathappenedtotheextrafundsgiventhatmillitself 15. 14. See Redhead, 13. ibid. 12. Thissemi-continuousmediumwidthmillwasinstalledatNovosibirskMetallurgicalPlant, 11. F.C.BiggertJr.,‘Developmentsin4-highRollingMills’PaperreadbeforeIronandSteel 10. . OntheEbbwValemillseeT.J.EssandD.Kelly,‘Richard ThomasandCo.,Ltd.InstallsFirst 9. plant. before thewarbrokeout.Perhaps theFrenchstandwasconsideredforNo.1openhearth halted bywar.TheMestaslabbing millstandhadbeenordered,builtandshippedfromthe US 1970 suggestsSummersconsidered buyingasecondhandslabbingmillfromFrance,but was consignment 12,box5,locationF10763,fi Reversing SlabbingMill,withSparePartsandRolls. . . Summers &Sons,Limited,Shotton,Chester,England. . . MechanicalEquipmentforOne40” Proposal fromMestaMachineCompany,Pittsburgh,PA, USA 1November1938toJohn location F10763,fi with SparePartsandRolls. . Sons, Limited,Chester,England. . . MechanicalEquipment forOne40”ReversingSlabbingMill, Proposal fromMestaMachineCompany,Pittsburgh,PA,USA 30June1937toJohnSummers& fail John Summers&SonsLtd,July1940). F10765 (subsequentlypublishedwithillustrationsbyJonathanCapeandcirculatedprivatelyfor Sons, 1940.TypescriptextractatGroupSecretaryCode426,Consignment12,Box7,location The millwasinstalledinadisusedbayattheMarshMills. therefor’, applicationdate(inUnitedKingdom)16July1936,andpublished17January1938. op. cit.(2)pp.264–265conveystheconfusedstateofSummers fi fi million andbacktoapublishedvalueof£5.4(seefi considerable creativity.Thebalancesheetvalueofthefi cost £2.9million.DraftaccountsforJohnSummersandSonsin1937,example,show Iron andCoalTradesReviewop.cit. oldest survivingwideHSMinEurope’, strip milltoRussiaaroundthattime,seeL.I.Odyn,Y.N.BelobrovandD.A.Marchenko,‘The Machine Company,Pittsburgh,24August1977(inauthor’scollection).Unitedalsosoldawide of ContinuousandSemi-ContinuousHotStripMills Novosibirsk City,Russiaandstartedupin1941.SeeMestaMachineCompany, United Effort Division ofAmericanSocietyMechanicalEngineers,Youngstown,Ohio,November10,1927. Hereford: RecordPrintersforCorus,2002. RT1-15 andB.Caswell,J.GaydonM.Warrender, Wide StripMillinUnitedKingdom’, strip millpatentsintheUKJanuary1939. location 4748).ThisrevealstheyagreedtopayArmco£16 250 forthelicencetouseArmco’s payment madetoArmcoof£126 250 (GroupSecretaryCode425,consignment36, Sons LimiteddealingswithArmco,seekingadviceonthetaxtreatmentoffi John Summers&SonsLimited,exparteTheCompany nancial accounts,GroupSecretaryCode425,consignment36,box6,location4747).Tolliday , Boston,Mass.:HarvardBusinessSchoolPress,1997. op. cit. op. cit. , vol.7(11),November1927,pp.3–7. op. cit. (3),pp.18–19. (2)p.147. le C. The BritishIronandSteelSheet Industrysince1840 (1i),pp.130–131andTolliday,op.cit.(2),chapter10. op. cit. The NewMill The Innovator’sDilemma:Whennewtechnologiescausegreat fi SHOTTON WIDESTRIPMILL (6). , PressandInformationCode426,consignment12,box5, (1d),p.10. , 1940 Iron andSteelEngineer Steel Times le C. , Typescript,HawardenBridge:JohnSummersand Case forCounseltoAdvise,17March,1941re. , typescriptsuppliedbyP.H.Fassett,Mesta , vol.229(7&8),July/August2001,p.262. , atypescriptaccountofJohnSummers& rm oscillatesfrom£5.3millionto£6.1 Ebbw Vale:‘TheWorks’1790–2002 le , PressandInformationCode426, , Supplement,16(7),July1939, 1938 JS&SLtd nances exactly. , London:G.Bell&Sons, , manuscriptof Reference List nal severance nal rms to rms 83 , Published by & (c) The Newcomen Society 84 6 J.H.Pearce, 46. 45. A. Reith Gray, 44. Sheet Trade Board, 43. File 21. T. J. Ess, 42. BrownisC.J.or‘Brownie’,oneoftheUScommissioning crewwhosephotographis 41. 23. Badlam, StephenBadlam,‘Developmentsinrolling fl 22. 40. Richard F. Summers, 39. A. ‘Mesta’smisstepssevertiestoproudpast’, Reith 38. Gray, ‘Steamshipbatteredingale’, 37 26. Andrew Reid, TheInternationalConstructionCompanyLtd., 25. 24. Ess, 6 SirGilbertTMorganandD.Pratt, 36. 35. Richard F. Summers, F.E.Leahy,‘Comparisonofsoakingpitdesigns’, 33. Demagdidnotbeginmarketingtheir‘FirststraightcontinuousWideStripMillinEurope– 32. 31. T. J. 30. ibid. Ess, 29. 28. T.J.EssandD.Kelly,op.cit.(9). 27. 7 RuggeroRanieri,‘Learningfrom America:TheremodellingofItaly’spublicsectorsteel industry 47. 34. Strip Mill fi consignment 13,box2,location3527. 1938 425 consignment36,box3,location4744. Pittsburgh Post-Gazette,op.cit. included inthepublishedversionofRichardF.Summers,op. cit.(6). Engineer vol. 28,pp.118–122. London: EdwardArnold,1938,pp.208–209. 1938, GroupSecretaryCode425,consignment40,box1,location8055. Summers &SonsLtd,toTheBankers’IndustrialDevelopmentCo.Ltd hot stripmills. fi August 1941,GroupSecretaryCode426,consignment12,box 7,locationF10765,fi Mesta contractandanote‘Lossondollars’probablydated30August1939. some 85sheetslistingeverysinglepaymenttodate.Therearealsodetailsofpaymentsunderthe Secretary Code425,consignment36,box6,location4747,handwrittenmanuscripts,including July 1939.Blisssoldonehotmill–thefi German product’untilsummer1939.Seetheiradvertisementin Development Association,1959. Electric MotorsandControls,ElectricityProductivitySeries,no.3 6758. Ovesen, StripMillEngineering,Departmentalcode426;consignment4;boxno.3;location Requisition, TheEnglishElectricCo.Ltd.,Stafford,4November1937 Transfer ofUSManagementModels in the1950sand1960s’Matthias KippingandOveBjarnar(eds.), 426, Consignment12,box7,location F10765,fi 1941, pp.64–71. le 10. Summaries ofResultsOctober1942toDecember1944 op. cit. , SouvenirCopy,Marked‘PrivateandConfi , 16(1),January1939,31;T.J.Ess,op.cit.(21),millvol.24,pp.96–101. op. cit. The ModernStripMill op. cit. (21),p.7,table3. l including le Report onSalesOrganisation op. cit. John Summers&SonsLimitedGeneralWorksReport,August1941 (22),pp.33–34andT.J.Ess,op.cit.(21)section1discussthelayoutofUSwide op. cit. (21),p.162. Report ofDelegationtotheUnitedStatesAmerica–16March to11April, (16);JohnSummersandSonsLimited,op.cit.(1c). op. cit. op. cit. (39). Strip MillConstruction,revisedestimate28February1940 SHOTTON WIDESTRIPMILL (6). (6). Liverpool DailyPost (38),p.21. , Pittsburgh:AssociationofIronandSteelEngineers,1941,mill , London:RoutledgeStudiesin BusinessHistory,1998. nishing trainforFairfi British ChemicalIndustry:ItsRiseandDevelopment Pittsburgh Post-Gazette . Typescript9August1941,Group SecretaryCode at steelproductsin1937–38’, le boxE,fi Report onStripMillInstallationforMessrs.John , 30March1939,5&12. Iron andSteelEngineer dential’, GroupSecretaryCode425, (actually1946),Typescriptfi le 10. eld, Alabamaworks. Iron andSteelEngineer , 10January1984,pp.1,21. , London:BritishElectrical The MarshallPlanandthe , London:22December Requestedby:MrH. vl 1() August vol.18(8), , Iron andSteel . Typescript9 le boxE, le, Code , Group , 16(7), , Published by & (c) The Newcomen Society 57. British Steel Corporation, 56. George Watkins, 55. 9 Foradetailedaccountofwartimeallocationsteeloutput,pricefi 49. 48. Richard Preston, 58. Gordon Smith, 4 IronandCoalTradesReview,‘New 42-in. slabbingmillatShotton’, 54. IronandSteel,‘Increasingstripcapacity— new coldreductionreversingwidestripmill’, 53. 2 W.H.Corlett,S.A.LewittandR.Gray,‘Someengineeringoperatingproblemsinthe 52. Warrenop.cit(20)suggeststhatagreementwasmadewithRichardThomasatEbbwValefora 51. 50. A. Reith Gray, Derbyshire: Moorland,1979. Mills, Typescript,GroupSecretaryCode426,consignment13,box12,location11/2577. I. J.F.Buchanan(ServiceS.&S) York: PrenticeHall,1991. Secretary Code426,consignment12,box15,locationF10773. collection. Review and Steel London: TheIronandSteelInstitute,1960,SpecialReport67,pp.112–117. conversion ofatandemcold-reductionmilltohigherspeed’,in in Wales. their addressas‘HawardenBridge,Chester,England’eventhoughtheworksisclearlylocated centre ofgravitywasoverwhelminglyPittsburgh.NotethatJohnSummersandSonsalwaysgave degree ofcooperativerunningthetwomills.Wefi Papers the supplyofscrapseeG.D.N.Worswick,‘BritishRawMaterialControls’, , vol.6,April1942,pp.2–14. , 8December1959,pp.869–873. , 19(3),December1946,pp.793–795. op. cit. Hot rollingendsatShotton American Steel:HotMetalMenandtheResurrectionofRustBelt The SteamEngineinIndustry—2,MiningandMetalTrades (39),p.2. SHOTTON WIDESTRIPMILL Shotton Brochure (circa1972),CareerreminiscencesofthePlantManagerHot , manuscriptpressrelease,23May1980,Group , London:BSC,circa1972,mimeo.,author’s nd noevidenceforthis.ForSummers,the Production ofWideSteelStrip xing andarrangementsfor Iron andCoalTrades Oxford Economic , Ashbourne, , New Iron 85 , 6. Development of computer applications in the iron and steel industry

