DIGITAL DESIGN Tools, Techniques, Perspectives

SPECIAL ISSUE

March 2016 The EngineerVolume 94 | Issue 3 The ﬂ agship Structuralpublication of The Institution of Structural Engineers

REFLECTIVE THINKING

ART OF PARAMETRIC DESIGN

OPTIMISATION METHODS

CITY OF DREAMS

VIRTUAL BY DESIGN

ENGINEERING IN A DIGITAL ENVIRONMENT

DIGITAL DESIGN Tools, techniques, perspectives

TSE51_01 Cover.indd 1 24/02/2016 11:34 p02_TSE.03.16.indd 2 19/02/2016 15:58 › www.thestructuralengineer.org Contents TheStructuralEngineer 3 March 2016

PAGE 34 OPTIMISATION METHODS PAGE 69 QATAR FACULTY OF ISLAMIC STUDIES PAGE 94 DIGITAL ASSEMBLY? TheStructuralEngineer Volume 94 | Issue 3

Upfront 44 What is your structural model not telling you? Opinion 48 What is the need for verifi cation, validation and 5 Editorial 92 Viewpoint: Structural engineering within a digital quality assurance of computer-aided calculation? 6 Institution news environment 52 Integration of hand calculations with computational 94 Viewpoint: Digital design, fabrication and assembly 8 Industry news output – a good practice summary 96 Viewpoint: Should we now say “design and Introduction Project focus analysis” and not “analysis and design”? 10 Insight not numbers – a brief history of computing in 98 Book review: Advanced Modelling Techniques in 56 City of Dreams, Macau – making the vision viable structural engineering Structural Design 69 Digitally designed – Qatar Faculty of Islamic Studies 99 Verulam Methods and practice Education 14 Time to refl ect: a strategy for reducing risk in At the back 78 Should students be introduced to analysis and structural design 101 Diary dates design software? 19 How to choose the right structural engineering 103 Spotlight on Structures software? 80 What are the benefi ts of exposing students to structural analysis and design software? 105 And fi nally… 24 Structural workfl ows and the art of parametric 106 Products & Services design 88 Virtual by design 108 Services Directory 34 An introduction to engineering optimisation methods 109 TheStructuralEngineerJobs

Front cover: ©ZAHA HADID ARCHITECTS

The Structural Engineer www.thestructuralengineer.org

PRESIDENT ADVERTISING EDITORIAL ADVISORY GROUP © The Institution of Structural Engineers. All non-member authors Alan Crossman are required to sign the Institution’s ‘Licence to publish’ form. CEng, FIStructE, FICE, DISPLAY SALES Project focus: Allan Mann, FIStructE Authors who are members of the Institution meet our requirements MCIWEM Patrick Lynn Features: Don McQuillan, FIStructE under the Institution’s Regulation 10.2 and therefore do not need t: +44 (0) 20 7880 7614 Technical: Chris O’Regan, FIStructE to sign the ‘Licence to publish’ form. Copyright for the layout and CHIEF EXECUTIVE e: [email protected] Opinion: Angus Palmer, MIStructE design of articles resides with the Institution while the copyright Martin Powell Professional guidance: Simon Pitchers, MIStructE RECRUITMENT SALES of the material remains with the author(s). All material published in Paul Wade The Structural Engineer carries the copyright of the Institution, but EDITORIAL t: +44 (0) 20 7880 6212 Price (2016 subscription) the intellectual rights of the authors are acknowledged. e: [email protected] Institutional: £390 (12 issues incl. e-archive, p&p and VAT) HEAD OF PUBLISHING The Institution of Structural Engineers Personal: £125 (12 issues incl. p&p) Lee Baldwin International HQ DESIGN Personal (Student Member): £40 (12 issues incl. p&p) 47–58 Bastwick Street EDITOR SENIOR DESIGNER London EC1V 3PS Single copies: £35 (incl. p&p) Robin Jones Craig Bowyer United Kingdom t: +44 (0) 20 7201 9822 t: +44 (0)20 7235 4535 e: [email protected] CREATIVE DIRECTOR e: [email protected] Mark Parry Printed by EDITORIAL ASSISTANT Warners Midlands plc The Institution of Structural Engineers PRODUCTION Ian Farmer The Maltings, Manor Lane Incorporated by Royal Charter t: +44 (0) 20 7201 9121 PRODUCTION EXECUTIVE Bourne, Lincolnshire PE10 9PH Charity Registered in England and Wales number 233392 and in e: [email protected] Rachel Young United Kingdom Scotland number SC038263

TSE51_03 Contents.indd 3 24/02/2016 11:34 KLOECKNER METALS UK WESTOK Cellular Beam Specialist

ftwa ngin ufac So re e E eer an tur lu in M e a g V

E C s e n e e st F l gi g llu li re ab ne ta l ia e il e y s ar, ec De Ava ring an Pla Sp sign Software Support at te and USFB

• Clear span solutions • CE Marking to BS EN 1090-1 Exc 4 • Optimised ﬂoor cross-section • Revit® integration • Ability to offer the right mix of cellular and plated beams • Free Westok design, vibration and BIM software • Flexible service integration • Free design service: Scheme, detailed and value • Pre-cambering provided free of charge engineering

ENTRUSTED TO DESIGN AND DELIVER MULTI-METAL SOLUTIONS

MARWE.FEB2016 e: [email protected] | w: www.kloecknerwestok.com

p04_TSE.03.16.indd 4 19/02/2016 15:59 › www.thestructuralengineer.org Upfront TheStructuralEngineer 5 Editorial March 2016 Upfront Digital design

Peter Ayres Guest Editor

Not so long ago, a journalist asked me an interesting question: “Do (page 14). My AECOM colleagues, Jon Leach et al., provide an in-depth you believe the work of the structural engineer can ever be replaced description of the practical application of parametric design methods by artifi cial intelligence”. I think she was somewhat taken aback when I to optimise complex geometrical solutions (page 24), while Peter answered “Yes”. Debney of Arup off ers a highly accessible introduction to the theory But before the esteemed readership of this magazine fl oods Verulam behind various optimisation methods, from simple quasi-Newtonian with missives of indignation, let me explain that I qualifi ed my answer; methodologies to state-of-the-art artifi cial neural networking (page 34). I postulated that while almost all the technical work undertaken by As examples of advanced digital design in practice, we are treated structural engineers at every level could, in theory, be overtaken by to two fascinating project case studies: BuroHappold’s City of Dreams artifi cial intelligence (and that it would be highly complacent of us Hotel in Macau (page 56), and Arup’s Qatar Faculty of Islamic Studies as a profession to assume our more “left brained” tendencies were in Doha (page 69). irreplaceable) the art of the structural engineer would always remain. A recurring concern expressed by many practising engineers is how Which begs the question, as structural engineers, what do we really student engineers can learn the fundamentals of structural engineering mean by design? When I was at university over 30 years ago, much of in a world where much of the work traditionally undertaken by our course work was taken up learning the hard, number-crunching graduates can be supplanted by technology. So in our fourth section, ways of analysing structures, while “design” lessons generally involved we explore how students should be exposed to software as part of practising the use of codes and standards to select and detail their development, rounded off by an inspiring piece from Institution structural elements. For the 21st-century structural engineer, these Past President Tim Ibell on how the digital revolution should allow a new are processes which can now be almost entirely automated. Our real kind of creative talent to emerge in the world of engineering (page 88). value comes in understanding when and how to apply the increasingly We conclude with a series of short opinion pieces, including a complex tools at our disposal to deliver value and creativity to our typically thought-provoking contribution from Tristram Carfrae of Arup clients and stakeholders. on the creative possibilities opened up by allowing engineers to “play” So in this special issue of The Structural Engineer, we set out to in a digital world (page 92). describe how far our profession has come, and where it might be going, I hope you will enjoy this special issue; I have certainly enjoyed in the development of digital design tools, and what this might mean for editing it. I truly believe we are a creative industry. Technology is a tool, structural engineers of the future. By way of introduction, we start with not an end in itself, and by choosing our tools wisely, we can continue a historical insight into the use of computers in structural engineering to create and sustain better world. by Allan Mann of Jacobs, including some fascinating recollections of predictions from the past (page 10). Peter Ayres is a structural engineer and multidisciplinary team We go on to present a collection of papers from across our leader at AECOM who has delivered an extraordinary range of high- profession exploring the methods and practice used by leading performance buildings in recent years, from the multiple award-winning practitioners today, including some salutary lessons from Iain MacLeod Halley VI Antarctic research base to world-class sports venues in of The University of Strathclyde on the importance of refl ective thinking Russia, Brazil and the Middle East.

The Structural Engineer The Institution The Structural Contributions published in The Structural Engineer are provides structural engineers and related has over 27 000 members in over 100 countries Engineer (ISSN published on the understanding that the author/s is/are professionals worldwide with technical information around the world 1466-5123) is solely responsible for the statements made, for on practice, design, development, education and is the only qualifying body in the world concerned solely published 12 the opinions expressed and/or for the accuracy of training associated with the profession of structural with the theory and practice of structural engineering times a year by the contents. Publication does not imply that any engineering, and off ers a forum for discussion on through its Chartered members is an internationally IStructE Ltd, a statement or opinion expressed by the author/s these matters recognised source of expertise and information wholly owned reﬂ ects the views of the Institution of Structural promotes the learned society role of the Institution concerning all issues that involve structural engineering subsidiary of Engineers’ Board; Council; committees; members by publishing peer-reviewed content which advances and public safety within the built environment The Institution or employees. No liability is accepted by such persons the science and art of structural engineering supports and protects the profession of structural of Structural or by the Institution for any loss or damage, whether provides members and non-members worldwide engineering by upholding professional standards Engineers. It is caused through reliance on any statement, opinion with Institution and industry related news and to act as an international voice on behalf of available both in or omission (textual or otherwise) in The Structural provides a medium for relevant advertising structural engineers print and online. Engineer, or otherwise.

TSE51_05 Editorial v1.indd 5 24/02/2016 11:35 ›

6 TheStructuralEngineer Upfront March 2016 Institution news

Institution transforms accessibility to Institution Chartered membership Fellow elected to National The Institution of Structural Engineers is options, so candidates can choose the transforming accessibility to Chartered route that is right for them. The format and Academy of membership and its sought-after “MIStructE” standard of the Institution’s interview and Engineering post-nominal by simplifying the routes exam remain unchanged. candidates follow to its Chartered Membership Once qualifi ed, all members of the Exam. The Institution is making the change Institution continue their development to expand opportunity for talented engineers, through programmes of mandatory career- regardless of background or circumstance. long learning. Candidates for Chartered membership Chief Executive, Martin Powell, said: have to demonstrate Masters level knowledge “Our new approach will transform and understanding in order to sit the exam. accessibility to Chartered membership, Previously this was demonstrated through an while leaving the rigour of our professional accredited MEng qualifi cation or via Further competence assessment unchanged. In Learning – “top-up” academic qualifi cations or this way we can provide new opportunities assessments like MSc degrees or a Technical for talented individuals irrespective of Report Route (TRR). background or circumstance. Masters level standard remains the “The Institution believes that this will be benchmark, but from 2016 candidates with of continuing benefi t to the public’s trust in a BEng(Hons) degree can demonstrate their structural engineers, who are the guardians Masters level learning, gained through their of public safety in the built environment.” postgraduate experience, via the Chartered Darren Byrne, Director of Membership Membership Exam. This change recognises and Education, said: that Masters level knowledge gained in “Expanding access to the exam to professional practice can be assessed through BEng(Hons) holders helps us better align the exam. UK applicants with those from other nations, The Professional Interview and rigorous while maintaining our interview and exam as exam – widely acknowledged as exceeding one of the most demanding assessments of Masters standard – will test levels of knowledge, understanding and professional competence and knowledge gained. Masters competence in the world.” level courses and the TRR are still valid fi b Bulletin 76 now available Congratulations to Professor John Burland FIStructE, who has been elected a foreign member of the US National Standards for specifying and ensuring the induced corrosion presented in fi b Bulletin 34: Academy of Engineering. John was one durability of new concrete structures are Model Code for Service Life Design (2006), of 22 foreign members announced by commonly of the prescriptive kind. The new fi b Model Code for Concrete Structures 2010 Academy President C.D. (Dan) Mote Jr. on fi b Bulletin 76: Benchmarking of deemed-to- and ISO 16204:2012. The work compares the 8 February 2016, bringing the total foreign satisfy provisions in standards – Durability of calculated reliability ranges thus determined membership to 232. reinforced concrete structures exposed to with the target reliabilities proposed by current Election to the Academy is among the chlorides presents the benchmarking of a specifi cations and, based on the comparison, highest professional distinctions accorded number of rules for chloride-induced corrosion off ers a proposal for the improvement of to an engineer. Academy membership as given in national codes such as European, deemed-to-satisfy rules and specifi cations. honours those who have made outstanding US and Australian standards. fi b Bulletin 76 presents and discusses contributions to “engineering research, This new benchmark determines the in detail the input data for the examined practice or education, including, where reliability ranges in the chloride-induced model parameters and off ers an extensive appropriate, signifi cant contributions to depassivation of rebar if the deemed-to- annexe documenting the values of the the engineering literature” and to “the satisfy rules of diff erent countries are taken individual parameters used in the analyses. pioneering of new and developing fi elds of into consideration. It does not only involve It thus provides a reliable database for the technology, making major advancements (probabilistic) calculations using input mainly performance-based probabilistic service-life in traditional fi elds of engineering, or based on short-term and rapid laboratory- design of concrete structures exposed to developing/implementing innovative test data but also involves input based on chlorides, be they in the form of salt fog, sea approaches to engineering education.” an independent assessment of existing water or de-icing salts. Individuals in the newly elected class will structures. For more information, visit www.fi b- be formally inducted during a ceremony The reliability analyses are carried out using international.org/benchmarking-of-deemed-to- at the Academy’s Annual Meeting in the probabilistic design approach for chloride- satisfy-provisions-in-standards. Washington, D.C. on 9 October 2016.

TSE51_6-7 Institution news.indd 6 24/02/2016 11:36 www.thestructuralengineer.org 7

Enter the Young Structural Engineering Professional of the Year award BIM offering for DELTABEAM® and connections

Peikko offers 3D modeling tools for Revit and Tekla Structures users.

Revit families Entries are now open for the Young Structural Engineering ® Professional of the Year award, which recognises outstanding young • DELTABEAM profiles, structural engineers. Anchor Bolts, PCs Corbels The competition is open to any member of the Institution aged with needed supplementary 30 years or under. The winner receives a £1500 cash prize, along reinforcement etc. with two tickets to The Structural Awards 2016 ceremony – the • Families include correct industry’s most prestigious awards event and a fantastic networking dimensions, colour codes, opportunity. The runner-up receives a £500 cash prize. names and weights for To enter, candidates must submit a 250-word opinion piece on what structural engineering means to them, with accompanying products poster and video presentations – plus an up-to-date CV and an • User interface offers more endorsement from their employer. information and links to Winning the award is an internationally recognised mark of Peikko product website professional achievement, and a unique opportunity to be marked out as “one to watch” by the Institution’s global membership. The entry deadline is Tuesday 5 April 2016. Full details about Tekla Structures tools the award are available at www.istructe.org/young-structural- engineering-professional-award. • Profiles, plugins and custom components for wide selection of Peikko products Award for Excellence in • 3D models of products include correct dimensions, Structural Engineering colour codes, names and Education: deadline weights for products extended • Tutorial videos available We constantly develop new and improved 3D modeling tools for designers based on user feedback. The Award for Excellence in Structural Engineering Education recognises academics who demonstrate the highest standards of teaching and a drive to develop innovative techniques to Design Service improve student learning. The deadline for submissions has Peikko UK’s engineering team is ready to help you with been extended to 24 March 2016. DELTABEAM® solutions integrated to connection components Award winners will be invited to the Institution’s annual People to produce an efficient cost effective slim floor design. and Papers Awards lunch in June. Selected entries will also feature at the Institution’s Academics’ Conference in September, and will be considered for publication in The Structural Engineer. Submissions may cover any aspect of learning/teaching of structural behaviour on any bachelor’s or taught master’s Contact info and modeling tools download course. www.peikko.co.uk Details about how to apply for the award are available at www. istructe.org/events-and-awards/people-and-papers-awards/ excellence-in-structural-engineering-education/how-to-apply.

TSE51_6-7 Institution news.indd 7 24/02/2016 11:37 ›

8 TheStructuralEngineer Upfront March 2016 Industry news

New sentencing guidelines for health- Revised code of and-safety offences take effect practice for safe use of cranes published Revised guidelines for punishing health-and- impact on the organisation’s ability to improve safety off ences in the UK came into force on conditions. 1 February. The new guidelines, published by The guidelines are intended to ensure that BS 7121-1:2016 Code of practice for safe use the Sentencing Council in November 2015, punishments are proportional to the severity of cranes. General has been revised to refl ect allow fi nes of up to £20M for corporate of the off ence, as well as the culpability and changes in other parts of the BS 7121 series manslaughter and will apply to all cases means of the employer. They also include a and changes in the application of risk-based sentenced from 1 February, regardless of when range of mitigating factors which allow for planning to lifting operations. the off ence took place. voluntary, positive action to remedy a failure BS 7121-1 is the essential standard for The guidelines allow courts to tailor fi nes on the part of off enders to be refl ected in those planning and carrying out lifting to an off ender’s specifi c circumstances, sentences. operations with cranes in the UK and gives taking into account a number of factors. For further details, see www. recommendations for the safe use of cranes These include an organisation’s turnover, sentencingcouncil.org.uk/news/item/new- permanently or temporarily installed in a work profi t margin, the impact on employees, or the sentencing-guidelines-come-into-force. environment. For more information, see http://pages. bsigroup.com/e/35972/EWS-BUILD- Engineering contractor issues warning Newsletter-BUYS-1602/hgwt4g/305563815. over unaccredited structural metalwork CROSS Newsletter No. 41 now available Some metal fabrication companies are acting products because they were the ones illegally, placing their customers at risk of responsible for ensuring the structures prosecution, according to John Grenville, they purchased were procured only from an The latest newsletter from Confi dential managing director of leading multi-specialist accredited company. Reporting on Structural Safety (CROSS) engineering contractor ECEX. The structural components covered by EN features reports, together with expert He warned: “It has been a criminal 1090 include structural steel and aluminium comment, on: off ence since July 2014 to supply structural components, kits, steel components used metalwork unless it conforms to EN 1090, a in composite steel and concrete structures • risks from off -site manufacture and hybrid three-part European standard that regulates and structural cold-formed members and construction the fabrication and assembly of steel and sheeting. • wind problems in city centre aluminium structures.” Structural components are defi ned as • roof constructed without structural input However, he added, many smaller those “to be used as load-bearing parts • fl aws with partial encasements around steel fabrication companies had failed to become of works designed to provide mechanical columns EN 1090 accredited: “Some are probably resistance and stability to the works and/or • shear failure risk during demolition ignorant of the legislation; others believe it’s fi re resistance, including aspects of durability • wrongly designed safety system disproportionately costly and time-consuming and serviceability which can be used directly • failure to check designs produced by software for a small business to meet the accreditation as delivered or can be incorporated into a requirements.” construction work”. To download the newsletter or make a This, said Grenville, was bad news for For more information, see www.ecex.co.uk/ confi dential report, visit www.structural-safety. buyers of structural steel and aluminium warning-over-metal-fabricating-law-breakers. org.

Victoria and Albert Museum to celebrate including previously unseen archival materials, shown alongside recent projects by Arup, design philosophy of Ove Arup the global engineering consultancy. There will be large-scale prototypes and building components as well as digital animations and London’s Victoria and Albert (V&A) Museum approach to design that has defi ned the way models. will explore the work and legacy of Ove Arup engineering is understood and practised The V&A Engineering Season will highlight this summer in a new exhibition forming part of today. This exhibition will focus on the design the importance of engineering in our daily lives its Engineering Season. The exhibition, entitled philosophy of Ove Arup, revealing his ideas of and consider engineers as the “unsung heroes “Engineering the World: Ove Arup and the collaborative working, total architecture and of design, who play a vital and creative role in Philosophy of Total Design”, will run from 18 design as a humanistic and technological tool the creation of our built environment”. June to 6 November 2016. for social responsibility. For more information, visit www.vam.ac.uk/ Ove Arup (1895–1988) was the most On display will be designs for some of content/exhibitions/exhibition-engineering- infl uential engineer of the 20th century Arup’s fi rst projects such as Sydney Opera the-world-ove-arup-and-the-philosophy-of- and the pioneer of a multidisciplinary House and the Penguin Pool at London Zoo, total-design.

TSE51_08 Indusrty v1.indd 8 24/02/2016 11:37 CHANG E YOUR PERSP ECTIVE.

Introducing NEXUS®

The new two-part system that creates stunning brick features quickly and easily. Steel-lined brick modules ﬁx directly to brick support angles to provide a light attractive sofﬁt in seconds. Brought to you through the collaboration of two industry experts, Ibstock and Ancon.

For further information on NEXUS®, please contact: t: 0161 480 2621 e: [email protected]

See NEXUS® on Ancon’s Stand No. E3080 Ibstock.com/nexus Ancon.co.uk/nexus

p09_TSE.03.16.indd 9 19/02/2016 16:00 ›

10 TheStructuralEngineer Introduction March 2016 History of computing

Insight not numbers – a brief history of computing in structural engineering

Allan Mann BSc(Eng), PhD, CEng, FREng, FIStructE Senior Consultant, Jacobs

A glance at this special issue’s contents will reveal how much computer technology Figure 1 Ferranti Mark 1 – underpins much of today’s practice. What where it all began is perhaps surprising is that today’s status has been reached from a zero start over the working life of engineers now retiring – and almost explosively over perhaps half a working life, from the mid-1990s. The challenges of coping with today’s technology, let alone what is going to happen in the next decade, are formidable.

Where did it all begin? We can trace the beginnings back to the end of World War II, when Alan Turing (of Bletchley Park code-breaking fame) and some colleagues came to Manchester University and built a line of computers that could store programs. These ﬁ rst ran in 1949, with their ﬁ rst commercial version (the Ferranti Mark 1) running in 1951 (Figure

1). In January 1956, R.K. Livesley (who was GETTY an engineering lecturer) published “The application of an electronic computer to some problems of structural analysis” in with the right skills; moreover, programmers detailing. Bensasson3 explains that there was a The Structural Engineer1; his text was based tended not to understand engineering and good choice of analysis programmes available, on elastic structural analysis using the engineers did not understand computers. with some progressing onto member design Manchester computers. In a later paper2, So communication, then as now, was via bar selection at discrete locations. But Livesley describes developments from 1946 problematic. programmes for detailing and scheduling were to 1966, crediting the key contribution made Take-up was most rapid in aeronautics, scarce, probably for reasons of mathematical by the Ferranti machine. Livesley himself where accurate analysis paid dividends in complexity, the need to handle a large amount suggests that the fi rst structural calculation weight savings and where aircraft design of data, and the development cost. was probably carried out in the Mathematical companies were large enough to aff ord Nevertheless, it was a beginning, and in Laboratory at Cambridge University in 1952, computer use, an advantage unavailable to 1979 The Structural Engineer published more and he includes the frame that was solved, most small structural engineering fi rms and a papers on computer-aided design (CAD). reproduced here as Figure 2. drawback not really overcome until nearly 30 The steelwork industry was keen to progress Livesley’s 1956 paper was wide-ranging years later. Thus, in that era, in our profession, and Weller4, also writing in 1979, gave a full and in it can be found the seeds of much computer development was largely left up to description of the current status, describing of the future development of computing. universities and enthusiasts. programmes for analysis and design, albeit However, even he could not foresee the still limited by machine capacity. Ongoing revolution that would follow from mass 1960s onwards work to expand into connection design, production of microchips later in the century. A review of papers published in The Structural with an objective of producing drawings Livesley describes several contemporary Engineer from the 1960s to the 1970s gives which could be used directly for fabrication hindrances to more widespread computer the impression of “nothing much happening”; information, was described. Putting aside use. In the early days, computers lacked what publications there were suggested that technical comment, the paper off ers insights storage, which is where the Ferranti machine the computer’s main role was seen as solving into usage issues that still face us today. It scored (it was able to store 200 000 complicated equations. Yet in 1978, The reminds us that: numbers). Another hindrance was that all Structural Engineer published what may be its programmes had to be written in machine fi rst special issue on computing, concentrating • although computers can solve complicated code and there were few people available on applications for concrete design and mathematics, “no ‘exact’ analysis is possible

TSE51_10-12 Intro Brief history v1.indd 10 24/02/2016 11:37 www.thestructuralengineer.org 11

for a practical building structure. The best best practice in design and fabrication. Today, are slower and more demanding on hardware we can do is to approximate its behaviour, modelling programmes are available that resources. Users will fi nd the fastest route to allowing for as many variables as possible, replicate entire steel structures in virtual reality the required solution. For this reason there with such features as erection stresses and to aid “visual” design and facilitate automatic has been a rediscovery of DOS. The suppliers foundation settlement being only estimated” generation of items for fabrication. have responded by providing Windows like • member selection remains an art and interfaces to applications running under DOS”. “frequently the lightest section may be 1990s Things don’t appear to have quite worked out rejected for a number of important reasons”, Returning to the beginning of the decade, the like that (not according to Bill Gates, anyway). as well as emphasising that “the engineer Institution was keen to ensure that computers On the other hand, and prophetically, Taff s must control any automated procedure and and IT were put to good use; it therefore mentions that “Electronic mail is now coming ensure that the fi nal result is suitable for the set up a working party in 1990. Kaethner6 of age” and predicts that “This will be a growth structure in hand” reports from this that: “At present, despite area over the next few years”. Never was there • output should be presented in a way that the shortage of traditional drafting staff and a truer prediction. an engineer “can inspect and modify the joint the sophistication of existing CAD packages, The Institution conducted an updated using a visual display unit before obtaining the surprisingly little use is made of CAD. (Perhaps survey in 1995 and a diff ering picture fi nal drawing” this is because of the overall cost of installing certainly emerged10. The survey noted the • we must ensure that any output is intelligible CAD – the training of staff being a signifi cant increasing availability of cheaper hardware and in a form suitable for designers and factor)”. and software and recorded that almost all checking engineers to understand A key paper published in The Structural engineers now had access to computers, Engineer in 1992 was the Gold Medal address with about half having sole use of a machine. All these points can be summarised as a by Professor Zienkiewicz: “The fi nite element CAD was becoming more widespread, plea to ensure that a human hand remains in method: its genesis and future”7. It’s a fair bet but seemingly remained the province of charge. Weller was not backwards looking; that this was the fi rst time most engineers had larger organisations which could aff ord the rather he looked forward to the undoubted even heard of the technique. As Zienkiewicz investment. Interestingly, the survey reports benefi ts advancing computational power points out, mathematical relaxation techniques not the de-skilling of draughting, but increasing would bring. The steelwork industry had on meshes had been around since the 1940s, dependence on skilled operatives, with a promoted this for a long time, looking towards but it was only the advent of computer power potentially widening gulf between engineers complete integration of design, detailing and that suggested a way of producing results and the drawing process. fabrication information via projects such as cheaper and faster. Would anyone even 30 There was increasing use of computers CIMsteel (computer integrated manufacturing years ago have predicted where we are today to facilitate management and administrative in constructional steelwork), which was with fi nite-element analysis (FEA)? Practical tasks and a distinct trend for engineers to introduced in The Structural Engineer in applications were reported on by Charlton in a do all their own typing. Thus, as with CAD, 19985. This was a Europe-wide collaborative paper the following year8. computers were having a noticeable eff ect eff ort aiming to meet industry challenges by Meanwhile in 1994, The Structural on the skills and employment balance across integrating the supply chain and providing for Engineer published a paper by Taff s on “The offi ces. The survey also reported an increase structural engineer and IT”9. This portrayed of data exchange, albeit stating that this was an increasingly complicated world, but Taff s still mostly by disc or tape: the internet age Figure 2 still noted that “CAD means, to most people, had yet to arrive. SOne of fi rst structural frameworks to be analysed on computer2 computer aided draughting. For many years it And, as ever, the theme of misuse was was used to impress rather than show tangible strong, the fear that the introduction of cost benefi ts. This watershed has fi nally been complicated software would eventually passed. If used correctly for simple draughting diminish the role of engineering judgement with some repetition, it will be cost eff ective”. and detract from proper application of basic Note the emphasis on simple draughting. principles. Clearly CAD was not yet a tool for universal That theme was strongly echoed in a paper use. by Professor Iain MacLeod in 199511, and it More positively, the paper also reported recurs periodically as a warning, e.g. in 2000 that “exchange of data between all parties from Dobson12: “Although technology has is becoming commonplace”. And there was advanced, there is increasing evidence that mention of word processors; cheap Amstrad analysis programs are being used without any word processors had been in production since understanding of the actual behaviour of real around 1985. structures and an unrealistic confi dence in the The 1990s was still a world where DOS results. So said the Steel Construction Institute existed and Taff s writes: “The migration from in 1995. Five years on and the issues remain DOS to Windows has been accelerating. The the same, but have we made any progress?” initial wave of enthusiasm was followed by Probably we have, in some ways. Certainly one of pragmatism. Windows implementations computer programmes now enable us to

TSE51_10-12 Intro Brief history v1.indd 11 24/02/2016 11:37 ›

12 TheStructuralEngineer Introduction March 2016 History of computing

simulate behaviour that previously could only be described. A good engineer to understand what programs can off er and to interpret their example is found in the fi eld of fi re engineering13. output. It also brings with it the danger of the “black box” mentality that all too many have warned about. Turn of the century As Livesley concluded, back in 1968, “…the prospective user of these By 2000, the profession had reached a stage where most engineers sophisticated and exciting new tools should take note of the motto had their own PC with direct access to sophisticated design and at the beginning of R.W. Hamming’s well-known book on numerical analysis packages (even FEA) and where drawings were prepared analysis: The purpose of computing is insight, not numbers”2. using CAD software with automatic scheduling. Many projects were almost working back to front from traditional practice, whereby building a complete model of the structure was fast and cheap, and thus the References starting point for a variety of proposes from analysis through to design and detailing, rather than being the end product or by-product of the E 1 Livesley R.K. (1956) ‘The application of an electronic design process. What only a few years before had been space age was computer to some problems of structural analysis’, now mundane, and the world of information and communication had The Structural Engineer, 34 (1), pp. 1–12 been completely revolutionised by the internet. We can see from papers included in this special issue that computing E 2 Livesley R.K. (1968) ‘The pattern of structural power will now enable us to design structures with increasing computing: 1946–1966’, The Structural Engineer, 46 confi dence, deploying methodologies that older engineers could never (6), pp. 177–182 have dreamed of. But all this presents the profession with many new challenges. As E 3 Bensasson S. (1978) ‘A state-of-the-art review of the 1995 survey foretold, the skill levels demanded of the profession computer programs for the detailed design of are increasing rather than decreasing. We have yet to fully arrange the reinforced concrete’, The Structural Engineer, 56A (10), management of the design process to take proper economic advantage pp. 275–277 of the benefi ts machines off er us: to use computers not just as tools for sophisticated design, but as instruments of economic effi ciency. E 4 Weller A. (1979) ‘Review of progress in computer uses That is a challenging goal because it is increasingly hard for the general for the design of structural steelwork’, The Structural Engineer, 57A (9), pp. 279–281

E 5 Garas F.K. and Hunter I. (1998) ‘CIMsteel (computer integrated manufacturing in constructional steelwork) Corporate Approved Inspector – delivering the promise’, The Structural Engineer, 76 (3), pp. 43–45 (AI) and Chartered Building E 6 Kaethner C.R. (1990) ‘Bringing technical software out Surveying Practice of the twilight zone’, The Structural Engineer, 68 (25), pp. 39–40 London based but geographical territory in and around London. E 7 Zienkiewicz O.C. (1992) ‘The ﬁ nite element method: Highly regarded Corporate Approved Inspector its genesis and future’, The Structural Engineer, 70 providing private Building Control Administration (20), pp. 355–360

and Chartered Building Surveying services. E 8 Carlton D. (1993) ‘Application of the fi nite element Allied Services: Structural engineering, Fire method to structural engineering problems’, The Risk Assessment, Party Walls, Town Planning Structural Engineer, 71 (4), pp. 55–59 Applications. E 9 Taff s D.H. (1994) ‘The structural engineer and IT’, The Turnover in excess of £306k, EBITDA £147k. Structural Engineer, 72 (9), pp. 146–148 Operates in a strong and thriving market with Building Control a mandatory requirement. E 10 Gardner P.J. (1995) ‘IT in structural engineering (results of the members’ survey on IT and computers)’, The No single account more than 9% of overall revenue. Structural Engineer, 73 (21), pp. 374–379 Operating within an industry expected to 11 MacLeod I.A. (1995) ‘A strategy for the use of experience fast, sustained and very signifi cant E computers in structural engineering’, The Structural growth over the next decade given Central Engineer, 73 (21), pp. 366–370 Government housing targets. Highly visible offi ce presence – Offi ce available - E 12 Dobson R. (2000) ‘Powerful analysis software needs skilled users’, The Structural Engineer, 78 (12), pp. 11–13 Freehold, Leasehold or rental options. Change of Lifestyle prompts sale. E 13 Bailey C.G., Burgess I.W. and Plank R.J. (1996) ‘Computer simulation of a full-scale structural fi re test’, Off ers invited by calling 07711 424051 The Structural Engineer, 74 (6), pp. 93–100

TSE51_10-12 Intro Brief history v1.indd 12 24/02/2016 11:38 › www.thestructuralengineer.org TheStructuralEngineer 13 March 2016 Methods and practice

Articles focusing on tools, techniques and approaches applicable to modern computer- aided analysis and design.

14 Time to reﬂ ect: a strategy for reducing risk in structural design 19 How to choose the right structural engineering software? 24 Structural workﬂ ows and the art of parametric design 34 An introduction to engineering optimisation methods 44 What is your structural model not telling you?

48 What is the need for veriﬁ cation, validation and quality assurance of computer-aided calculation? 52 Integration of hand calculations with computational output – a good practice summary

TSE51_13 M&P opener.indd 13 24/02/2016 11:39 ›

14 TheStructuralEngineer Methods and practice March 2016 Reﬂ ective thinking

Time to reﬂ ect: a strategy for reducing risk in structural design

Iain A. MacLeod PhD, BSc(Eng), CEng, FIStructE, FICE, Professor Emeritus of Structural Engineering, University of Strathclyde, UK

Introduction important issues for large-panel buildings, The use of computers has resulted in particularly how the panels should be immensely benefi cial changes for structural connected, were not addressed in the engineers, both at the operational level of code and hence in many, if not most, of the designing and at the conceptual level of designs. making us think more carefully about the Consequences of this lack of refl ective processes that we use and how they should thinking included: be used. However, there is much disquiet about • a major structural failure causing four the risks involved in computer use. Hazards deaths (the Ronan Point collapse; Figure 1) that may lead to faults, or even disasters, • the high cost of retrofi tting existing include: buildings that were found to be unsafe • a long-term pause in the use of a • diminished control of the design process construction method that has advantages for by engineers when relying on software some types of building (much of the process may be automated, requiring little input from the user; designers Sleipner oil platform collapse become deskilled; their understanding of In 1991 a large concrete oil recovery platform design contexts is reduced) (Figure 2a) was close to completion in a • misplaced confi dence in the potential of Norwegian fjord when a loud bang was the software to produce outcomes that will heard and the structure sank to the sea be fi t for purpose fl oor – a total loss. The fault was traced to • the potential for innovation to be stifl ed shear failure in a “tricell” wall (Fig. 2b) at a the bottom of the structure. The system had A main strategy for guarding against such Figure 1 been modelled using three-dimensional (3D) risk is to use what is called the “refl ective NFailed Ronan Point building fi nite elements (FE) (Fig. 2c), the results approach”1. This implies that one adopts a from which were used for assessment of the degree of scepticism about all received and and the potential role of refl ective thinking in shear strength of the wall. Questions that generated information; one is open to ideas; avoiding them, are discussed here. could/should have been asked include: one poses and seeks answers to questions; one makes personal assessments and Large-panel construction for buildings • Is the model that I am using more reassessments and seeks advice from In the 1960s, structural designers in the UK complex than is appropriate? The others, especially from experts; second or used Code of Practice CP114: The structural designers of the Sleipner platform assumed more opinions are sought if appropriate; use of reinforced concrete in buildings for that because they were using a complex when faults are found or improvements can the technical assessment of large-panel model it would necessarily give adequate be made, action is taken; an appropriate buildings. Many of these designers did not predictions. In the case of the tricell wall, amount of resource is allocated to seek to ask the question: “Does CP114 address all bending theory would have resulted in much ensure reliable outcomes. the issues that need to be considered for better predictions of bending moment and Use of refl ective thinking is fundamental the design of large-panel buildings that are shear force than from the model used. to good engineering practice. Computer use constructed using precast wall and fl oor • Is the mesh of elements used adequate does not diminish the need for it. panels?” If they had, the answer to that for the purpose? A review of the FE mesh question would have been a resounding in the area that triggered the collapse (Fig. Case studies of failure “No”. CP114 was written mainly for cast 2c) by a person experienced in FE modelling Two cases of failures that illustrate the risks, in situ beam-and-column structures; could have prompted an investigation of the

TSE51_14-18 Method reflective.indd 14 24/02/2016 11:40 www.thestructuralengineer.org 15

Figure 3 SApplicability of codes of practice

1960s did not realise that they were working in a context exempliﬁ ed by Fig. 3b. They were not asking the right questions.

Design process At its most basic level, the design process includes the following activities/stages:

• Inception: where information about the context is gathered and the requirements are established. a) Standard context b) Uncertain context • Conception: where a set of conceptual designs/options are identifi ed and assessed against the requirements, leading to a decision about the design solution. accuracy of the mesh for predicting bending Design context • Production: where the information needed actions. The Ronan Point collapse illustrates the to create the entity is established. • Can I do a simple check calculation? Fig. importance of considering the relationship • Review: the refl ective activity that is 2d shows a calculation, on the back of an between the design context and the scope pervasive in the process. envelope, based on treating a 1m depth of of the code of practice being used (Figure the tricell wall as a beam withstanding a 67m 3). A “standard context” can be considered In structural engineering this process can hydrostatic pressure head. The predicted as one with which the design team is wholly be applied to the structural system as a shear is three times the allowable. familiar and in which there is no innovation; whole or to a part or to details. codes of practice fully apply. An “uncertain While the process may be mainly linear, as The consequence of failing to ask and context” is one which involves any degree of shown in Figure 4, it can be deeply iterative, respond to such questions was a fi nancial innovation and/or issues that are unfamiliar e.g. in product design, a prototype may be loss of over $700M. to the design team. Safety-critical situations manufactured, tested, modifi ed, re-tested are also in the uncertain domain. The and so on. Figure 2 designers of large-panel buildings in the Traditionally “structural design” meant Sleipner oil platform collapse the use of code of practice rules to ensure that the system and its parts would perform

a) Platform prior satisfactorily. The term is now used for to collapse the process of synthesising and assessing the whole of the design information for a structure. It is better to refer to assessment using codes etc. as “technical assessment”. Technical assessment is sometimes required at the concept design stage to assess options, but it is mainly used in the production stage of design. It is said that if bad decisions are made at the concept stage, no amount of good detailing can rescue the situation. The b) Plan of part of FE mesh need for reﬂ ective thinking at all stages of the design process is therefore evident. However, it is errors in technical assessment c) FE model that have the greatest potential to result of system in major failures/disasters. Therefore this area is of the greatest concern in seeking to ensure that computer use is satisfactory.

