Continuous Development of Recovery Boiler Technology – 50 Years of Cooperation in Finland

SUOMEN SOODAKATTILAYHDISTYS FINNISH RECOVERY BOILER COMMITTEE

Continuous development of recovery boiler technology – 50 years of cooperation in Finland

PROCEEDINGS

INTERNATIONAL RECOVERY BOILER CONFERENCE Tampere Hall, Tampere, June 11, 2014 2 3 SUOMEN SOODAKATTILAYHDISTYS contents FINNISH RECOVERY BOILER COMMITTEE

50th ANNIVERSARY INTERNATIONAL Welcoming words 5 RECOVERY BOILER CONFERENCE Timo Merikallio

Tampere Hall, Tampere, June 11, 2014 Recovery boiler research highlights - 10 steps forward 7 Mikko Hupa CONFERENCE CHAIRMAN Esa Vakkilainen, Lappeenranta University of Technology Recovery boilers - history and future 19 PROGRAMME COMMITTEE Esa Vakkilainen Klaus Niemelä, VTT Mikko Hupa, Åbo Akademi University Understanding of black liquor sprays 33 Keijo Salmenoja, Andritz Oy Ari Kankkunen Esa Vakkilainen, Lappeenranta University of Technology Timo-Pekka Veijonen, Stora Enso Oyj Recovery boiler sootblowers - history and technological advances 47 Markus Nieminen, Pöyry Finland Oy Päivi Lampinen, Pöyry Finland Oy Honghi Tran and Danny Tandra

CONFERENCE MANAGEMENT Review of recovery boiler superheater material studies 61 Finnish Recovery Boiler Committee Martti Mäkipää and Pekka Pohjanne Markus Nieminen Päivi Lampinen Utilization of process historical data in recovery boilers 77 Song Won Park and Gustavo Matheus de Almeida Finnish Recovery Boiler Committee P.O. Box 4, Jaakonkatu 3 FI-01621 Vantaa XXL recovery boilers - how we developed them 95 FINLAND Kari Haaga Telephone + 358 10 3311 www.soodakattilayhdistys.fi High energy recovery boilers 111 Marja Heinola and Keijo Salmenoja Title: Continuous development of recovery boiler technology – 50 years of cooperation in Finland Editors: Esa Vakkilainen, Päivi Lampinen, Markus Nieminen CFD modeling of kraft recovery boilers - a retrospective 121 ISBN 978-952-93-3984-6 (hardcover) Andrew Jones ISBN 978-952-93-3985-3 (pdf) Entrained flow black liquor gasification - 127 © Copyright Finnish Recovery Boiler Committee, 2014 review of pilot scale research 2004-2014 All rights Reserved Rikard Gebart

Cover photos: Scenery of pulp mill at Stora Enso Oyj, Kaukopää, Imatra Tampere Hall, Tampere