Jonathan Aylen, "Megabytes for Metals – The development of computer applications in the iron and steel industry", Ironmaking and Steelmaking, vol. 31, no. 6, 2004, pp.465- 478

Megabytes for metals: development of computer applications in the iron and steel industry

Jonathan Aylen*1

The steel industry pioneered the use of computers for process control. By the mid 1960s, almost a fifth of the world’s process control computers were installed in the steel industry. The present paper documents the development of direct digital control with emphasis on hot strip mill control, notably the installation at Llanwern using a GE 412 computer. Early applications of computers in areas such as electric arc furnace control and order handling are identified. Archive sources, government documents, interviews, correspondence and technical papers show the leading role of steel in developing online control. Marked differences in adoption rates are identified. Two- thirds of the early steel installations were in the USA. Britain and Italy were also early adopters. Jones & Laughlin and Inland of the USA, the Steel Company of Wales and Italsider were among the leading innovators.

Keywords: Steel industry, Computers, Development, Process control I&S/1883

Control computers may someday be applied in possible. This extends to current developments: notably almost all of the processes found in an integrated through process modelling of the whole sequence of steel mill (Stout and Roberts, writing in 1960)1 steelmaking processes and online quality prediction. Precursors Computers and steelmaking: a symbiotic Just as there were precursors to modern steelmaking, relationship such as crucible steel in the 19th century, there were also Steel and computers grew up together. Steelmaking is practical precursors to computing, notably Jacquard older than is often supposed, and computers are looms, invented in 1801, and mechanical developments certainly older than their modern image suggests. To by Babbage from 1822 onwards. Hollerith’s punched put things in perspective, computers are the same age as cards and automatic tabulating machines were used for the radical innovation of oxygen steelmaking that the American census of 1890. Zuse, the son of a Berlin rapidly became the predominant steelmaking process civil servant, built a mechanical digital computer in his during the 1960s. parent’s living room in 1938. In 1943, Turing’s team at Bletchley Park built the Colossus electronic computer, The post-Second World War boom in steelmaking 2 coincided with the rise in practical computing. The steel using vacuum tubes, for code breaking. industry played a pioneering role in developing applica- It is hard to say where computer control of steel tions for the new computer technology. New capital manufacture began. Punched cards were used for mill equipment called for new control techniques, forcing control before computers. The first ‘computer con- developments in machine drives and controls, sensors trolled’ installations used one punched data input card per slab. This card contained all the information and data acquisition technologies, all vital to exploita- required by the computer to roll one slab. Arguably tion of the growing power of electronic computing. the first card controlled mill was the universal slabbing Computer based innovation continues in steel, as cheap mill feeding the Aliquippa continuous hot strip mill of computer power has made large scale modelling and Jones & Laughlin.3 This pioneer installation had three comprehensive data capture, analysis and storage distinctive features: the use of digital control; storage of the rolling schedule in a memory; and the first use of

1 transistors in steel mill operation. Westinghouse Electric PREST, Harold Hankins Building, Manchester Business School, Booth Street West, Manchester M13 9QH, UK were responsible for both the drive motors and the The paper was written while the author was at the Centre for Manufacture, control system. University of Manchester Institute of Science and Technology (UMIST). It is based on a keynote address to the Sheffield Metallurgical and The same approach was later used at Armco’s plate Engineering Association conference on ‘Automation in metals processing’, mill at Houston, but here a computer stored the rolling held at Sheffield University, June 2002 model equations which predicted separating force, mill *Corresponding author, e-mail [email protected] spring and screwdown settings for each pass using a

ß 2004 Institute of Materials, Minerals and Mining Published by Maney on behalf of the Institute Received 13 July 2004; accepted 19 August 2004 DOI 10.1179/030192304225019324 Ironmaking and Steelmaking 2004 VOL 31 NO 6 465 Aylen Megabytes for metals

1 British Iron and Steel Research Association (BISRA) promoted computer use in UK steel industry from 1957 although their initial focus was on greater business efﬁciency: courtesy Corus Colors

Westinghouse Prodac 4449.4 Part of the efficiency gain Pegasus 1 suitable for operations research purposes was from computer operation arose from meticulous produc- quoted to the Steel Company of Wales at £49 450 in July tion planning and scheduling imposed by the time taken 1959.10 to load punched cards and the need to follow a preset Among the first business computers in world steel was sequence within each batch of cards. This installation was the Leo II started at Stewarts & Lloyds Ltd, Corby in widely replicated, for instance at Italsider’s Taranto June 1958.11 This was the first Leo II computer to be works,5 the Sollac, Se´re´mange slabbing mill in France installed outside Lyons. Its main function was payroll and Dofasco’s Hamilton works slabbing mill. calculations, but it was also used for stress calculations United States developments in card control were on tubes and operations research applications in iron watched closely elsewhere in world steel. They alerted ore mining.12 It ran until 1971. steelmakers to the potential of digital control. A Steel By 1958, there were a dozen different types of Company of Wales letter responding to the press computer available in the UK, more than in the USA announcement of the Aliquippa slabbing mill by and Europe.13 At that stage, business applications Westinghouse says, ‘If we hurry we might at least be predominated, accounting for 47 of the 125 computers the first to put it on the hot strip mill.’6 installed in the UK (Table 1). Typical industrial applications included the statistical analysis of sales Pioneer computer use in UK and the allocation of can production across factories using linear programming at Metal Box;15 the statistical As early as 1953, the Ferranti computer at Manchester analysis of Customs and Excise import data for the University was used by the British Iron and Steel Cotton Board on a Ferranti Pegasus at Imperial Research Association for statistical analysis of blast Chemical Industries (ICI) Dyestuffs, Harpurhey, furnace behaviour.7 In 1957 the British Iron and Steel Manchester;16 and recording prices for paints at ICI Research Association (BISRA) issued a detailed 24 page using an IBM 650.17 publicity booklet Computers and Steel to steelmakers to explain how computers work in simple terms (Fig. 1).8 This promoted the new BISRA Computer Applications First process control computers in steel Section, set up within the Association’s Operational The iron and steel industry was at the forefront of the Research Department, and reported that BISRA had application of computers for control purposes. The USA decided to buy a Ferranti Pegasus computer which was led the way, reflecting the dominance of the American delivered in 1957. The main focus of the booklet was on steel industry in the 1960s and the availability of standard payroll applications pioneered elsewhere in the UK, at computers suitable for industrial application. Approxi- Lyons Bakeries.9 At the time, the price of a Ferranti mately two-thirds of installations listed in Table 2 were

Table 1 Early applications of UK built computers,* 1958

Use Installed in UK ‘Installed overseas’

Engineering design 32 3 Other mathematical work, research statistics and university mathematical laboratories 27 8 Service bureaux, testing of customer programs, etc. 19 … Business applications: accounting, payrolls, stores records and management statistics 47 1 Total UK computers 125 12 *Source: Gearing,14 Table 2.

466 Ironmaking and Steelmaking 2004 VOL 31 NO 6 Aylen Megabytes for metals

Table 2 World stock of process control computers for steel applications,*{ January 1964: total number 62