Technical assessment The main processes in technical assessment are:

• Model development process: here the “model” is the set of rules that need to be d) Back-of-envelope calculation addressed (normally code of practice rules). The main reﬂ ective question at the model

TSE51_14-18 Method reflective.indd 15 24/02/2016 11:40 ›

16 TheStructuralEngineer Methods and practice March 2016 Reﬂ ective thinking

Figure 4 SBasic design process

of technical assessment. The validation question is: “Is the analysis model capable of satisfying the requirements?” In traditional engineering education, analysis modelling, i.e. structural analysis, was treated as a dominant issue. Students learned to do time-consuming calculations by hand. Now engineers do not do complex hand calculations. Some people are of the opinion that this results in a decline in understanding of behaviour, but the development stage is: “Have all the relevant can be considered to be states. The system accuracy and effi ciency of computer issues been addressed and are the rules model is the full set of information about processing means that complex hand used adequate for this purpose?” This is the entity being designed. The engineering calculations are in the past. We must the validation question. In standard design model is that part of the system model which therefore look to other sources for contexts not much resource needs to be requires technical assessment. understanding of structural behaviour. applied to model development, but with There is concern in the profession that Computer use is seen as the problem; in innovation, posing the validation question is for the implementation of many Eurocode reality, it is the solution. Responding to a key issue. It was here that the designers provisions, such as complex combinations refl ective questions using analysis software of large-panel buildings in the 1960s made of load cases, computer processing is the can signifi cantly enhance understanding of their main errors. The fault that caused the only feasible strategy. Some people see this behaviour. collapse of the Sleipner platform lay in the necessity as leading to a dangerous “black Suppose, for example, that you are decision about what analysis model to use. box” mode of operation where designers accustomed to analysing braced frames for • Solution process: i.e. doing the become dissociated from the calculations. buildings; you may observe that a frame of calculations. Refl ective questions here Such a mode of operation is unacceptable. this type, under uniformly distributed lateral include: “Is the software reliable?” and “Are A refl ective ethos must be adopted. Rather load, tends to defl ect as in Figure 6a. Then the input values correct?” than see computational power as a threat, you solve for a similar, but moment-resisting, • Output assessment: here the main it must be harnessed to support improved frame with no diagonal bracing and fi nd refl ective question is: “Has the model design in, for example, answering “what that the lateral defl ection is as in Fig. 6b. been correctly implemented?” This is the if?” questions and better support in the A natural reaction is to think that you have verifi cation question. investigation of options. made an error. To respond to this refl ective question, you do some experimentation by Figure 5 is a diagram of the predictive Analysis modelling varying the stiff nesses of the members of modelling process. Technical assessment The analysis model is the mathematical the frame. This leads to the conclusion that rules are predictive in the sense that they representation of the behaviour of the there is a fundamental diff erence in how the are used to assess future performance of structure. The analysis modelling process1–4 two types of frame resist lateral load. You the structure. In this diagram the rectangular has the same form as illustrated for technical fi nd out what characterises the diff erence. boxes are sub-processes and the oval boxes assessment in Fig. 5. It is a sub-process Your learning about behaviour improves signifi cantly. Using such knowledge then informs improvement in design decisions. Many people say that in modern practice there is little time for such refl ection. But if a structural designer is operating outside the standard envelope (Fig. 3b), the risk in not being refl ective is unacceptable.

Structural failures While the scope of what we deﬁ ne as structural design is, as it should be, much wider than in the past, the risk of failure is still a dominant issue. The track record for preventing major failures in the UK is very good, but the risk will always be present. Reﬂ ective thinking is a key strategy in controlling such risk. Table 1 lists seven major failures and their root causes. None of the causes listed are related to errors in doing the calculations. The faults all occurred at the model-development stage of

Figure 5 design; in each case, the validation question Modelling process was not properly addressed. Very few major failures are attributable to

TSE51_14-18 Method reflective.indd 16 24/02/2016 11:41 www.thestructuralengineer.org 17

Table 1: Root causes of some major failures Incident Year Reason for failure Tay Rail Bridge collapse 1879 Inadequate provisions for wind loading, neglect of known good practice for detailing of connections Ronan Point building (London) 1968 Code of practice used did not address collapse important issues for type of structure

Cleddau Bridge (Milford Haven) 1970 Buckling of diaphragms not included in analysis collapse model

Hartford Civic Center collapse 1978 Analysis model neglected buckling and eccentricities Hyatt Regency Hotel (Kansas 1981 Error in assumption about how loads were City) collapse distributed

Sleipner oil recovery platform 1991 Inadequate FE mesh a) With diagonal bracing collapse

Ramsgate Walkway collapse 1994 Error in assumption about moment on stub axle

Figure 7 Computer and brain power have complementary roles to play in structural design

b) No bracing

Figure 6 NLateral displacement of frames GETTY

errors in doing calculations. However, this 2) or that all parties will access a central Conclusions does not mean that errors in calculations are model (BIM Level 3). In doing this, diffi culties A computer can do calculations and repeat not made nor that improvements in methods must arise, for example, in controlling them much more effi ciently and accurately for identifying them should not be sought. revisions to the design. The brain has a than is possible using brain power. Computer While we are very aware that structural special ability to make associations and power also beats brain power at processing calculations need to be carefully checked, identify anomalies that is not replicable by logic, especially when the rule set is the lesson to be learned from Table 1 and software. While seeking to use software in complex. other major failures is that all the sub- managing the processes, the brain must On the other hand, the deep associativity processes of the modelling process (Fig. 5) continue to be used as a main source for of knowledge and other features of the brain need to be assessed in a refl ective way. controlling uncertainty. The same refl ective allows us to: identify patterns, make subtle ethos as outlined in this paper needs to be inferences, understand, have hunches, ask BIM applied to all processes whether or not the penetrating questions, generate ideas, etc. Building Information Modelling (BIM) work is in a BIM environment. BIM software Computer technology is a long way from is leading to automation in structural must be such that it is possible for designers replicating the phenomenal power of the design. More importantly, it is an enabling to get answers to questions such as: “What brain to “think” (Figure 7). technology for interdisciplinary working. assumptions were made for the analysis If a process can be defi ned as a formal A fundamental aim is that each design model?”, “What is the defl ected shape?” and algorithm, to implement it on a computer is discipline will be able to access the models “What is the bending moment diagram for the sensible approach. Software should seek of the other disciplines involved (BIM Level that beam?” to help us in our thinking – e.g. it would have

TSE51_14-18 Method reflective.indd 17 24/02/2016 11:41 ›

18 TheStructuralEngineer Methods and practice March 2016 Reﬂ ective thinking

References

been possible for the software used for the Sleipner E1 MacLeod I.A. and Weir A. (In press) Principles for computer analysis platform model (Fig. 5) to have an embedded rule of structures, London, UK: IStructE Ltd that fl agged up the fault that led to the collapse – but harnessing the thinking ability of the brain must E2 The Institution of Structural Engineers (2002) Guidelines for the use remain a central activity in the design process. of computers for engineering calculations, London, UK: Institution of This should work in partnership with formal quality Structural Engineers management systems – as discussed by Rogers elsewhere in this issue (pages 48–51). E3 MacLeod I.A. (2005) Modern Structural Analysis: Modelling Process The low incidence of major structural failures and Guidance, London, UK: Thomas Telford Ltd in the UK indicates that we do have good checks and balances to prevent them. But the key strategy E4 Borthwick A., Carpenter J., Clarke B., Falconer R. and Wicks J. of refl ective thinking, which is at the core of good (2013) ‘The importance of understanding computer analyses in civil engineering practice and is very important in engineering’, Proc. ICE Civ. Eng., 166 (3), pp. 137–143 controlling risk, tends not to be explicitly addressed in education and training. Failure of structures is, E5 MacLeod I.A. (2014) ‘The ethos of professional engineering’, Journal of course, just one of the risks that need to be IESIS (Transactions), 154, pp. 7–12 [Online] Available at: www.library. controlled by such thinking. iesis.org/2014/IESIS-trans154-paper1665.pdf (Accessed: January I believe that that the ethos in which we operate, 2016) i.e. the thought processes that guide our thinking and actions5,6, is as important as technical knowledge in E6 Lucas W., Hanson J. and Claxton G. (2014) Thinking like an engineer: the pursuit of successful engineering outcomes. Implications for the education system. Summary report, London, UK: Structural engineers who have not developed a Royal Academy of Engineering [Online] Available at: www.raeng. refl ective ethos in their work must seek to move in org.uk/publications/reports/thinking-like-an-engineer-implications- this direction. In education, teachers must seek to summary (Accessed: January 2016) instil a refl ective ethos in student project work.

BIM Products for AUTODESK® Revit®: From 3D Model to 2D R/C Drawing

Architectural Structural 3D Rebar 2D R/C Model Model Model Drawing

For further details contact +44 (0)345 370 1500 www.excitech.co.uk [email protected] www.soﬁstik.com · info@soﬁstik.com

TSE51_14-18 Method reflective.indd 18 24/02/2016 11:41 › www.thestructuralengineer.org Methods and practice TheStructuralEngineer 19 Choosing the right software March 2016 How to choose the right structural engineering software?

Rastislav Bartek Ing, Eur-Ing, CEng, MIEI, MIStructE, Associate Director, BuroHappold Engineering | London Structures

Synopsis It is, however, fascinating to see how the diff erent packages on the same project. The choice of structural engineering perception of what structural engineering software today is very wide. However, software is and what it should do has changed. Modelling/editing a lack of comparison data makes Initially, structural engineering software tools Model generation, editing operations, model it diffi cult for engineers to make a were divided into three main groups: structural manipulation, etc. used to be diff erentiators well-informed decision about which analysis software, structural design software, in the past, but are now very similar and very and computer-aided design (CAD) software intuitive in the majority of packages. software tools would be the best fi t for the production of drawings. These three for their practice or for a particular groups still exist, but they are merging with Meshing project. each other, as well as merging with design The majority of packages can handle meshing This article considers some of the tools developed for architects and other of two-dimensional (2D) surfaces quite well. factors aff ecting choice of software, engineering disciplines. Complexity of 2D fi nite elements is still a big In 2002, Winzenhoeller5 described how new diff erentiator and it is important for engineers such as technical criteria, usability software tools which were being developed to fully understand the type of 2D fi nite and interoperability, and describes could enable better collaboration between element which the software is using and its a selection tool developed by architects and engineers in integrated limitations. For example, some programs do BuroHappold Engineering to enable digital models. Those ideas have become an not have 2D fi nite elements which can be used the fi rm’s project teams to make an everyday reality in the Building Information in non-linear analysis or which can be used for Modelling (BIM) models generated and shared shell structures experiencing combined axial, appropriate choice of software when by multidisciplinary design teams on projects in-plane and out-of-plane bending, or which embarking on a new project. today. can be used for tension-only membranes. The choice of structural engineering 3D fi nite elements are only available in high- Introduction software today is very wide. There is a lot of end, advanced analysis packages which are It is interesting to search the archives of information available in the form of individual often not practical and economic to use for The Structural Engineer for papers related software reviews, but there are not many the design of standard structures. However, to structural engineering software. Articles software comparisons or software use the use of 3D fi nite elements is often essential from the 1950s to the present day vary from surveys published. for deep-piled rafts and similar structures if detailed descriptions of numerical methods Without comparison data it is diffi cult to one wants to reach economic solutions. adopted in solvers1, to warnings that structural make a well-informed decision about which engineering software should be used with software tools would be the best fi t for an Analysis type caution and that engineers should avoid engineering practice or for a project, as it The majority of structural analysis packages overreliance on structural analysis software2,3, would be very costly and time-consuming for can perform fi rst- and second-order linear and that the results of structural engineering companies to evaluate every one of them. static analysis, eigenvalue buckling analysis, analysis are only as good as the engineer modal dynamic analysis, harmonic analysis using it4. Technical selection criteria and response spectrum analysis. All of these topics and concerns are Structure type There are a substantial diff erences in the as true today as they were 20 years ago. Some structural systems (e.g. cable nets, quality and limitations of non-linear static Perhaps concern about overreliance on fabric membranes, infl atable structures, analysis. Only some packages support software is even more actual today, as complex-geometry reinforced-concrete incremental geometrically non-linear analysis, design and construction programmes are shells, post-tensioned bridge decks) can use of non-linear material models for structural often accelerated and software design only be properly analysed by a small range elements and connections, use of non-linear tools are more accessible and easier to use. of structural analysis packages. Hence, for support conditions, investigation of post- These issues will remain current as long as engineers with the ambition to design such buckling behaviour, etc. such software exists and it is important for systems, the choice of structural analysis Another big diff erentiator is staged analysis. engineering consultants and analysts to software is much smaller. Often these Some packages do not support this feature adhere to rigorous procedures and always specialised analysis packages cannot handle at all, some allow linear static analysis of verify their results using simple approximate more common structural elements, which construction stages with temporary support analogies. means engineers need to use two or more conditions, and only a very small number

TSE51_19-22 Choosing the right software.indd 19 24/02/2016 11:46 ›

20 TheStructuralEngineer Methods and practice March 2016 Choosing the right software

Future potential Figure 1 Software houses publish their future 8 Example design workﬂ ow utilising interoperability between various design, analysis and BIM packages development plans. It is a good idea for engineers to check these and see whether they align with their long-term business plans, to ensure that the investment in training and IT requirements is worth it.

Interoperability and BIM integration In the last 8–10 years, interoperability and BIM integration have become a very important requirement. This is predominantly due to a reduction in the time allowed for the design and construction stages on projects, as investors are looking for faster returns in order to reduce risks in the current, more volatile economic climate. The reduced time available means that coordination between architects and engineers, and between various engineering disciplines, needs to happen much faster than before and needs to be done more effi ciently. In addition, the reduced time available had led to a requirement for designers’ BIM models to be usable for procurement and fabrication, as there is not enough time for fabricators to produce completely new fabrication models from the construction documentation. There are numerous drivers for allow time-based non-linear analysis of the advantage is that they can also off er users interoperability, e.g. between: construction stages. structural optimisation, where the analysis The situation is similar for dynamic analysis and element design process is repeated until • diff erent analysis packages capabilities. Only a few packages support the user-specifi ed goal (limiting defl ection or • analysis packages and element design time-history analysis, damped modal analysis, specifi ed weight) is reached. software the use of various types of artifi cial damping • analysis/element design packages and data elements, boundary non-linear time-history General selection criteria processing software (e.g. Excel) analysis, etc. Accreditation • analysis/element design packages and word Other special functions are soil–structure For engineers working on overseas projects, processing software used for report writing interaction, rail track–structure interaction, and it is essential to understand whether the • analysis/element design packages and form-fi nding, which again are only covered by software they are planning to use is accredited geometry modelling software and/or CAD a minority of packages. in the region. Local authorities (e.g. the Hong software used for drawing production Kong Building Department6 or authorities in • analysis/element design packages and BIM Element design Russia and China) often require the software software As mentioned earlier, software packages for to be accredited/approved for use in the • structural and other engineering disciplines’ the design of structural elements started as region, or at least to be commonly used in design packages to allow multi-objective a separate group from structural analysis the region. Additionally, the use of specifi c design optimisation packages. Today there are still packages software might be part of the “Employer’s • the designers’ BIM models and models used which specialise purely in the design of requirements” for the project. for fabrication – provided that these are not structural elements. These can be very useful the same for one-off checks, but for large projects with User interface and usability a large number of structural elements subject An easy-to-use, logical and intuitive user Interoperability is supported either via to a large number of load combinations it is interface saves time and improves productivity various data-exchange fi le formats or the not effi cient to check elements one by one and effi ciency. application programming interface (API). The or group by group, and the trend is therefore latter requires engineers to use scripting or to automate the process. For this reason, Cost programming. This was perceived as a rare specialised element design packages tend Costs – e.g. the cost of the software licence, skill 8–10 years ago, but it has now become to off er good links to analysis packages to multiple licence deals, and maintenance, relatively common and is used by some allow automated import of internal forces/ training and helpdesk costs – are of course engineers on a daily basis. Use of the API is moments and to perform design checks as one of the key factors in selecting the right often the preferred way of data exchange bulk operations. package. It is important, however, that between programs, as it allows the data to However, the majority of analysis software engineers do not think only about the initial be preserved in its native format, rather than packages now support element design for cost, but also consider the potential cost translating it to a common exchange fi le the three mainstream structural materials: savings which the software will bring to the format which can lead to errors and is also steel, reinforced concrete and timber. Their design workfl ow. more time/resource-intensive.

TSE51_19-22 Choosing the right software.indd 20 24/02/2016 11:46 www.thestructuralengineer.org 21

Many complex and highly effi cient design single package for a single large project, e.g. a the software packages selected, depending on workfl ows have been described in project metro line or a large airport which may contain availability (Figure 2). reports published in The Structural Engineer7,8. a combination of complex buildings, bridges It is important to note that the objectivity These utilise interoperability to allow structural and tunnels, and be spread over a large area – and quality of this tool are only as good as the design optimisation as well as coordination the entire city in the case of a metro line. background data which were initially collated with architects and other engineering by members of the BuroHappold Engineering disciplines (Figure 1)8. Developing a selection tool Software Expert Community. As the tool is The fi nal part of the workfl ow is production BuroHappold Engineering Software Expert used, engineers can provide comments and it of the calculation reports. The majority of Community was recently asked to review the will be regularly updated. This should ensure its analysis and element design packages company’s structural engineering software objectivity and quality will improve over time. support automated generation of a calculation and design workfl ow. Several software report; however, only a few maintain a link packages were evaluated by offi ces in Conclusion with the word processing software (e.g. Word) diff erent regions. Some interesting structural engineering such that any changes in the analysis model For the reasons already discussed, the software surveys have been published: (e.g. the analysis results) are automatically conclusion was reached that it is not possible The Institution of Structural Engineers has refl ected in the calculation report which at present to identify a package which would published surveys of the use of computers has already been exported. Production of fi t all BuroHappold’s requirements and all in workfl ows9,10; in 1998 Modern Steel the calculation report can often be time- projects in diff erent regions. In fact, using a Construction published a table of 42 software consuming; therefore, any functionality single package could lead to the opposite packages scored according to 14 diff erent supported by the structural analysis/design eff ect and reduce agility and effi ciency. criteria (based on responses from 1000 software which makes this process more Instead there were clear technical and readers)11; and a report published by Forest effi cient can be an important diff erentiator in economic advantages identifi ed in the and Wood Products Australia contains a the software selection process. selection of “the right software for the software usage survey of 30 engineering fi rms project” at the early stage of a large project, across Australia12. Unfortunately, however, Is there one structural engineering rather than just using whatever software the these are either out of date or very region- software package which fi ts all team had used before. This is logistically specifi c. criteria? possible thanks to in-house and external The Institution has a large number of For some smaller engineering consultancies, interoperability tools which allow the project’s members around the world, involved in a there may be a “one-size-fi ts-all” solution. technical data to be stored in a “neutral” variety of projects. It would therefore be Equally, specialist designers working only format and imported to any of the leading very interesting to conduct a contemporary with a particular type of structure (e.g. tensile analysis and BIM packages. software usage survey and make the results structures, post-tensioned concrete structures To assist the project teams with selecting available to members. or timber frames) may fi nd a single package the right structural analysis software, an in- Perhaps the results of such a survey, along which covers all their requirements. house software selection tool was created. with data provided by software developers, However, for large consultancies this is This is available via the intranet to all could even be used to create a web- almost impossible due to their wide range BuroHappold structural engineers. based tool (along the lines of the internal of diff erent projects, with diff erent levels of The tool allows engineers to select features BuroHappold Software Selection Tool). This complexity, on diff erent continents, using a which are important for the project and to could become an important contemporary combination of diff erent materials, etc. In fact, defi ne their importance in the decision-making reference for structural engineers around the sometimes it is not even possible to use a process. The tool then evaluates and scores world when making software selections.

Figure 2 Extract from BuroHappold software selection tool (NB The example does not reﬂ ect a real case selection and is shown for illustration purposes only)

TSE51_19-22 Choosing the right software.indd 21 24/02/2016 11:46 ›

22 TheStructuralEngineer Methods and practice March 2016 Choosing the right software

References

E1 Brotton D.M. (1958) ‘The use of electronic digital free-form spatial roof structure’, The Structural Engineer, computers in structural engineering’, The Structural 90 (1), pp. 27–33 Engineer, 36 (9), pp. 302–309 E8 Pottinger A. (2013) ‘BIM on the Louvre Abu Dhabi’, The E2 Humphrey A.T. (2004) ‘Structural engineering Structural Engineer, 91 (11), pp. 72–77 modelling and analysis’, The Structural Engineer, 82 (3), pp. 28–32 E9 The Institution of Structural Engineers (1990) Information technology and computers in structural engineering, E3 Brady S. (2015) ‘Hartford stadium collapse: why London, UK: Institution of Structural Engineers software should never be more than a tool to be used wisely’, The Structural Engineer, 93 (8), pp. 20–22 E10 Gardner P.J. (1995) ‘IT in structural engineering (results of the members’ survey on IT and computers)’, The Structural E4 Dobson R. (2000) ‘Powerful analysis software needs Engineer, 73 (21), pp. 374–379 skilled users’, The Structural Engineer, 78 (12), pp. 11–13 E11 Anon. (1998) ‘Structural Engineering Software E5 Winzenhoeller J. (2002) ‘Pioneering software can ease Survey’, Modern Steel Construction [Online] Available the digital process’, The Structural Engineer, 80 (13), pp. at: http://msc.aisc.org/globalassets/modern-steel/ 19–20 archives/1998/01/1998v01_structural_engineering.pdf (Accessed: February 2016) E6 Hong Kong Building Department (2005) Pre-accepted Computer Programs [Online] Available at: www.bd.gov. E12 Dunn A. (2014) PRA215-1011: Structural Engineer’s Timber hk/english/inform/index_acp.html (Accessed: February Design Software Review and Recommendations [Online] 2016) Available at: www.fwpa.com.au/images/marketaccess/ PRA215-1011-Tech-Transfer-Strategy-Software-Review. E7 Epp L., Berry K. and Hart R. (2012) ‘Cairo Expo City – a pdf (Accessed: February 2016)

Looking to increase your customer base? Subscribe to FindanEngineer.com for only £10 a month

4HE)NSTITUTIONOF3TRUCTURAL%NGINEERSISOFTENTHElRSTPORT of call for home owners and organisations looking for a structural engineer.

FindanEngineer.com is a free search engine which allows UK homeowners and organisations to lnd structural engineers in their local area. For any structural ENGINEERINGPRACTICELOOKINGTOINCREASEITScustomer base, thiSWEBSITEWILLHELP. How do I subscribe? 3IGNINGUPISSIMPLE!LLWENEEDISAFEWDETAILSWHIch you caNlLLINONLINE 'OTOWWWlNDANengineer.com

TSE51_19-22 Choosing the right software.indd 22 24/02/2016 11:46 In the hands of Tekla users, architect’s drawings evolve into constructable 3D models, that develop into real buildings. That’s made possible by Tekla software and Trimble.

Together we are shaping a smarter future for construction.

Join the evolution. tekla.com/evolution

Aki Luntamo BIM Master, Sweco Tekla BIM Awards 2015 winner

Tekla software brought to you by Trimble The EVOLUTION OF THE LINE

p23_TSE.03.16.indd 23 19/02/2016 16:01 ›

24 TheStructuralEngineer Methods and practice March 2016 Parametric design

Structural workﬂ ows and the art of parametric design

Jon Leach MEng, CEng, MIStructE, MICE, Director, AECOM Rossella Nicolin PE, LEED AP BD+C, Principal Engineer, AECOM Figure 1 Matthew Burton MEng, CEng, MICE, Senior Engineer, AECOM Typical parametric workﬂ ow process

Introduction When considering lean design and construction in the building industry, we often draw inspiration from the manufacturing industry. But despite many positive moves in recent years, the construction industry – and building design in particular – is one which often requires bespoke, client- and site-specifi c solutions and not commodity production. Nevertheless, by dissecting and interrogating the whole process of creating building structures, we can still draw effi ciency and marginal gains at each step of the way. As design engineers, our focus is often on day-to-day problem-solving in relation to a particular project or engineering challenge. Perhaps less frequently do we consider the same application of engineering to the design process itself. This article will explore how we in AECOM Role of technology in holistic design To overcome these barriers, engineers at are currently using technology to improve optimisation and communication AECOM are encouraged to broaden their the effi ciency of the design process, Buildings are becoming ever-more complex horizons and make the eff ort to understand while at the same time empowering and packed with technology in their design, other disciplines’ requirements in a more the structural engineer to be more procurement, construction and operation. holistic and integrated manner. Common creative, and take a more central role on The technology in a modern building means “hubs” of intelligent data can be used not multidisciplinary projects. Case studies we have specialist groups and subject- only by the structural engineer but by range from large-scale stadium projects matter experts for every facet of the design, all disciplines. Key to this is that most of down to small, but complex, pavilions, adding more and more parameters and the work described here is planned and and how the methods can be applied to constraints that need to be considered and executed by our structural engineers, and other projects through a cultural shift that incorporated. not outsourced to third-party specialists or capitalises on the accessibility of digital A common issue in large design teams is geometry experts. technology. a “silo” mentality, with people focusing only Within this context, Building Information Parts of this article are based on on their core disciplines, most often due to a Modelling (BIM) is widely acknowledged to previous papers by the authors: “The combination of contractual constraints and have shown potentially massive benefi ts for future of optimisation in engineering of their own knowledge and comfort zones. the industry, but the “BIM” that many people sports structures”1 and “The use of digital However, even in the most collaborative of are getting excited about is essentially just workfl ows in the design of the 2013 and teams, one of the biggest barriers is the use a large database of information, which is 2014 Serpentine Pavilions”2, as presented of incompatible software platforms, which worthless without considered application. at the IASS 2015 Annual International reduces the amount of data and information It is how you apply the data, and more Symposium on Future Visions. that we can share as a design team. importantly utilise appropriate data in the

TSE51_24-31 Method Leach v1.indd 24 24/02/2016 11:47 www.thestructuralengineer.org 25

Figure 2 Grasshopper script for generating truss options from same centreline

Figure 3 WParametric structural concepts for long-span roof, created for multipurpose arena competition, generated and conceptually analysed in four days

Figure 4 EParametric structural roof options for same span and loading constraints using shape ﬁ lters

fi rst place, that results in smarter design. (Figure 1). The programming environment of the interoperability processes. This initial One of the most important changes that the is designed to transform initial inputs geometry defi nition process is based not advent of BIM has brought is to make people into a structural geometry, which is then on complex analysis but on experience, think diff erently about how they can use redirected into several output streams and simplifi ed models, engineering judgment, technology to spend less time on repetitive interoperable software for detailed analysis, load path predictions and rules of thumb, data handling and more time on applying multidisciplinary coordination, technical and it is key for achieving an integrated their experience and judgement to optimise documentation and scheduling. collaboration between architect and their solutions. The Grasshopper script is built with a engineer. In our team we have successfully applied central core defi ning the structural centreline A typical example is shown in Figure 2, an integrated digital workfl ow to a series of geometry, which then feeds into several illustrating the creation of roof trusses from both large-scale and small-scale projects. subcomponents, aimed at directing the an initial architectural curve. Within the parametric geometry into distinct output Grasshopper environment, we can develop a Parametric design workfl ows streams and other software. While the truss prototype by introducing variables such Large-scale projects – long-span roofs geometry defi nition can involve diff erent as the required structural depth at defi ned With stadium and long-span roofs, the levels of complexity and extra programming points, lacing pattern and curved profi les of structural geometry is deeply intertwined specifi c to each project, the interoperability truss chords. Other more complex examples with the expression of the architectural output streams are standard programming are shown in Figures 3 and 4, where shape, and round-trip parametric components which can be applied to from the same architectural surfaces and optimisation techniques are vital to allow any project and signifi cantly optimise our functional requirements we created diff erent various design options to be tested and workfl ow. structural roof options using a combination prototyped from concept design through to The structural geometry defi nition is the of basic shape fi lters and additional construction documentation stages. fi rst step of the process. It is where our geometry components. In order to respond effi ciently to these structural engineers elaborate the sketch Once the structural geometry is defi ned, processes, we have developed our own ideas, defi ne constraints and opportunities, the other components of the script are scripts, principally based around Rhinoceros and create a concept with structure driving used to direct the information into further 3D (Rhino)3 and its visual algorithmic plugin the form. This is the key stage in the processing, most importantly structural Grasshopper4. process, where the engineer establishes the analysis and documentation. These scripts act as data hub with a parameters which will inform the geometry For the detailed structural testing, we have multifunctional, multi-objective approach defi nition, and hence forms the cornerstone created a direct link between Grasshopper

TSE51_24-31 Method Leach v1.indd 25 24/02/2016 11:47 ›

26 TheStructuralEngineer Methods and practice March 2016 Parametric design

Figure 5 Grasshopper subcomponent for SCIA Engineer export

model intelligence through all stages of the process, we have recently been able to do a complete round-trip from updating primary geometry parameters to updating the full documentation of a complex major stadium roof (Figure 7), including all trusses, secondary steelwork and glued laminated timber (glulam) rafters in less than ﬁ ve days. The majority of that time was spent carrying out the structural analysis and engineering optimisation, including construction-stage analysis and connection checks; the modelling and processing takes only a fraction of that time – a matter of hours depending on the extent of the change. It is noted, however, that if the parameters themselves were to change, this would obviously be more time-consuming. Determining the key design parameters at an early stage is absolutely key to the success of this process.

Structural design of stadium bowls A stadium seating bowl design is governed by a huge array of factors and functional requirements, most of which are interrelated by complex geometrical parameters. In our experience, most stadium architects now use scripting of some form to generate the baseline geometry, and add layers of bespoke parametric tools with their own experience to rapidly test options and hone in on the optimum solution. This includes calculations of sightlines, seat and handrail positions, escape distances, net capacity for various events, optimising fl exibility, pitch growth and occupant comfort, and ensuring compatibility with the overall architectural vision. The product of those parametric studies is typically a basic surface, received in various formats, which defi nes the surface of the precast concrete terrace units or seating “bents”. Workfl ows are normally then split, with the architect developing the design above the surface (Figure 8) and the and our analysis software, SCIA Engineer5 described here to be more manageable and structural engineer developing the terrace (Figure 5). Structural analysis parameters stable. Only when the analysis software is unit and raker frame design below (Figure 9). such as analytical centrelines, local axes, being used to carry out intensive automated Traditionally, the most time-consuming eccentricities, releases, restraints and optimisation would we gain real benefi ts aspect for the engineer is the generation of loading are defi ned in the Grasshopper from a two-way link, and that process would the structural geometry, especially for bowls canvas, written as XML database fi les and adopt a diff erent workfl ow typically. with radial grids, parabolic section geometry exported into SCIA for design (Figure 6). Any By employing a common nomenclature and cookie-cutter top profi les. This is changes made in SCIA are controlled from and reference system for all nodes and particularly true when the architecture will the Grasshopper hub and then analytical members, one of the most complex tasks continue to evolve and change throughout results, such as connection forces and in the process, the same process can be the project duration, as stakeholders and limiting temperatures, are pushed to other applied to export geometry information brief requirements evolve too. As for the fl ows for inclusion in schedules. A two-way into Autodesk Revit6 or other platforms for roofs, we adopt Rhino and Grasshopper as a link between SCIA and Grasshopper for project documentation. data hub to take the basic bowl surface and data other than results (e.g. geometry) is In terms of the parametric round-trip defi ne the geometry of all of the supporting possible, but we have found the method process, whereby we can maintain the structure. This includes the precast

TSE51_24-31 Method Leach v1.indd 26 24/02/2016 11:47 www.thestructuralengineer.org 27

Figure 6 Examples of structural models for Rio 2016 Olympics

a) Tennis centre b) Velodrome

terracing, raker beams, raker top proﬁ les and Figure 7 primary columns. Al-Wakrah Stadium, Components are all sized from a pre- Qatar – Zaha Hadid and AECOM Sports Architects optimised data set using look-up tables for varying spans and loads. An optimisation script then rationalises the precast units in order to limit the number of diff erent components. For example, it is typically desirable to arrange precast units into groups of two to four rows in order to limit the number of diff erent moulds and jigs required; tolerances can be addressed through the use of shims and mortar packing on site. It is inevitable that a degree of manual post-processing is required, especially when developing the full documentation model; there is a point at which all stadia are subtly diff erent and the beneﬁ ts and effi ciencies of ever-more complicated scripting become less. But as other design disciplines embark on using these techniques, more effi cient automation will become possible.

Beneﬁ ts to multidisciplinary design Modelling in this manner allows concept schemes to be quickly analysed to determine feasibility, a swift response to architectural

Figure 8 EBowl geometry – architectural input

TSE51_24-31 Method Leach v1.indd 27 24/02/2016 11:47 ›

28 TheStructuralEngineer Methods and practice March 2016 Parametric design

Figure 9 Bowl geometry – “mapping” structural components to architect’s geometric surface option studies, and rapid production of design documentation. At the early stages of design, engineering models can be brought together with the architectural models either in native fi le formats or in software such as Autodesk Navisworks7, to provide high- quality coordinated visuals of feasible and credible design options. Upon agreement of the overall form and key parameters, the structural engineer then takes ownership of the controlling geometry to allow detailed design to commence. Later in the design process, the same federated models can be used for design coordination, clash detection and documentation. The basic “cleaned up” geometry can then be utilised by all design disciplines in various formats, including the construction of physical models for wind-tunnel testing, acoustic modelling, computational fl uid dynamics and energy-modelling packages for comfort and pitch growth studies, putting our engineers in a central role in the design process (Figure 10). Similarly, a combination of fi re-modelling, heat-transfer and clash-detection software can determine fi re-protection requirements in performance-based structural fi re the architectural vision of the cloud-like fragile and distressed looking structure, akin engineering design. space frame. Without the use of these to the papier mâché used in his initial small- Constructability can also be reviewed digital techniques, the design could not have scale models. We proposed glass-reinforced using the same techniques. We have been realised in the timescale available: plastic (GRP) as an engineering material that developed scripts in Grasshopper which use the Pavilion being a temporary structure could match the layered appearance of the preloaded information on generic lifting and designed in six weeks, built in six weeks and papier mâché, using diff erent numbers of access equipment including practical lifting in place for just four months. layers to provide texture to the surface and capacities, reach and headroom constraints. The original design was conceived on also to vary the opacity of the structure. By combining a simple rule set on lifting a 400mm cubic module using 15mm solid As with the 2013 Pavilion, our structural capacity and reach for a certain crane, for steel bars. Using parametric design scripts engineers led the creation of the parametric example, with the component weights taken the engineers could quickly assess both model to control the geometry in a manner from the structural model and 3D clash- the geometrical and logistical issues. Using that could allow rapid prototyping, exchange detection techniques, this information helps the parametric model as a visual tool, the of data with the architect, and direct insertion to inform the conceptual build sequence engineers were able to demonstrate that of the geometry into the contractor’s moulds where site constraints have been identifi ed omitting members to create an 800mm and fabrication tools. and allows us to consider these issues from cubic module did little to depreciate the With the architect working remotely from the early design stages in an integrated overall form and impact of the structure, the engineers, regular communication was manner (Figure 11). and in fact provided better connectivity very important. Value engineering led to to the surrounding views and landscape. geometry changes and reductions in surface Serpentine Pavilion 2013 Furthermore, by adding back members area, and options were presented to the The technologies and methods in question in selected areas, including 400mm and architect in an eff ort to justify the reduction are equally applicable to small-scale 200mm step modules for access, the of GRP as creating more “distress” in the projects, a perfect example being the architect could vary the density and visual fragmented structure. Serpentine Pavilion, an annual commission texture of the structure to create more We developed a process, in collaboration by the Serpentine Gallery in Hyde Park, refi nement and interest. This resulted in with Nemetschek, which can take defi nable London. a change from approximately 35 000 to panels from Rhino into SCIA Engineer (Figure The multiple award-winning 2013 27 000 20mm × 20mm steel members, 14) via a bespoke XML script that further Pavilion (Figure 12) was one of the most which made the scheme viable in the implemented the procedures established successful to date and the parametric timescale available. for the 2013 Pavilion. The application of this design processes that were developed provided control over the analytical meshing have become a fundamental part of our Serpentine Pavilion 2014 process, and created geometry which could day-to-day workfl ows. Designed with Sou The 2014 Pavilion design (Figure 13) joined be set out and eff ectively rationalised before Fujimoto Architects, the AECOM team used AECOM engineers with Chilean architect the fi nal shop design was carried out by GRP parametric design techniques to realise Smiljan Radic. Radic’s vision was to create a specialists Optima.

TSE51_24-31 Method Leach v1.indd 28 24/02/2016 11:47 www.thestructuralengineer.org 29

Figure 10 Figure 12 Data hub conceptual diagram SSerpentine Pavilion 2013

a) Photograph

b) Structural model

For both the 2013 and 2014 Pavilions Figure 11 E Constructability (and indeed the recent 2015 Pavilion), the study digital models developed by AECOM have been exchanged on a regular basis with the contractor, Stage One, and were used with their fabrication software to create full- scale node and connection modelling and mock-ups of all of the major components. The parametric workﬂ ow established by our engineers facilitated regular liaison between all parties to ensure that the materials and details met the necessary aesthetic and performance requirements.