2 3 welcoming words

timo merikallio metsä fibre oy

Recovery boiler co-operation in Finland is recovery boiler day and chief engineer day are celebrating its 50th anniversary this year. To important occasions for networking. celebrate the occasion this anniversary conference Today the FRBC has 23 members which include is arranged at the 2014 International Chemical pulp mills, recovery boiler manufacturers, a num- Recovery Conference (ICRC) in Tampere, ber of insurance companies, engineering compa- Finland. nies, research organisations and universities in Formal co-operation between Finnish recovery Finland. The FRBC publishes different guidelines, boiler users and manufacturers to solve problems recommends best practices and arranges meetings of the field was initiated in 1964 with jointly and conferences. Most of the active work is done sponsored recovery boiler corrosion study from in five subcommittees, which cover the most 1965 to 1968. A few years earlier several mills important areas around the recovery boiler. had experienced severe corrosion in furnace wall During the past ten years the FRBC has conducted tubes which had led to expensive maintenance two large publicly funded research programs. shutdowns. The first research project was partici- The Finnish Funding Agency for Technology and pated by all the Finnish sulphate pulp mills who Innovation (Tekes) was the main financier on both owned recovery boilers. This research project programs. The first program SOTU 2 (2003-2006) succeeded so well that the need to co-operate even concentrated on the possibilities to increase power more closely was recognised and other investiga- production from the recovery boilers. Second tions into various operational and constructional program SKYREC (2008-2012) developed the problems were added to the work programme. studies of SOTU and deepened the understanding Since the beginning, the Finnish Recovery Boiler needed in the design and operation of the high- Committee has been active in promoting safe, efficiency recovery boilers in the future. economic and environmentally friendly operation Today the future of pulp industry in Finland looks of recovery boilers and closely related processes. brighter than some years ago and pulp mills are The strength of recovery boiler co-operation in developed towards bioproduct mills with wider Finland has always been the open discussion on product portfolio and higher energy yield. This recovery boiler and related problems between also puts pressure on chemical recovery and people who have wide-ranging experience and recovery boilers. Therefore, the good research people with in-depth theoretical knowledge. The work on this field and co-operation between the main emphasis of the FRBC has been put on members of FRBC is just as important as it was research, although mutual exchange of experience 50 years ago. and distributing knowledge of recovery equipment and operation both at home and abroad has always It is a great pleasure to welcome you all to this an- been an important part in co-operation. Annual niversary conference. Learn, network and have fun.

4 5 Recovery Boiler Research Highlights – 10 Steps Forward

Mikko Hupa Åbo Akademi university Finland

Introduction

Kraft recovery boiler technology has advanced The 50-year anniversary seminar of the Finnish dramatically since the first installations in the Recovery Boiler Committee at the International late 1930s. The advancement is naturally visible Chemical Recovery Conference is an excellent in a number of technological aspects but maybe event to summarize some of this great research the most exciting development has dealt with and development work. This presentation aims the furnace process. A deeper and more detailed at a not-so-serious list of top-10 past research ef- understanding of the recovery boiler furnace forts, which have had a significant influence on process has been one key to a number of major recovery boiler design or operation. The focus is improvements of the technology. on research related to liquor combustion and the furnace process. This better understanding is a result of ingenious research efforts by talented people at a number The top-10 ranking is based on a longer initial of universities, research centres and industrial list of candidate research topics defined by this laboratories around the world. Today, the ar- author. (There is no reason to hide the fact that chived scientific literature contains around 8000 the topics selected do quite much reflect the papers dealing with black liquor. Around 1100 interests of the present author.) These suggested of the papers deal more specifically with black topics were then evaluated and ranked by a panel liquor combustion. The interest in black liquor consisted of four senior recovery boiler experts. combustion research has steadily increased since The four panel experts represent boiler users, the 1980’s. During the last years around 20 new boiler manufacturers and boiler consultants, papers have been published per year (Figure 1). and their total common experience of recovery boiler technology exceeds one hundred years. Two of the panel experts come from Europe, two from North America.

The panel was asked to comment the suggested research efforts and grade them on a scale from 1 to 5 according to the following: 5 = very significant/interesting/useful, 1 = only marginally interesting. The panel members did their evaluation independently of each other.

This presentation cannot be a comprehensive Figure 1. Number of scientific publications review of the history and details of all the re- dealing with black liquor combustion. Based search topics ranked. Each topic is only briefly on the Science Finder by April 2014. 6 7 described together with some comments from consists of a great number of studies in several a) the evaluation panel. In all cases reference is research groups, especially at the University of /hr)

made to a few literature sources by some of the 2 Toronto, Åbo Akademi University, Oregon State contributing research groups in the area. University and the Institute of Paper Science and Technology. Most of this advancement took The first five most significant place in the 1980s and early 1990s. research efforts This research has shown how the behaviour of