Firm, location Use Type of computer Manufacturer

Power generation (total 4) Bethlehem Steel, Lackawanna, USA Operating guidance GE 312 General Electric Italsider S.p.A., Taranto, Italy Utilities dispatch CEA 2801 Construzione Electro Meccaniche Annettoni (Italy) Bethlehem Steel, Burns Harbor, USA Operating guidance H 610 Honeywell Inc. Midwest Steel Corp., Portage, Indiana, USA Plant demand Bailey 756 Bailey Meter Co. Burden preparation and coke ovens (total 2) Youngstown Sheet & Tube, Sinter data logging GE 312 General Electric East Chicago, Indiana, USA S.A. Forges de la Providence, Belgium Sinter log ACEC 51 Ironmaking (total 4) Inland Steel, Indiana Harbor, USA Data logging IBM 1710 IBM United States Steel, Homestead, USA Data logging Fox 97600/PDP4 Foxboro/DEC Hoogovens, IJmuiden, The Netherlands Data logging Elliott Arch Elliott Automation Nippon Kokan KK, Japan Data logging Hokushin Hokushin Electric Steelmaking: electric arc (total 5) Lukens Steel Co., Coatesville, USA Power demand IBM 1710 IBM Steel, Peech & Tozer, Rotherham, UK Load management Argus 108 Ferranti Armco, Kansas City, USA Power demand Bailey 760 Bailey Meter Co. Interlake Steel, Ohio, USA Power demand Bailey 760 Bailey Meter Co. Timken Roller Bearing Co., Ohio, USA Power demand Bailey 760 Bailey Meter Co. Steelmaking: basic oxygen (total 13) Great Lakes Steel, Detroit, USA Guidance TRW 330 Bunker-Ramo Bethlehem Steel, Lackawanna, USA Control oxygen, lance GE 412 General Electric Sharon Steel, Farrell, USA Control oxygen, lance H 290 Honeywell Ford Motor, Steel Division, Dearborn, USA H 610 Honeywell United States Steel, Duquesne, USA Guidance TRW 340 Bunker-Ramo Fuji Iron and Steel, Japan TRW 330 Bunker-Ramo Nippon Kokan KK, Japan Guidance HOC-300E Hokushin Electric GKN, Lysaght’s, Scunthorpe, UK Guidance KDN 2 English Electric Usinor, Denain, France Guidance RW 300 Bunker-Ramo/CAE Italsider S.p.A., Taranto, Italy Guidance (charge CAE 510 Comp. Europe´ene calculation, data d’Automatisme logging) Electronique (France) Italsider S.p.A., Bagnoli, Italy Guidance CAE 510 Comp. Europe´ene d’Automatisme Electronique (France) Armco, Ashland, USA Guidance IBM 1620 IBM Jones & Laughlin, Cleveland, USA Guidance TRW 330 Bunker-Ramo Reversing hot mills (total 10, one nickel) Lukens Steel Co., Coatesville, USA Plate mill GE Directomatic General Electric Armco, Kansas City, USA Plate mill/slabbing mill Prodac P 4449 Westinghouse Crucible Steel, Midland, Pennsylvania, USA Rougher to strip mill Prodac P 4449 Westinghouse Republic Steel, Gadsden, USA Plate mill screwdown Prodac P 4449 Westinghouse Inland Steel, Indiana Harbor, USA Bloom/billet mill GEPAC 4000 General Electric United States Steel, Gary, USA Plate mill scheduling Prodac P 4449 Westinghouse Italsider S.p.A., Taranto, Italy Slab mill scheduling IBM 1460 IBM Italsider S.p.A., Taranto, Italy Plate mill control Prodac P 500 Westinghouse Armco Steel, Middletown, USA Blooming mill IBM 1710 IBM Inco Combination mill IBM 1710 IBM Hot strip mills (total 10, one aluminium) Great Lakes Steel, Detroit, USA Control Daystrom 136 Control Data McLouth Steel Corp., Trenton, USA Control setup finishing GE 312 General Electric Inland Steel, Indiana Harbor, USA Control Prodac 580 Westinghouse Wheeling Steel Corp., Control Prodac 500 Westinghouse RTB, Spencer Works, Llanwern, Wales Control GE 412 General Electric Hoesch AG, Dortmund, Germany Scheduling GE 412 General Electric Steel Company of Wales, Port Talbot, Wales Control GE 412A General Electric Bochumer Verein, Germany Scheduling GE 412 General Electric Bochumer Verein, Germany Control GE 412 General Electric Alcoa, Warwick, Indiana, USA Control Prodac 580 Westinghouse Cold strip mills (total 6) United States Steel, Fairfield, USA 6 stand control GE 412 General Electric Steel Company of Wales, Port Talbot, Wales 4 stand control TRW 330 Bunker-Ramo Inland Steel, Indiana Harbor, USA 3 stand data logging GE 312 General Electric Atlas Steel, Welland, Ontario, Canada Z mill IBM 1710 IBM Mannesmann AG, Germany Sheduling GE 412 General Electric United States Steel, Gary, USA 6 stand control Prodac 580 Westinghouse Processing lines and long product rolling mills (total 8) Inland Steel, Indiana Harbor, USA Galvanising line IBM 1710 IBM Jones & Laughlin, Aliquippa, USA Tinning line RCA 110 Radio Corp. of America Jones & Laughlin, Aliquippa, USA Tinning line GE 412 General Electric

Ironmaking and Steelmaking 2004 VOL 31 NO 6 467 Aylen Megabytes for metals

Table 2 World stock of process control computers for steel applications,*{ January 1964: total number 62 (continued)

Firm, location Use Type of computer Manufacturer

Jones & Laughlin, Aliquippa, USA Annealing line control GE 312 General Electric Kaiser Steel, Fontana, USA Tinning line GE 312 General Electric United States Steel, Pittsburgh, PA, USA Tinning line GE 412 General Electric Samuel Fox, Stocksbridge, UK Billet cutting Elliott 803 Elliott Automation Shelton Iron & Steel, Stoke, UK Beam cutting KDN 2 English Electric *Sources: developed from Kirkland,18 p.116; Control Engineering,19 pp.78–79; Journal of Metals.20 {Possible omissions include one Swedish process control computer, one French research computer and Italsider, Piombino research computer. Please point out other omissions. Status 1964. Some locations had already upgraded from earlier machines. Excludes hard wired card control.

located in the USA. Steelmakers employed second them in petroleum and chemicals, yet supplied less than generation computers which used transistors as a basis 10% of process computers in the metals sector. for reliable mainframe computing. They exploited The UK steel industry was also a leader in the magnetic core memory storage invented by Forrester application of computer control to steelmaking, despite in 1950. (Pioneering British computers relied upon the small size, fragmented nature and scientific bias of cumbersome delay lines – precisely machined mercury the British computer industry;25 UK steel accounts for filled tubes – for data storage.) seven of the 60 or so early steel industry computers listed The journal Control Engineering published a sequence in Table 2. The publicly owned Italian steelmaker, of surveys on process computer use at 18 monthly Italsider, played a pioneering role using Italian, French intervals in the 1960s, listing every digital process and American computers.20 In terms of companies, control computer in the world (Tables 3 and 4). There Jones & Laughlin and Inland of the USA, the Steel are gaps in their lists owing to the omission of a few Company of Wales and Italsider were among the leading International Business Machines (IBM) computers innovators, judging by the listings in Table 2. installed outside the USA,25 but the company was not The Japanese steel industry adopted computers later, the major player it became in the late 1960s – far from it. with early installations confined to three advisory These surveys show that four industrial sectors domi- systems for blast furnaces and oxygen steelmaking. nated early industrial computer use: power generation; For some reason, France stood on the sidelines, with petroleum and chemicals; metals, effectively steel; and a just one basic oxygen steelmaking (BOS) shop installa- set of miscellaneous high technology engineering section at Usinor by 1965 using American knowhow from tors.26 The USA led in industrial application of Bunker-Ramo,27 although the Institut Recherche computers with almost 400 by 1965, compared with 54 Siderurgie (IRSID) was active in experimental in the UK, 39 in France, 13 in Italy and 12 in Japan. Of research.28 As late as 1969, the collaborative research the world total, 100 were involved in iron- and institute Centre National de Recherche Me´tallurgique steelmaking, or 18% of the total market for control (CNRM) in Belgium also presented an explicitly computers. negative view of automation in steel.29 General Electric (GE) were the leading computer Evidently, adoption of computer control for the steel suppliers to steel. By 1965, GE had 25% of the world plant was not a forgone conclusion even by the end of market for process control computers in metals, the 1960s. Marked differences between Italian and followed by Westinghouse with 17%. Bunker-Ramo French adoption rates are striking. Both steel industries was pre-eminent across all sectors in process control pursued state sponsored expansion schemes charac- with over 100 installations worldwide by 1965, half of terised by heavy investment in new steelworks.

Table 3 Number of process control computers, all applications,* 1963–67

USA UK France West Germany Italy Japan World total

Sept. 1963 237 34 27 4 4 9 340 March 1965 396 54 39 9 13 12 565 Sept. 1966 761 129 95 22 14 71 1325 March 1967 866 132 122 58 ,20 99 1571 *Sources: Control Engineering.19,21–23 Table 4 Number of process control computers by industry,*{ 1963–68

Miscellaneous Petroleum, chemical, Metals (almost (e.g. astronomy, paper, food, cement all steel) Power nuclear, space) World total

Sept. 1963 92 55 117 76 340 March 1965 166 106 161 132 565 Sept. 1966 336 242 289 485 1325 July 1968 .632 397 504 781 2890 *Sources: Control Engineering.19,21,22,24 {Last survey date excludes minicomputers, but has wider coverage in miscellaneous category. No separate data available for food industry in 1968.