Documentation output and project delivery In addition to linking structural analysis and geometry, we have linked the Grasshopper output with Autodesk Dynamo8 in order to data can be pushed into the Revit model towards BIM delivery in the form of 3D create the structural components as native where it is deemed benefi cial, although data-loaded models. Those models can, Autodesk Revit objects, with parameters heavily data-laden models can still prove in turn, be cut and annotated to produce such as material grade, fi re protection, slow to handle. “traditional” 2D drawings, with the data limiting temperature, connection forces At present we adopt Grasshopper as scheduled as required to produce the and reinforcement quantities preloaded or opposed to Dynamo for the main data contract deliverables. pushed through from the structural analysis hub as it is currently a more developed With current technology and software to allow automated scheduling. toolkit, particularly in relation to geometry understanding in the industry, structural Thus, during the documentation phase, the generation and confi dence in cross- engineers are in a transition period where same hub is used to extract and hold data discipline model sharing. However, we consultants are frequently required to “over- from the analysis model. Ideate BIMLink9 is expect the use of Dynamo and visual deliver”; issuing both data-loaded 3D models then used to collate those schedules from scripting in general to become much more and “traditional” deliverables. This is time- the various sources. Extracting the data in prevalent as a user interface for all forms of consuming and often wasteful, particularly Excel format gives us fl exibility for how the generative design software in the next few for structural components such as steelwork data is presented and used, be it for pricing, years. where most major fabricators will now fabrication, logistics or otherwise. The same The methods described here are geared take the 3D model as an IFC fi le export

TSE51_24-31 Method Leach v1.indd 29 24/02/2016 11:47 ›

30 TheStructuralEngineer Methods and practice March 2016 Parametric design

Figure 13 Figure 14 Serpentine Pavilion 2014 Import of Rhino shell into SCIA Engineer

and use that model as the basis of their and joint details. shop geometry. If the model is produced This allows us accurately, the 2D general arrangement to move much drawings should serve little purpose other closer to having than fulfi lling dated contractual obligations. all deliverables This is something we are seeking to address contained inside, or in AECOM by creating a Digital Execution at least linked into, Plan with “Level of Detail” and “Level of the 3D model. Information” matrices for each project work- At AECOM, we stage intelligently linked back to a project have trialled the scope that resides in our appointment use of exclusively document. 3D deliverables on The workfl ows that we have established a number of small aim to give us the best of both worlds. projects, e.g. on the Loading data into the models allows the 2013, 2014 and 2015 more progressive contractors to use the Serpentine Pavilions. data directly from the model. However, This was very this organisation of data also means that successful and the extracting 2D schedules of the same data, project team agreed and 2D drawings referencing, for example, that the quality of node and member identifi ers, is relatively the models meant Figure 15 straightforward. The scope for human error there was no need Multidisciplinary parameters for long-span roof is drastically reduced and the impact of for traditional 2D change, in terms of the time taken to make deliverables. It is only a matter of time before on much larger developments where updates, is also signifi cantly less. this becomes a realistic prospect for much budget has allowed for the expression of Contractor-designed details such as larger projects too; as an industry we should the more organic shapes and forms that connections and ancillary items are often be demonstrating to clients and contractors tend to be produced. Focusing on material still delivered in 2D format, and are then that this is a feasible option. effi ciency alone is unlikely to yield the most designed and detailed in the fabricator’s economical or sustainable design. However, 3D shop model. However, as we build Design optimisation using parametric with the advent of technologies such as up libraries of details across numerous design 3D printing, the creation of more organic projects, setting member cutbacks and There are various structural optimisation forms and the application of material only applying standard connection details for techniques available using current software, where it is needed structurally becomes the purpose of showing the design intent from basic auto-design routines using pick- more viable. For example, these shapes is now a credible option and adds little lists of predefi ned section sizes, to more traditionally rely on heavy castings or heavily time to the modelling process. Intelligent complex methods such as topography in fabricated welded sections with signifi cant tagging and linking inside models can still solid fi nite-element modelling. wastage, but if we adopt a mindset of using allow external referencing to a level of detail These more complex methods are very additive fabrication methods as opposed to that is undesirable to be contained in the exciting. They have been used on a number traditional subtractive methods, we can think main model fi le, e.g. grouting, waterproofi ng of relatively small-scale projects, but also more about the viability of using these more

TSE51_24-31 Method Leach v1.indd 30 24/02/2016 11:47 www.thestructuralengineer.org 31

irregular and material effi cient forms in the judgement of an experienced engineer. What pavilions and, most recently, bridge design future. these algorithms can do, however, is give us in relation to highway and permanent-way Considering multidisciplinary design options and combinations that we may not (track) alignment variables. Once people are parameters in structural design, at a simple have previously considered, so they can still be comfortable with the technology, they see the level deeper steel beams may be lighter, but powerful creative tools. benefi t of using it day to day on any project building height and cladding area may increase that has numerous interdependent design and ceiling heights and natural daylight may Conclusions variables. reduce. The use of technology in the design process By interrogating all aspects of our At a more complex level, we have the case can make it hugely effi cient, but the use workfl ows, both internally within our own study of a long-span roof (Figure 15), with the of automated processes and powerful discipline and externally between other key design parameters including structural computational analysis is certainly not the end disciplines, we can fi nd effi ciency and focus on effi ciency, constructability, microclimate, of the skilled architect, engineer or technician. delivering better solutions. This is an exciting environmental comfort, acoustic performance Instead the roles are evolving, and technology development. It inspires our engineers to truly and, more subjectively, the architectural form allows us to spend more time on the important feel part of the creative process, and it helps of the building, which can play a fundamental issues – developing robust and deliverable to reinforce the essence of the good engineer, role in attracting people and revenue to the concepts, coordinated design, optimised one who takes an interest in all aspects of the building. design, pushing new innovations – and not design and not just their chosen “specialism” A number of academic groups have been on the more tedious process of manual data of structural engineering. investing signifi cantly in developing genetic management and geometry defi nition. In and machine-learning algorithms (e.g. work addition, it allows us as structural engineers to Acknowledgements by Delft University of Technology10,11), which play a key role in the shape generation and in AECOM structural software development: allow computers to determine the optimum the development of the architectural concept Ricky Feigin solution while considering numerous design with dynamic and informed input. AECOM project images: Tom Gaunt, Martin parameters and constraints, and looking at While the need for speed and effi ciency Fowler, Paul Witham, Thomas Webster, Harriet ways of using software to place the design has driven us at AECOM to use these Eldred, Gary McCarthy, Madalina Aghinitei parameters in a hierarchy to give the process methods on some of our biggest and most AECOM Sports stadium bowl scripting more direction. complex projects, we are now exploring how images: J Parrish In reality, we are many years away from these methods can be used on all building Multidisciplinary optimisation: Michela Turrin, artifi cial intelligence being able to replace the typologies, including small-scale, but complex, Delft University of Technology

References E1 Leach J., Nicolin R. and Webster T. (2015) ‘The future www.autodesk.co.uk/products/navisworks/overview of optimisation in engineering sports structures’, Proc. (Accessed: January 2016) IASS 2015 Annual International Symposium on Future Visions, Amsterdam, The Netherlands, 17–20 August E8 dynamobim.org (2016) Dynamo [Online] Available at: http://dynamobim.org/download/ (Accessed: January E2 Leach J., Burton M. and Webster T. (2015) ‘The 2016) use of digital workﬂ ows in the design of the 2013 and 2014 Serpentine Pavilions’, Proc. IASS 2015 E9 Ideate Software (2016) BIM Link [Online] Available at: Annual International Symposium on Future Visions, http://ideatesoftware.com/ideatebimlink (Accessed: Amsterdam, The Netherlands, 17–20 August January 2016)

E3 Robert McNeel & Associates (2016) Rhinoceros E10 Yang D., Sun Y., Turrin M., von Buelow P. and Paul J. 3D (Rhino) [Online] Available at: www.rhino3d.com (2015) ‘Multi-objective and multidisciplinary design (Accessed: January 2016) optimization of large sports building envelopes: a case study’, Proc. IASS 2015 Annual International Symposium E4 Robert McNeel & Associates (2016) Grasshopper on Future Visions, Amsterdam, The Netherlands, 17–20 [Online] Available at: www.grasshopper3d.com August (Accessed: January 2016) E11 Turrin M., Sariyildiz A. and Paul J. (2015) ‘Interdisciplinary E5 Nemetschek (2016) SCIA Engineer [Online] Available parametric design: the XXL experience’, Proc. IASS at: www.scia.net/en/software/product-selection/scia- 2015 Annual International Symposium on Future Visions, engineer (Accessed: January 2016) Amsterdam, The Netherlands, 17–20 August

E6 Autodesk (2016) Revit [Online] Available at: www. Further reading autodesk.co.uk/products/revit-family/overview Leach J., Nicolin R. and Webster T. (2015) ‘Structural engineers (Accessed: January 2016) as creative leaders – from stadia to pavilions’, Proc. IABSE Conference 2015: Structural Engineering – Providing Solutions to E7 Autodesk (2016) Navisworks [Online] Available at: Global Challenges, Geneva, Switzerland, 23–25 September

TSE51_24-31 Method Leach v1.indd 31 24/02/2016 11:47 ADVANCED STRUCTURAL ENGINEERING SOFTWARE

Today’s engineers need to do more, do it faster, and with a greater degree of accuracy than ever before. For structural engineers working on buildings, bridges and tensile structures, a comprehensive design program is needed – GSA is the right solution.

Download your free 30 day trial version from our site today

oasys-software.com +44(0) 191 238 7559 [email protected] Oasys Software @Oasys_Software TheOasysSoftware

p32-33_TSE.03.16.indd 32 19/02/2016 16:03 OASYS GSA - SOFTWARE FOR AWARD WINNING BIM DRIVEN PROJECTS The SSE Hydro was the winner GHȴQHWKHFRPSOH[URRIJHRPHWU\ZLWK of the IStructE Structural Awards Region the remainder of the structure developed As engineers and architects Groups Award for projects that have LQ%HQWOH\VROXWLRQVWKLVZDVWKHQDOO push the boundaries with new GHPRQVWUDWHGEHQHȴWVWRWKHFRPPXQLW\ brought together to create one model for FUHDWLRQV*6$R΍HUVDVROXWLRQ and had a positive impact on lives. The DQDO\VLVDQGGHVLJQ)URPWKHUHVXOWLQJ which is quick, accurate and multi-purpose indoor arena has a 12,500 GSA model, piling schedules and steelwork %Ζ0FRPSDWLEOHΖWLVVSHFLȴFDOO\ FDSDFLW\DXGLWRULXPWKDWSURYLGHVD centreline lengths were derived in order to designed to connect with a FRPELQDWLRQRIȴ[HGUHWUDFWDEOHDQG aid the calculation of the overall material range of other leading software removable seating to enable a versatile TXDQWLWLHVUHTXLUHG7KH%Ζ0PRGHOȇV packages, allowing GSA to play OD\RXWIRUDZLGHUDQJHRIHYHQWV7KH intelligence was used to communicate a key role on projects of all sizes structure was designed to provide the best information between the design team and and demands. YLHZRIWKHVWDJHIURPHYHU\VHDW the contractor/sub-contractors, reducing risk on the project and supported the All of the IStructE Structural 7KHPDMRULW\RIWKHGHVLJQRIWKHQHZ FOLHQWȇVȆWUDGLWLRQDOȇSURFXUHPHQWURXWH7KH Award 2015 winners were Oasys arena took place when BIM was still use of BIM was central to the success of this customers and we’ve highlighted emerging as an important process for the JHRPHWULFDOO\FRPSOH[EXLOGLQJZLWKLWVIXOO two particular winning projects HQJLQHHULQJLQGXVWU\DQGVKRZHGLWVYDOXH 3D co-ordination of the structure, M&E and that illustrate the versatility of even then. GSA parameters were used to architecture. GSA.

The Vegas High Roller won ZDVDȆRQHRILWVNLQGȇSURMHFWLQPDQ\ZD\V the IStructE award for Arts or Entertainment PHDQLQJWKHPDWHULDOVZHUHUHTXLUHG 6WUXFWXUHVIRUH[FHOOHQFHLQWKHVWUXFWXUDO to be custom made, the demand for a GHVLJQRIUHOHYDQWLQGXVWU\VWUXFWXUHV7KH successful BIM model and coordination in tallest observation wheel in the world and 3D was therefore vital. GSA was used to its pioneering features including double SURYHHDUO\GHVLJQFRQFHSWVXVLQJDEHDP glazed spherical cabins supported from a HOHPHQWPRGHOWKHJHRPHWU\IRUZKLFK single tube rim all came together with the was imported from a model created with support of GSA. %HQWOH\6\VWHPV*HQHUDWLYH&RPSRQHQWV Integration between GSA and a number of Clockwise: SSE Hydro Complex Coordination, SSE 7KHFRPSOH[LW\DQGSUHFLVLRQUHTXLUHGIRU RWKHUWKLUGSDUW\WRROVLQFOXGLQJ%HQWOH\ Hydro BIM Communicating Design (GSA model), such a project were developed, designed 6\VWHPV5KLQRDQG1DYLVZRUNVDOODVVLVWHG Las Vegas High Roller Observation Wheel model, DQGSRUWUD\HGXVLQJWKHPRVWDGYDQFHG ZLWKWKHGHOLYHU\RIWKHSURMHFW Las Vegas High Roller Observation Wheel BIM practices available. The High Roller

p32-33_TSE.03.16.indd 33 19/02/2016 16:03 ›

34 TheStructuralEngineer Methods and practice March 2016 Design optimisation An introduction to engineering optimisation methods

Peter Debney Dip(comp), BEng, CEng, MIStructE, Arup (Oasys)

Figure 1 Two-beam problem Synopsis Some engineering problems are simple, like linear analysis; others are diffi cult, like non-linear analysis; but there is a third group: those that are complex. Complex problems are those where there are many possible answers that have to be explored and assessed before a decision is made as to which is the best one. This article will discuss the principal concepts of design optimisation, then look at the various suitable techniques and make suggestions as to where they might be used by structural engineers. These methods include quasi-Newton, gradient, simulated annealing, Monte Carlo, exploration of the many options feasible within 100 options, and so on. Each location within 4 genetic algorithms, particle swarms, a project timeframe . this design space can then be assessed for a Just as there are many possible solutions number of criteria, such as design utilisation neural networks, form-fi nding, and to a particular design challenge, there are and defl ection, and the best one chosen. evolutionary topology optimisation. also numerous techniques to help us fi nd the In the case here (Figure 1) it would not take While the article will not be optimum design. As we will see later, there is a computer long to check every single option exhaustive (which would take several no one best method, but each has advantages in the design space for one or two beams, books), it will provide suffi cient and disadvantages for particular applications. but these exhaustive methods would quickly Determining which is which is part of the fall foul of what is known as a combinatorial examples and typical formulas so that challenge and pleasure of these techniques, explosion: the number of possibilities grows those interested can start to explore but then you must expect some complexity exponentially with the addition of every new this fascinating subject. in the solution when addressing complex member. problems. This is not a problem with all structures, Introduction because if you prevent the members from One of the hardest and most important Design space interacting then you can design them aspects of structural engineering is the The “design space”, which goes by a number individually, but you might get a more effi cient initial design, followed by the optimising of of names, including “search space” and design by allowing interaction. For example, that design to reduce material use, cost and “phase space”, is a useful way of thinking compare a pinned steel frame to a moment environmental impact1. Design optimisation about the range of possible solutions to a frame. Moment frames will usually result in a is complex though, as there are a plethora of problem. It has one dimension per variable, lighter design, but at the expense of bigger choices to be made2, with each improvement so that, for example, if there are 100 possible connections. possibly having a detrimental eff ect on other steel sizes on off er for a particular beam, then To optimise a design you need to decide aspects of the construction3. For example, the design space has one dimension with what it is that you mean by optimise. This fi xing a portal column base will lighten the 100 locations along it. Similarly, if there are might mean the least-weight or least-cost steelwork at the expense of increasing the two diff erent beams then the design space structure that carries the load, or something size of the foundation. Computers make becomes two-dimensional (2D) with 100 × more project-specifi c. You then need to assign

TSE51_34-41 Method Equations v1.indd 34 24/02/2016 11:49 www.thestructuralengineer.org 35

Figure 2 Figure 3 Proximity to unity for two-beam problem – reﬁ ned beam selection Proximity to unity for two-beam problem – all universal beams

Figure 4 Two-beam solutions

a qualitative value to any particular solution so too small will not provide a useful result. it to a linear bending structure would have that you can compare them. This also means increased the self-weight by an order of that you can think of the design space as a Design optimisation methods magnitude. terrain, where every point in the space has an As with any analysis, the fi rst stage in So what heuristics might we use to elevation related to how optimum it is (Figures any optimisation process should always address these problems? Let us start with 2 and 3), and thus the location that is most be a manual one: what is the problem the quasi-Newton method. optimum is the highest point. The problem is: and how can I simplify it? As Einstein is how do we fi nd that peak? To make it more famously quoted as saying: “Everything Quasi-Newton diffi cult for us, there are no maps and we are should be made as simple as possible, but The quasi-Newton method is an iterative totally blind. All we do know is where and how not simpler”5. You need to optimise the process where each repetition gets you high we are. How we explore the design space optimisation to get the most effi cient results. closer to the desired result8. Consider a for that peak is the essence of many of these What does this mean in practice? It may steel moment frame: analysing the structure optimisation heuristics (search methods). mean reducing the load cases down to just and then designing each element for the Another problem is that there can be many the signifi cant ones, or excluding parts of resulting forces and moments will give a peaks, so we must be aware of premature the structure that are far stiff er or more new arrangement of section sizes and thus convergence, where a method might fi nd a fl exible and thus can be ignored or fi xed stiff nesses. The now stiff er elements will foothill and ignore the mountain. as appropriate. If the structure has lines of attract more load, possibly requiring even Note also that the design space can be symmetry then use them to group members heavier sections. Eventually the sections more than just the range of beam sizes, as to have the same section, or even remove will not change from one iteration to the it can include material choices, structural those duplicated parts of the structure and next, and we will have thus converged on a layouts, and more. Establishing exactly what replace them with suitable restraints. solution. the design options to be searched are is a key Conversely, oversimplifi cation will harm In the two-beam example in Figure 4, let step in the optimisation process. Be warned the optimisation. For example, Expedition us look for the least-cost solution. To do this though: making the design space too big will Engineering’s 2012 Olympic velodrome roof we can take the latest steel prices from the greatly slow the process, possibly stopping achieved a weight of only 30kg/m2 as it was Tata Steel website9, then sort the universal you from fi nding the answer, while making it a non-linear tension structure6,7. “Simplifying” beam (UB) sections by cost per metre and,

TSE51_34-41 Method Equations v1.indd 35 24/02/2016 11:49 ›

36 TheStructuralEngineer Methods and practice March 2016 Design optimisation

1. Choose starting sections. 2. Analyse model. 3. Find lightest/cheapest sections that carry the load. 4. If sections from #3 are the same as at #2 then stop; else update analysis sections to match #3 and repeat from #2.

Mapping all the optimum solutions and the tributary areas (Figure 5) shows that there are ﬁ ve possible answers, the chosen solution depending on the starting point in the design space.

Gradient/hill climb As mentioned earlier, we can think of the design space as a terrain in total darkness. The gradient or hill-climb metaheuristic (story explaining the heuristic) says that the way to the peak is to head uphill until you cannot go any higher. With a simple problem, like fi nding the lightest steel beam that can carry a load, you can put the beams in order of weight, then start at the bottom (lightest) and work your way up, testing each beam, until you fi nd the fi rst one that works12. If you look at the terrain of UBs in weight order against, say, I-value or bending capacity, then you will see that the surface is far from smooth. For restricted problems, such as beam bending, there are sections that could be left out, reducing the number of sections to be checked (Figure 6). In the two-beam example, this reduction will also take the problem design space from very rough (Fig. 3) to reasonably smooth (Fig. 2). Figure 5 When you have a design space of two NQuasi-Newton tributary areas for dimensions or more, there is no obvious route optimum designs up the slope, so how do you determine the gradient? One option would be to choose a direction at random and see if that generates a better solution; another is to check all adjacent points to see which is the local best and move to there. When you have checked all adjacent points and found that none are better, then you must be at the top and you can stop. The gradient method is fi ne if the design space has a single, smooth hill, but imagine that it is more varied, e.g. like Great Britain. If you start somewhere fl at, like Norfolk, it is

Figure 6 diffi cult to determine which way is the direction WUniversal beam of the upward slope. Alternatively, if you start moment capacity/weight somewhere like London, then the hill climb may take you as far as the North or South Downs (hills), but nowhere near the mountains. comparing the bending capacity10, remove The quasi-Newton method starts at Even the simple two-beam problem introduced any sections where the capacity reduces, as one particular location in the design earlier has multiple peaks. Thus you might run these will automatically be suboptimum and space, and each iteration takes us to a a number of gradient methods, each starting thus never selected. Next we can analyse new point. This new point leads to the in a diff erent location, and choose the best of every size combination of the two beams next, until it converges (the required the best. and determine the new sizes to carry the sections are those used in the analysis), Another problem is that the method can moments. I used a Python script working sometimes in one or two iterations, struggle when the slope is not smooth. with the GSA application programming sometimes in many more. The process is Consider the beam weight versus stiff ness interface (API) to achieve this11. as follows: (Fig. 6). If you used a classic gradient method,

TSE51_34-41 Method Equations v1.indd 36 24/02/2016 11:49 www.thestructuralengineer.org 37

then it would stop as soon as it reached a design space to give an understanding of utilisation of a member, then the fi tness f of a downward slope. It would not be smart enough the infl uence of the various parameters. phenotype could be: to try bumping over the intermediate valley; it Other options could be to consider would only conclude that it had reached the a range of material stiff nesses, say peak without fi nding the solution, and thus specifi ed versus supplied, for a seismic report a failure. How can we overcome this? As analysis to check predicted behaviours, The parents of the next generation are mentioned earlier, we can sometimes reduce or to determine “typical” loads for a chosen based on their fi tness. A typical the number of sections under consideration to serviceability vibration analysis. Adding way is to use a weighted roulette wheel avoid these ineffi cient beams. Another option the eff ect of construction tolerances to process, where the chance of selection is is to use a more sophisticated method, such as a fi nite-element analysis (FEA) model is proportional to their fi tness. “simulated annealing”. another possibility. Having chosen two parents, you now need to breed them. We do this using a crossover, Simulated annealing Genetic algorithm which means that you randomly choose a Simulated annealing is a refi ned gradient The Monte Carlo method provides a point along the genetic string, then take the method, where the metaheuristic is one of snapshot of a variety of positions in the code before that point from one parent and annealing cooling metal. Mathematically this design space, but what if we could build the code after from the other. There is also is implemented as a “temperature” variable on what it reveals? Genetic algorithms a small random chance that any bit in the that gradually reduces over the time of the (GAs)16–21 do just that, taking the solutions genetic string is mutated, or fl ipped, from 1 optimisation. This has the eff ect of making the from a Monte Carlo distribution and literally to 0 or vice versa. jumps round the design space large to begin breeding a new generation of improved with; then as the temperature cools, the jumps solutions from the best of the previous Example: gradually get smaller13. You can also control generation. Parent #1 = 11111111 this temperature according to how successful GAs are based on Charles Darwin’s Parent #2 = 00000000 the heuristic is at fi nding better values: theory of evolution, as well as modern Crossover at point 5 gives: increase the temperature when there are lots genetics. In a GA you must fi rst describe Child = 11111000 of better points found and reduce it when it the variable aspect of the structure using struggles14. a binary “DNA”, or genotype. The exact Repeat with two new parents until you So, if you were using a simulated annealing form of this genotype will vary between have a complete new generation, then test method to fi nd the highest point in Britain, problems: e.g. in a study on possible again. Repeat the whole process until either you might start with jumps of perhaps 200 external bracing locations in a tall building the total fi tness exceeds an amount or you miles or more to test for better locations. Each facade, each digit might represent a complete a certain number of generations. subsequent jump would gradually get smaller, diagonal in one direction in a particular until hopefully you fi nd Ben Nevis rather than panel, showing the brace either there (1) Particle swarm optimisation Snowden, Scafell Pike, or an even smaller hill. or absent (0). An alternative might have The methods described earlier explore One way to use this is in the search for the fi xed groups of digits giving the selection the design space; the “particle swarm most effi cient arch geometry carrying both number from a list of possible sections. optimisation” (PSO) method takes that a constant load, say from a road or rail, and Whatever the form of the encoding, each metaphor a step further (Figure 7). The PSO its self-weight. You could do this by defi ning genotype will give rise to a particular is based on the way that social insects and the arch geometry by a series of nodes, then solution, or phenotype, that you can birds might explore a location looking for the adjusting each node in turn up or down to see analyse and check. best food to share22,23. In a PSO you have if it improves the effi ciency, and permanently Once you have a complete generation, a number of particles fl ying through the change the nodal coordinate when it does. To you need to assess their fi tness, which will design space, assessing each location that begin with the jumps would be large, as little be a numerical value of how close they are they land on, but also remembering the best is known about the ideal rise or shape, but the to your requirements, while penalising but location that they have found so far. Their jump size would reduce with each iteration. not rejecting solutions that include failing next landing point is derived from their xi(t+1) Thus the routine might start by varying the members. For example, if is the design current position and movement vector Ui xi vi, coordinates by as much as 1m or more at a time, and then gradually reduce to a minimum of perhaps 1mm for fi ne-tuning the arch. Figure 7 EParticle swarm Monte Carlo method optimisation The Monte Carlo method was invented by the Manhattan Project scientists, particularly Stanislaw Ulam and John von Neumann, who were faced with the challenge of predicting the expected yield of a nuclear warhead despite the huge number of variables15. Their solution was to pick a large number of points at random in the design space, or phase space as they called it, and then analyse the mass of results as per regular experimental data to determine typical behaviours. With structures the Monte Carlo method is perhaps best used to do a broad survey of the

TSE51_34-41 Method Equations v1.indd 37 24/02/2016 11:49 ›

38 TheStructuralEngineer Methods and practice March 2016 Design optimisation

modiﬁ ed by an attraction to their own best design solution pi and to the group’s best solution , plus weighting factors φ pg .

The end result is that while some particles spiral into the global best solutions found so far, others circle much wider, constantly looking for a possible better answer (Figure 8). Note that while the design space may be deﬁ ned in terms of discrete values, say beam sections, the PSO works best if you treat the space as continuous, then choose the closest point to be the one assessed. Figure 8 Particle swarm search Pareto front In a problem where you are only optimising for a single variable, such as weight, it is reasonably easy to determine when you have found a good optimum solution; but this is far more diffi cult when you are trying to optimise for multiple variables that can be conﬂ icting or independent, such as steel weight versus solar gain (Figure 9)24. Graphing the results from multiple runs of a multi-objective optimisation will reveal a line or surface (depending on the number of variables chosen) known as the Pareto front, named after the engineer and economist Vilfredo Pareto. Examination of the region might help to decide which is the best optimum compromise to select, or even suggest solutions that have not yet been explored25.

Pattern recognition – neural networks Neural networks, or strictly speaking, artiﬁ cial neural networks (ANNs), reproduce the way that the brain works and particularly the way that it recognises patterns to predict the best answer26–29. A unique feature of ANNs is that they are not deﬁ ned but instead need training; a phenomena known as “machine learning”. ANNs have a number of input nodes that take a range of data; each of these inputs are then multiplied by a weight, a positive or negative number, and then summed to form the inputs for a second, or hidden row (Figure 10).

Figure 9 Pareto front showing trade-off between lifecycle energy costs and structural ﬁ rst costs24

It is common practice to squash the weighted inputs to the next layer using a sigmoid function so that they are always between 0 and 1:

The process is then repeated for the results of the hidden row to generate the inputs for the output row. One of the output nodes will then give the maximum output, and this will be declared to be the result (Figure 11)30. The key to ANNs is in the training. First you generate the initial weights randomly. Next you take a reasonable number of training cases where you already know the answer. The relevant input Figure 10 Neural network weighting values are given and if the output is correct, then the relevant weights are boosted; if the wrong answer is given, then the

TSE51_34-41 Method Equations v1.indd 38 24/02/2016 11:49 www.thestructuralengineer.org 39

weights that lead to that answer are reduced. If the output on U is y, the target value is d, then then the error e on Figure 11 the output layer is: Neural network

If the training constant is η then the change to the weighting Δw is:

on the output layer

on the hidden layer

The training is then repeated for all training cases until they consistently give the right answer. For engineering applications, it may be best to use the hypercube approach to training cases. If, for example, there were two variables to consider, the design space would form a square (or at least a rectangle). You would then derive test cases for the four corners of the space, the four mid-sides, and one in the centre. For three variables there would be eight corners, 12 mid-edges, six mid-faces, and one centre case of the design cube. And so on into the upper dimensions of the hypercubes. Once you have the ANN trained, it should then recognise similar patterns and react accordingly. So how are ANNs used? Google uses them for picture recognition, your smartphone camera uses one to recognise and focus on faces, and speed cameras use them to read your car number plate. In engineering they might be used for scheme design31 or assessing damage to a beam due to cracking32.

Geometric optimisation While some of the methods discussed earlier can be used to determine the best geometry for a structure, there are methods dedicated to ﬁ nding the optimum shape for tension structures, along with the Figure 12 Force density form-ﬁ nding necessary prestresses. Tension structures are very sensitive to the balance of prestress, geometry, and support stiff ness. Out-of-balance pre-tensions will result in slack cables or waves in a fabric surface.

Force density form-fi nding The fi rst and oldest of these methods is the force density method, which is particularly suitable for cable-net structures, although it can also be used on fabrics. In the force density method, the length of 1D elements and the area of 2D ones are set proportional to the applied force. For example, the three-piece tie in Figure 12 has an axial load and each piece has a target force density set to be 1, 2 and 3 respectively. The end result is that the one with a force density of 1 is twice the length of that set to 2, and so on. The advantage of the force density method is that it is fast (and hence popular in form-fi nding programs aimed at architects); the disadvantage is that the pre-tensions might not be achievable and are not guaranteed to be in equilibrium.

Soap fi lm form-fi nding Soap fi lm is a more advanced form-fi nding method, where you specify the structure target prestresses, and possibly loads, and then remove all stiff ness from the elements. The eff ect is to model something very like a soap fi lm (hence the name) bounded by elastic bands. This results in minimum fabric surfaces and cable lengths, which are in static equilibrium with the prestresses and applied loads. It is thus a more suitable form-fi nding method for engineers than force density. Soap fi lm form-fi nding is often achieved with a non-linear static Figure 13 GSA evolutionary topology optimisation analysis method called dynamic relaxation, where the structural

TSE51_34-41 Method Equations v1.indd 39 24/02/2016 11:49 ›

40 TheStructuralEngineer Methods and practice March 2016 Design optimisation

Figure 14 Conclusion Evolutionary topology optimisation So, which is the best method for design steelwork connections optimisation? Alas, the “No Free Lunch” theorem states that all methods considered over all problems average out the same. Or to put it another way, certain methods are better for some problems, but no method is best for all problems. Deciding which method to use is a matter of judgment, and possibly experimentation. You may also fi nd it useful to combine methods. A GA or particle swarm might get you close to the answer, and then a gradient method could fi ne-tune the result. Or you might use a GA to determine the ideal number of hidden nodes in a neural network, and so on. It is also good to remember the principle ARUP of “satisfi cing”, as you will struggle to wring those last few improvements out of your model. One of the arts of engineering is stiff nesses are converted into eff ective structural forms for additive manufacturing, knowing when close enough is good enough. masses that are then accelerated by the or 3D printing as it is also known. Arup has applied loads and resultant element forces. used it to develop lightweight exposed Acknowledgements These masses then move until all forces are steelwork connections (Figure 14)35. This paper expands on an annual lecture that in equilibrium and the ideal structural layout is The University of Bath has also used ETOs I have given for a University of Leeds Masters achieved33. to teach about concrete cracking behaviour. degree design optimisation module. Both form-fi nding methods work well for Here they generate a massively redundant I wish to thank Professor Vassili Toropov tension-only structures; for compression- truss of bar elements that fi ll the space of a (now University of London) and Dr Osvaldo only the trick is to reverse all the loads, as concrete beam, then progressively remove Querin (University of Leeds) for the an arch is the mirror image of a hanging the elements with the highest tension36. opportunity and their support. chain. The hard part comes when you want to mix the two, such as for an inclined arch supporting a curved bridge deck with a cable net. You either have to iterate the References form-fi nding over the various structural parts E1 Nolan J. (2012) ‘Cost versus value – the role of the consulting structural (tension structure, compression structure), engineer’, The Structural Engineer, 90 (2), pp. 13–22 or just form-fi nd a small part of the structure before gradually adding in more and more with each successive iteration; alternately E2 Mitchell M. (2009) Complexity: a guided tour, New York, USA: Oxford you can use one of the other methods University Press described earlier. E3 Debney P. (2012) Braess’ Paradox – or Why improving something can Evolutionary topology optimisation make it worse [Online] Available at: www.oasys-software.com/ Evolutionary topology optimisation (ETO) is blog/2012/05/braess%E2%80%99-paradox-or-why-improving- a method akin to sculpting, where you start something-can-make-it-worse/ (Accessed: November 2015) with a large block of material and gradually 4 Carfrae T. (2015) Designing with computers (Institution of Structural removing the bits that are contributing E Engineers 2014 Gold Medal address) [Online] Available at: http:// least34. These are typically determined istructe.hosted.panopto.com/Panopto/Pages/Viewer. by looking at the ratio of maximum to aspx?id=e214dba0-ac69-4523-ba3b-f4d1f0f34b2a (Accessed: minimum von Mises stress or strain energy November 2015) in a 2D or 3D mesh, and removing any elements that fall below the threshold. This E5 Quote Investigator (2011) Everything Should Be Made as Simple threshold may start as low as 1% or even as Possible, But Not Simpler [Online] Available at: http:// 0.1% of the maximum, and is then increased quoteinvestigator.com/2011/05/13/einstein-simple/ (Accessed: incrementally each time the routine fails to November 2015) remove anything. This process then stops at a predetermined ratio, say 25%, or when E6 Wise C., Weir A., Oates G. and Winslow P. (2012) ‘An amphitheatre for an instability develops in the remaining cycling: the design, analysis and construction of the London 2012 structure. The end result is a highly effi cient Velodrome’, The Structural Engineer, 90 (6), pp. 13–25 form that is often organic in appearance (Figure 13). ETO is an excellent method for developing

TSE51_34-41 Method Equations v1.indd 40 24/02/2016 11:50 www.thestructuralengineer.org 41

E7 Expedition Workshed (2011) The London Velodrome E23 Li L.J., Huang Z.B., Liu F. and Wu Q.H. (2007) ‘A [Online] Available at: http://expeditionworkshed.org/ heuristic particle swarm optimizer for optimization workshed/the-london-velodrome/ (Accessed: of pin connected structures’, Computers & November 2015) Structures, 85 (7–8), pp. 340–349

E8 Weisstein E.W. (2015) Newton’s method [Online] E24 Flager F., Welle B., Bansal P., Soremekun G. Available at: http://mathworld.wolfram.com/ and Haymaker J. (2009) CIFE Technical Report NewtonsMethod.html (Accessed: November 2015) #TR175: Multidisciplinary process integration & design optimization of a classroom building [Online] E9 Tata Steel (2013) Advance® sections. Price extras: Available at: http://cife.stanford.edu/sites/default/ Pounds Sterling [Online] Available at: www. fi les/TR175.pdf (Accessed: November 2015) tatasteeleurope.com/static_fi les/StaticFiles/ Business_Units/CCandI/Products/Sections/ E25 Wijnbeld B. (2015) Optimisation of High-Rise Section%20Price%20List%201%201%200912.pdf Structures Using the GSA COM-Interface [Online] (Accessed: November 2015) Available at: www.oasys-software.com/ casestudies/casestudy/optimisation_of_highrise_ E10 The Institution of Structural Engineers (1989) Manual structures_using_gsa (Accessed: November 2015) for the design of steelwork building structures, London, UK: The Institution of Structural Engineers E26 Hawkins J. and Blakeslee S. (2004) On intelligence, New York, USA: St. Martin’s Press E11 Debney P. (2016) Exploring the Design Space [Online] Available at: www.oasys-software.com/ E27 Montague R. (2006) Your brain is (almost) perfect, blog/2016/01/exploring-the-design-space London, UK: Penguin Group (Accessed: February 2016) E28 Open University (2008) Neural networks (2008 ed.), E12 Maier J. (2016) Marina Bay Sands – Use of API Milton Keynes, UK: Open University [Online] Available at: www.oasys-software.com/ casestudies/casestudy/marina_bay_sands_use_of_ E29 Jenkins W.M. (2001) ‘A neural-network-based api (Accessed: November 2015) reanalysis method for integration with structural design’, The Structural Engineer, 79 (13), pp. 25–29 E13 Open University (2008) Symbolic Intelligence (2nd ed.), Milton Keynes, UK: Open University E30 Luger G.F. and Stubblefi eld W.A. (1997) Artifi cial intelligence (3rd ed.), Reading, USA: Addison E14 Belegundu A.D. and Chandrupatla T.R. (2011) Wesley Longman Optimization concepts and applications in engineering (2nd ed.), New York, USA: Cambridge University Press E31 Rafi q M.Y., Bugmann G. and Easterbrook D.J. (2000) ‘Artifi cial neural networks to aid conceptual E15 Dyson G. (2012) Turing’s cathedral, New York, USA: design’, The Structural Engineer, 78 (3), pp. 25–32 Pantheon Books E32 Jenkins W.M. (1997) ‘An introduction to neural E16 Rutten D. (2010) Evolutionary Principles applied to computing for the structural engineer’, The Problem Solving [Online] Available at: www. Structural Engineer, 75 (3), pp. 38–41 grasshopper3d.com/profi les/blogs/evolutionary- principles (Accessed: November 2015) E33 Debney P. (2013) Form-Finding: Fabrics, Cables, & Shells [Online] Available at: www.oasys-software. E17 Mitchell M. (1998) An introduction to genetic com/webinar/webinar/form_fi nding_with_GSA_2013 algorithms, Cambridge, USA: MIT Press (Accessed: November 2015)

E18 Open University (2008) Evolutionary computation E34 Debney P. (2015) Evolutionary topology optimisation (2nd ed.), Milton Keynes, UK: Open University with GSA [Online] Available at: www.oasys- software.com/blog/2015/07/evolutionary-topology- E19 Birch A. (2003) ‘Animal magic’, Building Design, 31 optimisation-with-gsa/ (Accessed: November 2015) October, pp. 16–17 E35 Galjaard S. (2015) Embrace new tools for design E20 Dias W.P.S. (2007) ‘Engineering as cyclic problem freedom [Online] Available at: http://thoughts. solving – some insights from Karl Popper’, The arup.com/post/details/462/embrace-new-tools-for- Structural Engineer, 85 (2), pp. 32–37 design-freedom (Accessed: July 2015)

E21 Jenkins W.M. (1991) ‘Structural optimisation with E36 Ibell T. (2015) ‘Pillars of our education system’, the genetic algorithm’, The Structural Engineer, 69 Innovate, create, inspire – structurally engineering a (24), pp. 418–422 modern world, Singapore, 3–4 September

E22 Open University (2008) Natural intelligence (2nd ed.), Milton Keynes, UK: Open University

TSE51_34-41 Method Equations v1.indd 41 24/02/2016 11:50 FEM-Design Waterman Group chooses Scandinavian ƋƵĂůŝƚǇĨƌŽŵ^ƚƌƵ^ŽŌƚŽĚĞƐŝŐŶƚŚĞĐŽŶĐƌĞƚĞ ĞůĞŵĞŶƚƐŽĨ<ŝŶŐƐŽƵƌƚ͘

ings Court is a new seven story reinforced concrete structure in the ŚĞĂƌƚŽĨŽǀĞŶƚ'ĂƌĚĞŶ͕ǁŚŝĐŚŝƐŽŶĞŽĨƚŚĞŵŽƐƚƉƌĞƐƟŐŝŽƵƐĂƌĞĂƐ KŽĨ>ŽŶĚŽŶ͘dŚĞƐĐŚĞŵĞǁŝůůĂĚĚŝƟŽŶĂůůǇŽƉĞŶƵƉĂƉĞĚĞƐƚƌŝĂŶƌŽƵƚĞ ƚŽƚƌĂŶƐĨŽƌŵƚŚĞĞǆŝƐƟŶŐƉƌŝǀĂƚĞĐŽƵƌƚǇĂƌĚŝŶƚŽĂŶĞǁƉƵďůŝĐĐŽƵƌƚǇĂƌĚ͕ named Kings Court.