Two research topics received full grades from Gain (g/m Weight the dust particles and deposits is closely con- all the four panel members. Interestingly, both nected to the melting properties of the dust. The were efforts done already several decades ago. Figure 2. Effect of Cr content on corrosion complicated phenomenon of partial melting of Maybe not so surprisingly, they are connected resistance of steels during short term tests the dust was described by four characteristic to the two long-term challenges of the recovery at 400 °C in a simulated lower furnace gas temperatures based on the share of liquid phase b) boiler technology, corrosion and fouling. (Moberg et al. 1974). at that temperature (Backman et al. 1987, Tran 1997): Effort 1 (Shared): Lower furnace wall This research paved the road to replacing of

corrosion the furnace carbon steel tubes with composite • First melting temperature, T0 • Sticky temperature, T The first one was the “Research in the mecha- tubes. The composite tube consists of two alloys 15 • Flow temperature, or radical deformation nisms of lower furnace water wall corrosion metallurgically bonded together. The outer alloy temperature, T and development of composite tubes – al- – Cr containing stainless steel – withstands the 70 • Complete melting temperature, T lowing higher boiler pressures (1970s)”. This fireside corrosion, while the inner – carbon steel 100 - is the ordinary approved pressure vessel mate-

research was one of the first recovery boiler K in BLS (wt-%) rial. The first composite tubes were installed in These characteristic temperatures are strongly related research activities done as an organized dependent on the dust composition and have co-operation project between the boiler users 1972 in Sweden. By 1982 there were already 30 a direct connection to the fouling or corrosion Figure 3. Sticky temperature of the condensed and manufacturers. composite tube furnaces in Scandinavia. Today composite tubes are the standard solution in all properties of the dust in the boiler. The research dust. The history of this research goes back to the high-pressure boilers. showed how the characteristic temperatures for a) as function of the chlorine-to-alkali and end of the 1960s and 1970s when water/steam a given dust composition could be determined potassium-to-alkali ratio in the dust (Tran 1997) pressures of the new recovery boilers were As mentioned, the panel members all gave based on phase equilibrium diagrams, or later by b) as function of Cl and K contents in the black increased for higher electrical efficiency. In the full five points and had several positive using advanced thermodynamic equilibrium cal- liquor solids (Hupa 2004). these new installations the problem of furnace comments: culations. The well-cited Figure 3a shows how wall tube corrosion became aggressive. The the sticky temperature of the fine condensed dust “A great example of problem-oriented research corrosion was concentrated in the lower furnace is dependent on the content of chlorine and po- that finally solved the issues and new compound “Very fundamental, today the whole world and it was clearly connected to the higher tube tassium (Tran 1997). Based on some additional materials were launched as a result of the knows the terms T15 and T70” material temperatures. The Finnish and Swedish assumptions, the fume sticky temperature could project…” Recovery Boiler committees initiated a common be later related to the Cl and K content in the liquor dry solids directly. Figure 3b shows the same Effort 3. Single droplet combustion research program to study the chemical details “Vital” of the corrosion and to find solutions to the sticky temperature of the dust as function of the characterization Cl and K contents in the black liquor solids. problem. “This was a most important thing. With the The research, “Determining black liquor burn- composite tubes the furnace corrosion was ing properties (liquor burning stages, swelling The studies included short-term corrosion expo- The panel member comments showed the sig- controlled. This way the reliability and safety tendency) by single droplet tests (1980-90)” sure of a variety of steel qualities in synthetic nificance of this work: could be dramatically improved. When these was started at Åbo Akademi University and at environments simulating the lower furnace con- fundamental things were fixed we could focus on “Used every day!” the Institute of Paper Chemistry in Appleton, ditions with reduced sulphur gases present developing other things” Wisconsin (later IPST, Atlanta) in the early (Moberg et al. 1974). Figure 2 summarizes the “Superheater damages and incidents have de- 1980s. Black liquor burning behaviour could most important result from theses studies. This Effort 1 (shared): Dust chemistry and creased dramatically since the 1990s. The major in an effective way be studied by suspending type of corrosion, sulfidation, steeply increased melting reason has been the better understanding of the single liquor droplets on thin Platinum wire for all low alloy steels when the material tem- fly ash behaviour and chemistry…” and let them burn in a laboratory furnace under perature increased above some 280-300 °C. The second effort, which received full grades controlled conditions. This measuring system However, with an addition of sufficient amount from all the panel members, was described “To be able to calculate and predict fly ash appeared to be very suitable to for the relatively of chromium in the steel (above some 13%), this “Research on dust fouling behaviour and behaviour in different parts of the RB helps to large size (several mm) black liquor droplets, corrosion was practically completely inhibited the role of potassium and chlorine on dust design the boilers and optimize heat transfer which typically require 10-30 seconds to burn up to temperatures far higher than 400 °C. melting properties (1980-90)”. This research surfaces more cost-effectively.” 8 9 out. By recording the combustion process four techniques to study the fate of sulphur, nitrogen paper by some words concerning the computing Some challenges were also pointed out: burning stages could clearly be identified and and sodium during the combustion process. times: “Convergence is a major problem. - About studied: The panel members commented this research: three months of CPU time were needed for the “Char bed still a black hole in CFD…” base case on a MicroVAX II”. A few years later, • Drying “Introduction of single-droplet furnace was a in 1996, Tom Grace wrote the first review on “A drawback in CFD is that different tools give • Pyrolysis and devolatilization big step towards the understanding of different computer modelling of black liquor recovery different results (WYWIWYG = what you want is • Char burning stages in BL burning” furnaces (Grace 1996). what you get)” • Inorganic reactions “Knowledge about black liquor in-flight burn- Later, many other groups have contributed to Effort 5. High solids firing Figure 4 shows results produced with this ing behaviour from nozzle to char bed is