468 Ironmaking and Steelmaking 2004 VOL 31 NO 6 Aylen Megabytes for metals

Contrasting experience across countries shows that The solid state circuitry of a GE 312 computer was adoption of computer control was not an automatic composed of 2500 diodes, 2500 transistors and 12 000 consequence of capital investment during the 1960s. resistors, but no magnetic core memory. There were 20 Rather, it required a conscious decision to develop binary digits (bits) per word or per instruction. All computer control in place of manual operation. arithmetic was fixed point. Numbers were 19 bits plus Management enthusiasm for computer control helps to the associated positive or negative sign, not a very big explain why UK steel rapidly adopted computerisation number range when expressed in decimal form, just at a time when the industry was otherwise slow to take 2524 287 to z524 287.37 The GE 312 was designed by up process innovations such as oxygen steelmaking and A. Spielberg of the GE Computer Department newly continuous casting.30 formed in 1957.38 He had been recruited from Radio Corp. of America (RCA) to head the Process Control 39 Supervisory role of early computers Engineering section of GE. Early computers were limited to an advisory role. Llanwern hot strip mill: pioneering Among the first applications of a computer to any iron- and steelmaking process was the use of a GE 312 installation computer for data logging on the sinter plant of the The first successful use of a computer for complete mill Youngstown Sheet and Tube Co. at Indiana Harbor in control was the new generation II hot strip mill at 1959.31 The system was handicapped by the problem of Richard Thomas and Baldwin’s Spencer steelworks, obtaining representative samples for analysis – an early Llanwern, near Newport in Wales, which introduced example of the recurring problem of gaining accurate direct digital control of the whole mill in 1964.35,40,41 and reliable signals from sensors. The preprogrammed The main functions of the process control computer GE 312 GARDE system sold by General Electric was were initial setup, active operation and adjustment explicitly designed not to implement control actions. Its during rolling, sequence control of slabs and coils supervisory role is emphasised by the GARDE acronym: through the mill and logging of production. The ‘Gathers Alarms Records Displays Evaluates’. Control computer was also meant to optimise mill performance, was left to human operators. but here it was less successful. In general, sensing devices were not reliable enough to The installation at Llanwern used an American built allow direct online process control at the ironmaking GE 412 digital computer to control the whole mill and steelmaking stage. Ambitious early attempts at blast (Figs. 2 and 3). The GE 412 was essentially the earlier furnace control in The Netherlands concluded that GE 312 computer with a core memory added.42 The 412 development of better measurement instruments was was specifically designed for process logging, monitoring crucial.32 A GE representative discussing the possibility and control in applications such as electric power plants of online control of blast furnaces said bluntly at the and manufacturing.43 Computer makers in the USA time, ‘measurements of internal conditions within the were far more focused on commercial and industrial furnace are not yet satisfactory.’33 applications at the time, and helped by a large domestic market for industrial equipment. The Llanwern computer was installed in a room First online computer control: McLouth alongside the mill pulpit in air conditioned cabinets. It at Trenton was a stored program computer with an 8192 word high speed magnetic core storage and a 57 344 word magnetic Online computer control of steelmaking processes drum ‘bulk storage’ backup. Again, it had solid state became a reality with the first use of computers on a circuitry. The limiting factor was the design and cost of hot strip mill in 1962. McLouth, at Trenton, Michigan memory. In common with early computers, it had a used a GE 312 computer for gauge control on the magnetic core memory based on thousands of doughnut finishing train of a semicontinuous mill. The aim was to shaped ferrite cores.44 It is a tribute to the software set up the initial roll gap and then establish correct programmers that a complete hot strip mill could be run gauge as soon as the head end of the strip emerged onto on a computer with processing power equivalent to a 34,35 the runout table. The finishing train started running modern pocket calculator. under continuous computer control on 1 November The use of core memory at Llanwern seems to have 1962. H. Oldfield, General Manager of the GE been crucial. The first practical attempt at computerised 36 Computer Department, recalls: setup of a tandem mill in the UK was at Port Talbot in ‘Probably the most exciting application of the GE 312 1962 on their four stand cold mill, with the installation was to the hot strip mill of McLouth Steel Co. in of a TRW 330 computer which had a 28 bit word length Michigan. It was a difficult design inasmuch as each step and a 48k drum, but no core store, the equivalent of in the process had to be varied on the basis of the random access memory (RAM) on modern computers.45 measured values of the previous step. This required Thus, all computations had to rely on reading the continuous high speed feedback to set the five different programs from a drum and then carrying out the hot stands with absolute accuracy and reliability being necessary manipulations of the data on the drum also. essential; an error at one point could be magnified at the Much thought was given to trying to optimise the next, causing the entire process to go out of control. location of programs and data on the drum to minimise Fortunately, the GE 312 met the challenge.’ the latency time while waiting for the drum to rotate. As Hence, direct digital control of wide strip mills a result, there were numerous commissioning difficulties. became the first successful full scale application of Similar problems dogged the first use of computers on computer control in steel production. a process line. Jones & Laughlin installed a computer on

Ironmaking and Steelmaking 2004 VOL 31 NO 6 469 Aylen Megabytes for metals

2 General Electric (GE) 412 computer was mainstay of hot and cold strip mill computer control worldwide during early 1960s: courtesy K. Morgan

3 GE 412 computer that ran whole of Llanwern hot strip mill from 1964 relied on careful programming to overcome low computer power: courtesy K. Morgan

470 Ironmaking and Steelmaking 2004 VOL 31 NO 6 Aylen Megabytes for metals their Aliquippa works continuous annealing line in 1960. Contract details of these early computers have not This was a GE 312. Its functions were material tracking, been traced. It is known that General Electric sold its process monitoring, process control and production simpler 312 GARDE system for $US275 000 at the time, reporting. Progress in control was hampered by an plus $US25 000 for the basic peripherals.50 General inadequately sized memory drum and closed loop Electric also sold 8192 word location memories for control was not achieved until 1963.46 $62 400 in 1964. Hence, it is reasonable to suppose that Hence, a key breakthrough at the Llanwern hot strip the GE 412 computer for Llanwern cost Richard mill was the ability to respond to signals immediately Thomas and Baldwins in excess of $US400 000 after using core memory. The man responsible for adding the paying import duty – some £143 000 at the ruling core memory to the GE 412 was Spielberg, prior to his exchange rates. Cartwright51 gives a budget of £2.4m for departure for IBM in September 1963. (He has since the whole Port Talbot scheme, but two-thirds of this was become more famous for coaching his son in the art of allocated for upgrading the mill with a new edger, making home movies.39) loopers, screwdowns and drive control on the hot strip The aims of the Llanwern computer were: mill to improve control. (i) to express rolling relationships mathematically South Wales had one of the densest concentrations of (ii) to set up the mill using computer signals computer control in world steel by the end of the 1960s. (iii) to track slabs as they progressed through the This was no accident. The cluster of computer applica- mill tions owes its success to two factors. First, there was an (iv) to respond to sensors while the mill was rolling interchange of expertise and personnel. Knowledge of (v) to give output corrections to actuators process control techniques and programming skills spread (vi) to log production. rapidly in this environment as key personnel moved from 52 Computer control at Llanwern was comprehensive: It not job to job. As another veteran, A. Foss, expressed it, ‘In only controlled mill setup and operation on the roughing those days few people knew about computers and and finishing train, but it also controlled the reheat everyone was self-taught.’ Hence, networks of likeminded furnaces and coiler (Fig. 3). Sensors tracked the position people helped diffusion of the new technology. of material as it flowed through the mill, and measured Second, senior management were champions of speed, gauge, width and temperature. In response, roll computerisation. The Chief Executive of the Steel gaps, edger positions and side guides, mill speed and Company of Wales, F. W. Cartwright, took a personal cooling sprays were controlled. Recall too, generation II interest in computers. He chaired internal meetings mills were designed for zoom rolling whereby the strip promoting computer use. Taking one well documented was progressively accelerated as it passed through the instance, the Steel Company of Wales (SCOW) held an finishing train once the head end had been threaded onto internal afternoon conference on electronic digital the coiler. All in all, the computer received 200 analog and computers on 10 March 1959 for senior staff, chaired 75 digital signals and maintained 300 output contacts and by Cartwright.53 The participants watched films on Leo 150 other various digital outputs. and Pegasus computers and heard four short talks on Commissioning took place in stages. The computer equipment, construction, programming and the eco- started running in February 1963. The automatic crop nomics of computer application followed by questions shear started in July 1963 and gauge control in October to an expert panel. This was a typical event, led by the 1963. But it took another year before slab tracking, directors, to spread knowledge of computers through the logging and all the mill setups and temperature controls company. Someone with specific responsibility for were fully operational, by October 1964. automation gave regular monthly reports to the Computer hardware proved surprisingly reliable. SCOW board. Cartwright himself gave the fourth During the first 18 months of operation the computer annual lecture of the UK Automation Council in 1964.51 was available for more than 99.9% of operating time. South Wales spawned a further breakthrough in Instead, it was the conventional electromechanical computer control. In 1966, a consortium was established features of the mill that caused problems, along with to develop a rolling mill model involving the Ministry of shortcomings in the pioneering software. Technology, the electrical company Associated Electrical Industries (AEI), Imperial College and the Steel Company of Wales. A successful model was developed between 1968 Computer developments in South Wales and 1971 and applied to the revamped cold tandem mill at Port Talbot works.54 The model was then used at the new The Llanwern installation was replicated at Port Talbot using a GE 412A.47 Full computer control commenced Shotton five stand tandem mill in North Wales, and 48 eventually on almost every hot and cold strip mill in the there in September 1966, apart from mill pacing. The 55 only key difference between the two installations was British Steel Corp. The team included Professor that Llanwern tracked the slabs through the reheat G. Bryant and J. Edwards at Imperial College and furnaces, whereas computer control at Port Talbot D. Harvey, K. Edwards and R. Griffiths at the Steel began with the roughing train. A participant of the Company of Wales. A former joint chief executive of time, K. Morgan,49 recorded his experiences developing Corus observes that this initiative was ‘quite a remarkable exercise in transforming the automation of rolling mills – ‘hands free’ computer control at Port Talbot. The 56 system operated using GE 412-PAL – the General not just in the UK but worldwide’. Electric ‘program application language’. Data for the rolling schedule for a complete shift was input on punched cards. The same team of computer personnel Shortcomings of early computers then progressed to the new tinplate works of the Steel Computer control worked well in the UK. The GE 412 Company of Wales at Trostre. computer was a very successful system which performed