The reinforced concrete New Version ƐƚƌƵĐƚƵƌĞ ŝƐ ƉĂƌƚ ŽĨ ĂŶ ĂŵďŝƟŽƵƐ FEM-Design 15 development with a mix of new and refurbishment buildings to regenerate the area, establishing Moving Loads ŶĞǁŚŝŐŚƋƵĂůŝƚǇƌĞƐŝĚĞŶƟĂů͕ƌĞƚĂŝů and restaurant space in Central London. The project is designed by Waterman Group in London ǁŚĞƌĞ ^ƚƌƵ^ŽŌ &DͲĞƐŝŐŶ ŽĨůŽĂĚĐŽŵďŝŶĂƟŽŶƐƌĞƋƵŝƌĞĚďǇ is used to design concrete Eurocode, it has been very useful structures according to Eurocode to analyse the structure with Ϯ͘ ͞dƌĂĚŝƟŽŶĂů ƌŝƟƐŚ ^ƚĂŶĚĂƌĚ ŽŶůǇ ƐĞůĞĐƚĞĚ ůŽĂĚ ĐŽŵďŝŶĂƟŽŶƐ ĐŽĚĞƐ ŽĨ ƉƌĂĐƟĐĞ ĂƌĞ ďĞĐŽŵŝŶŐ ĐŽŶƐŝĚĞƌĞĚŝŶĞĂĐŚĂŶĂůǇƐŝƐƌƵŶ͘͟ obsolete, so we are trying to tŝƚŚ ^ƚƌƵ^ŽŌ &DͲĞƐŝŐŶ ŝƚ ŝƐ Steel Connections promote the use of Eurocode on all possible to perform all the three the new projects. For this reason ĚŝīĞƌĞŶƚ ŵĞƚŚŽĚƐ ŽĨ ĂŶĂůǇƐŝƐ it has been very important for us indicated in Eurocode 2 to ƚŽ ƌĞůǇ ŽŶ ^ƚƌƵ^ŽŌ͛Ɛ &DͲĞƐŝŐŶ ĐŽŶƐŝĚĞƌ ƚŚĞ ϮŶĚ ŽƌĚĞƌ ĞīĞĐƚƐ͘ ĐŽŶĐƌĞƚĞƉĂĐŬĂŐĞ͕͟ƐĂǇƐDĂƐƐŝŵŽ The choice between the three ͛/ŐŶĂǌŝŽ͘ methods is important because The design of concrete elements the general method is usually not is very simple and clear with ĐŽŶƐĞƌǀĂƟǀĞ͕ďƵƚƐŝƚƵĂƟŽŶƐĐŽƵůĚ ^ƚƌƵ^ŽŌ &DͲĞƐŝŐŶ͘ ͞tĞ ŚĂǀĞ arise where the general method appreciated how easy it is to check ŝƐ ĐŽŶƐĞƌǀĂƟǀĞ ĐŽŵƉĂƌĞĚ ƚŽ ƚŚĞ the results both with the detailed ƐŝŵƉůŝĮĞĚ ŵĞƚŚŽĚƐ͕ ĞƐƉĞĐŝĂůůǇ 3D Soil ƌĞƐƵůƚƐĂŶĚǁŝƚŚƚŚĞĐŽůŽƵƌŵĂƉƐ͘͟ with large moments where the &Žƌ <ŝŶŐƐ ŽƵƌƚ͕ ^ƚƌƵ^ŽŌ ƐƟīŶĞƐƐ ŝƐ ůĂƌŐĞůǇ ƌĞĚƵĐĞĚ ďǇ &DͲĞƐŝŐŶ ŚĂƐ ďĞĞŶ ƵƟůŝƐĞĚ cracking. to analyse the overall structural ĚĚŝƟŽŶĂůůǇ͕^ƚƌƵ^ŽŌ&DͲĞƐŝŐŶ behaviour of the building and to has been used to evaluate the design all the concrete elements ƵƉůŝŌ ĚƵĞ ƚŽ ƚŚĞ ŐƌŽƵŶĚ ŚĞĂǀĞ such as columns, beams, walls and which represents an important slabs. “Due to the large number issue to take into account in

p42-43_TSE.03.16.indd 42 19/02/2016 16:04 Swedish Software for the Building Industry Construction, Energy, Project, Production

FEM-Design Three-dimensional FEM platform for design and analysis of concrete, steel and timber.

WIN-Statik Simple program for ordinary structures such as beams, columns and frames.

IMPACT an environment where London BIM system for prefabricated elements. From ůĂǇ ŝƐ ƉƌĞƐĞŶƚ͘ ͞tŝƚŚ ^ƚƌƵ^ŽŌ design, to production and erection. FEM-Design, it has been possible to consider this behaviour ďĞĐĂƵƐĞ ƚŚĞ ƐŽŌǁĂƌĞ ŚĂƐ Ă ƉŽǁĞƌĨƵů ŶŽŶͲůŝŶĞĂƌ ĐĂůĐƵůĂƟŽŶ method and allows the user to set PRE-Stress ĨƌĞĞƌŽƚĂƟŽŶĐŽŵƉŽŶĞŶƚƐ͘͟ Analysis and design of prestressed beams throughout their lifetime. tƌŝƩĞŶďǇ͗DĂƐƐŝŵŽ͛/ŐŶĂǌŝŽ Ͳ^ƚƌƵĐƚƵƌĂůĞƐŝŐŶŶŐŝŶĞĞƌĂƚ tĂƚĞƌŵĂŶ'ƌŽƵƉΘEŝĐŽůĂĞEŽŶƵͲ ^ƚƌƵĐƚƵƌĂůŶŐŝŶĞĞƌ͕^ƚƌƵ^ŽŌ< VIP-Energy Dynamic energy balance calculations for all building types.

Dennis Kristensen ^ĂůĞƐĂŶĚ^ƵƉƉŽƌƚ͕^ƚƌƵ^ŽŌh< нϰϰϮϬϳϳϮϰϴϯϬϮ ĚĞŶŶŝƐ͘ŬƌŝƐƚĞŶƐĞŶΛƐƚƌƵƐŽŌ͘ĐŽŵ www.strusoft.com Photo rendering by Waterman Group©

p42-43_TSE.03.16.indd 43 19/02/2016 16:05 ›

44 TheStructuralEngineer Methods and practice March 2016 Finding hidden modelling errors

using the displacements. In this article we ﬁ rst present a short What is your structural discussion on how inaccuracies arise in structural analysis results while solving problems on computers, and demonstrate how modelling errors can amplify these model not telling you? inaccuracies. We then present two novel techniques that engineers can use to detect hidden errors in their FE models and to detect Ramaseshan Kannan Arup, UK potential inaccuracies in force and moment calculations. These tools are being used by Stephen Hendry Arup, UK engineers to gain conﬁ dence in the robustness Chris Kaethner Arup, UK of their analysis results.

Synopsis model, to identify inaccuracies or errors in Solving problems numerically on Finite-element analysis involves the model, and to know what confi dence they computers inherent approximations and can have in the results. In order to gain the A structural analysis model built and analysed benefi t of the more detailed modelling of the on an Intel or AMD processor-based computer numerical errors. In addition to these, structure, without the problems associated (x86 processor architecture) using a modern the increasing size of structural with complexity, the engineer needs new software package both stores real numbers models and the use of automated techniques that give greater insight into the and performs arithmetic on them in fi nite workfl ows for creating them can lead model. precision. Such a representation is called a to hidden user errors in these models. The use of the FE method for structural “fl oating point” number and is governed by the 2 In order for an engineer to have analysis involves approximations at various IEEE 754 standard . In simple terms, any real levels. These are: number x is represented as a fl oating point confi dence in the analysis results, it is number FL(x) given by the relationship3 necessary to be aware of how these • idealisation of the structural behaviour errors manifest themselves in models, • discretisation of the governing equations of what impact they have on analysis motion and of equilibrium results and, most importantly, how • storage and manipulation of real numbers on where δ is the round-off error whose computers using fi nite precision arithmetic magnitude is bounded by a constant that they can be detected. We present depends on whether the software uses single novel numerical techniques that Each layer of approximation introduces or double precision. (For our discussion in the analyst can use to “debug” their errors in the computed solution. Techniques subsequent sections, we shall assume that our models and verify the accuracy of their for understanding, analysing and bounding computations are done in double precision, i.e. analysis results. These techniques these errors have developed apace with the we have about 16 decimal digits of precision.) method itself. On the other hand, user errors Based on this representation, an arithmetic have been implemented in software (i.e. errors that are caused by incorrect use of operation such as the addition of two fl oating and have been successfully used by the software), though pervasive, have received point numbers will introduce an error in the practising engineers working on real- much less attention in academic literature. result as given by life projects. Examples of such errors include: . • lack of connectivity: adjacent elements that Notation are supposed to share a common node but The arithmetic model presented can be Capital letters are used to denote matrices are connected to diff erent nodes that are easily extended to other operations, such as and lower case to denote vectors and scalars. coincident, resulting in one of the elements multiplication or exponentiation. Therefore, acquiring insuffi cient restraints it is easy to see that a general function f Introduction • failure to idealise member end connections introduces an error of Δy in the result y when Since the fi rst use of computers for structural correctly, which can occur if a beam is free to it acts on x analysis1, the computational power available rotate about its axis although in the physical to engineers has increased dramatically. In model there is a nominal restraint against . turn, engineers have been quick to exploit rotation these capabilities and fi nite-element (FE) • modelling beam elements with large sections The best we can hope for is for Δy to be models have grown signifi cantly both in size and/or very small lengths, often the result of of the same or smaller order of magnitude as and in complexity. Other developments, such importing FE assemblies from computer-aided Δx. The error Δy can be aff ected by both the as the adoption of automated workfl ows design (CAD) models algorithm used to compute f and the sensitivity to generate analysis models from Building of f to changes in x. Information Modelling (BIM) and parametric Irrespective of whether the error arises To illustrate how errors in x propagate to design packages, have also contributed to the from approximation or from erroneous large errors Δy in y, we present two examples. aforementioned growth. input data, it can lead to inaccuracies Consequently, the engineer is more distant in results such as displacements. Such Example 1 from the actual calculation and so fi nds it errors can propagate to responses, such The fi rst example is of a function whose more diffi cult to establish the integrity of the as forces, that are subsequently computed computed solution is outside its analytical

TSE51_44-47 Method Kannan debugging v1.indd 44 24/02/2016 11:50 www.thestructuralengineer.org 45

range. Consider But it is easy to verify, by means of solving is said to be well conditioned. The largest the equation by hand, that the actual solution theoretical value of κ is infi nity and a matrix of the equations is with this condition number is singular, i.e. it has . no unique solution. In fi nite double precision, . however, the largest value of the condition number before a matrix is deemed singular is With some analysis, it can be seen that the y Therefore, our computed solution not 1016. is bounded by 0 and half for all values of x, i.e. only had the wrong sign, it had a relative The condition number also gives us a rule of 0 < y < 1/2. error larger than 100%! The reason for the thumb for the worst-case accuracy of u. There We wish to compute y(x = 1.2 × 10-5) erroneous answer lies in the sensitivity of our can be as few as 16 – log κ digits of accuracy on a hypothetical computer with 10 digits of matrix to the inversion function. (To solve the in the computed solution. A matrix with κ equal precision, i.e. any number with more than 10 linear system Ax = b, we eff ectively invert the to 1016 can therefore result in a solution with decimal digits gets rounded to one with 10 matrix A). The fi x in this case is simply to use not a single digit accurate and the results from digits or fewer. (This is purely for illustration. more precision. Repeating the same solution a matrix with κ > 1011 must be treated with x86-based computers support up to 16 digits with six digits of precision gives us the same caution. of precision.) On such a computer, cos (1.2 answer as the exact solution. The presence of user errors in FE models × 10-5) = 0.999999999 and 1 – cos x = can lead to a large condition number. Such 0.0000000001, which gives us Numerical errors and structural errors, examples of which we provide in analysis problems the Introduction, can result in zones of high . So, what is the connection between the stiff ness or low stiff ness in structural models. discussion presented and structural analysis In large and complex models, not only are they Clearly, our computed value is outside the problems? diffi cult to spot, it is hard to know whether they range of the function! As mentioned in the Introduction, there are exist at all. The error is due to the way we chose several sources of approximations leading to to compute the function. To fi x it, we must errors in structural analysis results. Central Estimating conditioning of stiff ness change the algorithm used to compute y. to the computation of these results is the matrix Rewriting it as solution of a linear system of equations given To ensure the robustness of results, it is by Ku = f for a stiff ness matrixK , loading f necessary to ensure the stiff ness matrix is well and displacement u. (In this article we limit conditioned. Most textbooks on FE analysis our discussion to linear/elastic structural prescribe computing κ using its defi nition as the analysis problems, although much of it can ratio of the largest and the smallest eigenvalues will yield the correct result. also be generalised to non-linear problems.) of K The challenge in doing so is that the These linear systems arise in several types of extremal eigenvalues are expensive to compute; Example 2 analysis, which include: in fact, it is more expensive to compute κ than to The second example comes from the use solve Ku = f! However, advances in numerical of Gaussian elimination, a common method • static: solve Ku = f for displacements u and analysis5 allow κ to be estimated to good for solving a linear system of equations on static : solve for an initial set of accuracy for a fraction of the cost required to P-Δ Ku1 = f1 a computer. We wish to fi nd the solution of loads and then solve for compute it. We implemented the algorithm in f1 (K + Kg)u2 = f1 the following simultaneous equations on a , where the geometric stiff ness matrix is Higham and Tisseur5 in the structural analysis u2 Kg computer with three digits of precision formulated using element forces computed package Oasys GSA6. from Our approach was to estimate the condition u1 • modal dynamic: solve Ku = λMu for natural number of the matrix each time a linear, elastic frequencies √λ and vibration mode shapes analysis was executed and report it as part of and u, where M is the mass matrix of the FE the analysis log. The reporting gives engineers assemblage feedback on when their model has potential . • buckling: solve and for issues that could aff ect the accuracy of their Ku1 = f1 Kv = λKgv Writing the system as a 2 × 2 matrix buckling modes v and buckling load factors λ, results. In some cases, it is simple to identify where is the same as in analysis the cause of the ill-conditioning warning; Kg P-Δ for instance, a missing global restraint that When solving a linear system of equations can set up a rigid body motion for the whole involving K there is an error in the computed model. In most cases, however, it is tedious we try to introduce a zero in the fi rst column displacements u. The error depends not only and time-consuming to fi nd the cause. What of the second equation. To do this we multiply on errors in K but also on the sensitivity of the the engineer wants is a way to quickly pinpoint the fi rst column by the ratio 0.780 / 0.913 and solution to small changes. The latter is referred where there are potential problems, to simplify subtract the result from the second column to to as the “conditioning” of the problem. The the fi xing of modelling errors. give conditioning of a problem determines the maximum (i.e. worst-case) change in the New technique to detect causes of ill- . solution in response to a small change in the conditioning input. For our stiff ness matrix, the conditioning Our investigation of ill-conditioning issues

is measured by a numerical property called the led to a new numerical analysis technique for Doing so leads us to the solution condition number of a matrix4, often denoted detecting their cause. Our method, which is by κ. presented in Kannan et al.7 and implemented . The smallest value of κ is 1. Such a matrix in Oasys GSA, has been used successfully by

TSE51_44-47 Method Kannan debugging v1.indd 45 24/02/2016 11:50 ›

46 TheStructuralEngineer Methods and practice March 2016 Finding hidden modelling errors

Figure 1 Figure 4 Figure 5 Small portion of Element virtual energy Loss of accuracy in element facade model visualisation force/moment results

engineers to correct their models. We present conditions, nodal connectivity or cross-section was connected to a diff erent node that was a summary description of the method in this properties) in the vicinity of the said elements in the geometric vicinity of the spring element section and refer the reader to the original would reveal the anomaly that causes ill- (Fig. 3). As a result, the plane elements were paper for further detail. conditioning. Once the anomaly is fi xed, the unsupported and free to fl ap about – an error Errors that cause ill-conditioning belong to engineer re-runs the analysis to ensure the that could lead to potentially incorrect results. one of the following two categories: condition number reduces and, if not, runs a Fixing the nodal connectivity error brought model stability analysis again. the condition number down to 108 and the • lack of stiff ness: this can arise, for instance, We now present an example of the use model was analysed without the ill-conditioning when parts of the model are insuffi ciently of model stability analysis on a model that warning. restrained or nodes are incorrectly connected returned ill-conditioning warnings during its The example described is one of the many • disproportionately large stiff ness: these analysis. Our example is set in the context of situations where model stability analysis allows can typically occur when certain elements GSA but the underlying numerical analysis is the detection of modelling errors that would have large second moments of area but small not specifi c to a software package. otherwise be hidden from the modeller – for lengths Figure 1 shows a portion of a larger model of more examples, we refer the reader to Kannan a facade cladding of a structure that consists et al.7 While both categories can generally result of 32 000 elements and 21 000 nodes resting from any user error, the latter is particularly on pinned supports. The glass facade panels Accuracy of element force calculation common when models have been imported are modelled using four-noded plane-stress In an FE analysis, element results are based using automated processes from BIM or CAD elements. Each panel rests on the grid of on strains which are a function of relative packages. beams through springs at each of its four displacements. Hence, even if the displacement Our method is called “model stability corners, as shown in the zoomed-in view in results are accurate enough, there can be analysis” and it uses the eigenvectors of the Figure 2. The element connectivity is illustrated inaccuracies in computing strains in elements. stiff ness matrix to formulate “virtual energies” in Figure 3a. By comparing the element displacement with for elements in the model. It can be shown An early version of the model triggered the element distortion the number of signifi cant mathematically that elements with large virtual a “large condition number” warning with a fi gures lost in the calculation of element results energies pinpoint parts of the model that either condition estimate of 1012. Model stability can be assessed. Displaying this graphically lack stiff ness or are disproportionately stiff . analysis was run and it identifi ed two elements allows the engineer to identify parts of the model When a model generates an ill-conditioning in the model with large virtual energies where the results are relatively less accurate. warning, the engineer can run a model stability (Figure 4). On investigation we found the analysis and graphically display elements following error in the nodal connectivity. The that have large relative virtual energies. corner node of the said plane elements did An examination of the model (e.g. support not share a node with the spring; instead it

Figure 2 Figure 3 Close-up Nodal connectivity errors for element with view of element large virtual energy connectivity

NB Blue quadrilaterals model facade panels, green lines represent beams, and springs are drawn as coils. Gaps between the elements are a a) Closer view of element 13828 b) Corner node connectivity graphic view setting and are only for visual clarity

TSE51_44-47 Method Kannan debugging v1.indd 46 24/02/2016 11:51 www.thestructuralengineer.org 47

If we consider a simple bar element (with axial eff ects only), the force problems when the engineer is unable to properly assess how well their in the bar can be calculated, provided we know the strain, from model represents reality. In order to have confi dence in the modelling, it is important that the engineer can detect problems and have confi dence in the results. . While mathematical techniques can be used to make some of these And assuming constant strain in the element assessments, it is important that the results of these assessments can be presented in a way that is meaningful to the engineer. Ultimately, the best engineering solution depends on the engineer understanding the limitations of the model and knowing what confi dence to place in the analysis. It is the authors’ hope that what is presented here is a step in where l is the length of the bar and the strain depends on the relative that direction. displacement u(2) – u(1). To illustrate this, consider a 1m steel bar with a stress of 200N/mm2. Acknowledgements The corresponding strain is 1 x 10-3, meaning there is an elongation The authors would like to thank Chris Hughes from Arup Manchester of 1mm. If this is part of a large structure, there may well be a rigid for his comments on the article. body displacement of the bar as a whole of, say, 100mm, meaning that in calculating the force (stress) in this element we are losing two Contact details signifi cant fi gures. Obviously, the shorter the element and greater the To contact the authors. email [email protected]. rigid body displacement, the greater the loss of accuracy in the force or stress calculation. The rigid body displacement is the average displacement of the uR element References

. E1 Felippa C.A. (2001) ‘A historical outline of matrix structural analysis: a play in three acts’, Therefore the displacement causing straining at the nodes is: Computers & Structures, 79 (14), 1313–1324

. E2 IEEE (2008) IEEE 754-2008: Standard for Floating-Point Arithmetic, New York, USA; IEEE If the rigid body displacement is large compared with the distortional displacement, there can be a loss of signifi cance in computing u(D,i). In E3 Higham N.J. (2002) Accuracy and Stability of the worst case, the relative error can be as large as Numerical Algorithms (2nd ed.), Philadelphia, and therefore we could lose up to n signifi cant fi gures given by USA: Society for Industrial and Applied Mathematics

E4 Kannan R. (2014) Numerical Linear Algebra . Problems in Structural Analysis (Doctoral thesis), Manchester, UK: School of Mathematics, The This loss of accuracy can therefore propagate to the force and University of Manchester [Online] Available at: moment results for the element. http://eprints.ma.man.ac.uk/2195/ (Accessed: This can be generalised for an element with more than two nodes, February 2016) each with six (or fewer) degrees of freedom in a straightforward 5 Higham N.J and Tisseur F. (2000) ‘A block manner8. We shall omit the derivation to favour brevity, but note that in E algorithm for matrix 1-norm estimation, with an a more general case, we have multiple values of “loss of accuracy” for application to 1-norm pseudospectra’, SIAM. J. each element. However, as this is essentially a qualitative analysis, rather Matrix Anal. & Appl, 21 (4) 1185–1201 than assigning the values to nodes it is suffi cient to simply record the maximum loss of accuracy value for the element. E6 Oasys (2016) Oasys GSA [Online] Available at: This algorithm has been implemented in Oasys GSA. For the simple www.oasys-software.com/products/ model shown in Figure 5 the diameter of the contour blobs indicates the engineering/gsa-suite.html (Accessed: February “loss of accuracy”. It is noticeable that the short (and therefore stiff er) 2016) elements have, as expected, a greater loss of accuracy. The closer n (the digits of lost accuracy) gets to 16, the greater the likelihood of E7 Kannan R., Hendry S., Higham N.J. and Tisseur inaccuracy in the forces. F. (2014) ‘Detecting the causes of ill-conditioning in structural ﬁ nite element models’, Computers & Conclusions Structures, 133, 79–89 As the analysis power available to engineers continues to increase, the size and complexity of models will continue to grow. While an engineer E8 Oasys (2015) Oasys GSA Help Guide [Online] should carry out simple checks to verify the suitability of their model, Available at: www.oasys-software.com/media/ there is a danger that the complexity means that analysis is seen as a Manuals/Latest_Manuals/gsa8.7_manual.pdf “black box” – especially with automated workﬂ ows driven from BIM and (Accessed: February 2016) CAD. Numerical conditioning and accuracy issues are under-discussed in standard FE material in both academia and industry. This can lead to

TSE51_44-47 Method Kannan debugging v1.indd 47 24/02/2016 11:51 ›

48 TheStructuralEngineer Methods and practice March 2016 Veriﬁ cation and validation What is the need for veriﬁ cation, validation and quality assurance of computer-aided calculation?

Chris Rogers MIStructE, Managing Director, CREA Consultants Ltd

Synopsys Quality management system Modern computer-aided calculations are a signifi cant advantage in the The primary tool for controlling the quality, design of structures, whether simple or complex. Software developers invest documentation and archiving of the work is the quality management system (QMS); in the development of tools that are increasingly simple to use; engineers then and engineering design around the world is use those tools to develop and, to a large extent, test the possible structural increasingly being performed to satisfy a QMS confi gurations. The software will then, in many cases, present design designed to comply with ISO 90011 and the calculations and code checks allowing a semi-automated documentation of NAFEMS Quality System Supplement (QSS)2. the design. Even if engineers do not work to a formal ISO The underlying questions are: how do we know that the results are reasonably 9001 QMS, it is likely that they will be working within the spirit of the code. A QMS only correct; and who is responsible if there is a failure? The answer to the second needs to be written down if it is to be formally question is that the Chartered Engineer leading the design project is responsible; audited. the answer to the fi rst point is quality assurance. For engineering analysis, the The primary purpose of the QMS is to major tools for quality assurance are the twin processes of verifi cation and provide the framework for an organisation’s validation (V&V). In very simple terms, verifi cation is the demonstration that the management process; a well-constructed QMS will not be overly prescriptive, it will be mathematics and numerics are correct; validation is the demonstration that the fl exible within the operational contexts of the idealisation of the physics is correct. business. As V&V underpin any quality assurance system, this paper discusses the A QMS written to conform to ISO 9001 will need for V&V and indicates who is primarily responsible for the two processes: address the following generic principles: software developer or design engineer. • customer focus • leadership Introduction • engagement of people Computer-aided calculation, analysis and techniques such as computational fl uid • process approach design are concepts familiar to all practising dynamics (CFD). These often complex • improvement engineers in all disciplines. The use of underlying computational processes are • evidence-based decision-making computer tools has revolutionised the design increasingly hidden from the user. • relationship management of structures; we now produce structures that These developments are all of great have shapes that would have been unthinkable benefi t to the profession and they do allow However, it is not necessary to provide a decade ago. All we have to do is look around more creativity, along with the capacity to procedures for all of these. If a principle does the major cities worldwide. test variations; the concept that mechanical not apply to the organisation or to a specifi c Software developers have invested heavily engineers would call virtual prototyping. The group within the organisation, then it can in making their tools easier to use; and question that does need to be addressed, be excluded from the QMS. If a principle is extending the capabilities of their programs. however, is who is responsible for the design: excluded, it should be positively excluded in This change in capability has gone from the the software house or the Chartered Civil or the Quality Manual, not simply omitted. analysis of load paths and resulting stress, Structural Engineer leading the design project The QMS, formal or informal, lays down the to the design of elements, to code checking (the “Engineer”)? Given that the Engineer is method of recording the progress of the work; and to draughting. The analysis phase of the person ultimately responsible for providing in the context of this paper, the engineering the process now very often revolves around a safe design to the client’s specifi cation, design. The aim is to have a means of tracing the fi nite-element (FE) method (or close there is a need to know how to assess and the workfl ow; demonstrating that the work has alternatives), including the use of advanced demonstrate the quality of the computations been critically assessed and comments acted techniques such as non-linearities and supporting the design, however highly upon; and fi nally archiving of the project. The dynamics. There are systems that also add automated the process may be. benefi ts of this lifetime record are that:

TSE51_48-51 Method Validation v1.indd 48 24/02/2016 11:55 www.thestructuralengineer.org 49

• there are records to demonstrate that the coordinated to be complementary and without Verifi cation is composed of two activities: analysis and design have been critically unnecessary duplication. One important code verifi cation and solution verifi cation. examined and have been shown to be fi t for consideration is that even though the work Code verifi cation is primarily the role of the purpose of both groups originates in the advanced FE developer, as it relates to the mathematics • it is possible to examine the records to guide and CFD analysis of safety-related nuclear, and numerics; this is encased in the program future work (a commercial as well as technical defence and aerospace industries, the output source code, which the user very rarely benefi t) is relevant to all levels of design. Care is taken has access to. To this end, the developer • in a case where there are disputes or, even not to include ideas specifi c to the “hi-tech” should provide a verifi cation manual clearly worse, failures, records exist to allow the case industries; or where it is necessary, to clearly demonstrating that the program produces to be examined in detail mark the thinking as such. mathematically correct results. This is usually achieved by documenting the verifi cation In the context of engineering analysis, the Defi nitions of V&V tests and identifying variance from the exact QMS is therefore the basic requirement for Committee V&V 10 examined the defi nitions solution to specialised test problems. It does assessing the competence of the software of V&V within the US nuclear and defence not mean that the source code has to be being employed; as it provides the means of sector and derived single common defi nitions provided; this is the developer’s intellectual reliably and consistently documenting the work for V&V. The AMWG separately examined property and is therefore a commercial and the outcomes. The common exclamation the defi nitions using the broad scope of the confi dence. Although the developer is primarily that a QMS adds to cost is correct as there is NAFEMS membership and concluded that responsible for the verifi cation of the software, an additional need for examination of the work. the V&V 10 defi nitions were applicable to the user is ultimately responsible because they However, if the QMS is a signifi cant barrier all engineering simulation (including hand use these complex software tools to produce to performing the work, then it needs to be calculation). The defi nitions were published in a design. The user should require demanding redesigned, as it should not drive the working the ASME V&V 10 guide3 in 2006 (a revised software testing and documentation of the process. edition is due for publication in 2016) and an testing from the developer. The primary The other exclamation that the QMS stifl es ASME/NAFEMS joint introductory text4 in weakness in existing software testing is that creative thinking is also raised as a reason for 2009. the observed order of convergence is not not using a QMS. The answer to this is that Verifi cation is the process of determining provided by the developers; this is true in both if blue-sky thinking is required, then it should that a computational model accurately solid mechanics and CFD software. be performed outside of the QMS so that represents the underlying mathematical model There are verifi cation tasks to be performed there are no constraints. Only those ideas and its solution. by the user; however, these primarily revolve that emerge from the “brainstorming” process Validation is the process of determining around examination of inputs, outputs and with potential to be used for production work the degree to which a model is an accurate application of design codes. need to be subject to the QMS. It is even the representation of the real world from the Validation is the role of the user, although case that a design group can have people perspective of the intended uses of the model. the V&V 10 restrictive view changes the who never work on production tasks; their role It should be noted that the forthcoming emphasis; this is discussed later. It is for the is the blue-sky thinking; and they can exist 2016 edition of the V&V 10 guide will revise user to ensure that the correct methodology outside of the QMS. the defi nition of validation as the examination has been applied for the intended use; and The QMS requires a tool to assess the of the results of a numerical simulation of a that the results have a degree of accuracy design process; this tool is largely the process physical test against the results of that test (or relevant to the end use. of verifi cation and validation (V&V). V&V are a experiment). The test must be a physical test, The validation tasks are therefore requirement of ISO 9001; they are tasks that not an alternative simulation. primarily the responsibility of the user. cannot be excluded from any QMS whatever V&V is also discussed in Guidelines for the Specifi cally, the user chooses various physics the working context. V&V form the basis of the use of computers for engineering calculations5, modelling options within a wide range of remainder of this paper. published in 2002 by The Institution of available options. The user is responsible Structural Engineers, although the publication for choosing those modelling idealisations Verifi cation and validation is now generally considered to be out of date that are appropriate and accurate for the The process of V&V is the signifi cant tool to be and is out of print. particular design that is being considered. used to demonstrate that the simulation used Verifi cation is therefore the examination of The secondary responsibility falls on the to inform the design is fi t for purpose; and that the mathematics and numerics and is seen developer in the sense that they can provide it has been performed using best practice. as being a precise examination. Validation validation test cases to show how well specifi c The process is equally important in hand is the examination of the application or the idealisations of the physical processes calculations and computer-aided simulation. physics and provides relative results; the pass/ represent simplifi ed physical experiments. fail criterion being based on the acceptable Validation is not as deeply established NAFEMS and ASME variance from the real-world behaviour. in the engineering design process as Two groups at the forefront of the verifi cation. The checking of calculations – development of thinking on V&V in engineering Responsibility verifi cation – has always been performed are the Analysis Management Working Group In the context of engineering software, who and is second nature to engineers. (AMWG) of the independent international is responsible for V&V: the developer or the Validation, on the other hand, is not intuitive; engineering simulation body NAFEMS, and the user? The responsibility to demonstrate an engineer using a particular computer American Society of Mechanical Engineers that V&V have been correctly applied lies program to solve a problem does not (ASME) Codes and Standards Committee V&V with the Engineer, as it shows best practice automatically ask if the tool is the correct 10: Verifi cation and Validation in Computational and compliance. The responsibility for the tool. Increasingly with modern software, Solid Mechanics. These two groups work execution of the two processes is split the complex calculation is completely closely together to the extent that there are between the software developer and the hidden from the user. The concept of the crossover members of the groups; and work is engineering designer. user choosing the physics modelling and

TSE51_48-51 Method Validation v1.indd 49 24/02/2016 11:55 ›

50 TheStructuralEngineer Methods and practice March 2016 Veriﬁ cation and validation

idealisations applies equally to the use of satisfying the requirement. The ISO calls for this is the analysis of seismic soil-structure advanced high-end software, such as FE “validation activities”, which implies a suite interaction. Given the variability in the soil solvers, and to bespoke design systems of processes that can measure the validity investigation, the resulting soil properties that design a single class of problem, e.g. of the work. The engineering processes cannot be assumed to be precise, as would be reinforced-concrete beam structures. In that provide this measure are validation and the case for a sample of steel; therefore, the the latter case, the validation task is to predictive capability assessment. Predictive analysis examines the soil with as-measured demonstrate that the bespoke program capability assessment is the working title for properties; then 50% and 200% of the shear is the correct tool to design a particular the alternative to true validation and the title modulus. With computer-aided techniques, suite of real-world structural elements. The may change when the mapping process is varying parameters to produce sensitivity software developer has no control over published. The ISO 9001 requirement can studies is logically straightforward, although the detailed use to which the user puts the therefore be satisfi ed by a demonstration there are practical considerations. program. of validity (for the test environment) or a An overall objective is to reduce the risk that The demonstration of validity of the tool predictive capability assessment for general the modelling process will give predictions not for the job is the user’s responsibility; how purpose work. refl ecting the reality of structural behaviour. therefore can ISO 9001 be satisfi ed? This therefore means that a critical Thus, key requirements of any structural assessment will be performed; the outcome validation are to assure that: ISO 9001:2015 validation being the result of the simulation and bounds of applicability. There are ways of achieving • any design assumptions are properly ISO 9001 8.3.4 Design and development this alternative to true validation so that the refl ected in the as-detailed structure controls ISO 9001 requirement is achieved. • predictions of adequacy are insensitive to The organisation shall apply controls to the Design code: this is by far the most any credible variations in design assumptions design and development process to ensure common method of providing a validated that: design. If the design is shown to be compliant Verifi cation will assure numerical … to the code of practice, and if the code of compliance, but the type of calculation must d) validation activities are conducted to ensure practice is a calibrated code that has been be the right one in the fi rst place – validation. that the resulting products and services meet rigorously developed, then validity to ISO The fi nal comment is that the validity is the requirements for the specifi ed application 9001 is implied. If, however, the design is to conditional: the simulation is valid for the or intended use beyond code criteria, including in many cases load regime that has been examined in to appendices to the design code, validity the validation testing; and it is valid within ISO 9001 therefore requires that work is to ISO 9001 cannot be implicitly claimed. It the bounds of the validation testing. If the validated; it is the task of the Engineer to should also be noted that, in many cases, the conditions change, then the validation testing ensure that this validation has been carried out code itself calls for validation of calculation, has to be repeated. correctly. The forthcoming V&V 10 defi nition of including examining error bounds. The other validation, discussed earlier, apparently leads consideration with respect to design to code is Conclusions to the situation where design of structures that of verifying that the correct code is being The underlying questions are: how do we know cannot be validated; e.g. it is not possible to applied. that the results are reasonably correct; and physically test a nuclear reactor core against a When designing to recognised codes of who is responsible if there is a failure? 1 in 10 000-year seismic event. practice, the error bounds are within the code The ultimate responsibility for the design, The AMWG is currently examining these calibration and the design engineer does not whether or not it is performed using computer- apparent confl icting perspectives in a need to explicitly examine these further, unless aided techniques, lies with the Engineer. project to map the ISO 9001 requirement to such an examination is a condition of the code Therefore, it is for the Engineer to determine the process of examining validity. The ISO itself. that the results are correct within the practical 9001 requirement is a statement that the Beyond code design: there are many limits of the design requirements. It is for the validity should be demonstrated; however, it design cases where the codes of practice at Engineer to ensure that the calculations have is a generic requirement and is necessarily best provide guidance while at worst there been adequately verifi ed and validated. general. It has to apply to a wide range of are no applicable codes. This includes many Verifi cation of work carried out using products, e.g. from children’s toys to space safety-related design cases where extreme computer-aided techniques is split between vehicles. The mapping will identify that the loading with return periods greater than 1000 the developer and the user. The developer ISO requirement can be met by either formal years are considered. Demonstrating ISO is responsible for the verifi cation of the validation, as defi ned by V&V 10, or an expert 9001 validity in this case is more involved. mathematics and numerics; since these are opinion assessment. The result is validation to Clearly, a part of the solution is using encased in the source code, the user has ISO 9001, even though the assessment that results from the examination of parts of the no access to these. As the source is the has been carried out is not formal validation. structure; these examinations can be code- developer’s intellectual property, the evidence This is work in progress; however, it is hoped compliant investigations; they can be simplifi ed of the verifi cation is usually provided by means that it will be complete in time for the issue of assessments. Another available approach of documentation. As the user is responsible the revised V&V 10 guide in 2016. is comparison to previous work and looking for the design, it is the user’s responsibility to at variance in the results. The previous work demonstrate that the developer verifi cation is Demonstrating validity would in itself have to be ISO 9001 valid, and in place. The user’s verifi cation responsibilities In general civil engineering it will be a rare great care is necessary if the examination is an revolve around examination of inputs, outputs event where a simulation of a physical test extrapolation of the previous work. and application of design codes. is carried out; therefore, validation to V&V The major tool for examination of validity Validation is the role of the user, although 10’s defi nition is not possible. To examine is to perform sensitivity studies, varying the V&V 10 restrictive view changes the this confl ict it is necessary to separate the parameters in the design to see how the emphasis. It is for the user to ensure that requirements of ISO 9001 from the means of results change. One extreme example of the correct methodology has been used

TSE51_48-51 Method Validation v1.indd 50 24/02/2016 11:56 www.thestructuralengineer.org 51

for the intended use; and that the results have a degree of accuracy relevant to the end use. Validation is not as deeply established in the References engineering design process as verifi cation. The checking of calculations – verifi cation – has always been performed and is second nature to engineers. Validation, on the other hand, is not intuitive and is not E1 British Standards Institution (2015) BS EN necessarily mathematically driven. ISO 9001:2015 Quality management systems. Responsibility for ISO 9001 validation of the application of the Requirements, London, UK: BSI results of a simulation lies with the end user, as the developer has no knowledge of how the program will be used. The exception would be E2 NAFEMS (2008) QSS:2008: Engineering simulation where the software takes the design from input conditions provided by – Quality management systems – Requirements, the user, through analysis, then to code-compliant design. At this point Hamilton, UK: NAFEMS the developer would also be required to verify that the design code assessment has been correctly applied. The end user would also need E3 American Society of Mechanical Engineers (2006) to run variations of the design to examine the variability of the outcome, V&V 10-2006: Guide for Verifi cation and Validation thus validating the overall design process. in Computational Solid Mechanics, New York, USA: If the end user cannot get access to the developer’s V&V, then the ASME [Revised edition due in 2016] process is to carry out ISO 9001 validation studies to bound potential variations to the design; this is equally the case where the software E4 NAFEMS and American Society of Mechanical analyses and designs. It is for the end user to demonstrate ISO 9001 Engineers (2009) What is Verifi cation and Validation? validity of their work, as it is for the end user to take responsibility for the [Online] Available at: www.nafems.org/downloads/ end product. working_groups/amwg/4pp_nafems_asme_vv.pdf The ISO 9001 requirement to perform “validation activities” is (Accessed: January 2016) satisfi ed by restrictive validation (of tests) or by examining the predictive capability of an analysis or calculation. E5 The Institution of Structural Engineers (2002) Guidelines for the use of computers for engineering Declaration calculations, London, UK: IStructE Ltd Chris is the chair of the NAFEMS AMWG and a member of ASME Codes and Standards V&V 10. This paper has, however, been prepared in his personal capacity; the opinions expressed are his own and should not be considered the view of NAFEMS or ASME.