vital this development including the University of Liquor solids content strongly influences the technique (Hupa et al. 1987). At the beginning to get proper combustion and char bed burning Toronto, Helsinki University of Technology (to- overall boiler efficiency. Consequently, the the technique was used to detect differences control” day Aalto University), VTT and Åbo Akademi evaporation plant technology has been in the in the burning properties of different liquors. University in Finland, the Royal Institute of “Fundamental research which also gave a basis focus of continuous improvement work since Especially the characteristic swelling of the Technology in Sweden and several company for CFD modeling.” the early days of the kraft recovery technology. burning droplet during the devolatilization stage groups. By now, some 200 papers concerning However, in the 1980s this development jumped was shown to vary significantly from liquor to numerical modelling of the recovery furnace One panel member further suggested additional to a new level when the technology allowed liquor (Frederick et al. 1991). This difference in have been published. Today, the worst bot- work on the root-causes of the different burning liquor solids contents to exceed 70%, and later the swelling tendency was shown to be signifi- tleneck is no longer the computing time. The behaviours (swelling, etc.). even 80%. cant from a practical point of view. At identical biggest challenges are connected to the detailed conditions the liquors with a heavy swelling ten- “Now when we are planning to burn lean black descriptions of the many physico-chemical pro- This “Research on development of high solids dency were shown to burn faster than the liquors liquors, this will be emphasized even more.” cesses taking place in the furnace, the so-called firing – for better furnace behaviour and lower that swell less. The swelling also influences the submodels to be included in the general flow SO emissions (1980s)” was initially strongly aerodynamic behaviour of the liquor droplets in 2 Effort 4. CFD based furnace modelling model. Such recovery boiler-specific processes driven by the boiler or evaporator manufacturing the furnace. that have required special development include: “Development of the CFD-based recovery fur- companies (Hyöty & Ojala 1987, Ryham 1989). Later many research organizations such as KCL, The single droplet technique has been thorough- nace modelling to simulate novel boiler designs • Liquor spraying (course droplets) Åbo Akademi University and VTT in Finland ly used to generate experimental information for (1990-2000)” has been a result of a long devel- • Liquor droplet burning behaviour (very differ- were involved in studying the many details and mathematical descriptions of the droplet burning opment process. Mathematical furnace models ent from any other fuels) effects of the high solids firing. process (Järvinen 2002). These droplet-burning based on computational fluid dynamics (CFD) • Fate of the molten inorganic chemicals re- models have been essential components in all have been developed since the early 1980s and maining after burn-out It was soon recognized that, besides the more general furnace process models (Mueller et a great number of research groups at universities • Formation/consumption of the char bed energy efficiency benefits, this “high solids al. 2004). Later the single droplet technique has and companies have contributed to the advance- • Gaseous chemistry of nitrogen and sulphur firing” changed the recovery boiler process been further developed with a variety of analysis ment of these modelling techniques. The main much more than expected. In fact, the changes development – including creation of several Detailed calculation models to predict the gas in the furnace and flue gas operation were commercial CFD codes – was initially focused flows and combustion in recovery furnaces have quite dramatic. Liquor firing and bed control on furnaces fired with gas, oil or pulverized coal. been a very important contributor to the devel- became significantly easier. However, the most However, black liquor recovery boilers have opment of the recovery technology. The models remarkable effect of the high solids firing was required a lot of additional creative development are today used as a standard tool when new the practically complete elimination of the SO due to the many specific processes that differ boiler design features are considered, or when 2 emission in boilers when liquor solids content significantly from the conventional combustion operational problems in existing boilers are exceeded some 73-75%. This effect was shown systems. addressed. The courageous dramatic scale-up of to be connected to the higher temperatures in the the boiler sizes that has taken place during the lower furnace caused by the smaller amount of This work, specific to recovery boilers, was very recent years would not have been possible water entering the furnace along with the liquor. started already in the late 1980s. The two first without thorough model studies using advanced The higher furnace temperature resulted in a publications on recovery boiler furnace model- CFD models. Figure 4. Average diameter and carbon dioxide ling were presented at the International Chemical dramatically better capture of sulphur dioxide formation (“carbon release”) of a black liquor Recovery Conference (ICRC) in 1989. One Panel comments: by sodium vapours released during combustion. droplet during combustion. The black dots originated in research from the Institute of Paper The elimination of SO from the flue gases indicate observations based on the simultane- Chemistry (today IPST), and the other was a joint “CFD has been of great help to the boiler manu- 2 caused further changes in the furnace and ous image recording (appearance of the flame, paper by Tampere University of Technology facturers…has helped to optimize air system and flue gas chemistry. The dust particles became maximum swelling, end of devolatilization, end together with the boiler manufacturer Tampella boiler dimensions” alkaline and sticky dust problems, caused by of char burning) (Hupa et al 1987, Fredrick et at that time (today Valmet) (Grace et al. 1989, “A tool to allow planning of bigger and bigger formation of acidic sulphates, disappeared. Also al. 1991). Uppstu et al. 1989). Tom Grace concluded his boilers” 10 11 the fate of chlorine changed, in high solids fir- of the black liquor nitrogen, typically 20-30%, The panel members also mentioned the unre- the first melting point, and at these conditions ing all chlorine stays in the solid dust and any was released as ammonia. This ammonia very solved challenge of the quantitative prediction the corrosion criteria are less well under control. purge of chlorine in the form of HCl in flue gases readily formed NO when the volatile gases were of NO emissions: Relevant new research results concerning the stopped. Interestingly, it was later shown that the oxidized in the upper furnace. corrosion effects of alkali chlorides have also high solids