Ironmaking and Steelmaking 2004 VOL 31 NO 6 471 Aylen Megabytes for metals

as planned, apart from the mill pacing function which did not work. Mill pacing was meant to maximise throughput without damaging either the mill equipment or the strip itself. The failure of mill pacing led to hardware reconfiguration and local development of software to improve mill output, reducing delay time between successive bars, for example. These problems were common to Llanwern and Port Talbot. However, otherwise, strip mill control was highly successful in Britain. By the end of 1969, five systems were installed on UK hot strip mills (Llanwern; Port Talbot; Ravenscraig; Lackenby coil plate mill for gauge control; and Brinsworth narrow hot strip mill). Computer control of hot strip mills was not such an unequivocal success in the USA. A key American survey was highly critical of the limited results achieved after the first 5 years of computer operation of hot strip 4 IBM 360 series was widely used for management tasks mills.57 There were a total of 10 attempts to introduce by late 1960s: this IBM 360 was installed in a purpose direct digital control on new generation II hot strip mills built steel building at John Summers and Sons, in the USA, but in contrast with the UK these enjoyed Shotton Works: courtesy Corus Colors only mixed success.58 In particular, there were failures in the area of furnace heating control, mill pacing and stages and produce the necessary documentation. A coiler control (e.g. Ref. 59). There was no evident purpose built, steel clad computer centre was erected to reduction in overall mill manpower, while maintenance house the computer in an air conditioned ‘clean room’ technicians and software programmers were specifically environment as well as its 50 associated staff. required for the computer. Similarly, Coheur29 reports Production control and scheduling was a priority also an American study which found that only 12% of US in Japan. The complexity and size of process flows at the steel companies questioned in 1965 were able to assess Yawata-Tobata works made production planning the profitability of their investment in either adminis- imperative.64 The software was based on backward trative or process control computers. Yet a detailed induction, a variant of the familiar ‘travelling salesman’ report on a very early installation for computer control problem once taught to Fortran programmers. To solve of billet cutting on a rolling mill at Stocksbridge in the the problem, it is necessary to start at the destination UK showed annual savings from yield gains to outweigh and work backwards to identify the ideal departure time the total initial cost of the scheme, implying a very rapid from the origin. In the same way, the loading operation payback on the initial investment in equipment and at Yawata-Tobata for monthly, weekly and daily programming.60 schedules simply reversed the product flow. Starting Nevertheless, by the end of the 1960s some 24 with anticipated orders on the rolling mill, the computer computers were in use or being installed on hot strip worked backwards to calculate the implied flow through mills worldwide, including five in the UK and 10 of the the preceding primary mills, and then allocated output 11 new second generation mills built in the USA.35 schedules to the steel plants, including two BOS shops, Progress was helped by reliable and accurate instru- an arc melting shop and a range of open hearth furnaces. mentation available for hot strip mills, notably for A similar installation using a HITAC 4010 computer (an thickness, width and temperature. RCA 3301 made under licence by Hitachi) was installed at Hirohata works in 1965.65 Order handling and scheduling Computing progressed at an extraordinary speed during Furnace control in Rotherham the 1960s. The IBM 360 series of mainframes was Process control computers were widely used in the launched in 1965 and quickly became the standard in electricity generating sector itself, so major consumers of international computing. Key features included software electricity were quick to appreciate their cost saving that was compatible across computers; a wide range of potential. Steel, Peech & Tozer’s Templeborough plant sizes and power across the whole 360 series; and was one of the largest electric arc melting shops in the suitability for both business and technical use, by world in the early 1960s, and hence they had much to allowing short and long word applications.61 gain from computer control of power usage. Moreover, IBM were supported by their technical The Central Electricity Generating Board provided leadership in magnetic disk drives developed at their power to Templeborough under a bulk supply agree- San Jose research laboratories between 1952 and 1956.62 ment which imposed local limits on consumption, with The steel industry was quick to spot the potential of costly penalties if these limits were exceeded at times of larger computers for production planning. John peak demand on the overall electricity supply network. Summers and Sons commissioned an IBM 360, model A Ferranti Argus 108 was installed to control power 30 with six disk drives at Shotton on 3 April 1967 for supply to the six furnaces in the melt shop.66 This was order handling and production control (Fig. 4).63 At the able to monitor and predict electricity use and cutoff or outset, orders for strip and coated coil were accepted on reduce the offtake of power at peak times in a way that the computer and process routings identified. The aim optimised furnace use. For instance, the first furnace to was to schedule orders through the various production be restricted would be the one that had most recently

472 Ironmaking and Steelmaking 2004 VOL 31 NO 6 Aylen Megabytes for metals commenced melting, since the rate of potential heat loss impressive as a pioneer given their role at Trostre and from stoppage increases with the progress of the melt, Velindre and their initial work on cold mill control.18 while loss of power during refining may adversely affect The only comparable plants in terms of breadth of metallurgical conditions. A very similar system was computer applications were Italsider at Taranto, or developed 10 years later in Krefield.67 Inland Steel in the USA. 3. Third, process computers were clustered geogra- phically in South Wales, South Yorkshire and the Billet optimisation at Parkgate Iron and former Colvilles works in Scotland. Steel 4. Fourth, strip mill applications dominated in terms Parkgate Iron and Steel at Rotherham developed two of both number and complexity over long products. innovative schemes for early rolling mill control during Then, as now, section mills were a ‘black art’ far 1963 and 1964. One scheme used an English Electric removed from computer control. KDN2 computer to optimise and control the cut to 5. The size, sophistication and extent of computer length shear on a bloom billet mill. Foss52 described the control varied from minor task to comprehensive KDN/KDF machines as ‘good fast processors with control. For instance, the Digital Equipment Corp. mnemonic code software that was easy to understand PDP 8 series computer at Bilston was a no frills, low and implement.’ This was online real time control using end, cheap computer performing a very simple opera- a computer with 4k core memory based on germanium tion. The PDP 8 was the first commercially successful transistors, ‘which was more than adequate for control minicomputer and sold for one-fifth of the price of a of major items of plant’, according to Foss. There was basic IBM 360: only $18 000 in the USA. Hence, it was also an English Electric KDF6 mill pacing computer. ideal for a simple process control task such as cutting This had an innovative alphanumeric display driven by a bars to length. It was also small – tabletop size – and had Marconi character generator. A digital memory based a video display terminal. control system also controlled the pass sequences for a 6. There was an extremely wide variety of computers reversing blooming mill. Finally, there was a KDF6 in use, partly due to the fragmented nature of the UK computer used for order processing using conventional computer industry at that time. Each computer required data cards. specific software which was incompatible with other machines. There was some degree of standardisation, such as the use of Elliott Arch machines in South Wales. Diffusion of process control computers Indeed, by 1968 the merged English Electric and Elliott Automation became the second largest supplier of By 1969 there were over 3000 computers used in various process control computers in the world ahead of industrial processes throughout the world (of a total IBM.71 The Ferranti Argus series was coming into world stock of computers then approaching 100 000). Of widespread use. Otherwise there was very little hardware these, some 400 computers were associated with iron- or software compatibility across the industry at the time. and steelmaking worldwide. Among them, approxi- However, in its day, the GE 412 was something of a mately 50 were in the UK by 1969 and around 80 by world standard for strip mills, with applications in the beginning of 1970.35 South Wales, Bochum, Dortmund and Duisburg, as well In 1970, the recently formed British Steel Corp. (BSC) as process lines and cold mills in the USA. (The GE 412 published a list of their industrial computers in was known as the AEG GEAMATIC 1005 in operation or awaiting delivery for a parliamentary Germany.) However, both Ferranti and GE were to enquiry (Table 5),68 although there are omissions from leave commercial computers, so even this degree of the list such as the Leo II at Corby. At the time, BSC standardisation was not to persist.72,73 operated y1.5% of the UK computer stock.69 The 7. Finally, in some areas of technical leadership, such Corporation had at least 48 business computers and 36 process control computers, although the distinction is as strip mill control, computer use diffused rapidly not clear cut. At least six of the business computers were across plants, firms and countries helped by the associated with technical R&D, operations research or marketing prowess of GE Computers. Yet the pioneer- statistical analysis. The number of business computers ing work on electric arc furnace load control at Steel, was then declining, as BSC rationalised its operations Peech & Tozer was confined to just one location in the around fewer, but more powerful, machines such as UK and not replicated for another 10 years. IBM 360/40s and 50s linked by telephone lines. By 1970, computers had been in use in the steel industry for a Retrofitting decade, and there was scope for rationalisation across the newly formed Corporation. By 1970, computer control had become an accepted A breakdown of process computer use in BSC in feature of rolling mill operation, though they were often Table 5 highlights the following points. limited to specific tasks such as setup and gauge control. 1. Some process areas were completely devoid of For example, a Ferranti Argus 500 was fitted to the computer control in the UK. There were no applications Brinsworth narrow strip mill finishing train between to burden preparation which was among the first 1970 and 1972. This had a 32k core memory (four times applications to be debated in the USA, USSR70 and the size of Llanwern), used a 24 bit word length and a France. The new technology of continuous casting was hard disk memory store of 0.6 m in diameter. Paper not computer assisted. tapes were used as the storage media for instructions. 2. Second, only one works – Port Talbot – had The finishing mill setup was derived from physical model computer control at all stages in the production process. calculations and temperature measurements. In all, 91 The former Steel Company of Wales was all the more devices were controlled, including the setting of roll gaps

Ironmaking and Steelmaking 2004 VOL 31 NO 6 473 Aylen Megabytes for metals

and motor speeds across six stands, plus looper of computers began to diffuse rapidly, especially as pressures and coiler settings. An important feature was steelmakers appreciated their ‘retrofit’ capabilities to a secure software structure to avoid crashes. The tweak the performance of existing aging facilities. Ferranti Argus was powerful, flexible and easy to use, Computers were also used for reheat furnace control but it was considered too expensive for general use.52 A where they were first employed by the British Steel Corp. General Electric Co. (GEC)/AEI Con Pac 4060 was in 1977.74,75 Furnace set points were selected from a retrofitted on the Lackenby universal plate mill stand ‘carpet diagram’ relating temperature to the type of and on the finishing train of the coil plate mill. The use slabs and throughput rate. The background diagram

Table 5 Use of process control computers in British Steel Corp.: status at 1 January 1970, installed or awaiting delivery*

Function Type of computer Location Serial no.