The Institution publishes a wide

range of authoritative manuals, reports and guides that provide essential, accurate information for structural engineers.

These publications are a valuable addition to any practising structural engineer’s reference library.

Every title undergoes an exacting peer review procedure and is reviewed by at least two external assessors (and, where appropriate, is subject to a legal review).

Members enjoy substantial discounts across all publications, including the bestselling suite of Eurocode manuals. Manual for the geotechnical design of structures to Eurocode 7 May 2013

Institution Structural Engineers shop.istructe.org

TSE51_48-51 Method Validation v1.indd 51 24/02/2016 11:57 ›

52 TheStructuralEngineer Methods and practice March 2016 Hand calculations Integration of hand calculations with computational output – a good practice summary

only, before introducing lateral loads arising Andrew Wright BEng(Hons), MIStructE, MICE, CertCII, Triton from wind and eccentricity. A basic check should be carried out for the lateral loads; then, once the model has been determined to be The design of structures using computer While reviewing this subject it has become performing correctly under the separate loads, software is prevalent throughout the industry. apparent that there is very little guidance load combinations can be considered. It provides quick analysis of structures that available from the various institutions and Following this, the engineer should consider can be readily incorporated into Building bodies that have an interest in the topic, and the transfer of loads through the structure Information Modelling (BIM) systems such as this should be addressed. However, in the and whether the loads are accumulating in Revit1 . However, while the use of such software meantime, what should an engineer consider the correct places and acting in the correct is essential in the present-day design offi ce, it when undertaking computer-aided design? direction. For example, a column supporting should be recognised that hand calculations a heavy masonry facade should not attract a and sketches are an important part of an Inputs negligible load, and a transfer beam supporting engineer’s toolbox in establishing a “feel” for If the data used to form the analysis in the eight storeys of building should have, at the the structure that they are designing. fi rst place are incorrect, it is inevitable that the least, a large bending moment or shear force, Without an engineer’s inherent results will be incorrect. Basically: “rubbish in, or both. understanding of what a structure can and rubbish out”. The engineer should also check the bending should be doing, which is generally established The results will identify input errors very moment and shear force diagrams for the through years of training and experience, quickly – a 3km long beam is going to defl ect structure. Are they acting as expected? For it is quite possible for design and analysis a lot no matter what depth it has! So, in the example, do the moments and shear forces software to produce serious errors that would fi rst place, the engineer should check that work in the correct direction and are moment otherwise be easily caught by the use of the dimensional data used in the model are joints attracting a moment? Similarly, are common sense and hand calculations. correct, the supports, nodes and connections pinned nodes transmitting axial compression This article aims to provide a “good practice” are performing as intended, and the correct or tension only? summary for engineers in the early stages of units are being used. The last item is their career, while also serving as a reminder particularly important when converting units Numerical checks for those with more experience. from metric to imperial or from one constant Once the engineer has confi dence in the to another. model and is satisfi ed that it is performing as Safety concerns The engineer should then consider the intended, they should consider undertaking The Standing Committee on Structural Safety loads applied to the model. Are they acting basic numerical checks on the results to (SCOSS) has raised concerns that a growing in the correct direction? How do they work ensure that they are correct and within the gap between the power of the software and in conjunction with other loads? Are the load expected order of magnitude. Key elements, the understanding of users may lead to the cases correct? Have all of the applicable load along with any issues highlighted by the construction of unsafe structures. Possible cases been identifi ed? “common-sense” checks, should be subject to misuse of computer software may arise from: basic hand checks. Results Global checks should be carried out on • inappropriate modelling of structures for Common-sense checks the structure to confi rm that the approximate analysis Before undertaking any interrogation of load applied to the structure is similar to that • the inability of engineers to use approximate the numerical output, the engineer should applied to the foundations. Furthermore, highly design methods effi ciently look at the graphical output and establish loaded or lightly loaded columns should be • too much trust being placed in computer that the results are performing in line with checked to confi rm that the foundation loads outputs with the result that the engineer is expectations. In the case of a model of are generally correct. distanced from the problems being addressed a multistorey building, say, is the global The extent of hand checks should be and unable to apply practical engineering defl ected shape of the structure sensible determined by a brief risk assessment thinking and of the correct order of magnitude? Any of the importance of the element under • an insuffi ciently fundamental checking odd defl ections of beams and columns or consideration. If the element is especially process excessive sway should be investigated and critical to the stability of the structure, or the • the limitations of the software not being corrected. initial checks have identifi ed a problem, more suffi ciently apparent to the end user It would be a worthwhile exercise to rigorous checks should be undertaken. • the software containing errors consider the model fi rst under gravity loads In the fi rst instance, approximate checks on

TSE51_52-53 Method Hand calc.indd 52 24/02/2016 12:00 www.thestructuralengineer.org 53

the bending moment and shear force directions should be made. These should then be backed up with checks on span/depth ratios of beams or reinforced-concrete slabs, and stresses in columns and other load- FOR SALE bearing elements. Consideration may also be given to the percentage of reinforcement in the cross-sectional area of the elements, among other things, to confi rm whether they are too small or too densely packed. Structural & Summary Architectural practice Many engineers – whether embarking on their career or more experienced – will consider the advice presented here normal practice and will have adopted these ideas already and developed them to suit • M25 location their own requirements. However, the temptation for some software users is to create their model, input the design loading and accept • Established 20+ years whatever comes out with no resort to interrogation or basic hand • £1M - £1.5M turnover checks. This may be acceptable for a single beam where the answer is expected, but for more complex analysis, the engineer should consider • Profi table how confi dent they are in the adequacy of the design that they are • Commercial client base proposing to construct. So, while design software is a wonderfully effi cient tool in the design • Good core staff offi ce, engineers at every stage of their career should remember that there is no substitute for the innate knowledge of structural design • 1 – 2 year workload confi rmed accumulated over many years of training and experience. The common • 2 – 3 year handover sense that we are so proud of is the most important item in our tool box!

Replies in conﬁ dence to: Reference m25structengbussale E1 Autodesk (2016) Revit [Online] Available at: www. @gmail.com autodesk.co.uk/products/revit-family/overview (Accessed: February 2016)

‘Engineering in a Dynamic Environment II’ Technical A new series of free technical lectures

Lecture A series of lectures examining topics, from Designing for Construction to Construction Methods for Concrete Bridge Decks.

All the events will take place at the Institution’s International Headquarters at

Series 47-58 Bastwick Street, London. Registration will be from 17:30 with the lecture starting at 18:00, unless stated otherwise.

PROGRAMME Thursday 31 March 2016 Temporary Works: Yours, Mine or Ours? In the context of CDM15 (Joint discussion and debate with the Temporary Works Forum) | David Lambert, Martin Thorpe, Phil Cantrell

Wednesday 20 April 2016 Best Construction Methods for Concrete Bridge Decks – Cost Data (Joint meeting with the Concrete Bridge Development Group) | Simon Bourne

Wednesday 27 April 2016 Designing for Construction | Angus Cormie

Thursday 12 May 2016 Engineering and Constructability | William Parker

For more information, please contact [email protected]. Registration is required online, please visit the events section of the Institution website, www.istructe.org

TSE51_52-53 Method Hand calc.indd 53 24/02/2016 12:00 p54_TSE.03.16.indd 54 19/02/2016 16:06 › www.thestructuralengineer.org TheStructuralEngineer 55 March 2016 Project focus

Peer-reviewed papers focusing on the structural engineering challenges faced during the design and build stages of a construction project.

56 City of Dreams, Macau – making the vision viable

This article describes how cutting-edge parametric-based engineering techniques have been used to achieve the detailed design of 2500 complex steelwork connections for the exoskeleton of the new City of Dreams hotel in Macau. It discusses the tools, methodology and strategy employed by the engineering team to automate the diffi cult and time-consuming process of creating, verifying and documenting the geometrically challenging, large-scale steel connections using ﬁ nite-element methods within an ambitious timescale of just 12 months.

69 Digitally designed – Qatar Faculty of Islamic Studies

This paper describes the digital parametric design and fabrication optimisation that was carried out on the recently-completed Qatar Faculty of Islamic Studies building in Doha, Qatar. The processes, tools and techniques used are described, with background given about the issues and decisions that led to their development and implementation.

TSE51_55 _Project focus opener.indd 55 24/02/2016 12:00 ›

56 TheStructuralEngineer Project focus March 2016 City of Dreams

City of Dreams, Macau – making the vision viable

Emidio Piermarini EI, BEng, MEng, Engineer, BuroHappold Hong Kong Hayden Nuttall MSc, DIC, BEng, CEng, FIStructE, MHKIE, Director, BuroHappold Hong Kong Rob May CEng, MIStructE, PE, MHKIE, MHKIBIM, Associate Director, BuroHappold Bath Victoria M. Janssens PhD, PEng, Senior Structural Engineer, BuroHappold Hong Kong

Synopsis Introduction This article describes how cutting- An extraordinary building is taking shape edge parametric-based engineering in the City of Dreams entertainment techniques have been used to resort in Macau (a Special Administrative achieve the detailed design of 2500 Region of the People’s Republic of China). The 42-ﬂ oor twin-tower construction complex steelwork connections incorporates an irregular-form, aluminium- for the exoskeleton of the new City clad structural exoskeleton with of Dreams hotel in Macau, China. It connections of such scale and complexity discusses the tools, methodology and that they are possibly the most analytically strategy employed by the engineering and geometrically challenging large-scale steelwork connections ever to be built team to automate the diffi cult and (Figure 1). time-consuming process of creating, The project for Melco Crown verifying and documenting the Entertainment by Zaha Hadid Architects geometrically challenging, large-scale and BuroHappold is under construction steel connections using ﬁ nite- (Figure 2). When it opens in 2017 it will provide the City of Dreams development element methods within an ambitious Figure 1 with a dramatic landmark building to ECity of Dreams timescale of just 12 months. hotel – architect’s complement the existing complex of hotels rendering ARCHITECTS HADID ZAHA www.thestructuralengineer.org 57 ›

58 TheStructuralEngineer Project focus March 2016 City of Dreams

Figure 3 SStructural system

Figure 2 City of Dreams hotel – current progress (January 2016)

a) Concrete cores VADIM ISMAGILOV VADIM

and entertainment facilities on the “Cotai In structural terms, the steel exoskeleton Strip”. Housed within its 150 000m2 of and the two internal concrete cores act fl oor space will be a seven-storey atrium, together to provide lateral load resistance, 780 hotel rooms, suites and villas, various sharing wind and seismic loads in restaurants, luxury retail outlets, gaming proportion to stiff ness. The gravity system and event facilities, and a sky pool. comprises composite beams and slabs BuroHappold carried out the general that span between the exoskeleton and the structural design of the building, together cores with minimal internal columns (Figure b) Exoskeleton with the detailed design and construction 3). documentation of all the steelwork There are approximately 2500 connections. The structural design work connections in the exoskeleton. The faced engineering challenges arising from members and connections are fabricated the typhoon wind climate, seismic design from steel plate up to 150mm thick requirements, complex load paths and using grades up to S460. Many of the highly irregular geometry of the building, connections incorporate “off shore-quality” but it is the uniquely complex problem of plate to BS EN 102251 in order to ensure the detailed design and documentation of adequate ductility and strength in the the thousands of dissimilar and irregular through-thickness direction. Members are steelwork connections of the exoskeleton generally bolted together at connections – and the innovative methodology used in the fl at regions and site-welded in the to solve it – that are the subject of this free-form central zone and the corner fi llets article. (Figure 4).

Structure Methodology The design concept for the City of Dreams With such complex and irregular geometry hotel is of a striking exoskeleton which it was clear from the outset that traditional wraps around the two concrete cores, code-based methods and standard c) Total system bringing them together with a ﬂ owing mid- drawing software would not be suffi cient section featuring three irregular-shaped to design and document the exoskeleton curved openings. Inside the building, the connections. Instead, the BuroHappold option to verify their structural adequacy. free-form steel framework continues, team decided that the complex stress It was also clear that standard software curving high above a huge atrium space states that exist where members merge packages would not have the functionality that is echoed by that of the sky pool into the connections meant that ﬁ nite- required to create the construction above. element (FE) analysis was the only viable documentation, especially for the free- www.thestructuralengineer.org 59

Input Parameters Grasshopper Output Model (variables) Deﬁnition/Script Viewed In Rhino 3D

Figure 4 Figure 5 Figure 6 NZones of exoskeleton NParametric deﬁ nitions using Grasshopper visual programming for Rhino 3D SDesign process

form central region. (3D) visualisation of every connection complex 3D forms quickly and accurately To complicate matters further, since the throughout. The approach allowed the using visual programming techniques and, exoskeleton would be clad in aluminium, engineering team to focus on the quality crucially, to make changes to the geometry all connections and associated plates of the engineering solution, rather than on by changing the parameters (Figure 5). and bolts would have to be located cumbersome data handling and repetitive They could literally “see what they were within the cladding zone defi ned by the number-crunching tasks, resulting in doing” in each step of the programming architect. This would inevitably constrain signifi cantly faster and reliable output. As logic and in the corresponding geometry and limit options for the geometry of the a result, the entire detailed design and as it was being created, making the code connections and necessitate non-planar documentation process was completed on debugging process much easier and solutions. schedule, in a fraction of the time that the quicker than it would have been using Finally, the timescale for detailed team estimated would have been required traditional practices. design and documentation of all 2500 using a more conventional methodology. Rhino 3D was also used to model exoskeleton connections was just 12 the outer surface geometry as a clash- months. Put another way, the team would Tools detection study to show that the need to complete an average of 50 Parametric design is a process in which connections were within the cladding connections per week. problem parameters are defi ned as zone. Autodesk’s Robot Structural Analysis In response to this seemingly impossible variables and a series of functions applied (RSA)6 software was used to create task, BuroHappold drew on its expertise in order to fi nd the solution(s). By varying the local FE models for each unique in parametric engineering and structural these parameters, many variations of connection type. optimisation developed on previous the same problem can be solved. In this In this context, it is worth noting that projects, including the Aviva Stadium in case, the problem was FE analysis of the the size and complexity of the structure Dublin, Ireland2, and the Louvre Museum many and various steel connections in the meant that the global analysis model for in Abu Dhabi, United Arab Emirates3. exoskeleton. the building, which was created using Essentially, the team’s solution was to The modelling software Rhinoceros MIDAS structural analysis software7, took create a unique, bespoke computational 3D (Rhino 3D)4 and its plug-in module over 12 hours to run. Hence, it was not approach using application programming Grasshopper5 were chosen as parametric viable to create and insert FE models of interface (API) techniques to allow design tools for the speed and accuracy all the connections into the global model, effi cient processing of the huge number they would bring to the task. as this would have increased the analysis of FE models required and, critically, The combination allowed the team to time even further, possibly by three or four corresponding three-dimensional create the geometry for a large number of times. Similarly, if the models were inserted ›

60 TheStructuralEngineer Project focus March 2016 City of Dreams

Figure 7 Grasshopper script to identify similar connections

one at a time, there would be at least a 12-hour wait each time the team wanted to investigate alternative arrangements for a connection. The only practical alternative was to create separate “local” FE models of the connections and transfer, or map, onto them the corresponding moments and forces from the global model results ﬁ le, for all 105 load combinations. To maintain the tight programme, almost every aspect of the local model generation and analysis was automated using bespoke Visual Basic scripts that linked MIDAS, RSA and Excel with Grasshopper via their APIs. In achieving the solution to this ambitious project, the BuroHappold engineering team found themselves at the cutting edge, using the software in ways that had not been done before, sometimes working at the limits of the products’ capabilities. The team maintained frequent dialogue with all the software companies’ technical support teams throughout, which proved to be highly productive for both parties.

Strategy The engineering team’s strategy was to break Figure 8 Visualisation of data for similar connections this immense problem into fi ve key steps (Figure 6). The fi rst four months of the project were spent developing bespoke Grasshopper scripts for every step. This signifi cant time investment was justifi ed many times over by the huge time saving made in the subsequent analyses and generation of documentation. It is important to understand that bespoke programming, however skilled it might be, does not replace engineering expertise. Rather, it augments it by handling large amounts of data effi ciently and releasing engineers to focus on optimising the design. Accordingly, visualisation, manual verifi cation and

a) b) c)

Figure 9 EDeveloping connection arrangement RUPERT INMAN RUPERT www.thestructuralengineer.org 61

acceptance were considered essential and a total number of over 2500 connections, this The connection designs also had to built into the process throughout. reduced the number of unique types to about accommodate extraordinary architectural The fi ve steps were: 400. constraints. Zaha Hadid Architects had provided a Rhino 3D model of the inner 1. Identify similar connections Step 2: Develop the connection arrangement surface of the cladding zone that all the steel 2. Develop the connection arrangement Next, the principles of the connection were elements and connections had to fi t inside 3. Find and map forces from global analysis developed through engineering judgement (Figure 10a). For simpler connections, with model based on the load paths (Figure 9a) and a little or no curvature, the connections and 4. Analyse the connection Grasshopper script was created to allow the associated plates and bolts were similar in size 5. Generate analysis reports and construction designer to rapidly conceive a connection’s to the steel elements; therefore, clashes were documents geometry to meet architectural and fabrication relatively easy to manage. However, in the free- constraints before sending the connection to form central zone, multiple members typically Step 1: Identify similar connections be analysed. meet with high curvature, leading to complex The fi rst task was to confi rm the number Mindful of the fabrication and erection intersections (Fig. 10b). For this reason, the of unique connection types required by challenges that such massive connection connections are necessarily signifi cantly identifying those that were similar, in order to nodes would present, 3D study models of larger than the individual members. Given the reduce fabrication and erection time. With up each connection were created using Rhino architectural envelope was not only tight but to nine elements connecting at each node, 3D to ensure that the connections could be also varied in depth where it was in double and each element potentially having a diff erent readily fabricated. The models show the “plate- curvature or warped, clashes were a real section shape, section size and/or curvature, by-plate” fabrication sequence for appropriate possibility (Fig. 10c). this was not an easy task. clearance at every stage, including edge Since the constraints of fabrication To identify unique connection types, a distance tolerances and room for site welding, would often oppose those presented by Grasshopper script was written to interrogate testing and bolt tightening (Fig. 9b). the architecture, the team realised that the exoskeleton member geometry Rhino a) 3D fi le created previously to help build the global structural analysis model. It contained the member centreline geometry and the associated section shapes and sizes. The script was used to search this fi le for all intersections of centrelines (to locate the connections) and to collect and organise the relevant geometric data, such as the number of intersecting members, whether members are straight or curved, the member shapes and sizes and the angles between adjacent members. Thus organised into a programming library, the data could be easily and accurately compared to determine similarity of the intersections, allowing for cases where the geometry is handed (Figure 7). b) c) Once the unique connection types had been identifi ed, the script visualised the geometric data from the Rhino 3D fi le, allowing the team to verify the similar connection information easily (Figure 8). From

Figure 10 Working with architectural envelope ›

62 TheStructuralEngineer Project focus March 2016 City of Dreams RUPERT INMAN RUPERT

Figure 11 NParametric connection deﬁ nition and fabrication connection

fi nding an optimal solution meant being multiple connections. These allowed parts applying them to the local connection able to explore the design space for of the parametric scripts for one unique models was signifi cant. Once again, each connection rapidly. To address this, connection to be copied or developed for Grasshopper’s capability as a tool BuroHappold engineers programmed the application to others. for creating and visualising geometry geometry of each connection using a For example, as a general design off ered a number of benefi ts in terms of parametric script with variables defi ned principle, a 25mm edge distance tolerance speed and reliability. for all dimensions that were likely to was allowed for members being site- With a script similar to that used in need further study to meet architectural, welded to the connections, to account for Step 1 to identify unique connections, fabrication and construction constraints erection tolerances. However, increasing data including connection geometry, (Figure 11). The more complex the the thickness of a connection node in bar/node numbering, section sizes and geometry of the connection, the more order to maximise edge distance for member orientations were transferred complex the parametric script became, but site welds would make it more likely that to Grasshopper from the global analysis some guiding principles were common to the connection would clash with the model and mapped for each connection architectural envelope. In order to under consideration (Figure 13). The Figure 13 Global analysis model explore this, the thickness was corresponding forces/moments were defi ned as a parameter within the then also extracted. The volume of Grasshopper script. The value data this created was so large that it could then be adjusted until the was split into 55 separate fi les, each edge distance tolerance of 25mm containing up to fi ve million sets of bar was achieved. forces/moments. Thanks to Grasshopper’s The bespoke scripts allowed powerful visualisation, all these designers to search for any set of changes occurred graphically and forces/moments from the entire data in real time as the designer moved set and instantly visualise them on the slider value up and down screen. In-built vector transformation (Figure 12). If the 25mm tolerance tools could then be used to map the could not be achieved because forces/moments onto the local model. of a clash with the architectural The task would have been much more envelope (as in the example diffi cult and time-consuming without the shown), the designer could rapidly powerful visualisation functionality that determine what value would Grasshopper provides, allowing as it did optimise the edge distance while for “visual debugging” of the script. remaining within the architectural Even with Grasshopper’s power envelope. vector tools, mapping and translation of the forces/moments from multiple fi les Step 3. Find and map forces from was susceptible to error, so the team global analysis model used a two-step verifi cation process With 105 load combinations and comprising visual and numerical checks up to nine members in a single to ensure the extracted data were connection, the process of fi nding correct (Figure 14). the correct forces/moments in For the visual check, the connection the global model and correctly was displayed in 3D together with www.thestructuralengineer.org 63

Figure 12 SUsing parametric deﬁ nition to meet multiple constraints

vectors showing the magnitude and direction of the applied forces/moments. This quickly displayed any missing data and veriﬁ ed that the forces were acting in the correct direction. Additional analytical information from the global model, such as bar/node numbers, section properties, gamma angles and local axes, could be displayed as well to ensure proper mapping of bar information. For numerical veriﬁ cation, an equilibrium check was performed for all load combinations to ensure no “out-of- balance” forces/moments existed. Any questionable load combinations or nodes were then displayed graphically and further interrogated.

Step 4: Analyse the connection The accurate prediction of the resultant stresses where multiple members intersect was a major concern. Consideration of even a simple cruciform example illustrates the importance of accurately predicting stresses where members merge (Figure 15). At the start of the project, the BuroHappold team had determined that neither established code-based methods nor bespoke ﬁ rst-principles methods would readily capture the complex stress states that exist in the many and varied connections of the exoskeleton where individual plates intersect and overlap, especially in locations where multiple plates up to 750mm wide merged into a single plate. Given the geometric complexity and sheer size of the connection nodes, an FE approach was the only viable method for verifying the adequacy of the connections. Almost every step of the connection Figure 14 analysis process was semi-automated N Visualisation and numerical check of mapped forces ›

64 TheStructuralEngineer Project focus March 2016 City of Dreams

Figure 15 SInteraction of in-plane principle stresses and theoretical von Mises envelope for simple cruciform connection a) In Rhino 3D

b) Connected live to RSA

NB In both cases, the stress levels and for the i ncoming members of the cruciform are set at the yield stress of the material σ1 σ2 . When and are both positive or negative (right-hand case), the maximum stress in the overlapping region does not (py) σ1 σ2 signiﬁ cantly increase. However, when and have opposite signs (left-hand case), the maximum stress in the overlapping σ1 σ2 region reaches . This phenomenon is predicted by inspection of the well-known “von Mises failure envelope”. √3 × py

to reduce set-up and processing time, Figure 16 EConnection using bespoke scripts to link the various model software programs to Grasshopper though their APIs. The scripts were used to

• generate the local FE model • add extension bars and apply the forces surfaces at the centre of the plates, which This again mitigated errors associated • apply analytical links and boundary could be planar or curved, and converted with manual processes such as copy and conditions these surfaces into RSA objects. It then pasting tabulated data. • run the analysis and extract results asked RSA to create the FE mesh of 2D Under a conventional approach, the shell elements from these objects. Since defi nition of the analytical links between At every stage, the engineer could the FE mesh would be generated inside the bars and the shell elements in RSA, employ visual checks to ensure the correct RSA, the geometry of the surfaces created and the defi nition of boundary conditions data were being used. Once the scripts in Rhino 3D needed to be of suffi cient (analytical supports) would both have had been created, these local analysis accuracy to avoid meshing problems, been time-consuming manual operations. models took just a few minutes to run which can occur when the meshing Here, they were both scripted to happen (compared to 12 hours for the global algorithms cannot determine the intended automatically, saving considerable time analysis using MIDAS), allowing the team common boundary between adjacent for the project. The nodes of the FE mesh to run them as many times as they needed surfaces. Since the Rhino 3D geometry were imported into Grasshopper, which to, in order to match plates’ thicknesses to was defi ned parametrically, the overall applied a script that used geometric stress levels and optimise the connections. geometry could be altered as necessary search algorithms to fi nd the appropriate The FE models were based on 2D shell until the connection was optimised and nodes to which the bar elements should elements that incorporated all plates in the the various fabrication/architectural be connected. This information was then connection together with an appropriate constraints had been met. sent back to RSA and used to create portion of the incoming members. Beyond Once the 2D shell elements were the analytical connections. The script this, bar elements were added to match generated, the script automatically added also automatically applied the required those in the global model and the mapped the bar elements to the model. The bar boundary conditions to the local RSA forces/moments from the global model geometry was extracted directly from the model in predetermined locations. were applied to these. Since the geometry global analysis model and placed in the After the forces/moments for all load and the forces/moments in the local and same virtual position in the local model. cases had been applied, the models were global models should match, it was easy As the bar forces had been mapped batch-processed. Finally, the sum of each to check these visually and numerically inside Grasshopper, and the bar numbers reaction was checked to ensure they all against one another. generated in the local model matched equalled zero before the results were The fi rst step was to generate the local the global model, the load combinations prepared for extraction. analytical model in RSA (Figure 16). The and bar forces/moments could be To avoid unnecessary handling of large script fi rst created a Rhino 3D model of 2D automatically applied using Grasshopper. and cumbersome data fi les, and to speed www.thestructuralengineer.org 65

Figure 17 RSA von Mises stress plot and fabricated connection

Figure 18 could be a laborious task. Since all the WExample of calculation output visual data available to the designers during the design process were created in Grasshopper, the logical solution was to transfer this to an Excel template after the analysis was complete. By creating a tool to automate this task, the team made considerable time savings and provided a comprehensive visual record of all steps Figure 19 of the design process, ensuring that any 3D documentation for largest free-form connection independent party could easily follow the assumptions made and data used for the design of each connection (Figure 18). While documentation was not a primary objective of the process, the Grasshopper scripts generated rich and coordinated data that could be easily extracted to provide accurate and relevant information for the fabricator. After careful consideration of the options, it was agreed with the contractor that the construction information would be issued in the form of 2D drawings for the connections in the fl at-sided areas and curved corners, where the geometry could be readily defi ned using conventional drawing software, and as 3D digital models for the free-form areas to assist the fabricator in understanding the connection geometry (Figures 19 and 20). This was because the design intent for up the process, a script was developed to results were achieved. the connections in the free-form area was extract stresses in batches to determine Finally, the results were all individually more diffi cult to communicate using 2D the governing load cases. Stress maps reviewed by BuroHappold engineers as drawings. Since the 3D information was of these connections were interrogated part of the verifi cation and acceptance readily available, it seemed illogical to using a scale based on the maximum plate process. convert this to conventional 2D drawings thickness for a given selection of plates that would have required multiple views, (Figure 17). The stress maps were then Step 5: Generate analysis reports and sections and coordinates to defi ne the visually inspected to establish whether the construction documents shape, position and orientation relative to stresses in any areas were unacceptable. It was recognised early in the project the fi nished structure. Rather, using the If necessary, the plates’ thicknesses, that, given the large number of unique Rhino 3D surface models that had already arrangements or grades were changed and nodes, the generation of engineering been created for the clash-detection the whole process re-run until satisfactory documentation for each connection studies, 3D models that were geometrically ›

66 TheStructuralEngineer Project focus March 2016 City of Dreams

Figure 20 SExample of complex 3D documentation

a) Assembly details b) 3D setting-out

Figure 21 N3D documentation via digital model and construction

accurate in every sense (plate thickness, connections and 105 load cases. The carefully designed to avoid being a so- plate geometry, plate hierarchy at plate whole process was run using bespoke called “black box” set of tools, but rather intersections, actual position/orientation parametric Grasshopper scripts, which an extension of the engineer’s hand; cutting in the building) were provided for the successfully integrated MIDAS, RSA, out mundane tasks and allowing more time fabricator, who simply transferred them into Rhino 3D and Excel. Due to the number of to focus on problem-solving. their own 3D construction model. unique arrangements, their highly irregular The initial decision to spend the ﬁ rst The approach was mutually beneﬁ cial shapes and the complex stress states four months of the 12-month programme as it saved time for all parties in what was that exist where the members merge, the developing the process and writing/testing already an aggressive schedule and helped exoskeleton connections are possibly the parametric scripts was a bold one, but to minimise fabrication errors (Figure 21). the most analytically and geometrically one which paid off later when some of the challenging large-scale connections of connections were being created, analysed Conclusion any building constructed to date (Figure and documented in less than one hour. To meet the aggressive construction 22). There was inevitably periodic updating programme for the City of Dreams The Grasshopper scripts not only of the scripts throughout the project, hotel project, BuroHappold needed to allowed the engineering team to process but the majority of the development was develop a state-of-the-art approach to vast amounts of data quickly; importantly, completed in this early stage. Once set up, the complex design and documentation they also incorporated “on-screen” visual this innovative design approach achieved of the exoskeleton connections. This checks at all stages of the process to huge savings in man-hours and allowed involved full FE analysis of more than 2500 help eliminate errors. The scripts were BuroHappold to consistently deliver ahead www.thestructuralengineer.org 67

Figure 22 ENode fabrication in Guangzhou, China of schedule. Structural engineering in the modern era is challenged by projects of increasing complexity, falling fees and faster construction programmes. The profession will not meet these competing challenges successfully without harnessing the best available technology. The construction industry is now largely a “digital” industry, with the leading design teams, contractors and manufacturers increasingly creating and sharing digital information. For structural engineers, parametric and computational design are the tools that will enable them to embrace this complexity, avoid getting bogged down in ever-increasing amounts of data and devote more valuable time to what they do best – engineering.

References

E1 British Standards Institution (2009) BS EN 3D [Online] Available at: www.rhino3d.com 10225:2009 Weldable structural steel for ﬁ xed (Accessed: January 2016) off shore structures. Technical delivery conditions, London, UK: BSI E5 Robert McNeel & Associates (2016) Grasshopper [Online] Available at: www.grasshopper3d.com E2 Shepherd P. (2011) ‘Aviva Stadium – the use of (Accessed: January 2016) parametric modelling in structural design’, The Structural Engineer, 89 (3), pp. 28–34 E6 Autodesk (2016) Robot Structural Analysis Professional [Online] Available at: www.autodesk. E3 Shrubshall C. and Fisher A. (2011) ‘The practical co.uk/products/robot-structural-analysisoverview application of structural optimisation in the design (Accessed: January 2016) of the Louvre Abu Dhabi’, Taller, Longer, Lighter: Proc. IABSE–IASS Symposium, London, UK, 20–23 E7 MIDAS Engineering Software (2016) midas Gen September [Online] Available at: http://en.midasuser.com/ product/gen_overview.asp (Accessed: January E4 Robert McNeel & Associates (2016) Rhinoceros 2016)

A concrete shield A serious fire can damage business. Concrete’s inherent properties mean it doesn’t burn or give off noxious fumes.ADVERT Its slow rate of thermal conductivity also provides an effective fire shield, protecting lives, livelihoods and the environment. Choose concrete for built-in fire resistance. Visit www.concretecentre.com to find out more. fact SCISC COURSES

sted Expertise www.steel-sci.com Tru for s & 30 se Y ur ea o r C s

SCI provides technical in-depth courses to keep engineers up to date. /HDUQDERXWWKHODWHVWVWHHOGHYHORSPHQWVWRGHVLJQFRPSHWHQWO\HIÀFLHQWO\DQGVDIHO\GVDIHO\VDDIIHIHO\ Check website for current schedule.

EC4 Composite Design Steel Connection Design Steel Building Designn too EC3 EC 14 April 2016, Bristol 28 April 2016, London 10 May 2016, Reading • Composite construction as an 15 September 2016, Glasgow 27 September 2016, Warrington HIÀFLHQWRSWLRQ • Nominally pinned connections • Loading and load combinations • Design procedures for simple • Moment resisting connections • Frame stability composite beams • Bracing connections • Brittle fracture • Design procedures for composite slabs • Bases, splices & welded joints • Member resistance (axial load and 6KDOORZÁRRUDQGSUHFDVWVROXWLRQV • Non-standard connections lateral torsional buckling) • Examples of good practice • Connection design SCI Members £220.00 + VAT Non Members £280.00 + VAT

The SCI is committed to helping members meet their design, For further details contact manufacture, construction and commercial objectives. +44 (0)1344 636500

@SCIsteel steel-construction-institute SCI Course List www.steel-sci.com/courses

p68_TSE.03.16.indd 68 19/02/2016 16:07 › www.thestructuralengineer.org Project focus TheStructuralEngineer 69 Qatar Faculty of Islamic Studies March 2016

Digitally designed – Qatar Faculty of Islamic Studies

Ben Lewis BEng (Hons), CEng, MIStructE, Associate Director, Arup

Synopsis This paper describes the digital Figure 1 Completed QFIS building parametric design and fabrication optimisation that was carried out on the recently-completed Qatar Faculty of Islamic Studies building in Doha, Qatar. The processes, tools and techniques used are described, with background given about the issues and decisions that led to their development and implementation. The benefi ts we believe these tools brought to the project are also discussed, from the point of view of both the building’s construction and Arup’s internal workfl ow. QF/ASTAD/CCC The paper focuses on the building’s Foreword both the fundamental thinking behind the roof structure, which is completely The culture of digital parametric design process of digital design and the technical free-form, architecturally defi ned in the building industry is evolving rapidly. skills required, such as writing code by a 3D doubly-curved NURBS Originally pioneered by architects who (scripting), to undertake their own project surface. The roof’s surface area wanted to build complex curving forms, studies. This paper presents a project that digital parametric design is now being was designed on the cusp of this new digital 2 is approximately 13 500m and applied to more mainstream projects. While age, and helped spur on the development of is constructed from curved plate parametric design is not a new paradigm in a parametric design programme at Arup. girders, each one unique. The roof’s engineering or architecture – an example geometry ranges from areas with a from over a hundred years ago being the Introduction radius of curvature of approximately form-fi nding work of the architect Antoni The Qatar Faculty of Islamic Studies (QFIS) Gaudí for projects such as la Sagrada is located on the outskirts of Doha and is 500m, therefore relatively fl at, to Família in Barcelona – digital tools available part of the Education City development. The areas of high curvature with a radius today have accelerated its potential building comprises teaching and research of 5m or less. exponentially. facilities, with a fl oor area of 35 000m2 over Digital parametric design, or just four fl oors. Figure 1 shows a bird’s-eye view parametric design as it will be described of the completed building, while Figure 2 from this point on, is a process whereby the shows a close-up of the mosque area roof, computer takes on part of the design role, the mosque having a clear internal span of following a procedure (algorithm) developed approximately 50m. Design work began on by the designer. The computer makes the project in 2008 and the building was decisions based on the rules and parameters opened by Her Highness Sheikha Mozah bint created by the designer to perform a task. Nasser in March 2015. The building received Often these tasks are highly repetitive and the Religion award at the World Architecture time-consuming to perform manually. Festival 2015. At Arup, we are fostering a culture of While the project posed many engineering digital exploration, teaching our engineers challenges – such as its two minarets (one

TSE51_69-76 Project Focus Qatar v1.indd 69 24/02/2016 14:01 ›

70 TheStructuralEngineer Project focus March 2016 Qatar Faculty of Islamic Studies

Figure 2 Mosque area roof QF/ASTAD/CCC

beneath. Models were regularly shared with the architects, who used the same software, allowing our structure to be easily visualised by them and rapidly tested in their architectural models. Apart from visualisation purposes, our Rhino model also enabled us to easily create the faceted centreline geometry required for our structural analysis software. The modelling of the structural geometry quickly developed into a parametric process using the Rhino plug-in Grasshopper2*. This happened when it became apparent that there were a number of parameters driving the design of the structure, most of which were yet to be conﬁ rmed by others in the design team. These included the fundamental off sets for the roof cladding depth and the facade width. We wanted to

ARUP be able to progress our detailed modelling studies but to retain the fl exibility to adapt and update the model quickly and easily Figure 3 Curvature analysis of architect’s design surface, complex pattern formed of a tessellation when these parameters were fi xed. with contours at 200mm vertical intervals of three triangular panels, the largest of The roof geometry was given to us by the which has a side length of 1.8m. The panels architect as a 3D doubly-curved NURBS are connected via a secondary framework (non-uniform rational basis spline) surface 80m and the other 90m tall) and the post- of aluminium rails, with the overall cladding in Rhino, which represented the outside tensioned mosque slab, which cantilevered build-up being 565mm. The span of the face of the cladding and was referred to almost 15m – this paper focuses on the primary plate girders ranges from under 10m as the “design surface”. The geometry design of the geometrically complex roof to just over 50m, with a typical span being can be defi ned as free-form, having no structure. around 20m. The purlins span approximately simple mathematical defi nition. Initially, we The roof has a completely free-form 6m and are set at about 3m centres to considered working with the architect to geometry with a developed surface area of support a 100mm deep profi led roof decking. rebuild the free-form geometry into surface approximately 13 500m2, this being about The purlin spacing and decking type was patches with rational geometry, e.g. spherical 12 250m2 in plan. The structure comprises chosen to allow the contractor the possibility or conical. After consideration, we decided singly-curved primary plate girders with of self-curving the decking to the required that this would be time-consuming and straight purlins and a profi led metal decking. profi le on site. ultimately unnecessary as we believed we The cladding is a rain-screen system, could absorb the complexity of the surface comprising open-jointed, triangular, glass- Geometry geometry in the geometrical defi nition of the reinforced concrete (GRC) panels fi xed to a From the outset, structural concepts and structure and cladding. conventional standing seam build-up. options were modelled in three dimensions To understand the architect’s geometry, The panels are typically equilateral (3D) using the popular software Rhinoceros1 we initially carried out a number of studies triangles with an edge length of 1.3m, (Rhino). This was the case not just for the in Rhino. This ranged from cutting cross- although the more visible areas have a roof structure but also the building structure sections at regular intervals through the

*Grasshopper is a graphical-based programming language which enables parametric and associative models to be created. It is characterised by components that are wired together to perform a variety of geometrical operations. The wires represent the ﬂ ow of data from one function to another. Due to its graphical nature, it is intuitive and easy to learn, with no programming skills required.