firing could well be realized without “Some NOx chemistry has been incorporated been obtained by groups working with super- any significant increase in the NOx emissions. The truly surprising finding was that some part, into CFD models, but still an unresolved area in heater corrosion in boilers fired with solid bio- typically 20-40% of the black liquor nitrogen black liquor combustion chemistry…” fuels. In these boilers the key compound causing The panel comments were unanimous: could end up dissolved in the smelt. This nitro- corrosion is potassium chloride. The corrosion “…models are not able to predict N behavior in gen was identified as sodium cyanate, NaOCN. mechanisms of the two alkali chlorides appear the furnace” “High dry solids firing has been one of the major Some of the first mill measurements showed that to be quite similar (Enestam et al. 2013). The development steps in RB history” the distribution of the nitrogen into NO or smelt very first step in the corrosion of stainless steels cyanate can be strongly influenced by boiler Effort 7. Alkali chloride induced corrosion by alkali chlorides appears to be formation of “It solved the SO emissions, increased the ef- 2 operation (Saviharju & Aho 2006). The maximum steam temperatures in recovery alkali chromates, which causes the break-up of ficiency of the boiler, etc.” boilers are just above 500 °C. This is significantly the protective chromium oxide layer (Petterson In the dissolving tank and the green and white “The dramatic reduction in SO emissions also lower than in boilers burning conventional fuels; et al. 2005). 2 liquors this cyanate gradually converts to ammo- in these boilers the steam temperatures can be forced us to try to understand the Na-S chemistry nia. This ammonia was shown to be a potential during combustion. Nobody understood why SO above 600 °C. The lower steam temperature The understanding of the corrosion mechanisms 2 source of uncontrolled emissions – depending results in lower power production efficiency. has increased significantly. With this knowledge disappeared from flue gases with high dry solids on the details of the recovery process of the mill. firing.” Consequently, there is a great interest in the pos- the reasons for corrosion patterns in boilers can in most cases be explained. However, the re- Interestingly, the European panel members sibilities of increasing the steam temperatures in search has not yet produced any direct solutions Interestingly, one panel member questioned if, graded this research clearly higher than the recovery boilers. The higher steam temperatures to allow clearly higher steam temperatures in an or to what extent, this development was a result North American members. have been limited by high-temperature corrosion of targeted research efforts: induced by the fireside deposits containing alkali alkali chloride environment. “NOx chemistry is an important topic, since chlorides. Also in this research the European evaluation “Significant thing for the control of sulphur stringent emission limits are also applied to was more positive than the North American: emissions. However, it probably had advanced recovery boilers and combustion control is still “Research on alkali chloride induced superheater gradually along with practical experiences as the only BAT (Best Available Technology)” corrosion mechanisms to pave the road for high- well.” er steam temperatures (2000-2010)” has been a “Important to the control of superheater cor- “Complicated chemistry and kinetics play a topic for research for long and many research rosion and increase of the power production. Results can be immediately applied.“ The next five highest-ranking major role in NOx formation.” groups have contributed with various studies. Oak Ridge National Laboratory, Paprican (today research efforts “Important research. Based on this the NOx “This together with the ability to predict fly ash FPInnovations), VTT, University of Toronto and emissions could be predicted better than before. behaviour is essential in increasing the power Effort 6. NOx formation Åbo Akademi University have been the most ac- The results were directly applicable in practice.” output from recovery boilers.” tive ones. Many research techniques have been Formation mechanisms of nitrogen oxide emis- applied. Laboratory exposure furnaces have sions in black liquor combustion were shown to Some questions were also raised: been used to systematically study the influence differ from combustion of other fuels in many of various factors on the salt-induced corrosion. respects. The “Research on the fate of BL nitro- “Why is 515 °C still the upper limit for steam?” Air-cooled probes have been used to test materi- gen in combustion and theoretical basis for in- als in the superheater section of boilers. “Why do austenitic steels behave well in super- furnace NOx reduction strategies (2000)” started heaters, but low-alloyed steels do not?” already in mid 1990s but became very active in Recent work has come up with a number of the 2000s. interesting new findings. Even extremely small Effort 8. Acid sulphates amounts of chloride in the alkali salt deposit has Even if the nitrogen content in the liquors is quite been shown to dramatically initiate corrosion on The formation of acid sulphates may make the low, typically around 0.1%, it appeared that the a variety of steel qualities (Skrifvars et al. 2008). flue gas dust sticky and corrosive in the cooler absolute majority of the nitrogen oxides, NOx, The deposit first melting temperature has been part of the flue gas channel. Acid sulphate for- formed in a recovery furnace originated from shown to give the upper limit to the material mation is a well recognized problem in boilers that. The reaction pathways of this black liquor temperature in the superheater, which has given with a high SO2 content in the flue gases. The nitrogen in the combustion became a highly inter- one useful guideline for the design of the mate- research here entitled “Exploring the factors esting and demanding research topic. The single rial temperatures in the superheater. governing the formation of acid sulphate in the droplet burning tests at Åbo Akademi University Figure 5. Reactions of the black liquor nitrogen However, the presence of chlorides may cause dust for control of sticky ash (1980-90)” was ini- shed light on the issue (Aho et al. 1994). It was during combustion of black liquor droplets (Aho corrosion also at material temperatures below tiated already in the late 1970s in Sweden by the shown that, during the devolatilization, a part et al. 1994, Forssén et al., 2000). 12 13 Swedish Pulp and Paper Research Institute STFI Helsinki University of Technology (now Aalto (today Innventia) and ÅF, and in Finland by the University) have been the pioneers (Empie et al. boiler manufacturer Tampella (today Valmet). 1995, Kankkunen & Helpiö 1995). Sprays of real The initial work was connected to boilers burn- black liquor are an extremely challenging topic ing sodium sulphite liquor. In these sodium to study experimentally. The liquor is very dif-