Burden preparation and coke ovens (None) Ironmaking Control of blast furnace Ferranti Argus 500 Port Talbot 12 Control of combustion in stoves Hawker Siddeley DCC2 Redbourn 13 Steelmaking Electric arc Control of power input Ferranti Argus 100 Templeborough, Rotherham 1 Basic oxygen Data logging and control of steelplant English Electric KDN2 Normanby Park 4 Control of steel plant English Electric/Marconi Myriad Port Talbot 14 II with touch screens Control of steelmaking English Electric KDN2 Ravenscraig 30 Open hearth (None) Bessemer Control of furnace Ferranti Argus 350 Workington 2 Chemical analysis Steel analysis from Quantovac output English Electric KDN2 Sheffield 3 Continuous casting (None) Rolling mills Slabbing and blooming mills Control of cutup at shears English Electric KDN2 Rotherham 10 Control of rolling rate English Electric KDN2 Rotherham 11 Control of soaking pits Ferranti Argus 108 Redbourn 15 Control of soaking pits IBM 1800 Bilston 26 Control of soaking pits English Electric KDF7 Ravenscraig 31 Control of slabbing mill English Electric M2140 Ravenscraig 32 Hot strip mills Control of hot strip mill General Electric 412 Llanwern 16 Control of hot strip mill General Electric 412A Port Talbot 17 Mill setup and control Ferranti Argus 500 Brinsworth 6 Slab pacing from reheat furnace to coilers Ferranti Argus 104 Brinsworth 7 Gauge control on coil plate mill finishing train GEC/AEI Con Pac 4060 Lackenby 28 Control of hot strip mill English Electric M2140 Ravenscraig 33 Cold strip mills Control of 5 stand tandem mill GEC/AEI Con Pac 4060 Port Talbot (converted from 4 stand, see Table 2) 18 Data logging on reversing mill Ferranti Argus 400 Newport 19 Data logging on 3 stand mill GEC/AEI Con Pac 4060 Trostre 20 Data logging on 5 stand mill Elliott Arch 1000 Velindre 21 Plate mills Gauge control on universal plate mill stand GEC/AEI Con Pac 4060 Lackenby 27 Control of reversing plate mill English Electric M2140 Dalzell 34 Section mills Section cutup and tracking Ferranti Argus 500 Scunthorpe 9 Control of saws on section mill English Electric KDN2 Shelton 35 Control of section mill English Electric M2140 Colvilles 36 Billet, bar and rod mills Data logging and information display Elliott Arch 1000 Stocksbridge 5 Control of billet cutting to length Elliott 803A Stocksbridge 8 Control of hot bar cutting Digital Equipment PDP 8/S Bilston 29 Process lines Control of electrolytic tinplate line Elliott Arch 9000 Ebbw Vale 22 Quality recording on tinplate line Elliott Arch 1000 Velindre 23 Quality recording on tinplate line Elliott Arch 1000 Trostre 24 Quality recording on tinplate line Elliott Arch 1000 Trostre 25 *Source: derived from ‘Computer policy’,68 Appendix B.

474 Ironmaking and Steelmaking 2004 VOL 31 NO 6 Aylen Megabytes for metals was based on extensive data logging of actual furnace computer at the top of the hierarchy. As the price of behaviour. The computer simulated these initial set hardware fell, minicomputers or programmable logic points and then modified them in the light of events, controllers (PLCs) became feasible for localised control including the actual measured state of the reheat of processes. These could be bolted on in an almost furnace, progress on the rolling mill and any unantici- ad hoc way leading to quick upgrades and the pated delays. The system was initially installed on two progressive spread of computerisation across an increas- slab reheating furnaces of 180 and 220 t h–1 capacity at ing range of process control tasks. The third automation Lackenby, and brought fuel savings averaging 15%/ system of the Port Talbot five stand cold mill was week. Ultimately, two GEC 4080 computers were used undertaken using one online computer, a standby to run the system. computer and various GEM 80 microprocessors.55 The difficulty was that PLCs were becoming more powerful New plant automation and widely used, but still did not achieve the very fast response times required for process control operations. By the 1970s new plant was designed for computer Moreover, links between them were slow. control from the outset (e.g. Ref. 76). In Finland, the number two plate mill at Rautaruukki Oy, Raahe was the first in the world to use hydraulic gauge control. This Process control at heavy end was controlled by an Interdata 7/16 computer with 16k Ironmaking and steelmaking themselves lagged behind of memory installed in 1976, which calculated and set up in terms of process control, partly owing to sensor the pass schedules and operated the hydraulic gauge problems and partly because comprehensive thermo- control during each pass online in real time.52 The chemical models do not always characterise the complex computer had to set the roll gap, count passes, carry out aspects of furnace operation very well. In steelmaking, mill reversal and determine width and thickness of the early computers were limited to offering advice on plate through a schedule of 7–17 successive passes. The charge weights and blowing times, given targets for end computer used a magnetic core memory without any carbon and tapping temperature. For example, NKK backup hard disk. Hydraulic gauge control has a used a Hokushin Electric computer to advise the particularly fast response (10 ms) compared with con- operators of two 42 t BOS vessels. This used results ventional mechanical screwdowns. Again, the computer from model calculations from one heat as the starting needed secure software to avoid catastrophic problems point for initiating calculations for the next.77 A similar under load (with a rolling force up to 5000 t). Here too exercise on Kaldo converters at Sharon Steel was unable the approach was to look up stored tables of schedules to achieve dynamic control, despite the slow pace of for either straight through or broadside passes as refining in the revolving vessels.78 appropriate. In principle, calculation of rolling equa- Blast furnaces represented an even more complex tions provided a neater solution, but stored data had the modelling problem, but the payoffs in terms of energy pragmatic advantage of accuracy. efficiency were substantial. There were isolated pioneers Interdata computers with 16k memory came to be such as NKK Kure.79 Computer based artificial widely used for flatness control systems in steel and intelligence models for blast furnace control were aluminium from 1977 onwards, starting with the shape developed simultaneously in Finland and the UK and control system at the SIDAL in Belgium. These had the marketed widely. modern feature of a visual flatness display with a refresh rate of 50 ms. Shape control systems became widely applied on cold strip mills for aluminium and steel Integration of process control and during the 1980s. The system controlled roll bending quality: Linz hot strip mill and cooling sprays on the basis of signals from a flatness measurement roll. This not only brought flatter strip, By 1990, the operations of scheduling, process control but more stable operation allowed faster rolling speeds. and data logging by computer had become universal, at The main system was developed by Davy, but a rival least on flat product rolling mills and finishing lines. system was later developed by Asea. Now shape control During the 1990s large scale computer models were is a conventional feature of all mill control systems. developed which allowed integration of scheduling and In due course, shape control became integrated with process control with quality assessment. This requires a hydraulic gauge control on cold mills, but this required real time model of metallurgical transformation during more powerful computers such as the PDP 11 series of the rolling process. Developments by Voest-Alpine Stahl computers. By 1991, direct digital control of individual and VAI at Linz allow immediate predictions of quality hydraulic capsules was implemented at Iscor in South for the whole length of a rolled coil. Luger and 80 Africa using VME Motorola equipment. The operating Hubmer reported experience with the VAI-Q system 81,82 system on the computer was set up to achieve 1 ms at the Linz strip mill. This is one of the oldest response times. surviving hot strip mills in Europe, yet it makes a range of demanding products to high standards and sells profitably to sophisticated customers, especially for Move to distributed computing automotive applications. At four million tonnes a year, Dividing computer programs into a sequence of the Linz mill has a formidable output level for a subroutines became commonplace among software semicontinuous mill of this design. programmers during the 1970s. It was logical to suppose The starting point at Linz is a physical–metallurgical that computer hardware could be similarly distributed model to predict strip quality in terms of tensile strength, across individual tasks to provide local control of yield strength and elongation. This modelling project particular processes under the supervision of a central began in November 1995. Once it was established that