TSE51_69-76 Project Focus Qatar v1.indd 70 24/02/2016 12:10 www.thestructuralengineer.org 71 ARUP

Figure 4 WDiagram illustrating process of curve-to-arc algorithm

design surface, to creating contour lines on the surface. Both of these techniques were used to understand how the surface sloped and in which direction. More sophisticated curvature analysis was also undertaken to understand the radius of curvature of the surface and how it varied, as this had the potential to infl uence the spacing of the steelwork, the specifi cation of the roof decking, and possibly the manufacturing process of any curved structure (limitations of rolling). Figure 3 shows the design surface with a colour contour render illustrating surface curvature, and contour lines at 200mm vertical intervals. The curvature Figure 5 analysis showed that the surface had Feedback from Grasshopper parametric model curvature ranging from approximately 500m (red areas), therefore virtually fl at, to approximately 5m (blue areas), therefore highly curved. To put this into context, considering a typical span of 20m, the deviation from straight with a 500m radius would be 200mm. Moving to Grasshopper also enabled us to take full advantage of Arup’s plug- in, Salamander. Salamander has a set of components for directly exporting a parametric model to the analysis software used at Arup, Oasys GSA3. This streamlined our workfl ow even further and allowed us to progress our structural analysis with the ability to make geometric changes with signifi cantly less impact on our progress.

Optimisation Initially, the geometry of the curving plate girders was generated in the parametric

ARUP model by intersecting vertical planes with the design surface. This resulted in 142 ϭϮϬϬ unique members. Due to the surface being Figure 6 free-form, the resulting intersection curves Number of output arcs plotted against deviation from were also free-form, meaning they had ϭϬϬϬ original curve (blue line) and input curves (142; red line) constantly varying curvature and could not be deﬁ ned with simple mathematical rules. ϴϬϬ At the beginning of the detailed design phase of the project, we met with a number of specialist steelwork fabricators to get ϲϬϬ feedback on cost and technical issues. While most had the technology to fabricate such a ϰϬϬ geometrically complex structure, there was

EƵŵďĞƌŽĨƌĐƐKƵƚƉƵƚ a consensus that if we could simplify the geometry there would be a signiﬁ cant cost ϮϬϬ and programme beneﬁ t. We therefore decided to embark on Ϭ a rationalisation study, developing the Ϭ ϮϬ ϰϬ ϲϬ ϴϬ ϭϬϬ ϭϮϬ parametric model we had already begun in ĞǀŝĂƚŝŽŶ;ŵŵͿ order to facilitate this new process. The idea

ARUP was to rebuild the free-form curves into arcs

TSE51_69-76 Project Focus Qatar v1.indd 71 24/02/2016 12:10 ›

72 TheStructuralEngineer Project focus March 2016 Qatar Faculty of Islamic Studies

and compound arcs, taking advantage of the rain-screen cladding system to absorb the geometrical deviation such a process would result in. Therefore, the design surface specifi ed by the architect would, in principle, remain unaltered and the cladding zone would become a variable. While Grasshopper is a fl exible parametric modelling tool, it would not have been possible to defi ne a set of its standard components that could perform the iterative process required to rebuild the curves. Therefore, we used a Grasshopper component which allowed the user to write bespoke code (script) within it, and which could be incorporated easily into the rest of the parametric model. The algorithm would divide the curve into a large number of segments and then begin to draw arcs from the fi rst point to each subsequent point, maintaining tangency to the initial curve at its start point. As each arc was generated,

it would be tested to see if its deviation ARUP from the original curve was within a defi ned Figure 7 limit, the deviation being the principal input Mosque area under construction showing varying curved plate girders parameter for the process. When it found an arc that was outside of the deviation limit, it would stop creating arcs. It would then choose the penultimate arc in the series as the best-fi t geometry to the original curve. The algorithm would then begin the iterative process again and cast another set of arcs through the points on the original curve, this time from the end point of the previous best-fi t arc, maintaining tangency to that arc and not the curve as it had originally. It would repeat the process until it reached the end point of the original curve. Figure 4 illustrates the process, which could be undertaken on all of the curves simultaneously and took a matter of seconds to run and then re-run with a modifi ed deviation parameter. Although the number of segments the initial curve was divided into was not an explicit parameter, we did fi nd that it had an eff ect on the outcome; therefore, a high number was chosen, e.g. 200. This tool gave us the ability not only to rationalise curves into arcs, but also to CCC optimise the resulting geometry based on the allowable deviation. A study was Figure 8 Installation of GRC cladding panels undertaken where the allowable deviation was increased from 10mm to 100mm in order to see the resulting number of arcs were able to agree with the facade engineer end points. output. The parametric model was able and the architect that a deviation of 100mm With an initial geometrical deviation of to give instant feedback on a number of (±50mm) from the original free-form curve 10mm, the parametric model output 870 geometrical characteristics of the resulting was possible, and that this could be taken up arcs from the 142 input curves, which was geometry, such as the total number of by the cladding brackets. unrealistic from a fabrication point of view. arcs created, the maximum and minimum Our fi nal step was to consider fabrication By increasing the deviation from 10mm to length of the output arcs, and the maximum and erection tolerances, and for members 100mm, the number of arcs was reduced number of arcs any one curve was rebuilt with a rise of less than about 15–20mm to 317, a reduction of almost 60%. Figure 6 into. Figure 5 shows the output from the (approximately length / 1000) the arc was illustrates the number of arcs output against Grasshopper model. Following this study, we further rebuilt to be straight between its two the deviation in millimetres, and the number

TSE51_69-76 Project Focus Qatar v1.indd 72 24/02/2016 12:10 www.thestructuralengineer.org 73 ARUP

Figure 9 Revit model

of input curves. distribution ductwork through the plate on site to ensure the panels were located Of the geometry output from the girders gave a shallower overall depth than in the correct position. Figure 8 shows the parametric model, 67 were defi ned as single passing them below. This had the additional GRC panels being installed in an area with arcs, and 28 of these were subsequently benefi t of signifi cantly reducing the the equilateral triangular panels. rebuilt as straight sections due to their defl ections, due to the beams being slightly large radii. Of the compound arcs output, deeper than required for their span to Delivery the maximum number of arcs in any one accommodate large openings. Nonetheless, The drawings were produced from a 3D section was seven, with the average being it is believed a precamber could have been Revit model (Revit Structure4, version 2.2. Figure 7 shows the mosque area – the included in the parametric process. The 2008), with the geometry of the roof most curved location on the roof – under plate girders were also sized to avoid the steelwork exported from the parametric construction. requirements for web stiff eners around Rhino model as simple AutoCAD line work openings, stiff eners increasing fabrication (.dwg or .dxf format). This stage broke Design complexity and therefore cost. Typically, the link in the parametric workfl ow and The cladding system is a fairly heavyweight plate girders were 1.0–1.2m deep with up to modelling the beams in Revit was time- build-up, with the GRC panels being 600mm deep openings. consuming, even with the geometry defi ned. approximately 15mm thick. Therefore, the The geometry of the cladding panels was This meant that careful consideration had dead load defl ections were signifi cant. based on the design surface. Construction to be given to how “frozen” the architectural Thankfully, the plate girders were not tolerances and deformations of the geometry was before undertaking this required to be precambered, which structure were considered by the facade stage of the process. Revit was chosen might have ruined our plan for a simple engineer when specifying the joint between as the best all-round software available at geometrical defi nition. panels, which was specifi ed as 40mm with the time for modelling steel and concrete This was due to the structure and a tolerance of ±10mm. Furthermore, the structures, and for its ease with drawing services zones being combined. Following secondary framework the panels are fi xed production. Figure 9 shows the complete a study with our building services too allowed for some adjustment, both Revit model of the building. It is also worth counterparts at Arup, running the air vertically and laterally, by the contractor noting that the building services engineers

TSE51_69-76 Project Focus Qatar v1.indd 73 24/02/2016 12:10 ›

74 TheStructuralEngineer Project focus March 2016 Qatar Faculty of Islamic Studies

Figure 10 Illustration of data export from Rhino to Revit

modelled their primary ductwork and pipework in Revit, and the two models were combined for coordination purposes. We had numerous discussions with the client about how we would produce the tender information for the roof and whether we would produce drawings of all of the unique girders, as the client was not comfortable with just issuing the 3D model for tender. We convinced the client that we would produce a drawing describing the typical geometrical characteristics of the curving girders and then produce a schedule describing the geometry of each a) Roof model in Rhino

individual member. ARUP / QF When we looked into Revit’s scheduling functionality, we found it was lacking and b) Member schedule script infl exible. In addition, the fact that some of the members were made from more than one arc section compounded this problem. The solution was to go back to our Rhino model and extract the data directly from it. The data would then be imported into Revit and placed on a drawing sheet. While this could have been done manually, there were quite a few girders and quite a lot of data was required; therefore, this process would have been time-consuming and prone to error. In response, we developed a script (using the RhinoScript language, based on the Microsoft VBScript language) whereby it was possible to select all of the beam centrelines at once – these being the same centrelines originally exported to Revit, the script would then analyse each arc and write all of the geometrical information directly to an Excel spreadsheet. All that was required next was to format the spreadsheet, export it as a .dxf fi le and import it into Revit on a drawing sheet. Although writing the script took a few days of research and development (and some late nights), it saved a signifi cant amount of time and reduced the number of drawings required to describe the roof’s geometry. It also paid off when we updated the geometry and the exercise of documenting it needed to be done all over again. We produced two schedules: a member schedule and an assembly schedule. Each required a diff erent script. The member schedule contained the information regarding the overall girder, e.g. whether it was straight or curved; whether it was made from a single arc or multiple arcs; if it was made from multiple arcs, how many;

its curved length and straight-line length ARUP / QF

TSE51_69-76 Project Focus Qatar v1.indd 74 24/02/2016 12:11 www.thestructuralengineer.org 75 QF / ARUP / QF

c) Scripted member schedule in Excel

between end points; and the 3D XYZ them in the parametric model. Figure 10 areas of the building to route additional coordinates of the two end points. shows the scheduling process from Rhino, pipes under the girders. The assembly schedule took the beam illustrating part of the script and the script This was all possible because we references from the member schedule and output to Excel. maintained a strict modelling protocol, added a suffi x for each of the arcs that a A similar exercise was also carried out whereby there was a single master Rhino particular girder was made from. It output on all of the inclined columns with another model. This model contained the input the following: radii, arc length, chord length, script that could analysis them and geometry for the Grasshopper parametric and sector angle. It also gave a direction for output a schedule of: length, inclination, model and the output geometry, and was the section, to represent whether the arc orientation, and top and bottom 3D XYZ used to export geometry and data to both was to face upwards or downwards in the setting-out points. Bespoke scripts were Revit and GSA. Figure 11 shows a fl ow completed member. also written to undertake other repetitive diagram illustrating our modelling workfl ow, All 142 members were recorded on a tasks, such as putting beam tags on starting from the architect’s design surface single A1 drawing, with the 289 arcs on drawings to represent their connection and ending in the drawing package. The another drawing. This documentation types based on data output from the GSA blue rectangles represent the input and process was not directly linked to the structural analysis model, or measuring outputs from the parametric model, the geometrical defi nition or optimisation the distance from the underside of the orange rectangles represent the parametric process; the scripts were standalone girders to the ceiling. The latter was to see modelling design processes, and the green tools as it was felt unnecessary to include whether there was enough room in some rectangles represent the fi nal outputs.

TSE51_69-76 Project Focus Qatar v1.indd 75 24/02/2016 12:11 ›

76 TheStructuralEngineer Project focus March 2016 Qatar Faculty of Islamic Studies

Model Export Structural Analysis (Salamander) (GSA)

Design Surface Centreline Geometry (Rhino) (Rhino/GH)

Geometry Export (.dwg)

Drawing Production (Revit)

Geometrical Data Data Format (RhinoScript) (Excel)

Figure 11 Workﬂ ow diagram ARUP

Conclusion Without the tools presented in this paper, should be considered at the beginning Project team I believe we would not have achieved of each project. Consideration should Client: Qatar Foundation such a refi ned and optimal solution for be given to the particular aspects of the Project management: ASTAD the QFIS roof structure. To rebuild all of building you are designing and the people Architecture: Mangera Yvars Architects / the 142 curves into arcs manually would you are collaborating with, to achieve the RHWL have taken a signifi cant amount of time most worthwhile outcome for yourself and Lead consultant – structural engineering, to do just once, let alone multiple times to your client. building services (MEP), facade engineering: optimise the outcome. If we had attempted Arup to undertake this exercise manually, we Discussion Contractor: JV Consolidated Contractor might have undertaken a study on a few It should be acknowledged that there have Group and Teyseer Contracting Company curves and applied what we had learnt been some signifi cant advancements in Steel fabricator: Eversendai Engineering on these to the rest. Alternatively, we the software mentioned in this article since might have left the contractor to fabricate the design work for QFIS was undertaken the complex free-from members initially in 2009–10. These include a plethora generated from the parametric model. of plug-ins for Grasshopper and Revit References Furthermore, we saw signifi cant benefi ts (e.g. Geometry Gym and Dynamo) which in our delivery process using these tools go some way to addressing issues of and techniques. We reduced the number interoperability and geometry generation E1 Robert McNeel & Associates of drawings required to describe the roof that we found challenging. (2016) Rhinoceros 3D (Rhino) geometry, as well as reducing the potential However, it should also be [Online] Available at: www. for errors from manually transcribing the acknowledged that while there are many rhino3d.com (Accessed: setting-out information from the model to pre-packaged tools now available, there February 2016) the drawings. In conclusion, I believe we will always be situations where what is achieved savings for the client on project needed cannot be done with these tools. E2 Robert McNeel & Associates construction cost and savings in our This is not to say that it cannot be done. (2016) Grasshopper project delivery costs. It is therefore my belief, particularly when [Online] Available at: Before embarking on developing a dealing with more complex projects, that www.grasshopper3d.com parametric model, I believe it is important a broader understanding of geometry (Accessed: February 2016) to think carefully and decide what you are and modelling techniques is required so trying to achieve and what parameters that bespoke tools can be developed if E3 Oasys (2016) GSA [Online] are worthwhile. It can be all too easy to required. In summary, what one might call Available at: www.oasys- generate an extremely large and complex a “fi rst-principles” approach can always be software.com/products/ model where absolutely everything is used to solve any problem. structural/gsa/ (Accessed: governed by parameters of some sort or February 2016) another, or even a single parameter. This Acknowledgements type of model becomes cumbersome and As with all projects of this size and E4 Autodesk (2016) Revit [Online] slow to update, ultimately defeating the complexity, the design and delivery was Available at: www.autodesk. objective of creating it in the fi rst place. a team eff ort, although I would like to co.uk/products/revit-family/ Designing the design process is just acknowledge Toby Clark (a geometry overview (Accessed: February another aspect of structural engineers’ specialist at Arup) for his work writing the 2016) work on more complex buildings, and code for the curve-to-arc algorithm.

TSE51_69-76 Project Focus Qatar v1.indd 76 24/02/2016 12:11 › www.thestructuralengineer.org TheStructuralEngineer 77 March 2016 Education

Articles exploring the education of structural engineering students in an increasingly digital environment. 78 Should students be introduced to analysis and design software?

80 What are the beneﬁ ts of exposing students to structural analysis and design software?

88 Virtual by design

TSE51_77 Education opener.indd 77 24/02/2016 12:11 ›

78 TheStructuralEngineer Education March 2016 Teaching software Should students be introduced to analysis and design software?

Some expertise in the use of technical software, often developed in a ﬁ nal-year project, is a necessary skill for graduates in civil or structural engineering, argues Professor Roger Johnson, although this poses several questions.

Introduction Project work would certainly be included in BEng and MEng degree courses in the UK, even if the Joint Board of Moderators did not require it. The provision of appropriate software for students’ use raises questions of policy and practice. This article is based on the author’s experience at the University of Warwick of setting and teaching ﬁ nal-year projects with structural engineering content; this is assumed to be in Year 3 for BEng degrees and Year 4 for MEng degrees. The article will discuss:

• the extent to which use of software should be Figure 1 encouraged or required ETwo-member space frame for teaching use of • teaching its use and checking its outputs structural analysis program • the choice of software, which is assumed here not to be governed by its cost

It is assumed that students entering their fi nal year are familiar, to whatever level they cannot be based on what they may meet sections etc. – still a valuable skill – and use of need, with programs such as Word, Excel and later. Employers expect a basic competence, output charts from whichever PSA is chosen. PowerPoint. The subject here is programs applicable widely. The teaching of the use of Iain MacLeod’s book Modern Structural that create drawings, analyse structures or software should therefore focus on generic Analysis: Modelling Process and Guidance1 is a do detailed design of members following skills: how to model a structure in a way useful reference. Eurocodes or other standards. suitable for input to a drawing or analysis program; the methods used to defi ne axes, Final-year projects Why use software? loadings, properties of materials, members, The many types of fi nal-year project have Programs for structural analysis (PSAs) have etc.; the interpretation of outputs; and the need diff erent software needs. The types discussed to be used in a way that enhances (rather than for independent checks on them. here are: replaces) understanding of how structures respond to loading. This requires tuition. Drawings • Type 1: open-ended research-type projects Simple hand methods are needed at concept Sketching by hand, almost a lost art, must – often done by individual students, and best stage and for checking, and must be taught be encouraged, but is not suffi cient. For the suited to high fl yers too. Computer analysis is essential for any period before “civil” and “mechanical” degree • Type 2: design projects specifi ed as from realistic design project, for most research courses diverge, it is likely that the software a client – requiring several concepts and an projects, for illustrating three-dimensional chosen will relate better to mechanical outline design, analysis and construction plan, (3D) behaviour of framed structures and for engineering than to civil (rules for projected usually done in teams with 3–6 members, all parametric studies that illuminate behaviour in views, etc.). SolidWorks is an example. We taking the same “civil” degree course. The a way that older methods never could. have not found time in a fi nal year to add to task is often issued as the client’s brief for a So many programs are in use in practice “civil” students’ skills anything beyond hand realistic bridge that the choice of software for students’ use drawings to scale of plans, elevations, cross- • Type 3: projects that are a blend of types 1

TSE51_78-79 Education Johnson 2 v1.indd 78 24/02/2016 12:12 www.thestructuralengineer.org 79

and 2 – done by multidisciplinary teams of 6–8 Some new staff members have experience “borrow” the geometry and loading fi les members, and so needing solutions to many of SAP2000 software3, which will be used created by another member of the team and, engineering aspects of the problem in 2016/17. No one program seems to be in with help, use them. Employers will fi nd that general use by students in higher education. as with mathematics, graduates’ IT profi ciency When the fi rst one or two years are We have not found any entry to the varies widely, and are warned that it can be common across engineering, as at Warwick, fi nite-element world simple enough for impracticable to deduce, from a student’s the knowledge of structural materials and class teaching in the time available. Bridge report, whether software is being used mechanics that MEng students have at the decks are analysed by a combination of correctly. start of their fourth year is inevitably greater computer and hand methods which is found than that of the BEng students starting to develop understanding of their behaviour Conclusion their third year. Also, the expected learning under simplifi ed loadings. Where a student’s In summary, the use of appropriate technical outcomes and levels of achievement of the design would be impracticable at detailing software is an essential component of a two groups are diff erent. The BEng group are level (e.g. because reinforcement would fi nal-year project. However, this raises three usually set projects of type 2. These have a be over-congested or uneconomic), we questions: link with practice that improves motivation, and require examination of that problem by a provide opportunities for structural modelling hand method and sketching to scale, but • How best to develop students’ motivation to and analysis that enhances understanding do not use the “checking”, “detailing” or master software that is far more complex than of the content of courses on materials “resistance of members” modules of any is needed? and structures, which are being taught program. They are too code-oriented and • How to engender suspicion of outputs and concurrently. MEng students do a project can include unstated assumptions that impair how to check them, without any “it looks of type 1 or 3, selected from a list, and an students’ understanding. Students have to wrong” skill? individual project in Year 3. check resistances by hand methods, to gain • How best to teach the use of IT in the very In or before Year 3, all the students need the experience of their use. limited time available, to a class where the same fairly simple PSA as there is no time to prior expertise of students can range from teach the use of more than one; ideally one in Graduates near-zero to beyond that of the teacher? which all of the supervising academics have A feature of work by students’ teams, which competence. This is a problem, because of the are selected to have a range of abilities, is Author biography wide range of PSAs in use in higher education. that many students become profi cient at Roger Johnson Each of us prefers the program that we know providing drawings or pictures by SolidWorks, MA, FREng, FIStructE, FICE best. Students doing MEng projects may and at modelling a proposed structure for Roger’s career began with three also need more specialised software; usually analysis, sometimes with 100 or more nodes years as a site engineer, followed programs suggested and taught informally by and members. Others struggle bravely. Some by design work with Arup. Since then he has researched and their supervisor. MEng students master unfamiliar software taught structural engineering, with little help. Checking and interpretation fi rst at Cambridge and now at Choice of software for structural of output are less good. A few students try Warwick. His work on Eurocodes, begun in 1971, analysis and design to avoid the subject altogether, and then continues. He is a Gold Medallist of the Institution. No PSA aimed at mid-course undergraduates has been found; programs are written for the much larger professional market. Their complexity makes it diffi cult to teach the few References per cent of capability needed for a fi nal-year project. We regularly review commercial E 1 MacLeod I.A. (2005) Modern Structural Analysis: Modelling Process and programs, asking in each case how much Guidance, London, UK: Thomas Telford teaching, trialling and soaking-in time would be 2 ATIR Engineering Software Development (2015) STRAP [Online] Available needed to enable a student at second-class- E at: www.atirsoft.com/webportal/AboutStrap (Accessed: January 2016) honours level to analyse with understanding the 3D staircase structure shown in Figure 3 CSI (2016) SAP2000 (v18) [Online] Available at: www.csiamerica.com/ 1. We aim to achieve this in just two lectures E products/sap2000 (Accessed: February 2016) in Year 3, followed by practice and (for BEng students) tutorials. We currently use STRAP2.

TSE51_78-79 Education Johnson 2 v1.indd 79 24/02/2016 12:12 ›

80 TheStructuralEngineer Education March 2016 Teaching software What are the beneﬁ ts of exposing students to structural analysis and design software?

There is a strong case for exposing undergraduate students to structural analysis software, argues Jon Carr of The University of Sheffi eld. While the case for structural design software is less compelling, this also off ers potential learning beneﬁ ts.

Introduction In the period since the author graduated Figure 1 from university and started working as a Typical multistorey building analysis model consulting engineer in 1988 (when there was just one computer in the entire offi ce, which you had to book a slot for), the advent of powerful computer hardware and software has meant that the way in which structures are analysed and designed in practice has changed signiﬁ cantly. However, anecdotal evidence suggests that the way in which the analysis and design of structures is taught at universities has, in general, not kept pace with the “real world”. Some academics may argue that exposing students to computer software is tantamount to training (as opposed to educating) them, and therefore has no part in a formal engineering education. Unlike pure science, however, structural engineering is about the application of appropriate knowledge and skills in order to solve real-world problems. Hence, it is essential that universities prepare students for their future careers by equipping them with a broad range of relevant knowledge and skills. While the intention should not be to train students to use any particular piece of software, the ability to accurately and effi ciently model, analyse and design real structures, using TUOS / BROWNBILL TOM appropriate tools, is an essential skill for today’s structural engineer. And, as when Table 1: Continuous beam design coeffi cients to EC24 learning a language, it is something which At outer Near middle At ﬁ rst interior At middle of At interior the author believes is better learnt sooner support of end span support interior spans supports rather than later. Furthermore, if used eff ectively, structural Moment 0 0.09Fl –0.11Fl 0.07Fl –0.10Fl analysis software can reinforce prior learning Shear 0.45F – 0.60F – 0.55F of traditional structural analysis methods, Notes

develop conceptual understanding of how F is the total design ultimate load (1.35Gk + 1.5Qk) for each span, and l is the length of the span. a diverse range of structures behave, and No redistribution of the moments calculated from this table should be made. promote creativity.

TSE51_80-87 EducationCarr 2 v1.indd 80 24/02/2016 12:16 www.thestructuralengineer.org 81

This paper looks at the current use of structural analysis software at The University Figure 2 of Sheffi eld (TUoS) in order to: Typical pedestrian footbridge analysis model

• demonstrate the beneﬁ ts of exposing undergraduate students to analysis software • consider the challenges of exposing undergraduate students to analysis software • present general guidance on “good practice” for building analysis models (see Appendix A for more detailed advice)

Finally, the paper considers the potential for more ambitious uses of analysis software in the university curriculum.

Current use of structural analysis software

In the true spirit of the “Constructive TUOS / KONG YU KA Alignment” approach to learning and teaching, which essentially advocates that the most Figure 3 eff ective way to learn how to do something is Typical long-span roof analysis model to actually do it1, the use of analysis software is typically embedded into design projects at TUoS, most notably the “Integrated Design Project” (IDP). The IDP, which received a commendation in The Institution of Structural Engineers’ Excellence in Structural Engineering Education awards in 2014, is a full-time module lasting 15 weeks, taken by third-year undergraduates. It essentially comprises a ﬁ ve-week concept design stage, during which only hand calculations are permitted, followed by a more detailed eight-week scheme design stage, during which computer analysis and design

are introduced, with appropriate guidance on TUOS / SCOTT MATTHEW validation and verifi cation provided. At the start of the scheme design stage, students attend a three-hour workshop in a structural analysis models. in the case of a simply designed braced steel computer lab, in which they are introduced to Always remember the following pieces frame (i.e. where joints are assumed not to the Oasys GSA structural analysis package by of wisdom: “Structural engineering is the develop moments adversely aff ecting either a range of university teachers and graduate art of moulding materials we do not wholly the members or the structure as a whole), teaching assistants (GTAs) who have recent understand, into shapes we cannot precisely beams and columns can generally be designed experience of the software, as well as analyse, so as to withstand forces we as individual elements, with any bracing frames engineers from the local Arup offi ce. Guidance cannot really assess, in such a way that the and shear walls/cores designed separately. on good practice in building analysis models community at large has no reason to suspect Avoid “fl oors” in order to avoid “fl aws”: is also covered in this session, as described the extent of our ignorance”2. Put another way, Floor slabs can signifi cantly complicate in Appendix A. Other sources of guidance it is better to be vaguely right, than exactly analysis models for little if any benefi t, and can include user manuals and video tutorials. wrong3. often be excluded, with fl oor loads applied Several subsequent workshops are held, in 3D or not 3D, that is the question: To to supporting beams as required. There which students can go through their models determine how refi ned their analysis model are exceptions to this, such as reinforced- on a one-to-one basis with university staff and needs to be, students need to consider concrete fl at slabs (commonly analysed practising engineers, and with their peers. For the nature of the output they require, how using fi nite-element analysis methods) and examples of the models produced in just a few accurate it needs to be, and how sensitive it composite steel beams. However, even these weeks, including multistorey buildings, large- is to changes in input parameters. While the can be modelled as a series of 2D beam-and- span roof structures and pedestrian bridges, natural tendency for students is to build three- column frames in the case of fl at slabs, and see Figures 1–3. dimensional (3D) models containing every as isolated members in the case of composite structural element (e.g. as shown in Fig. 1) this beams. “Good practice” for building structural is often unnecessary. Indeed, for reasonably Carry out a sensitivity analysis, particularly analysis models regular/repetitive framed structures, a series of on complex, non-standard, long-span or large- As part of the initial workshop described 2D frames may suffi ce. In some cases, a series deformation structures, or where the reliability above, students are given the following of sub-frames may be perfectly adequate (e.g. of key parameters/data is uncertain. One general “good practice” guidance on creating using the tabulated coeffi cients in Table 1) or, example of this would be to see how varying

TSE51_80-87 EducationCarr 2 v1.indd 81 24/02/2016 12:16 ›

82 TheStructuralEngineer Education March 2016 Teaching software

Figure 4 SBending moment diagram

a) Using same section for columns and rafter the cable diameter and Young’s modulus aff ects the behaviour of cable-stayed bridges. Similarly, the sensitivity of structures on varying ground conditions could be assessed by using a range of spring stiff nesses to model the foundations. Carry out parametric studies to understand how changing member properties aff ects continuous/rigidly jointed frame behaviour. For example, it can be seen from Figure 4 how changing the relative member stiff nesses of a portal frame aff ects the bending moments. While only the second moment of area has been changed in this example, changing materials (i.e. Young’s modulus, E) and member lengths (L) would also aff ect frame behaviour. Computer software only does what you tell it to do, so any errors are almost certainly due to the user, as opposed to the software. Hence, it is essential to validate input data and b) Using larger rafter verify output (see “Useful resources” section). Adopt the following general process when developing a model:

• Think before you start. • Don’t over complicate models. • Build models systematically/iteratively. • Carry out simple checks throughout.

Benefi ts of exposing students to structural analysis software The use of analysis software at TUoS has resulted in signifi cant learning benefi ts, not all originally envisaged. Parametric studies are used to show how varying member properties aff ects the behaviour of rigidly jointed frames. Specifi cally, students can quickly vary section sizes, member lengths and materials (i.e. changing member stiff nesses, which are proportional to EI/L) and see how this aff ects c) Using larger column bending moment diagrams and defl ected forms, for example. This can be a very eff ective way of consolidating prior theoretical learning in this area. Furthermore, rather than being detrimental to student learning in the area of structural analysis, the use of analysis software encourages students to dig out their lecture notes and apply appropriate hand analysis methods (e.g. moment redistribution and the method of joints) in order to carry out preliminary analysis, and check computer output. Students develop an enhanced understanding of how member end releases aff ect overall frame behaviour and individual member forces. While students generally have a good understanding of how diff erent support conditions aff ect the behaviour of

TSE51_80-87 EducationCarr 2 v1.indd 82 24/02/2016 12:16 www.thestructuralengineer.org 83

a structural frame, not many have a clear address is students becoming over-reliant on, packages, so increased hardware costs understanding of how member end releases and over-trusting of, output produced by a should not be incurred. aff ect overall frame behaviour, or the forces computer (the “black box”). To help address present in individual members. In 2D frames, this issue, before students are taught the Staff expertise for example, they often fail to appreciate that stiff ness method, they are provided with Many students pick up analysis software members can only be subject to in-plane a MATLAB script which implements this relatively quickly and are often quite ambitious forces (specifi cally major axis bending, major method, which they are required to modify in terms of the models they build (e.g. lamella axis shear and axial force), but that out-of- (MATLAB is a high-level language and gridshells) and the types of analysis they run plane buckling still needs considering. interactive environment used by millions of (e.g. dynamic analyses of footbridges). Hence, Students develop a feel/intuition for scientists and engineers worldwide)6. it is helpful to employ suitably experienced how structures behave conceptually/ It is also essential that students have a academics and/or practising engineers when qualitatively. As noted by Tristram Carfrae in clear conceptual understanding of how the using analysis software. his Institution Gold Medal address in 20155, structure should behave before producing analysis software can be a very eff ective tool a computer analysis model, so that they Staff time for learning about structural behaviour. can check the resultant bending moment, Helping students to resolve modelling Students develop an enhanced shear force and defl ected form diagrams. problems can be time-consuming for staff , understanding of what diff erent types of Further, students should also carry out hand especially when students have not followed support and member end connections calculations to determine critical forces and “good practice” or have been particularly look like in reality, and how they should serviceability conditions in the structure, ambitious. The use of GTAs, as well as be designed and detailed. While students from which preliminary member sizes can be encouraging peer support, can reduce this generally have a good grasp of the theoretical determined, prior to producing a computer potential burden. concept of pinned, fi xed and semi-rigid joints, model. they do not always appreciate the practical Useful resources implications of the fundamental assumptions Resolving diff erences between hand Useful resources for checking input and they make, in terms of the design and detailing calculations and computer output output from structural analysis software (and of joints, as well as how their assumptions It is not uncommon for students to fi nd for conceptual structural design) include: aff ect foundation designs. For example, many signifi cant diff erences between computer students are surprised to learn that partial output and the results from their preliminary • student lecture notes and even full-depth end-plates to steel beams hand analysis, particularly for complex • Understanding Structural Analysis7 can be classed as nominally pinned. Similarly, structures. While there may be a number • Modern Structural Analysis8 many students are surprised at the increased of reasons for such diff erences, the time • Seeing and Touching Structural Concepts9 foundation sizes required for columns with required to identify the causes of the • Structural Engineering: Art and fi xed bases, which are generally avoided diff erences, and to subsequently rectify the Approximation10 where possible in industry. issue, should not be under-estimated. Typical • Structural Engineers’ Pocket Book11 It encourages, and facilitates, greater reasons include: • Rigid frame formulas12 creativity. While this seems counterintuitive, • the Institution’s “Technical Guidance Notes” experience suggests that students use • errors in hand calculations (published in The Structural Engineer between analysis software to explore a more diverse • errors in the computer model (e.g. not January 2012 and June 2014) range of options, in terms of both structural applying the correct loads) • the Institution’s design guides to Eurocode 24, form and materials, and, again as noted • limitations of hand calculation methods Eurocode 313 and Eurocode 514 by Tristram Carfrae, to “locate structural (particularly for complex structures) • the Institution’s Structural Behaviour Course15 opportunities”5. Indeed, the range of structures • failure to consider key structural eff ects/ • Expedition Workshed website16 which can practically be analysed by hand behaviour (a classic example of which is • “The basics of bending moment and shear is relatively limited, so computer analysis cable-stayed bridges which, when analysed force diagrams” video17 arguably results in greater creativity. While using computer software, generally defl ect • online tutorials produced by software users it remains essential that students carry out signifi cantly more than students expect, due and developers (e.g. Oasys)18 checks on computer output, it is easier to to cable elongation not being considered) check the general forms of bending moment, • inappropriate modelling of supports and/or Potential for more ambitious use of shear force and defl ection diagrams, and to member releases in the computer model (it analysis software in curriculum calculate their maximum values, than to do a is not uncommon for students to confuse the While time constraints mean that analyses full structural analysis by hand. two, and provide supports at every node) carried out for the IDP are typically limited to fi rst-order elastic analyses of structures Challenges of exposing students to Software and hardware costs subject to static loads, opportunities exist structural analysis software Depending on the analysis package to cover second-order, plastic, dynamic While there are inevitably challenges involved used, there may be capital and/or annual and fi nite-element analyses, subject to the with exposing students to analysis software, maintenance costs. However, several capabilities of the analysis package used. This these are considered to be signifi cantly software providers make their products would clearly require additional instruction and outweighed by the benefi ts already described. available either free, or at a signifi cant support over and above that currently provided discount, for educational use. It is likely that (expertise which may not be available “in “Black box” syndrome the computers already available in universities house”), and would need to be accommodated Probably the most signifi cant issue to will be adequate for running most analysis within an already packed curriculum.

TSE51_80-87 EducationCarr 2 v1.indd 83 24/02/2016 12:17 ›

84 TheStructuralEngineer Education March 2016 Teaching software

Figure 5 Clearly, if more ambitious uses of analysis WUse of second- order analysis to software are to be adopted, these should be explain P-delta eff ect and EHFs considered in conjunction with the relevant theory and codifi ed guidance. For example, getting students to carry out second-order analyses could be used to reinforce the theory underpinning P-delta eff ects, and to explain the use of equivalent horizontal forces (EHFs) as specifi ed in EC219 and EC320. Indeed, it would be an interesting exercise for students to run a series of fi rst- and second-order elastic analyses of geometrically perfect and imperfect frames, subject to various combinations of vertical and horizontal loads (including EHFs), in order to appreciate the importance of second-order eff ects (Figure 5).

Structural design software While a strong case has already been made for exposing undergraduates to analysis software, consideration should also be given to exposing them to structural design software, although the case for this is perhaps less strong. Proprietary design software is used extensively in industry, and it is clearly important for students to understand the limitations of such software and the importance of inputting appropriate Figure 6 WSample output data. However, it can be argued that the use of from MATLAB software such software does not require the same level of skill as does building an analysis model. Further, there appears to be limited learning beneﬁ t associated with using some design software, the danger being that students will adopt a “plug and chug” approach, blindly following a “recipe book” procedure without understanding the implications of their actions. Such students would subsequently struggle to develop solutions to problems not covered by a pro forma. Perhaps a better approach for developing a deeper understanding of structural design, and the design process, would be to encourage the use of software such as Excel and MATLAB, through which students could develop their own

Figure 7 tools (e.g. pre- and post-processors, optimisation SSample routines/algorithms) to solve a diverse range of output from Excel spreadsheet structural problems (Figures 6 and 7).