sulphite recovery boilers the SO2 content in the ferent from all other liquid fuels. It is extremely flue gases is very high and the acid sulphate viscous and at the presently common high solids formation was recognized first. Later, the acid contents it is strongly non-Newtonian. For sulphates were found in kraft boilers as well. pumping and spraying it has to be preheated to Run 1 Run 2 Run 3 6.1 kg/s, 112 °C Åbo Akademi University and the University of a high temperature. The relevant droplet size of 2.3 kg/s, 111 °C 3.6 kg/s, 111 °C Toronto continued the initial work, which gradu- the liquor in practical boiler furnace operation is ally revealed the mechanisms of acid sulphate quite large, several millimetres. formation and factors affecting the formation (Backman et al. 1984, Poon et al, 1993). Different experimental techniques have been developed to characterize real black liquor The sticky ash is formed when low melting sprays by advanced video recording etc. Tests

sodium bisulphate NaHSO4 is formed on the have been performed in special spray chambers surface of the dust particles. The bisulphate is but recently also in furnaces (Miikkulainen et Run 2 Run 4 Run 5 formed via a reaction with sulphur trioxide, SO3 al. 2008). Significant new information has been (and water) in the flue gases. Sulphur trioxide, produced on factors affecting spray behaviour 3.6 kg/s, 111 °C 3.6 kg/s, 120 °C 3.6 kg/s, 130 °C again, may form when some small fraction of and droplet formation mechanisms. Droplet size Figure 6. Characterization of black liquor sprays under different flow velocities and liquor tempera-