Ironmaking and Steelmaking 2004 VOL 31 NO 6 475 Aylen Megabytes for metals

an offline model gave accurate predictions of the Acknowledgements mechanical properties of hot rolled coil, the model was used to control actual setup and cooling on the The application of computers to process control in steel finishing train from January 2000 onwards. Strip is an unwritten history. The pioneers seldom recorded varies throughout its length. Therefore, the model their contributions before moving on to their next tracks each segment of the strip so that microstructure project. Rapid turnover in technology has left few traces can be predicted and modified during final rolling and of original equipment. Little has been preserved. The cooling. Optimum rolling and coiling temperatures are author is especially grateful to those who gave their time crucial for high strength low alloy (HSLA) steels, for to develop the story. M. Steeper of VAI persuaded the example. author to write this paper and provided much material. There are substantial commercial advantages from Both K. Morgan, curator of Trostre Works Cottage and being able to predict the quality of each coil straight Industrial Museum of Corus Packaging Plus, and M. away. The coil can be passed on for further processing Dean were invaluable through their contributions to the immediately. The whole length of the coil is ‘checked’ by conference on ‘The history of the wide strip mill in inference, whereas conventional measurement is Europe’ held in Manchester in April 2001. Comments restricted to samples from the head and tail ends, which from participants at ‘Automation in metals processing’, may be untypical anyway. There are substantial savings held at Sheffield University, June 2002, especially A. in alloying costs as it is possible to optimise mechanical Foss, added to the story. Dipl.Eng. Dr A. Luger of properties such as tensile strength, rather than relying on Voest-Alpine Stahl and Dr G. Hubmer of VA Tech overkill with expensive manganese additions at the brought the author up to date on recent developments. steelmaking stage just to make sure. R. Wells granted access to the John Summers and Sons The aim of computer modelling in steel now is to archive at the Corus Colors Records Centre, and R. develop ‘through process models’, first to predict the Holtho¨fer and his team tracked down requests for influence of one process stage on the next and then to material on Shotton and South Wales. Dr F. Fitzgerald optimise the whole sequence of processes from start to proved a mine of information. Dr T. Lunt and A. finish. Through process modelling of cold rolling and Roughley provided insights into Ferranti, and J. Carroll annealing has received less attention than hot rolling. led the author to early applications of Leo computers. Bodin et al.83 discussed development of the ‘toolbox’ R. Ranieri alerted the author to the Italian pioneers. approach at Corus for through process modelling of J. Bryant developed the Port Talbot story. N. cold rolling, annealing and temper rolling of low carbon Hopkinson gave valuable advice. The author alone steels. The aim is to predict the final mechanical remains responsible for errors, and offers help with access to source material. properties of a finished cold rolled coil by combining models of cold rolling, furnace and temper rolling Funding for this research was partly provided by the processes. Hence, the cold rolling stage is concerned British Academy as a Small Research Grant held with a with predicting deformation resistance using the colleague, Dr R. Ranieri, on ‘Comparative strategies for the introduction of wide strip mills in the UK and Italy Bergstrom model to determine dislocation density. (1935–1958)’. Models at the annealing stage focus on recrystallisation, precipitation and grain growth. For example, calculating free nitrogen helps to predict recrystallisation in a References continuous annealing line. As with hot strip mills, 1. T. M. Stout and S. M. Roberts: ‘Some applications of computer the task is to gather the data, calibrate a model and control in the iron and steel industry’, Iron Steel Eng., 1960, 37, then use it to predict behaviour for out of sample steel 101–110. grades. 2. ‘Technology timelines: who, what and when since 1500’, Everyday Evidently computer modelling is a way to make new Pract. Electron., May 2000, insert. 3. A. W. Smith: ‘Card programmed control system applied to hot products, and achieve higher capacity utilisation and 84 strip reversing mill’, Iron Steel Eng., 1956, 33, 164–165; and yield. Winkler et al. reported an optimisation package ‘Memory storage unit to operate automatic steel roughing mill’, for sequencing strip through continuously linked pickle New York Times, 4 November 1956. lines and cold mills. Application in a US plant brought a 4. D. R. Jones and A. W. Smith: ‘Combination slab and plate mill 5% increase in annual throughput with more even rolls under computer control’, Iron Steel Eng., 1965, 42, 134–141. 5. R. Bazuro and L. A. Oliva: ‘Taranto: plate rolling with on-line pickling speeds and a smoother flow through edge computer control’, J. Met., 1965, 17, 1110–1113. trimmers. This is exactly the same pacing problem that 6. H. H. Ascough: Letter from H. H. Ascough, Mills Superintendent, computer pioneers found so difficult to solve on hot Steel Company of Wales to L. N. Bramley, British Iron and Steel strip mills during the mid 1960s. Research Association, requesting help with project to implement card control of Port Talbot hot strip mill continuous roughing Ultimately, the aim is a business model which train, 23 November 1956, Corus Colors Record Centre, optimises a complete production programme subject to Department 378, Consignment 4, Box 55, Location 041472, File constraints of equipment, manufacturing times and ‘Hot strip mill 1956’. transport. In this way it should be possible to evaluate 7. G. D. Spenceley and P. H. Scholes: ‘British Iron and Steel Research costs and profitability of a wide range of production Association – the immortal memory’, Part 2, 24; January 2004, Teesside, UK, Corus (privately published CD-ROM). choices. This kind of comprehensive computer simula- 8. D. G. Owen: ‘Computers and steel’, London, British Iron and Steel tion and control is not likely to arrive before the end of Research Association, 1957, Corus Colors Record Centre, the present decade. By then, computer applications in Department 378, Consignment 4, Box 8, Location 041425, File steel will have reached their 50th birthday. This ‘Computers 1957/59’. 9. G. Ferry: ‘A computer called Leo: Lyons Teashops and the world’s achievement will unify the two distinct strands in first office computer’; 2003, London, Fourth Estate. computer development apparent since the earliest days, 10. ‘Proposal for the purchase of a Ferranti Pegasus 1 computer by namely business applications and process control. SCOW’, Chief Accountant to Mr W. F. Cartwright dated 24, July