TSE51_80-87 EducationCarr 2 v1.indd 84 24/02/2016 12:17 www.thestructuralengineer.org 85

Summary less compelling, the use of software such Author biography It is apparent that students can develop as Excel and MATLAB can help students Jonathan Carr signiﬁ cant skills in creating and analysing develop a deeper understanding of BEng, MPhil, CEng, MIStructE, FHEA models using structural analysis software structural design, and the design process. Jon worked for Anthony Hunt Associates in just a few weeks. Further, if used from 1988–2010, specialising in sports eff ectively, analysis software can reinforce Acknowledgements and educational facilities. Having taught multistorey building design at TUoS on a prior learning of traditional analysis This paper has been produced in part-time basis since 2007, he formally methods, develop students’ conceptual consultation with Professor Ian Burgess, joined as a University Teacher in 2011, and understanding of how a diverse range of Professor Buick Davison, Professor was instrumental in setting up the IDP. structures behave, and promote greater Matthew Gilbert, Richard Harpin and Paul Jon also runs Jon Carr Structural Design creativity. Hulbert from The University of Sheffi eld as a sole practitioner. He sits on the While the case for exposing students Department of Civil and Structural Institution’s Education Committee and is Yorkshire Regional Group Chair for 2016. to proprietary structural design software is Engineering.

References

1 Biggs J. and Tang C. (2011) Teaching for quality E E 11 Cobb F (2015) Structural Engineer’s Pocket Book (3rd learning at university (4th ed.), Maidenhead, UK: ed.), Boca Raton, USA: CRC Press McGraw-Hill Education E 12 Kleinlogel A. (1952) Rigid Frame Formulas, New York, E 2 Dykes A.R. (1978) ‘Bees in his bonnet’ [Verulam], The USA: Frederick Ungar Structural Engineer, 56A (5), pp. 150–151 E 13 The Institution of Structural Engineers (2010) Manual E 3 Read C. (1898) Logic, deductive and inductive, London, for the design of steel building structures to Eurocode UK: Grant Richards, p. 351 3, London, UK: IStructE Ltd

E 4 The Institution of Structural Engineers (2006) Manual E 14 The Institution of Structural Engineers (2007) Manual for the design of concrete building structures to for the design of timber building structures to Eurocode 2, London, UK: IStructE Ltd Eurocode 5, London, UK: IStructE Ltd and TRADA Technology Ltd E 5 Carfrae T. (2015) Institution of Structural Engineers Gold Medal Address: Designing with Computers E 15 The Institution of Structural Engineers (2015) [Online] Available at: http://istructe.hosted.panopto. Structural Behaviour Course [Online] Available at: com/Panopto/Pages/Viewer.aspx?id=e214dba0- www.istructe.org/resources-centre/structural- ac69-4523-ba3b-f4d1f0f34b2a (Accessed: December behaviour (Accessed: October 2015) 2015) E 16 Expedition Workshed (2015) Expedition Workshed E 6 MathWorks (2015) MATLAB [Online] Available at: [Online] Available at: http://expeditionworkshed.org/ http://uk.mathworks.com/products/matlab/ (Accessed: October 2015) (Accessed: October 2015) E 17 Ibell T. (2012) The basics of bending moment and E 7 Brohn D. (2005) Understanding Structural Analysis shear force diagrams [Online] Available at: www. (3rd ed.), Kingsbridge, UK: New Paradigm Solutions youtube.com/watch?v=Hd8w_7s_78A (Accessed at: October 2015) E 8 MacLeod I.A. (2005) Modern Structural Analysis – Modelling Process and Guidance, London, UK: Thomas E 18 Oasys (2015) GSA Video Tutorials [Online] Available Telford Ltd at: www.oasys-software.com/gsa-tutorials.html (Accessed: October 2015) E 9 Tianjian J. and Bell A. (2006) Seeing and Touching Structural Concepts [Online] Available at: www.mace. E 19 British Standards Institution (2004) BS EN 1992-1- manchester.ac.uk/project/teaching/civil/ 1:2004 Eurocode 2. Design of concrete structures. structuralconcepts/home.htm (Accessed October General rules and rules for buildings, London, UK: BSI 2015) E 20 British Standards Institution (2005) BS EN 1993-1- E 10 Morrison H. (2014) Structural Engineering: Art and 1:2005 Eurocode 3. Design of steel structures. Approximation, London, UK: Paragon Publishing General rules and rules for buildings, London, UK: BSI

TSE51_80-87 EducationCarr 2 v1.indd 85 24/02/2016 12:17 ›

86 TheStructuralEngineer Education March 2016 Teaching software

Appendix A. Building a structural analysis computer model: a quick reference guide for students

This reference guide, which should be read in conjunction with the accompanying paper, presents a step-by-step Figure A1 SStep-by-step approach (Figure A1) intended to help students build consistent and reliable structural analysis models using approach for building computer software. It should be noted that the approach presented is generic, and should be applicable to most, consistent and reliable structural analysis if not all, commercially available structural analysis software packages. models

• Think about what is required from your model before you start Before you start • Don’t over complicate models

• Note that if you have a multistorey building in which all the ﬂ oors are essentially the same, or very similar, you Input model geometry should ideally just build one ﬂ oor initially, as shown in Figure A2 • Add member properties • Add supports, but no releases

• Apply the self-weight load case and then run the analysis, and check your output • If there are errors, check your input (remember that computer software only does what you tell it to do, so it Apply self-weight is likely that any errors are due to incorrect input / operator error) load case • Check that the bending moment (BM) diagram, shear force (SF) diagram and deﬂ ected form are as expected in principle, and that the maximum values of BM, SF and deﬂ ection are of the right magnitude • Check the reactions, and that the model is in equilibrium

• Release the supports as appropriate (e.g. if pinned or roller supports are being used) – do this in a logical Release fashion, and re-run the analysis at each stage • Note that structures modelled in 2D have three degrees of freedom at each support and at the ends of each supports member, as shown in Figure A3, while structures modelled in 3D have six degrees of freedom at each support and at the ends of each member, as shown in Figure A4

• Add member end releases as appropriate (e.g. for members which are pin-ended as opposed to continuous) Add member end releases and in a logical fashion (e.g. release the primary beams, then the secondary beams, then the columns, then the bracing members etc.) and re-run the analysis at each stage

• At this stage, error messages are often displayed, commonly due to one of the following reasons: a) all members connecting at a node have been released rotationally. However, even in 2D and 3D pin-ended frames and trusses, at least one member needs to be rotationally restrained at every node. Note that this will Debug any not generate any moments, since moments cannot be transferred from the member with rotational restraint errors to the other, pin-ended members at the node b) members in 3D frames cannot have torsional releases at both ends, otherwise they will "spin" (similar to a wall-mounted toilet roll holder)

Add other load cases • Now add the rest of the loads, and re-run the analysis at each stage

Add load combinations • Now add the rest of the load combinations, and re-run the analysis at each stage

• Make a global assessment of dead, live and wind loads (as a minimum), and compare this to the total reaction 2 Carry out "sense checks" • For example, if you have a single-storey building which is approximately 1000m on plan and the unfactored dead and live loads are equal to 0.3kN/m2 and 0.6kN/m2 respectively, then you would expect the total support reaction due to dead and live loads to be 900kN (or very close to that value)

Add rest of frames/ﬂ oors • If all is OK, add the remaining ﬂ oors or frames, as shown in Figure A5, and re-run the analysis at each stage

• Do not ignore errors and warnings. These are often singularities in the stiff ness matrix, and they do matter Warnings & errors • If in doubt, ask for help!

TSE51_80-87 EducationCarr 2 v1.indd 86 24/02/2016 12:17 www.thestructuralengineer.org 87

Figure A3 Degrees of freedom in 2D structures

Figure A2 Initial model of ﬁ rst storey of repetitive multistorey building MIKE PURVISMIKE

Figure A4 Degrees of freedom in 3D structures

Figure A5 Final model of all storeys Acknowledgements of repetitive multistorey This guide has been produced using training material originally building developed by Mike Purvis (now at Ramboll) while working for SKM Anthony Hunt Associates. MIKE PURVISMIKE

Structural Behaviour Course A free online course for our Academic Community and Student Members. This exciting new course offers 200 questions which assess elements of structural behaviour, with 20 questions presented to you at random each time you log in. You can take the course, free of charge, as many times as you like. Go to www.istructe.org/resources-centre/structural-behaviour

A year’s access to the Course is also available to other members of the Institution for just £5 in the UK (or less, depending on VAT outside the UK).

TSE51_80-87 EducationCarr 2 v1.indd 87 24/02/2016 14:02 ›

88 TheStructuralEngineer Education March 2016 Embracing creativity

Virtual by design

Institution Past President, Tim Ibell, welcomes the creative possibilities of the digital revolution in structural engineering and urges universities to widen the appeal of the profession to school-leavers with interests other than maths and physics.

Real pace then constructed through profound reliance the left-hand side – which is the heart of the I was reminded recently that smartphones on computational power. These possibilities profession as we move forward into the digital have only been around for eight years or so. are fairly new, and our profession is changing revolution – to be adequately embedded. It But do you know anyone who doesn’t own rapidly to keep up with these developments. will not be. Our education system simply must one or, at least, use one from time to time? But, the education underpinning our embrace the left-hand side in ways it has thus Technology is moving at an extraordinary profession has an even greater requirement far chosen not to do, in general. pace, it is becoming entirely pervasive in to move with the times, and to do so rather everyday life, and we are all continually rapidly. I believe that academia is far behind Breadth adapting to such advances. where we need to be to launch our profession When we were children, we were interested Consider the time spectrum in Figure 1 profoundly into the digital age. in absolutely everything and we asked all for what might be considered to be a To me, the digital age means that the sorts of questions. Then we were sent to typical project progression for a structural right-hand side of the spectrum in Fig. 1 school, and we were told that these sorts of engineering designer. The project starts on is becoming less critical as an immediate interests could be pigeon-holed into various the left, and from here to just beyond the educational target zone in the formative subjects. It then dawned on us that our middle of the spectrum are the fun words creation of a structural engineer. The progression was judged based on our ability which describe the creative side of our left-hand side is where the future of our to write examinations in such subjects. For profession. One might argue about the choice profession, and of its education, lies. To me, many of us, advice was given about which or ordering of the words which appear here, this is profoundly exciting. I am not suggesting subjects to stick to, based on examination but that is less important to me. The key that the right-hand side is irrelevant. Far from ability and “aptitude”, and which subject issue is the right-hand side. At this end of the it, as exploration of ideas always requires areas to consider giving up. At A-level (age spectrum are the detailed mathematical and a strong skillset from the right-hand side. 16–18), for many of us examination success computational aspects of our profession. However, I am indeed suggesting that we or failure led to three or four subjects being Most of us reading this article will have had can no longer simply aim our education at the chosen ones. Universities carry with a formal education which was aimed squarely the right-hand side and magically expect them the danger, certainly in engineering at the right-hand side of this spectrum. This education, of still further narrowing people. education focus has served the profession This delusional obsession with academic adequately, at best, I would suggest, but is it narrowness must change. fi t for purpose even at an adequate level as Breadth of outlook, interest in the world we move forward in the digital revolution? I "If an incoming around you and a deep motivation to learn suggest not. Emphatically not. are the key fundamental talents which we The digital revolution is changing fi rst-year student should be looking for in future in order to fuel everything. Software is being used routinely the profession through the digital revolution. to form-fi nd major structural systems using doesn’t yet know If an incoming fi rst-year student doesn’t yet powerful parametric design principles. how to integrate know how to integrate sin x, so what? They’ll Analysis software can check itself through learn. Quickly. But only if they are inspired to physics-based routines. Major software sin x, so what? do so. This is where the importance of the suppliers are developing virtual design left-hand side is critical. If students can be spaces where a designer can play, connect They’ll learn" exposed to the real fun of engineering right elements together, bend elements in a virtual from the start of their degree programme, world into appropriate shapes, and thereby they will indeed be inspired to learn all the build structures which can be checked and right-hand side for themselves. The left

TSE51_88-89 Education software 2 v1.indd 88 24/02/2016 12:18 www.thestructuralengineer.org 89

Figure 1 Time spectrum on project for typical structural engineering designer

drives embedment of the right. It is true that of our profession (in the UK). Can it possibly The digital revolution will provide us with there are many engineers who might say be true that not a single one of the other the basis to allow exploration of ideas as that the right leads to the left (“fundamentals 620 000 school-leavers has the aptitude to never before. This, in turn, will allow us fi rst, application later”), because this is the be a great structural engineer, just because at to immerse our students in the creative only way in which it has been possible thus the age of 17 they are not yet able to explain aspects of our wonderful profession. To be far, in the main, to experience structural a particular quantum mechanics issue? It’s a great structural engineer requires breadth engineering formation from university to bonkers, isn’t it? in outlook so that creativity is stimulated, practice. The “left leading to the right” has If a student is terrifi ed of mathematics, which, in turn, leads to deep learning. The really not been given universal exposure, that student will never be an engineer. That is digital revolution will do nothing less than except by a few. But to me it is a profoundly clear. But if a student enjoys mathematics but open up the attractiveness of our profession obvious step which our global education simply prefers other subjects even more and to a much larger proportion of the 650 000 system must take. is forced to make diffi cult choices because school-leavers each year. This is a true We need to drive educational change by of the A-level narrowing issue, then surely revolution, and simply fabulous for our future. understanding what makes a student want to we should rethink our approach to these learn anything at all. Deep learning takes place students? Some universities are ahead of the Author biography when a student is happy, motivated and in a game in this enlightened approach, and the Tim Ibell creative mindset. We have to teach a student rest of us need to take note. It is not good FREng, CEng, BSc(Eng), PhD, FIStructE, FICE, when they are uninspired. Learning and enough to scoff about this issue any longer. FHEA teaching are anything but synonymous. In the How easy it is for a broad-minded student Tim Ibell was President context of structural engineering education, I to learn how to transpose a matrix. How of The Institution of Structural Engineers in believe fi rmly that deep technical learning can diffi cult it is for a narrow-minded student 2015. He is Associate only thrive when students are immersed in the to learn how to think holistically and with Dean for Research left-hand side. Otherwise, teaching is required. broad understanding. No one ever did sums in Engineering at the Thus, breadth in those aspiring to enter our to check something that no one else had University of Bath, UK, profession is essential. I cringe when I see imagined might happen. If you cannot imagine and a Fellow of the Royal statistics which show how leaky the “pipeline” possibilities and then check them, you are Academy of Engineering. He believes that creativity doomed to always do what has been done of 650 000 UK school-leavers is. This only lies at the heart of great provides us with 30 000 possible engineering before – or to learn the hard way. And in our structural engineering, students as the mathematics-and-physics profession, learning the hard way is just not including in its research A-level pool from which to choose the future acceptable. and education.

TSE51_88-89 Education software 2 v1.indd 89 24/02/2016 12:18 p90_TSE.03.16.indd 90 23/02/2016 17:20 › www.thestructuralengineer.org TheStructuralEngineer 91 March 2016 Opinion

Letters or longer articles written from a personal perspective, on topics of current interest that off er a particular opinion and often encourage further discussion and/or debate. 92 Viewpoint: Structural engineering within a digital environment

94 Viewpoint: Digital design, fabrication and assembly

96 Viewpoint: Should we now say “design and analysis” and not “analysis and design”?

98 Book review: Advanced Modelling Techniques in Structural Design

99 Verulam

TSE51_91 Opinion opener.indd 91 24/02/2016 12:19 ›

92 TheStructuralEngineer Opinion March 2016 Exploring the virtual world Viewpoint

Structural engineers should embrace the creative possibilities off ered by a virtual, digital world, believes Tristram Carfrae, if they are to transcend the limitations of their own experience. Structural engineering within a digital environment

Traditionally, as structural engineers, we conceive structures in our heads and draw them on the backs of envelopes or on napkins in restaurants. We communicate our ideas with sketches and discuss them verbally with our collaborators and clients. And there is absolutely nothing wrong with this methodology… but it is not the only way to approach structural design. The design process is fundamentally about seeking opportunities within the constraints of what will physically work (is it robust, safe, buildable etc.?). If we do all of our exploring within our heads, we tend to

be limited by what we already know; what we ARUP AND UP THINK OF COURTESY have experienced; and what we can mentally Figure 1 conceive. NDesign process Alternatively, if we explore possibilities answer, or at least an outline of the answer, within a virtual world that correctly before we begin. Furthermore, there is no represents (at least approximately) the guarantee of success; we may continuously physical behaviour of the real world, then "If we do all of meet dead ends and never identify a better we may look into areas in which we have solution. During this experimental design no previous experience, and examine our exploring process we make ourselves vulnerable; we possibilities that we cannot, at least initially, don’t know where we are going or where we mentally comprehend. In doing so, we will within our heads, might end up. learn more from our virtual experience we tend to be Unfortunately, as structural engineers and thereby think of new ideas for further we are likely to have spent most of our exploration. limited by what we formative years being taught to ﬁ nd the only This sort of design process is an solution available for increasingly complex exploratory, meandering adventure. It cannot already know" problems. We have been presented with be fully planned and the outcome is unknown precisely the right amount of information when we begin. This tends to make logical from which we can derive the correct engineers nervous; we want to know the answer, singular. We become uncomfortable

TSE51_92-93 Viewpoint digital 1.5 v1.indd 92 24/02/2016 12:21 www.thestructuralengineer.org 93

when faced with a blank sheet of paper, individually concerned with their own be looking for a conceptual modelling even though this represents the moment of discipline to the exclusion of all others. It environment in which we can get instant maximum opportunity, and would rather wait requires all of the team to consider, and feedback on physical performance? for someone else to start constraining the care about, the project as a whole; even if, For example, if we try to make and test challenge until we can attack it using our individually, we are only responsible for a structures within the physics engine used by deeply ingrained problem-solving skills to get part of its performance. computer games, they will either stand up to the “right” answer. Personally, I am most comfortable or otherwise. This is, of course, only a crude It is the more non-linear, creative exploring structural ideas using the software approximation of the real world, but the processes that defi ne humanity. The sort of tools that I am most familiar with; often ones benefi ts of speed may outweigh the benefi ts process exercised when given a blank sheet that I helped to write. Younger generations of accuracy, at least initially. We may learn of paper. Purely logical systematic deduction are likely to be using much more powerful more using this approximate approach than may be relatively easily replaced by artifi cial parametric tools and processes that I am through using more accurate ones, as it may intelligence (as has already begun). less familiar with. This doesn’t make them encourage more exploratory design. Design in the real world is of course more wrong! Instead, it obliges me to try and Whichever way we each wish to approach complicated than this. It is not just about understand what they are doing, so that structural engineering, it is not wrong for structural engineering. The whole project my experience and knowledge can be some of us to learn from exploring the design process can be represented by the brought to bear on their work rather than virtual, digital world. diagram in Figure 1. forcing them to use my simpler, outdated This shows that we are not conceiving our methodology. I need to learn enough about Author structures in isolation, but helping to best their techniques and software to properly biography satisfy a more complex brief. We are not engage with their processes. Tristram Carfrae RDI simply minimisers of the risk of structural At the moment, I see people using Tristram Carfrae, failure. Ultimately, we usually seek the parametric geometric models linked to recipient of the best experience for the users of the built full-fat analysis tools and creating Building Institution’s Gold Medal environment at the least possible cost in Information Modelling (BIM) output even in 2014, is Deputy Chairman of Arup. terms of time, money, resource consumption, in the early stages of conceptual design. Appointed as a Royal CO emissions, environmental impact etc. This complex sequence is capable of all 2 Designer for Industry in (assessed over the full lifecycle of the project). the detailed requirements and accuracy of 2006, Tristram is famous The potential of the whole project is fi nal design, but it may not be the best way for his innovative structures that have often been rarely maximised by a team who are only to explore new ideas. Should we instead conceived in a digital world.

The Award Winner: YOUNG – £1,500 cash prize STRUCTURAL The Institution of Structural Engineers’ Young – Two tickets to the industry’s most prestigious 20 Structural Engineering Professional Award awards ceremony, The Structural Awards in ENGINEERING recognises outstanding new talent at work in the November 2016 PROFESSIONAL structural engineering profession and is open to AWARD 16 any member, worldwide, under the age of 30. Runner-up: – £500 cash prize The prizes will be presented at the People and Papers awards ceremony in June 2016. Winners will also be invited to act as ambassadors of the profession on behalf of the Institution, participating in Institution activities and providing media comment where appropriate. ©Photo credit: Serpentine Pavillion by Adam Bowie, Flickr

How to enter:

All entries, which includes a brief CV, a report, and a video application, should be submitted via the online form by Tuesday 5 April 2016.

process please visit istructe.org/ysepa or contact the Events Team: [email protected]

TSE51_92-93 Viewpoint digital 1.5 v1.indd 93 24/02/2016 12:21 ›

94 TheStructuralEngineer Opinion March 2016 Automating the future Viewpoint

Tim Lucas of Price & Myers envisages a not-too-distant future in which automation is extended to assembly on site, giving engineers full control of the building process and licence to explore their wildest design dreams. Digital design, fabrication and assembly

bandstand’s vastly more complicated great- granddaughter – Richard Wilson’s Slipstream sculpture (Figure 2) at Heathrow Terminal 2 – in 2014. Back in 2001, I modelled the bandstand by hand in Autodesk 3ds Max, sectioned the form, found it was self-intersecting, and traced it manually in AutoCAD to produce the fabrication proﬁ les. Slipstream was hand- modelled around an animation of a tumbling stunt plane, but relied on lots of scripting to generate the 35 000 unique coordinated parts that make up the sculpture.

Creative opportunity Figure 1 Bandstand at De La Warr Pavilion, Bexhill-on-Sea Scripting – computer programming – is NICK KANE NICK moving forwards in the design process and the ability to fabricate accurate things is moving closer too. At Price & Myers we I recently went to visit an old project, the after 15 years by the sea and Michael now have a workshop in our basement. For bandstand at the De La Warr Pavilion (Figure McHugh, who made it, bought my original competitions we build models on our laser 1) in Bexhill-on-Sea on the south coast of 2m long drawings with their curves and grids cutter that are almost facsimiles of the real England, which we designed and built with marked up with his notes from the time. It thing, made in the same way that the fi nal architect Niall McLaughlin in 2001. It was was handmade from those drawings. If we project will be. a seminal one for me because it was one were building it now, we would design it for These design and production tools mean of the fi rst projects for which I worked out production on CNC (computer numerical that we need, as engineers, to develop a tacit the shape and drew every bit so that the control) routers and laser cutters. We knowledge of all aspects of materials and the fabricator could build it from my drawings. It might design it for easy assembly with tools that might form them into projects. We was bespoke and unusual, but being willing self-jigging notches, codes carved into also need to remember that programming to draw each part and take responsibility for the timber to say where things should go scripts is not everything and that drawing and it made it happen for not much money. and pilot holes for screws. This is how we manual modelling need to come fi rst. The bandstand needs a refurbishment produced the fabrication information for the I am confi dent that this ability to control

TSE51_94-95 Viewpoint Lucas 2 v1.indd 94 24/02/2016 12:21 www.thestructuralengineer.org 95 DAVID LEVENE DAVID

"Automation… off ers the opportunity to perhaps build what we always dreamed of"

Figure 2 E Slipstream sculpture at Heathrow Terminal 2

automatically how a line is drawn or how a tool moves will unleash a great deal of creative opportunity in our profession. The wilder extremes of what we can imagine can now be made and built using our own digitally driven tools.

Automation Automation for many industries, such as taxi driving, is fearsome and implies job losses. For making the built environment, I think it

off ers the opportunity to perhaps build what CSI we always dreamed of, but never thought we’d fi nd a client with the money for. Automation of design goes hand in hand with digital fabrication of parts. These are already here and working together well. / KLEINHOUT MAARTEN The fi nal hurdle is automation of assembly. This has been cracked for repetitive objects in factories, but not unique and bespoke many others, a Masters degree in Design for economic timeframe, both in design and objects on messy building sites. Perhaps it Manufacture. We have the robots and we construction time. We’ll have to reach for the will start with factory-made modules that have space to see if we can digitally design, “Blue Book” less and less. come together in the diffi cult environment manufacture and, all importantly, assemble of a site, like modules for the International the built environment there. Watch this Space Station. I think this is the fi nal required space. Author part of the puzzle of being able to build as an I see a future, not that far away, where we biography engineer with full control of the process. will – more and more – fi nd ourselves able Tim Lucas At the Bartlett School of Architecture in to bring together assemblies of material MEng, CEng, MIStructE Tim Lucas is a Partner University College London, where I teach, into structural, useful and effi cient elements in Price & Myers we have leased a huge factory space in of the built environment; specifying and and Lecturer at the Hackney Wick to go alongside the School controlling how they are made and built, to Bartlett School of in Bloomsbury. We plan to teach, among fi t together as a structure in a reasonably Architecture, London.

TSE51_94-95 Viewpoint Lucas 2 v1.indd 95 24/02/2016 12:22 ›

96 TheStructuralEngineer Opinion March 2016 A&D or D&A? Viewpoint

David Sanderson, Product Development Manager, Engineering Structures Division, Trimble, gives his perspective on developments in structural engineering software and asks whether it is time to starting thinking in terms of “design and analysis”, if not just “design”. Should we now say “design and analysis” and not “analysis and design”?

Over the last 30 years, since computers fi rst started appearing in structural design offi ces, structural engineers have referred to the software packages that run on these computers fi rst as analysis packages and then later as analysis and design (A&D) packages. This article sets out the case that it’s now time for structural engineers to think again and start referring to them as “design and analysis” packages. Around 50 years ago, Gordon Moore predicted a doubling of computational power every year; this has held more or less true. This continued increase in power, and ever- increasing aff ordability of PCs, has had a signifi cant eff ect on the construction industry. In more recent years, analysis has come from being expensive and time-consuming to being readily available and able to be run

on even the most basic PCs; even “physics Figure 1 games” on mobile phones can now solve a Generation of multiple analysis constantly changing system in real time. models A&D software, underneath the covers, uses the ﬁ nite-element (FE) method. Interestingly, this has hardly changed over time: new elements are created, new techniques Where we have seen the biggest change This is a major step change from adopted, but the majority of commercial is not in the analysis elements, nor in the where we were 10–15 years ago, where software packages still rely mainly on the analysis techniques themselves, but the ever- traditionally the engineer started with Timoshenko beam and Mindlin–Reissner plate increasing complexity and ingenuity of how an analytical model, with no concept of formulations. these features are used to better approximate the design model. Software suppliers the physical structure. tried, with varying degrees of success, to Paradigm shift In recent years, there has been a paradigm retrospectively derive a design model on A&D software is rapidly advancing, with the shift in the role of A&D software; previously, the back of the analytical model in order analysis component just one of many tools engineers had over time come to model, to aid the engineer, but this was often a used by design software. This, along with analyse and design a structure all within a cumbersome interactive process with other productivity-driven tools for automated single software package. However, with an limitations. The actual physical model itself wind loading, automated pattern loading, load ever-increasing focus on Building Information was nowhere to be seen. combination generation and complex mesh Modelling (BIM), it is becoming more and more With the advent of BIM, complex generation, is now available in most modern common for a model to begin life outside of physical models are easily created in A&D packages. the A&D software. products such as Tekla Structures and

TSE51_96-97 Viewpoint.indd 96 24/02/2016 12:22 www.thestructuralengineer.org 97

Figure 2 Physical modelling of simple concrete frame

a) Traditional centreline model b) Designing physical length of members c) Making connection zones rigid

Revit Structures. These physical models are checks, and a second-order building analysis physical dimensions is an innovation which is then routinely imported into the latest A&D for strength checks. For concrete structures, becoming more popular, and will continue to software packages, which are able to interpret individual substructures, typically for each become more widespread. This is achieved by the complex physical model and create a ﬂ oor, are generated and analysed to facilitate considering the structural elements’ physical design model from the physical model. The beam and slab design using results combined size and location, and modelling these, rather design model can then be decomposed to from building analysis, FE chase-down and than the usual centreline “stick” models. one, or more, analytical models. grillage chase-down (Figure 1). Physical modelling can produce far more Even for relatively small structures, the economic design compared to the stick model Multiple analysis models number of analysis models generated can be centreline approach. In order to design a structure eff ectively, it high. The requirement to create all of these Consider the simple concrete frame in is necessary to generate multiple analysis multiple analysis models in a single design Figure 2. Traditionally, the beam and columns models from this single design model, to pass is driven by the design process. are considered to run between nodes (Fig. 2a); ensure that all of the design eff ects are taken however, in reality, the beam exists only to the account of. Tekla Structural Designer, for Physical modelling face of the columns. By designing the beam example, does this and uses planar analysis It’s not only the number of analysis models for this shortened length (Fig. 2b), the peak models to accurately decompose loading on which has increased. It’s also the complexity hogging moments are reduced. By modelling two-way slabs to the supporting structure: a of these models, and ensuring that they these connections as rigid elements (Fig. 2c), ﬁ rst-order building analysis for serviceability are suitable for design. Modelling using the moment can also be redistributed. Physical modelling can also be considered in the connection of columns and slabs. By Figure 3 creating a column “cut-out” in an FE mesh, SPhysical modelling of columns and slabs peak moments are drastically reduced (Figure 3). Physical modelling also creates a much more stable and accurate analytical model as a singularity is avoided. We should always remember, however, that no matter how sophisticated the design process has become, it is still important that the structural engineer remains in control and is able to visualise the generated analytical model(s), in order to verify how members are

a) Slab deﬂ ection with and connected, degrees of freedom, ﬁ xity, etc. without physical member This must always remain transparent to the modelling user of any these design packages.

Conclusion So, is analysis now subservient to the design; b) Slab moment contour plot with and should we refer to packages as “design and without physical member modelling (very high peak moments at column are not analysis” instead? present when using physical modelling) Or maybe renaming it design and analysis software doesn’t go far enough; maybe we should just call it design software!

Author biography David Sanderson First with CSC, now Trimble, David has spent the last 30 years creating and managing the development of market-leading analysis and design software, including Fastrak Building Designer, Fastrak Portal Frame Designer and Tekla Structural Designer.

TSE51_96-97 Viewpoint.indd 97 24/02/2016 12:22 ›

98 TheStructuralEngineer Opinion March 2016 Book review Review This introduction to advanced modelling techniques will likely be of interest to senior engineers seeking background knowledge on the use of analysis software for specialised structural design, writes John Lyness.

which must comply with Eurocode or other Advanced standards, several conditions must be satisfi ed by the geometric data, material property data and load data. References are provided at Modelling the end of each chapter that go some way to provide this. However, many questions Techniques in remain, for this reviewer, about proving the suitability of the FE analysis techniques and software that have been chosen. For such Structural Design structural forms, load cases and material behaviours, involving the use of proprietary analytic software, the quality assurance (QA) compliance sequences and procedures are surely paramount. For more completeness it Author: Feng Fu would have been useful if a description of the Publisher: Wiley-Blackwell appropriate design QA procedures relevant Price: £79.95 to these modelling steps in real-life structural ISBN: 978-1-118-82543-3 design could have been included. I am thinking of competence certifi cation, software certifi cation, ISO compliance, design standard This book covers the analysis, for design The introductory chapters list the compliance, error bounds, tolerances, purposes, of more unusual structural forms, proprietary geometric modellers, CAD statistical confi dence, risk assessment etc. load cases and material behaviours. The software and FE software that are described, Perhaps citing references to the QA for book seeks to describe the use of upstream later, in the applications. The proprietary the relevant FE techniques, the QA for the geometrical modellers as an effi cient source analysis software that is listed covers most of relevant proprietary FE software and the of geometric data for the analysis model. the popular software currently in use for “one- possible assessments of accuracy, following The book contains 10 chapters. Two are off ” structures and scenarios. The widespread adaptive re-meshing, error bounds, etc., could introductory chapters and then eight follow routine design offi ce use of structural analysis, be considered by the author in future. that cover the use of geometric and computer- design and detailing software for more I think that this book’s principal users will aided design (CAD) modellers with diff erent common structural forms is not described. be senior engineers who want speedy access proprietary fi nite-element (FE) and other Within each chapter a “nutshell” introduction to topical background knowledge on the use analysis packages. The objective of the use is given describing the appropriate mechanics of currently accepted analysis software for of the geometric and analysis modellers is to theory and the theory for the analysis method, specialised structural design, and information provide information on predicted stresses and together with some useful references. The on using their upstream 2D and 3D CAD displacements for diff erent, generally extreme preparation of the geometrical data, using modellers. I also think that the book will be scenarios and load cases due to static, geometric modellers, is followed by use of of interest to specialised structural analysis dynamic, environmental and thermal loadings. an appropriate numerical analysis method. practitioners who would like to broaden their The analysis and scenario chapters cover The use of the selected proprietary analysis practical knowledge of the scope of other tall buildings, earthquake analysis, progressive software is outlined by describing the design providers of currently available acceptable collapse, blast and impact loading, structural load cases, constraints, material properties analysis software for more specialised fi re analysis, space structures, bridge etc. The set-up procedure for the analysis structural analyses and scenarios. structures and foot-induced vibrations. is written up as a sequence of sentences The author is obviously an analysis with some commentary. The selection of the practitioner who has been engaged in appropriate analysis output and its useful ensuring the appropriate data and data format (graphical, vector etc.) for the structural Dr John Lyness preparation needs for the eff ective use of design process is illustrated and discussed. MIStructE analysis software, and who has been faced The written English is variable in parts. John is a Reader in Civil Engineering at The University of Ulster. He previously with the practicalities of optimising the use Several of the screenshot diagrams are worked as a structural engineer in of fi le exchange between the upstream illegible and most of the diagrams would have Northern Ireland and in off shore geometric (CAD) modellers and the analysis benefi ted from annotations. engineering in London. modellers. For the use of the output in design,

TSE51_98 BOOK REVIEW v1.indd 98 24/02/2016 12:23 › www.thestructuralengineer.org Opinion TheStructuralEngineer 99 Letters March 2016

Send letters to… All contributions to Verulam should be submitted via email to: [email protected]

Contributions may be edited on the grounds of style and/or length by the Verulam Institution's publishing department. Topics of importance openly discussed

“Reasonably done to satisfy the duty or requirement, or that there was no better practicable means Web buckling… practicable” than was in fact used to satisfy the duty or requirement.” and bearing – a need for Note the redefi nition of the meaning of the phrase and the reversed burden of guidance? proof: unless the accused can prove that Colin Taylor (who knows a great they did everything practicable that could deal about steel codes) writes in possibly have been done to prevent the with further feedback on Alasdair Verulam can always rely on Alasdair unfortunate event that happened, they Beal’s conundrum about steel beam Beal for a comment! are deemed to be guilty. It is a draconian web buckling (January 2016). requirement, far removed from Lord Justice Asquith’s view. Colin Mitchell’s quotation (Verulam, February In civil cases like that quoted by Mr Further to David Brown’s response (February 2016) of Lord Justice Cyril Asquith’s Mitchell, Asquith’s defi nition still applies. 2016) to Alasdair Beal’s query on web 1949 defi nition of “as far as reasonably However, with regard to the CDM buckling, it should be noted that it is also practicable” is helpful and, dare I say Regulations, which defi nition applies: is it necessary to check web bearing. In BS it, reasonable. However, although Lord Asquith or is it H&SW Clause 40? Most 449 this is based on dispersion through the Asquith’s defi nition may still hold good in published commentaries assume the fl ange and root radius at 30°; see clause 27e. most areas of law, anyone unfortunate former, but the Regulations themselves For the beam in Alasdair’s example, the BS enough to be prosecuted under the Health do not say and the Health and Safety 449 web bearing value is 156kN, far less than and Safety at Work Act (H&SW) will fi nd that Executive has not published clear guidance. either of the web buckling values quoted. under Clause 40 of this legislation a very This is not a “nebulous” question: it is a In end connection design to BS 449, web diff erent defi nition applies: serious ambiguity in the CDM Regulations bearing typically governs rather than web “In any proceedings for an off ence under which aff ects a number of important buckling. This is probably why problems any of the relevant statutory provisions clauses. In the interests of fairness and were not experienced in practice. On the consisting of a failure to comply with a duty justice it needs to be clarifi ed by the other hand, it is important that designers do or requirement to do something so far as relevant authorities. check both failure modes, whatever code is practicable or so far as is reasonably they use. practicable, or to use the best practicable Indeed it is diffi cult to judge what Incidentally, the original web buckling rule means to do something, it shall be for the is “reasonably practicable” and the in BS 449 for buildings was the same as accused to prove (as the case may be) that interpretation would vary from person to given in BS 153 for bridges. It was changed it was not practicable or not reasonably person. One part of the test must surely in BS 5950:1985 as the result of preliminary practicable to do more than was in fact be what current professional practice fi ndings from research at Aston University by would deem to be appropriate. A starting Alan Astill, which sadly remained unfi nished point as well must be for parties to have following his untimely death. The new "IT IS A DRACONIAN at least formally considered what hazards version in BS 5950:2000 arose from the REQUIREMENT, FAR are likely to apply and recorded those and development of Eurocode 3. REMOVED FROM LORD recorded what action they considered JUSTICE ASQUITH’S VIEW" appropriate at the time to reduce risks. This chain of correspondence reinforces As Alasdair reminds us, this is a topic the need for those drafting codes to which can be of immense importance. document why published rules are what Views please! they are.