476 Ironmaking and Steelmaking 2004 VOL 31 NO 6 Aylen Megabytes for metals

1959 and related papers, Corus Colors Record Centre, Department 40. A. S. Brower: ‘Controlling a complete hot strip mill’, Control Eng., 378, Consignment 4, Box 8, Location 041425, File ‘Computers 1963, 10, (10), 57–63. 1957/59’. 41. R. Meredith, J. T. Fisher and R. W. Kirkland: ‘Computer control 11. LEO COMPUTER SOCIETY: http://www.leo-computers.org.uk/ of hot strip mills at Spencer Works’, Iron Steel Eng., 1965, 42, (12), leo2s.htm (accessed 4 March 2004). 73–81. 12. G. Ferry: ‘A computer called Leo: Lyons Teashops and the world’s 42. H. R. Oldfield: ‘General Electric enters the computer business – first office computer’, 153–155; 2003, London, Fourth Estate. revisited’, IEEE Ann. History Comput., 1995, 17, (4), 45. 13. H. W. Gearing: ‘A brief survey of the installed computer capacity 43. M. H. Weik: ‘A fourth survey of domestic electronic digital in the United Kingdom’, summarised in Steel Company of Wales, computing systems’, Report no. 1227, January 1964, Maryland, ‘Computers in operation’, Port Talbot, 3 October 1958, reporting Aberdeen Proving Ground, Ballistics Research Lab., reproduced at conference at Northampton College of Advanced Technology, July http://ed-thelen.org/comp-hist/BRL64-g.html (accessed 5 March 1958, 4–6, Corus Colors Record Centre, Department 378, 2004), 112. Consignment 4, Box 8, Location 041425, File ‘Computers, 1957/ 44. NAVAL EDUCATION AND TRAINING COMMAND: 59’. ‘Magnetic cores’, in ‘Digital computer basics’, Chap. 6, Rate 14. H. W. Gearing: Ref. 13, Table 2. Training Manual, NAVEDTRA 10088-B; revised 1978, 15. H. W. Gearing: Ref. 13, 6. Washington, DC, Government Printing Office, at http://ed- 16. A. Roughley: Interviews with former Economic Assistant in thelen.org/comp-hist/navy-core-memory-desc.html (accessed 12 Statistics Department of Cotton Board, Manchester, UK, 16 and March 2004). 17 September 2003. 45. M. Dean: ‘The development of wide cold mills – a challenge to the 17. J. E. Mantell: ‘The applications of a computer to commercial control engineer?’, Proc. Conf. on ‘The history of the wide strip mill problems in the paint industry’, summarised in Steel Company of in Europe’, Chancellors Conference Centre, Manchester, UK, Wales, ‘Computers in operation’, Port Talbot, 3 October 1958, April 2001, University of Manchester. reporting conference at Northampton College of Advanced 46. R. W. Kirkland: ‘Process computers – their place in the steel Technology, July 1958, 19, Corus Colors Record Centre, industry’, Iron Steel Eng., 1965, 42, 124. Department 378, Consignment 4, Box 8, Location 041425, File 47. H. D. Morgan and R. W. Kirkland: ‘The application of auto- ‘Computers 1957/59’. matic gauge control and computer control to an existing hot strip 18. R. W. Kirkland: ‘Process computers – their place in the steel mill’, in ‘Digital computer applications to process control’, (ed. industry’, Iron Steel Eng., 1965, 42, 115–124. Instrument Society of America), 167–206; 1965, New York, Plenum 19. ‘Digital computer control list lengthens, market matures’, Control Press. Eng., 1963, 10, (9), 73–83. 48. I. S. Scott-Maxwell: Assistant Managing Director, Engineering 20. ‘Italy’s ultra-modern steel plant in the South: Taranto’, J. Met., Research and Automation, monthly reports to the Board of the 1965, 17, 1081–1120. Steel Company of Wales, Corus Colors Record Centre, 21. ‘Process computer scorecard updated’, Control Eng., 1965, 12, (3), Department 378, Consignment 3, Box 23, Location 041328, 57–62. Sequence of Board Papers 1965 and 1966. 22. ‘Process computer scorecard’, Control Eng., 1966, 13, (9), 73–82. 49. K. Morgan: ‘The evolution of hands-free control of strip mills’, 23. ‘What’s doing overseas? Process computer scorecard’, Control Proc. Conf. on ‘The history of the wide strip mill in Europe’, Eng., 1967, 14, (3), 51–55. Chancellors Conference Centre, Manchester, UK, April 2001, 24. ‘A staff survey: on-line computer scorecard updated’, Control Eng., University of Manchester. 1968, 15, (7), 79–90. 50. M. H. Weik: ‘A fourth survey of domestic electronic digital 25. C. Freeman: ‘Research and development in electronic capital computing systems’, Report no. 1227, January 1964, Maryland, goods’, Natl. Inst. Econ. Rev., November 1965, (34), 40–91. Aberdeen Proving Ground, Ballistics Research Lab., reproduced at 26. S. F. Bush: ‘Computing developments and their impact on http://ed-thelen.org/comp-hist/BRL64-g.html (accessed 5 March engineering over the next 10 years’, Proc. Conf. Process 2004), Table 14. Development Technology 8, London, 22 October 1985, 51. W. F. Cartwright: ‘How iron and steel will be automated’, Fourth Institution of Chemical Engineers. annual lecture of the UK Automation Council at the Institution of 27. J. Auricoste and P. Westercamp: ‘Computer control of an oxygen Electrical Engineers, London, 6 October, summarised in Control steelmaking process’, in ‘Digital computer applications to process 1964, 77, 580–581; (78), 636–637. control’, (ed. Instrument Society of America), 143–165; 1965, New 52. A. Foss: Personal communication and notes on ‘Evolution of York, Plenum Press. hardware and software for process plant control’ and ‘Summary of 28. C. Staib and J. Michard: ‘On-line computer control for the blast projects’, Riverside Automation, Sheffield, UK, June 2002. furnace, Part II, control of furnaces with sinter and complex 53. File ‘Computers 1957/1959’, Corus Colors Record Centre, Code burdens’, J. Met., 1965, 17, 165–170. 378, Departmental consignment 4, Departmental box 8, Location 29. P. Coheur: ‘Automation in the iron and steel industry’, Proc. IISI 041425. Third Annu. Conf., Tokyo, Japan, October 1969, 118–133. 54. G. F. Bryant, I. G. Cumming, W. J. Edwards and J. H. Westcott: 30. J. Aylen: ‘Innovation in the British steel industry’, in ‘Technical ‘Research in tandem mill automation’, J. Iron Steel Inst., 1971, 209, innovation and British economic performance’, (ed. K. Pavitt), 869–875. 200–234; 1980, London, Macmillan. 55. S. M. Alston and M. Dean: ‘Automation systems for the tandem 31. R. W. Kirkland: ‘Process computers – their place in the steel cold strip mills within BSC’, in ‘Advances in cold rolling industry’, Iron Steel Eng., 1965, 42, 115–119. technology’, Book 354, 281–291, Table 1; 1985, London, The 32. J. Van Langen, A. Poos and R. Vidal: ‘Continuous control of the Institute of Materials. blast furnace by computer’, J. Met., 1965, 17, 1379–1384. 56. J. Bryant: Personal communication from former Assistant 33. R. W. Kirkland: ‘Process computers – their place in the steel Manager, Tandem Mill, Port Talbot, Wales, 5 August 2004. industry’, Iron Steel Eng., 1965, 42, 119. 57. H. E. Miller, R. A. Smith, J. F. McCarthy, H. Smith, Jr and N. J. 34. ‘Hot strip mill at McLouth Steel under computer control’, Control Hittinger: ‘Control and automation of hot strip mills from the Eng., 1963, 10, (4), 32. operator’s viewpoint’, Iron Steel Eng., 1969, 46, 67–75. 35. D. J. Ray: ‘Present and future trends in hot strip mill computer 58. T. J. Ess: ‘The hot strip mill generation II’, 96–100; 1970, control’, J. Iron Steel Inst., 1969, 207, 907–915. Pittsburgh, PA, Association of Iron and Steel Engineers. 36. H. R. Oldfield: ‘General Electric enters the computer business – 59. E. S. Lodics: ‘A computer runs a hot strip rolling mill’, Control revisited’, IEEE Ann. History Comput., 1995, 17, (4), 46–55. Eng., 1969, 16, (4), 89–93. 37. M. H. Weik: ‘A fourth survey of domestic electronic digital 60. R. H. Baulk, R. J. Jakeways and K. C. Padley: ‘Experiences with computing systems’, Report no. 1227, January 1964, Maryland, an on-line process control computer in a steel works’, in ‘Digital Aberdeen Proving Ground, Ballistics Research Lab., reproduced at computer applications to process control’, (ed. Instrument Society http://ed-thelen.org/comp-hist/BRL64-g.html (accessed 5 March of America), 1965, 207–232; New York, Plenum Press. 2004), 110–111. 61. K. Flamm: ‘Creating the computer: government, industry and high 38. PERSO CLUB: ‘The American operations – General Electric 1956– technology’, Chap. 4; 1988, Washington, DC, The Brookings 1970’, at http://perso.club-internet.fr/febcm/english/groupe__bull_ Institution. chronology.html (accessed 22 July 2004). 62. C. M. Christensen: ‘The innovator’s dilemma: when new technol- 39. H. R. Oldfield: ‘General Electric enters the computer business – ogies cause great firms to fail’, Chap. 1; 1997, Boston, MA, revisited’, IEEE Ann. History Comput., 1995, 17, (4), 54. Harvard Business School Press.

Ironmaking and Steelmaking 2004 VOL 31 NO 6 477 Aylen Megabytes for metals

63. ‘Schedule control by computer’, John Summers Rev., 1967, 2, (2), 74. R. D. Glatt and F. X. Macedo: ‘Computer control of reheating 7–9, Corus Colors Record Centre, Department 426, Consignment furnaces’, Iron Steel Int., 1977, 50, 381–396. 13, Box 10, Location 37998. 75. P. H. Scholes: ‘British Steel research achievements 1970 to 1990’; 64. H. Ogoshi and Y. Inoue: ‘Computer system used as production 2002, Middlesbrough, Corus RD&T, Teesside Technology Centre. planning tool at Yawata Works’, Iron Steel Eng., 1965, 42, 128– 76. W. Fabian, K Von Germersheim, F. Knorr, R. Kubel, H.-G. 133. Mo¨hring, G. Wiethoff and H. Wladika: ‘Neuere Erfahrungen mit 65. ‘Japanese computer to manage steel mill’, Control Eng., 1964, 11, einem Prozeßrechner fu¨r den Gesamtbereich eines (11), 25–26. Warmbreitbandwalzwerkes’, Stahl Eisen, 1977, 97, 1023–1028. 66. J. Tippett, A. Whitwell and L. H. Fielder: ‘Controlling megawatts 77. K. Katsura, K. Isobe and T. Itaoka: ‘Computer control of the basic in steelmaking’, Control Eng., 1965, 12, (6), 68–70. oxygen process’, J. Met., 1964, 16, 340–345. 67. H. H. Borggrefe, K. Hohendahl and G. Zingel: ‘Optimization of 78. J. E. Binning and D. R. Berg: ‘Computer control of Kaldo oxygen energy distribution in the operation of an electric melting shop’, in steelmaking’, J. Met., 1965, 17, 725–728. ‘Mathematical process models in iron- and steelmaking’, 57–63; 79. S. Horie: ‘High productivity blast furnace operations at Nippon 1975, London, The Metals Society. Kokan’, J. Met., 1965, 17, 243–245. 68. ‘Computer policy: memorandum by the British Steel Corporation’, 80. A. Luger and G. Hubmer: ‘In-line control of the mechanical Select Committee on Science and Technology of the House of properties of hot rolled strip based on complete process modelling’, Commons, Session 1969–70, 143–158, in ‘UK computer industry’, Proc. Conf. on ‘Modelling of metal rolling processes – through Vol. I, ‘Minutes of evidence’, HC 137, HMSO, London, February process modelling’, London, December 2001, The Institute of 1970. Materials. 69. K. Flamm: ‘Creating the computer: government, industry and high 81. G. Hubmer: ‘Modelling done with VAI-Q strip’, Proc. VAI-Q technology’, Chap. 4, Table 5.1; 1988, Washington, DC, The ‘Strip’ Symp., Linz, Austria, November 2000, 2–5. Brookings Institution. 82. J. Andorfer: ‘Looking back on VAI-Q strip as a project’, Proc. 70. T. M. Stout and S. M. Roberts: ‘Some applications of computer VAI-Q ‘Strip’ Symp., Linz, Austria, November 2000, 66–70. control in the iron and steel industry’, Iron Steel Eng., 1960, 37, 83. A. Bodin, J. Colijn, W. Kranendonk, J. Oerlemans, P. Van Popta, 104–106. J. Vrieze and M. Van der Winden: ‘Through-process modelling of 71. ‘A staff survey: on-line computer scorecard’, Control Eng., 1968, cold-rolling and annealing of low-carbon steels’, Proc. Conf. on 15, (7), Table 1. ‘Modelling of metal rolling processes – through process modelling’, 72. G. Tweedale: ‘A Manchester computer pioneer: Ferranti in London, December 2001, The Institute of Materials. retrospect’, IEEE Ann. History Comput., 1993, 15, (3), 37–43. 84. R. Winkler, G. Finstermann, H. Aflenzer and F. Mitter: ‘Strip 73. J. A. N. Lee: ‘The rise and fall of the General Electric Corporation sequence optimisation of continuous tandem mills’, Proc. Conf. on computer department’, IEEE Ann. History Comput., 1995, 17, (4), ‘Modelling of metal rolling processes – through process modelling’, 24–45. London, December 2001, The Institute of Materials.

478 Ironmaking and Steelmaking 2004 VOL 31 NO 6