TSE51_99 Verulam v1.indd 99 24/02/2016 12:23 › www.thestructuralengineer.org 100 TheStructuralEngineer March 2016 OpinionAt the

Letters or longer articles, written from a personal perspective on topics of current interest that off er a particular opinion and often backencourage further discussion and/or debate. 48 Basements and subterranean development: The home of diary dates, updates on Institution services and other miscellanea, plus products,It’s services time and to jobs. engage with duty of care and Party 101 WallDiary matters dates

10352 Verulam Spotlight (reader’s on Structures letters)

10554 Regional And ﬁ nally… Group Proﬁ le – Xxxxxxxxx

106 Products & Services

108 Services Directory

109 TheStructuralEngineerJobs

TSE51_100 At the back opener.indd 100 24/02/2016 12:24 › www.thestructuralengineer.org At the back TheStructuralEngineer 101 Diary dates March 2016

Unless otherwise stated, technical meetings start at 18:00 with refreshments from 17:30 - they are free of charge to attend, unless stated otherwise. Registration for evening technical meetings is required via [email protected]

History Study Group meetings start at 18:15 Diary with refreshments from 17:45. Registration is not required for History Study Group meetings except for the Annual business meeting held dates in January. Note that more current information may be available from the Regional Group Committee members should submit details of forthcoming Institution website: www.istructe.org/events-and-awards events to: [email protected]

MEETINGS AT HQ Tuesday 14 June Thursday 17 March 47–58 Bastwick Street, Historic gasholders in Annual General Meeting Midland Counties Northern Counties London EC1V 3PS, UK Italy Philip Nelson (Tees Centre) Barbara Berger Lancashire County Cricket Tuesday 22 March Wednesday 6 April Club, Old Traff ord, Talbot Emergency engineering Tuesday 8 March Young Researchers’ REGIONAL GROUPS Road, Manchester M16 0PX and President’s visit Brazil Olympics: design Conference 2016 18:00 for 18:30 Alan Crossman and Nick of water slalom Prof Ian Kinloch East Anglia Francis (Joint meeting with ICE) 09:00 for 09:30 start Wednesday 13 April Room G34, School of Civil Stephenson Building, (Registration required) Wednesday 16 March The Brian Bogle & Mechanical Engineering, Teesside University, Off shore wind – meeting Memorial Lecture: Etihad University of Birmingham, Middlesbrough TS1 3BA Wednesday 20 April the challenges Stadium extension, Birmingham B15 2TT 17:45 for 18:15 Best construction (Joint meeting with ICE) Manchester 18:00 for 18:30 Details: [email protected] methods for concrete Rob Mattholie and James Paul Hulme Details: prashant.sikka@ bridge decks – cost data Rinkel University of Bolton, Deane gmail.com Tuesday 12 April Simon Bourne The Oaklands Hotel, 89 Lecture Theatre, Deane Joining of composite (Concrete Bridge Yarmouth Road, Norwich Campus, Bolton BL3 5AB Secretary: Craig McFadyen materials Development Group) NR7 0HH 17:45 for 18:30 (email:craig.mcfadyen@ymail. Chris Worral 17:30 for 18:00 18:00 for 18:30 com) Stephenson Building, Tuesday 19 April Teesside University, HISTORY STUDY Secretary: Paul Wilson Design in the Historic Middlesbrough TS1 3BA GROUP (tel: 01603 614 834; email: Environment Northern Counties 17:45 for 18:15 Tuesday 15 March [email protected]) (Joint meeting with ICE (Tyne Centre) Details: [email protected] The new HMS Victory Merseyside) support structure Andrew Burns and Ray Tuesday 5 April Secretary: Trevor Little Richard Fenton Lancashire and Makin President’s visit followed (tel: 0191 255 7300; email: Cheshire Liverpool John Moores by ‘Skill and care [email protected]) Tuesday 19 April University, James Parsons vs compliance with Sir John Hawkshaw Tuesday 8 March Lecture Theatre, School of speciﬁ cations’ Martin Beaumont 125 years of steel bridges Engineering, Byrom Street, Allan Crossman and Robert Northern Ireland in Britain Liverpool L3 4AF Langley Tuesday 3 May Alan Hayward 18:00 for 18:30 Crowne Plaza Hotel, Friday 11 March The Roman Walbrook University of Manchester, Stephenson Quarter, Forth Chairman’s address and its legacy Renold Building, PO Box 85, Secretary: Ian Tickle (email: Street, Newcastle-upon-Tyne and visit by CEO Martin Tony Taylor Sackville Street, Manchester [email protected]) NE1 3PF Powell M1 3BB 17:45 for 18:15 Michael Shaw 17:45 for 18:30 Details: [email protected] Malone House, Barnett

TSE51_101-102 Diary v1.indd 101 24/02/2016 12:24 ›

102 TheStructuralEngineer At the back March 2016 Diary dates

Demesne, Belfast BT9 5PB environment 18:00 for 18:30 Secretary: Mahara 423-4030; web: www.dfi .org 17:30 for 18:15 Andrew Burns and Ray Register and details: www. Booshanam Makin ise-tvrg.eventbrite.co.uk (maharabooshanam@gmail. Tuesday 12 April Lecture Theatre M, Surrey com) London, UK Conservation and University, Guildford, Surrey Secretary: Parmindar Mann restoration of St Mel’s GU2 7XH (parmindarkmann@hotmail. 22 June Cathedral, Longford 18:00 for 18:30 com) Yorkshire CBDG Conference: (Joint meeting with IEI) Details: edward.bromhead@ Advances in Concrete Kevin Fay and Jonathan btinternet.com Monday 14 March Bridges Macauley Wales The Kelpies The Institution of Structural Room ASH/GM/001, Ashby Monday 11 April Tim Burton Engineers, International Building, Queens University Report writing Tuesday 15 March Bedern Hall, Bartle Garth, HQ, 47–58 Bastwick Street, Belfast, Stranmillis Rd, Brendan Brophy South Wales YMG 2016 York, North Yorkshire London EC1V 3PS, UK Belfast BT9 5AH John Galsworthy Building, Pub Quiz YO1 7AL Details: www.cbdg.org.uk 17:30 for 18:15 Kingston University, Penrhyn The Yard, 42-43 St Mary 18:00 for 18:30 Road, Kingston-upon- Street, Cardiff CF10 1AD Details: j.f.carr@sheffi eld. Tuesday 3 May Thames, Surrey KT1 2EE 18:00 (First students get free ac.uk Newcastle, UK Chartered Membership 18:00 for 18:30 drink on arrival) Examination seminar Pt 1 Details: edward.bromhead@ Wednesday 20 April Wednesday 26–Friday Room ASH/02/016, Ashby btinternet.com Secretary: Tim Bennett Highlights from the 28 October Building, Queens University (email: tim.bennett@arup. President’s Address Novel Structural Skins – Belfast, Stranmillis Rd, Secretary: Ruslan com) Alan Crossman Improving sustainability Belfast BT9 5AH Koutlukaev (email: rk@ Leeds Beckett University, and effi ciency through wcjeng.co.uk) The Rose Bowl, Portland new structural textile Tuesday 10 May Western Counties Crescent, Leeds LS1 3HB materials and designs Chartered Membership 18:00 for 18:30 Newcastle University, Examination seminar Pt 2 Thames Valley Tuesday 8 March Details: j.f.carr@sheffi eld. Newcastle-upon-Tyne, UK Room ASH/04/006, Ashby Structural behaviour – do ac.uk Details: Alison Bird Building, Queens University Thursday 7 April you know your hogging ([email protected]) Belfast, Stranmillis Rd, Routes to Membership from your sagging? Monday 16 May Belfast BT9 5AH Darren Byrne Dr A.J. Crewe New developments Holiday Inn, High Wycombe Queen’s Building, Bristol in cold-formed steel Brisbane, Australia Hon. Secretary: John M40 Jct 4, HP11 1TL University, Bristol BS8 1TH construction Hutchinson Details: Daljit Matharu (ise- Prior registration: http://bit. Dr Jurgen Becque and Dr Thursday 22–Sunday 25 (tel: 028 2602475; email: [email protected]) ly/ISE_TG_MAR_16 Iman Hajirasouliha November john@hannaandhutchinson. Register at: www.ise-tvrg. 18:00 for 18:30 University of Sheffi eld, Australasian Structural com) eventbrite.co.uk Western Road, Sheffi eld Engineering Conference 18:00 for 18:30 Thursday 17 March S10 2TN (ASEC 2016) – The roles Forensic engineering and 18:00 for 18:30 of structural engineers South Eastern Wednesday 25 May learning from defects, Details: j.f.carr@sheffi eld. Keynote speakers include: Counties Annual Dinner (BBQ) and failures and damage ac.uk Tristram Carfrae, Dr Sean President’s visit Colin Richardson Brady, Alan Crossman Tuesday 8 March Runnymede Boat House, Arup, 63 St Thomas St, Tuesday 14 June (IStructE President), S.P. Inaugural Meeting and opposite Magna Carta Tea Avon, Bristol BS1 6JZ Stanbrook Abbey Chiew and Rob Hayward Dinner Room, Windsor Road, Egham 18:00 for 18:30 Speaker: TBA Brisbane Convention & Alan Crossman and Martin TW20 0AE Thorpe Park Hotel, 1150 Exhibition Centre (BCEC), Powell 18:30 for 19:00 (Boat will Friday 13 May Century Way, Leeds Brisbane, Australia Croydon Park Hotel, 7 Altyre leave at 19:00 sharp! Return Annual Dinner LS15 8ZB Details: asec2016.org.au Road, Croydon CR9 5AA at 23:00. Free parking is Ashton Court Mansion, 18:00 for 18:30 Cost: £26/person provided) Long Ashton, Bristol BS41 Details: j.f.carr@sheffi eld. OTHER EVENTS (vegetarian option) Tickets: £25.00 per head 9JN ac.uk Friday 27 May Contact: Jeff rey.fi sher@ and guests are welcome. 18:30 for 19:00 Chartered Membership mottmac.com or tim.pope@ Please book by 11 May, Tickets: £35.00 up to 29 Secretary: Nick Buxton one-day preparation mottmac.com stating food option (non- February; £45.00 from 1 ([email protected]) course 18:45 for 19:30 vegetarian/vegetarian) March Peter Gardner and David Bookings: www.eventbrite. using form: www.istructe. Formal dress: Lounge suit INTERNATIONAL Lowe co.uk/e/istructe-serg- org/events/regional/ or black tie CONFERENCES Royal Institution of inaugural-dinner-2016- thames-valley/2016/ Naval Architects, 8–9 tickets-21425015799 thames-valley-regional- Thursday 19 May Chicago, USA Northumberland Street, group-social-evening-pr Four Elms, Cardiff London WC2N 5DA Secretary: Kiu Li Pat Ruddock and Tom Tuesday 7–Thursday 9 Cost: £250 (Group size limit Thursday 16 June Martin June 24 – fi rst come, fi rst served) CDM 2015 Arup, 63 St Thomas Street, Superpile ’16 Contact: David Lowe (01202 Surrey Dr Mike Webster Avon, Bristol BS1 6JZ The Westin, 909 North 748325)) Copthorne Slough-Windsor, 18:00 for 18:30 Michigan Avenue, Chicago, 09:00–16:30 Monday 7 March Cippenham Lane, Slough, Register at: http://bit.ly/ Illinois 60611, USA Designing in the historic Berkshire SL1 2YE ISE_APR_16 Further details: tel: +1 (973)

TSE51_101-102 Diary v1.indd 102 24/02/2016 12:24 › www.thestructuralengineer.org At the back TheStructuralEngineer 103 Spotlight on Structures March 2016 Spotlight on

In this section we shine a spotlight on papers recently published in Structures – the Research Journal of The Institution of Structural Engineers.

Structures is a collaboration between the Institution and Elsevier, publishing internationally-leading research across the full breadth of structural engineering which will beneﬁ t from wide readership by academics and practitioners.

Access to Structures is free to Institution members (excluding Student members) as one of their membership beneﬁ ts, with access provided via the “My account” section of the Institution website. The journal is available online at: www.structuresjournal.org

Impact statements In this issue, we launch a new feature for 1.05 for fl exural design is as strong as it was Soil–structure interaction analysis of a Structures. “Impact statements” are intended in 1989, helped by a change in the BS for FPS-isolated structure using fi nite element to highlight papers in which the work is likely reinforcement which means yield stress is model to have a more immediate impact on practice calculated on nominal area so variation in A. Krishnamoorthy and S. Anita and which will, consequently, be of greater actual area is no longer a factor. interest to practising engineers. Neither reliability theory nor the experience Concrete fi lled elliptical steel tubular We begin with a paper discussing the partial base for using it with the EN 1992 varying members with large diameter-to-thickness safety factor for reinforcement (published in angle truss approach for shear is as good. ratio subjected to bending Volume 5). The paper had its origins in work The work does, however, justify using it for Kojiro Uenaka and Hisao Tsunokake carried out by the late Professor Andrew other cases and, where not constrained by the Beeby. Sadly, Professor Beeby died before need for strict compliance with current codes, Behaviour of PVC encased reinforced the paper was complete. Author Paul Jackson this could be done immediately. This would concrete walls under eccentric axial explains: lead to a 9% saving in any fl exural or direct loading “I agreed to complete the paper. I tension reinforcement governed by ultimate Amr Abdel Havez, Noran Wahab, Adil Al- anticipated this being minor editing but found strength. Mayah and Khaled A. Soudki that a change in the BS for reinforcement had The work is being discussed in the relevant not been considered. This required redoing BSI committee and it is estimated that full Compressive behaviour and design of the analysis but made the case stronger. adoption of the recommendations would lead prestressed steel elements “Many members will remember Professor to a saving of some £7M a year in the UK. Jonathan Gosaye, Leroy Gardner, M. Ahmer Beeby with both aff ection and great respect. Wadee and Murray E. Ellen For me, it is an honour to be joint author of Volume 5 the last Andrew Beeby paper and I only hope The fi fth issue of Structures (Volume 5) is now Seismic performance assessment of the parts I had to change are as good as the available and features the following articles: self-centering dual systems with diff erent original.” confi gurations Geometrically and materially nonlinear Mehdei Kafaeikivi, David A. Roke and Qindan Partial safety factor for reinforcement creep behaviour of reinforced concrete Huang Andrew Beeby and Paul Jackson columns In 1989 the partial safety factor for Ehab Hamed and Cynthia Lai Partial safety factor for reinforcement reinforcement in BS 8110 was reduced from Andrew Beeby and Paul Jackson 1.15 to 1.05. The justifi cation for this using both Seismic risk assessment of low rise RC reliability theory and experience was covered frame structure Performance-based Seismic Design of an in a paper by Andrew Beeby which provoked A. Melani, R.K. Khare, R.P. Dhakal and J.B. Irregular Tall Building — A Case Study extensive discussion. Since then, the standard Mander Ali Ruzi Özuygur grade has increased to 500N/mm2. As the study justifying the previous change was for The sensitivity of bridge safety to spatial Eff ect of stay-in-place PVC formwork panel 460 grade, the factor was changed back to correlation of load and resistance geometry on fl exural behavior of reinforced 1.15. It was anticipated this would be temporary Donya Hajializadeh, Eugene J. Obrien and concrete walls until data was available for the new steel. Mark G. Stewart Benjamin Scott, Noran Wahab, Adil Al-Mayah In this paper the safety factor for the 500 and Khaled A. Soudki grade reinforcement is investigated using new An evaluation of EC2 rules for design of very extensive CARES data, some 17,000 compression lap joints Eff ect of concrete compressive strength on test results. It is found that the case for using John Cairns transfer length

TSE51_103-104 Spotlight.indd 103 24/02/2016 12:26 24/02/2016 12:26 structural formstructural design structural construction engineering innovation structural events extreme sustainability topics (that architectural performance)impact structural materials Papers are accepted Papers to, but not limited from, the following subject areas: ● ● ● ● ● ● ● ● mass–spring–dampercation of z model of walking humans walking model of Pavic, Aleksandar Erfan Shahabpoor, Vitomir Racic PVA- of Longitudinal shear resistance composite slabs ECC Muhammad Aswin, Bashar S. Mohammed, and Muhammad Harry Beatty Walden Haﬁ RC of Seismic Assessment Probabilistic I — UncertaintyBridges: Part Models Raimundo Delgado and Monteiro, Ricardo Rui Pinho RC of Seismic Assessment Probabilistic II — Nonlinear Demand Bridges: Part Prediction Raimundo Delgado and Monteiro, Ricardo Rui Pinho considering soil–structure interaction soil–structure considering Abdoul R. Ghotbi Identiﬁ Research JournalResearch of The Institution of Structural Engineers Structures

› the back At on Structures Spotlight ned With Carbon Fibres An experimental study on the eff ect of ect on the eff study An experimental exterior of on the behavior bers PET ﬁ subjectedRC beam-column connection loading cyclic to reversed Shembiang Marthong and Comingstarful Marthong and sand- smooth Bond of behavior (SMA) coated shape memory alloy in concrete rebar and M. ShahriaA.H.M. Billah Muntasir Alam RC Columns Behaviour of Cyclic Lateral Conﬁ Pedro Faustino, Pedro Frade and Carlos Frade Pedro Faustino, Pedro Chastre Analytical anchor of approach steel base plate ness and stiff rod calculation under tension Daniel Tsavdaridis, Konstantinos Mohamed A. Shaheen, Charalampos and Emad Salem Baniotopoulos Response sensitivity analyses of bridges with and without skewed TheStructuralEngineer March 2016 March

Editor-in-Chief: Professor Leroy Gardner (Imperial College London) Gardner Professor Leroy Editor-in-Chief: Collaboration between The Institution of Structural Engineers and Elsevier The Institution of Structural between Collaboration All published articles freely available to IStructE members via www.istructe.org/my-account to IStructE members available All published articles freely

www.structuresjournal.org ● ●

● MEMBER PORTAL NOW AVAILABLE under foam lled with polyurethane 104 Alberto T. Ramirez-Garcia, Royce W. W. Royce Ramirez-Garcia, T. Alberto Hale and J.R. Micah Martí- W. Floyd, Vargas GFRP bridge deck panels Behavior of inﬁ exposure environmental various Volz, Mohamed ery Jeff Hesham Tuwair, K. Mohaned Mohamed, ElGawady, Birman Victor Chandrashekhara and buckling of of On the improvement columns steel universal pretwisted Megahed and Abdulla H. Abed, Mai Farid Al-Rahmani Passive Application Intelligent of Based on Shape MemoryDevices Structures of in Seismic Control Alloys Neda Salari and Asgarian, Behrouz Behnam Saadati Functionally of Buckling and Vibration Material ColumnsGraded Sharing Cases Mode Shape, and New Duncan’s , Moshe Eisenberger and Isaac Elishakoff Delmas Axel TSE51_103-104 Spotlight.indd 104 › www.thestructuralengineer.org At the back TheStructuralEngineer 105 And ﬁ nally… March 2016

The place to test your knowledge and problem-solving ability. If you would like to submit a quiz or And ﬁ nally... problem, contact [email protected]

We continue this section with another steel quiz brought to you by the SCI. This month’s topic is bolts. Answers will be published in the April issue.

Question 1 Question 2 Is it appropriate to use grade 8 nuts for Why is tack welding of grade 4.6 bolts to nuts galvanised or sheradised 8.8 bolts? permissible, but should be avoided for grade 8.8 and higher?

Question 3 Question 4 Could galvanised bolts be used with a For practical reasons, countersunk high- polyester powder-ﬁ nished mild steel beam if strength friction-grip (HSFG) bolts are it is exposed to the external environment? proposed for use in a splice detail of a runway beam. What is the diff erence in performance between countersunk HSFG and standard HSFG bolts? SCI is the leading, independent provider of technical expertise and disseminator of best practice to the steel construction sector. www.steel-sci.org

Answers to February’s quiz

1. In order to satisfy tying requirements, is it acceptable to increase the thickness of an end plate? For simple joints, the bending resistance of the end plate is likely to be governing the tying resistance, so increasing the thickness of the end plate is a very tempting course of action. However, in order to keep the joint nominally pinned, the end plate needs to remain ﬂ exible, so increasing the end plate thickness should be avoided.

2. When can the point of contraﬂ exure be considered a point of lateral restraint in a portal frame? When the rafter is a rolled section, the purlins are not shallower than 1/4 of the depth of the rafter, and at least two bolts are provided in the purlin-to-rafter connection.

3. Do second-order eff ects have to be considered when a frame is 4. Can profi led steel decking be relied upon to restrain the top fl ange braced? of the beam during construction? Yes. Second-order eff ects may only be ignored if the increase in Yes, if the decking is spanning perpendicular to the beam. The shear internal forces and moments caused by deformation of the structure stiff ness of the steel decking is typically suffi cient to provide ectiveeff under load is small enough to be neglected and small displacement continuous restraint in the plane of the sheeting. To verify that suffi cient theory may be assumed. In other words, as structures become lateral restraint is provided, the sheer stiff ness of the sheeting should more slender and susceptible to deformation, large displacement satisfy the criterion given in BS EN 1993-1-1, Annex BB.2. This is true theory applies and the change in structural behaviour caused by for sheeting connected to the beam at every rib. If the sheeting is deformations must be accounted for in the design. This is true in connected to a beam at every second rib, the shear stiff ness required braced or unbraced frames in concrete and timber, as well as in steel. increases by a factor of fi ve.

TSE51_105 And finally.. v1.indd 105 24/02/2016 12:26 PRODUCTS & SERVICES › Telephone: +44 (0)20 7880 7633 Email: [email protected] 106 TheStructuralEngineer March 2016 Reinforcing bar couplers get HAPAS-approval Ancon’s MBT mechanically-bolted couplers are the ﬁ rst reinforcing bar splicing system to be approved by the Highway Authorities Product Approval Scheme (HAPAS). The company’s MBT ET-type couplers have been independently proven to provide 100% strength of the bars being joined and exceed ‘Class D’ fatigue strength requirements. HAPAS certiﬁ cation means that the necessary requirements for proprietary mechanical joints in reinforcing bars given in the Manual of Contract Documents for Highways Works are met. The couplers are used for replacing cut out corroded reinforcement as part of highway maintenance, where bar end threading or rotation would be impossible. They are available in sizes from 10-40mm and join straight deformed high-yield carbon steel bars (grade 500) of the same diameter. The two bar ends are supported within the coupler on two serrated saddles and are locked in place by a series of special lockshear bolts, the heads of which shear off when the predetermined tightening torque is reached. Further information: Ancon (tel: 0114 275 5224; email: [email protected])

Working with Eurocodes? You’re not still doing it like this, are you? To use Eurocodes you need to research and cross-reference them with various separate documents, including: the National Annexes, Execution Standards and Non-Contradictory Complementary Information (NCCI). But not just once, you will need to reference forwards, backwards and forwards again, within, across and between documents. But there is another way - Eurocodes PLUS digitizes all 15,000 pages of the Eurocodes, making them more accessible and easier to understand, helping you reduce the time and cost of implementation. Find out how Eurocodes PLUS can help you clear your desk. Further information: BSI Group (web: www.bsigroup.com/eurocodesplus)

CRA chairman outlines development strategy The Concrete Repair Association’s new chairman Keith Barrow has outlined his plans for future development of the Association, detailing a series of key objectives which aim to uphold its profes- sionalism. During his 2-year tenure as Chairman, Keith plans to develop further the training available to its members, ensuring their continued competence and advancing their skills and expertise. He also hopes to improve the beneﬁ ts and rewards of membership and encourage greater participation, particularly from smaller contractors. Highlighting that the CRA is an established professional association within this specialist sector of construction, Keith explains that its strict entry requirements mean that membership provides businesses with a marker of credibility. As well as being able to demonstrate a proven track record and overall proven ability, each full member is obliged to be QA accredited to ISO 9001, and to the environmental standard BS EN ISO 14001. Members must also comply with the Association’s stringent Codes of Practice and maintain health and safety standards. Further information: Concrete Repair Association (tel: 01420 471612; email: [email protected]; web: www.cra.org.uk)

Structural Concrete Alliance 2015 award winners Bersche-Rolt has won top award in the 2015 Structural Concrete Alliance Award for Repair and Re- furbishment, organised by The Structural Concrete Alliance, for its concrete repair and coating works to the Grade II listed Barry Island Eastern Shelter carried out in September 2014. The award was presented by broadcaster Huw Edwards at the Concrete Society Awards at the Grosvenor Hotel, Park Lane, London in November 2015. All the structural elements of the 1930s reinforced concrete structure showed signs of concrete defects, with cracking and delamination within the beams and columns, reinforcement corrosion and peeling and discoloured paintwork. Works included cleaning and priming the reinforcement, the application of a bonding bridge, erecting shuttering and repairing the concrete. The complete structure was then coated with a three-part protective anti-carbonation system in a decorative coloured ﬁ nish. Second place was awarded to Sika Ltd for its solution for Britannia House, a 1930s building with a concrete-encased steel frame in the heart of Bradford city centre. Balfour Beatty Concrete Repairs was awarded third place for its repairs to the Sherborne Footbridge in Salford. Further information: Structural Concrete Alliance (web: www.structuralconcretealliance.org.uk)

P&S_Mar16.indd 106 19/02/2016 16:09 PRODUCTS & SERVICES ›

Telephone: +44 (0) 20 7880 7633 Email: [email protected] www.thestructuralengineer.org TheStructuralEngineer 107 March 2016 Cool solution for cold store construction

McColgan Quality Foods in Strabane, Northern Ireland has completed a £7.3M investment in an advanced packing and despatch cold storage facility. The main contractor on the project was Woodvale Construction. Reliable and accurate temperature control for the building is helped by Sundolitt XPS extruded polystyrene insulation. Buried in the ﬂ oor of the new 8000ft2 factory extension the XPS not only gives high thermal performance, it is unaff ected by freeze/thaw cycles and is highly resistant to water absorption. Two grades of XPS were supplied to account for varying load requirements. Point loading of the ﬁ xed storage racking feet was120kN/m2, while under the central concrete base the more resilient XPS300 was installed for the long-term pressure of 140kPA. Sundolitt XPS is manufactured in a range of thicknesses and sheet sizes, the primary range having a thermal conductivity of 0.033 – 0.037W/mK (EN 13164) and according to use achieves outstanding compression resistance from 200 – 700kPa (EN826). Further information: Sundolitt (tel: 01786 471586; email: [email protected]; web: www.sundolitt.co.uk)

Adopting 3D BIM software opens new vistas Mitchellson Formwork has opted to use Tekla Structures 3D modelling software from Trimble to bring time and cost effi ciencies throughout the business and future projects. In recognising many of its customers and partners are adopting and benefi ting from 3D modelling and BIM, Mitchellson Formwork knew they had to adapt and lead with it. Tasked initially to trial Tekla Structures, Caroline Kilbane, BIM Manager at Mitchellson Formwork, said, ‘The fi rst thing that struck me was how quickly you could open, amend and save fi les - much simpler and more stable than other 3D programmes as the Tekla fi les are smaller and easier to use and manage.’ Tekla off ers a purpose-built solution for concrete construction. It can be used at any stage of the design and construction process and features easy-to-use tools to create data-rich concrete models that behave like real concrete. Because all information is created in 3D, it is easier to adapt to changes in design and visualise the building in the model both before and during construction. Further information: Tekla (web: www.tekla.com/uk) Mitchellson Formwork (web: w ww.mitchellson.co.uk)

WANT TO ADVERTISE YOUR PRODUCTS & SERVICES?

The Structural Engineer is the offi cial publication of IStructE and is the Institution’s principal means of communicating with its members. It combines news, opinion and the latest technical knowledge in structural engineering reaching 17,000 readers.

Contact Greg Lee on 020 7880 7633 or email [email protected] TheStructuralEngineer The ﬂagship publication of The Institution of Structural Engineers

P&S_Mar16.indd 107 19/02/2016 16:09 › SERVICES DIRECTORY

Telephone: +44 (0) 20 7880 7633 Email: [email protected] 108 TheStructuralEngineer March 2016

ANALYSIS & DESIGN STRUCTURAL DESIGN & CAD DRAWING

TONY DOYLE ASSOCIATES LTD COMPUTER AIDED DESIGN & DRAWING

A u t o C A D D r a u g h t i n g S e r v i c e s

ANALYSIS & DESIGN We are a team of professional technicians with ofÀ ces in the UK À STRAP software for analysis and and Ireland providing an ef cient, design of cold formed, concrete, steel on-demand service for our clients and pre & post-tensioned structures. - RC Detailing Specialists - Fine GEO5 geotechincal software for · soil structure interaction, slope stability, - Structural Drawing GA s & Details - earth pressure, sheeting & pile design. - Bespoke Drawings for all Applications - Trial downloads at www.sigma-x.net UK: +44 (0) 203 603 1442 Ireland: +353 (0) 86 85 45 425 Tel: 01934 808468 www.sigma-x.net www.tonydoyleassociates.co.uk

SOFTWARE ANALYSIS & DESIGN

QUIKPORT - MULTI SPAN + QUIKJOINT - COMPREHENSIVE £2695.00 Get the job done! Visit www.quiksoft.co.uk and download your free trial email: [email protected] call: 08456 250 280

REINFORCED CONCRETE DETAILING

R.C. Detailing Specialists WANT TO ADVERTISE YOUR Established 1980 Extensive experience PRODUCTS & SERVICES? Proven reliability Large capacity

Paul Benham Partnership Ltd Contact Greg Lee on 020 7880 7633 or email [email protected] Civil & Structural Engineering Draughting Service [email protected] www.paulbenhams.co.uk 01273 730956 South TheStructuralEngineer Coast & London OfÀ ces The ﬂagship publication of The Institution of Structural Engineers

DirectoryMar16.indd 108 19/02/2016 16:11 www.thestructuralengineer.org

further your career

Senior Structural Engineer Principal Structural Engineer Associate Godalming Cambridge Central London £45,000-£50,000 Plus Benefits £48,000-£52,000 Plus Benefits £60,000-£65,000 Plus Benefits Our client is seeking a degree qualified Our client is an award winning engineering An Associate structural engineer is sought engineer with at least 5 years’ experience practice who offer innovative and creative by this award winning firm in Central for this challenging and diverse role. You design solutions. Their office in Cambridge London. Their main strength is their client will have worked on commercial and requires a principal structural engineer to relationships, focusing on a cohesive bond residential developments and have the take a lead role within the buildings between client and all staff with Directors ability to take a lead role on projects, structures team. Their orderbook consists being approachable and available for setting an example for more junior of a diverse project range from private every project. A gregarious and technically members of the team. Confident to work residential through to large international talented engineer, you will have a history of with autonomy as well as under the projects as well as high end retail. This successful project delivery and team guidance of a principal engineer, you dynamic firm recognise that hard work leadership. You would be expected to run possess excellent communication and deserves rewards, and you will be offered multi million pounds building structure client facing skills. Varied role requiring an an excellent benefits package. projects, running a team of about 3-5 other adaptable engineer with a positive attitude. engineers and 1-3 technicians. You must be technically competent and able to Associate Director represent our client at meetings with Chartered Structural Engineer Middlesbrough architects, clients and contractors. Hemel Hempstead Excellent package Opportunity to become a full Director £58,000-£65,000 Plus Benefits within this young and thriving business. Our client is a globally recognised This industry leading national developer engineering consultancy who work on has an exciting role for a Chartered pioneering international projects. Their Associate Director Structural Engineer within their Hemel multi-discipline office on the outskirts of Leeds Hempstead office. Their award winning Middlesbrough is looking for an £65,000-£70,000 Plus Benefits projects are innovative, and they have experienced Associate Director to help an ongoing workload of high quality bring on the business and grow the Our client is a leading multi-discipline residential and commercial schemes structures team. Candidates who have consultancy offering services across the nationwide. Residential experience is local connections would be preferred as globe. Their Leeds office is seeking an vital, as well as preparing project reports you will carry out a business development Associate Director to head all aspects of and risk assessments and therefore role, and your project experience should project delivery. A major part of the role communication skills must be excellent. include residential and commercial as well will be to grow the team and bring on This is a dynamic firm who offer as leisure and healthcare. An exciting junior engineers, assisting them achieve outstanding career prospects. opportunity for a Chartered or Incorporated Chartered status. You will possess at engineer to join this blue-chip organisation, least 10 years’ post Chartered leading major schemes whilst mentoring experience, having worked in two of the Senior Associate and heading your team. following sectors; property and Central London buildings, transportation, water, energy £75,000-£80,000 Plus Benefits and resources. The role offers a varied Associate Structural Engineer workload and candidates who have This international multi-discipline retained their technical ability would be consultancy prides itself on its award Bristol £55,000-£60,000 Plus Benefits preferred. Excellent business winning cost effective engineering development experience required. solutions. Their Central London team is Our client is seeking a Chartered engineer seeking a first class Senior Associate for who has a proven track record of taking a their structures team. A dynamic individual lead role at Associate level on buildings with the ability to lead a motivated team structures projects. Fully competent in all Find more jobs online at and establish new client relations, whilst types of design, you will be a confident conradconsulting.co.uk retaining long term existing clients. A team leader with a strong desire to achieve Chartered engineer who has retained their high results in your career. A client facing For more information about any hands-on design capability and now wants engineer with the ability to oversee all of these positions, please contact to focus on a more managerial role. aspects of the project with both enthusiasm [email protected] Excellent opportunity to join a forward and drive, confidently delivering within thinking global firm. client expectation. or, for a confidential chat, call Graham on 0203 1595 387

Suffolk 01728 726 120 | Leeds 01132 805 840 | Manchester 0161 209 3246

TSE Rec Mar16.indd 109 19/02/2016 16:12 TheStructuralEngineerJobs Register to receive latest jobs by email - visit www.thestructuralengineer.org/jobs. Telephone 020 7880 6212 Email [email protected] www.thestructuralengineer.org

Chartered Structural Engineer – Associate (Partnership Opportunity) We are a structural and civil engineering design consultancy with offices in Wilmslow and Burnley and seek an experienced Chartered Structural Engineer looking for fast track career progression. You must be self motivated and have the ability to manage both the technical, contractual and financial aspects of projects and have a “hands on” approach to design and the CDM Regulations 2015. Francis Bradshaw While based at our Wilmslow office you will also be working on various projects Partnership LLP throughout the UK in most construction sectors and be able to develop new business with a view to becoming a Partner within 2 years.

Senior/Principal Engineer We seek a Chartered Structural or Civil Engineer with at least 3 years post Chartered experience to be based at our Wilmslow oﬃce. You must be able to demonstrate a proven ability to run projects from inception to completion and communicate Please apply in writing eﬀectively with Clients, Architects and other professionals. with your CV to: The work varies from undertaking structural appraisals of existing buildings and preparing reports, design of domestic extensions and alterations, through to multi Francis Bradshaw million pound industrial, commercial and heavy engineering projects. You must Partnership, Bank be prepared to produce calculations and CAD drawings themselves and be fully Chambers, 4-6 Church conversant with the CDM Regulations 2015. Street, Wilmslow, Experience in seismic design would be an advantage but is not essential. Cheshire SK9 1AU.

www.thestructuralengineer.org/jobs

TSE Rec Mar16.indd 110 19/02/2016 16:12 www.thestructuralengineer.org

Chartered Senior Structural Engineer Central London Ref: 50865 Up to £54,000 + Beneﬁts 130 strong leading premier consultancy based in London’s West End has a requirement knowledge based for a Chartered Senior Structural Engineer to join ALFRISTON SCHOOL SWIMMING POOL their London studio as it continues to expand BELIEVE IN BETTER BUILDING recruitment in working on new-build and/or refurbishment C projects up to £100million. Candidates C structural engineering will need to be Chartered with IStructE and/or ICE and must have gained good design and project- consultancy running skills in the UK.

STAGE BY THE SEA Chartered Civil Infrastructure Engineers W Greater London Ref: Various Up to £55,000 + Benefits Walker Dendle Technical has several positions across Central & Greater London for ELLIOTT WOOD ENGENUITI & ARUP ASSOCIATES Chartered Civil Infrastructure Engineers in various premier, niche and mainstream consultancies. Candidates should be Chartered with ICE, Structural Associate be educated to MEng/MSc level, have Project Engineer Director good civil infrastructure design and Central London Ref: 50669 Central London Ref: 50629 project-running skills in roads Up to £47,500 + Benefits Up to £65,000 + Benefits & drainage and associated infrastructure works. Niche consultancy based in Southwark has Leading international premier rapidly- a requirement for a Structural Project Engineer EXPEDITION expanding consultancy has a requirement for to join the expanding London studio working on an Associate Director to join the London studio as INTESA SANPAOLA TOWER exciting new commissions with top Architects OTKRITIE ARENA (SPARTAK STADIUM) it continues its expansion working on both UK and helping to develop the brand. Candidates & international commissions. Candidates will W will need to be a Graduate or just need to have extensive post-chartership Chartered member of IStructE and/or (IStructE) experience in design, project ICE as well as being educated to and team-running consultancy MEng/MSc in Civil, Structural or combined with an affinity with Architectural Engineering. high-profile architecture.

CENTRAL LONDON STONE STAIR Façade Design/ C Project Engineers Central London Up to £54,000 + Benefits Walker Dendle Technical has several roles EXPEDITION & STUDIO OSSOLA AECOM for Façade Engineers up to Chartered level across Central London within the façade arms of high-profile and niche consultancies. Candidates Senior Structural 4No Structural will need to be educated to MEng/MSC level Engineers Design/Project Engineers in Façade Engineering (Architectural or Central London Ref: 50859 Central London Ref: 50866-69 Structural) and will have gained good façade design skills (project- Up to £52,500 + Benefits Up to £42,500 + Benefits running skills at more senior 90 strong leading niche consultancy based 170 strong leading premier consultancy level) in the UK or Europe. in Clerkenwell has a requirement for several WEBB YATES ENGINEERS based in London’s West End has a requirement Senior Structural Engineers to join various teams for 4 Structural Design/Project Engineers to as it continues to expand rapidly. Candidates CHICHESTER FESTIVAL THEATRE join various teams as it continues to grow SAN BERNARDINO JUSTICE CENTRE will need to be near or recently Chartered organically. Candidates will need to be a with IStructE and/or ICE, have a MEng/ Graduate member of IStructE and/or ICE, MSc in Civil, Structural or Architectural have a MEng/MSc in Civil, Structural or Engineering and will have worked Architectural Engineering and will in another London-based have gained good design skills premier/niche consultancy. in another UK consultancy.

STRUCTURAL AWARDS WINNERS 2015 For the ﬁfth year running Walker Dendle Technical we were a proud sponsor of Recruitment would like to congrat- The Structural Awards by IStructE ulate Expedition on their two winning and this year we sponsored the projects featured with a W and Elliott “Award for Community or Residential PRICE & MYERS Wood, Engenuiti & Webb Yates Engineers SKIDMORE OWINGS & MERRILL Structures”. Well done to all the winners for their commended projects with a C in their and see featured iconic projects by some categories at the Structural Awards 2015. For of this year’s successful nominees and the 9th year running we had a table for clients of Walker Dendle Technical the night with guests from Conisbee, T 020 8408 9971 Recruitment. Eckersley O’Callaghan, Engenuiti, Expedition, Price & Myers, E [email protected] Sinclair Johnston, Techniker & IMAGES SHOW RECENT PROJECTS Webb Yates Engineers. UNDERTAKEN BY SOME OF OUR CORE CLIENTS uualkerdendle.co.uk

TSE Rec Mar16.indd 111 19/02/2016 16:12 Analysis Concrete Steel

Composite Timber Masonry

Retaining Walls Drafting Lite Suite

p112_TSE.03.16.indd 112 19/02/2016 16:14