sulphur dioxide, SO2, reacts with oxygen in the and velocity distributions have been mapped un- tures (Kankkunen & Helpiö 1995, Miikkulainen et al. 2008). cooling flue gases. Bisulphate is easily detected der a variety of conditions (Figure 6). Attempts by a lowered “dust pH”, i.e. pH of a water solu- are also made to further understand the liquor key publications came from the University of In spite of the creative research efforts made, the tion with a few per cent of dissolved dust. behaviour in the nozzles by advanced flow mod- Toronto. Later, among others Åbo Akademi boron technology has found only limited use in elling (Levesque et al. 2004). University, the Oregon State University and the practice. The panel also noted this: For this particular research the North American University of Luleå made a number of studies “This is a topic that only a limited number of panel members gave clearly higher grades: The panellists gave somewhat mixed comments: around boron autocaustization. mills were interested in.” The boron-based autocaustization implies “Very significant…used every day” “Black liquor spraying has been studied since “… did not actually solve the bottleneck prob- adding in the black liquor sodium metaborate, the 1990s and it is an important topic still.” lems at the mills.” The European panellists were more moderate NaBO2, which reacts with the smelt sodium and saw this research as mostly relevant to “One of the key issues for successful combustion carbonate forming a more sodium rich borate. “It lived for a while, but the interest died quite sulphite boilers only: of black liquor, especially in large furnaces.” When dissolved in the dissolving tank this borate rapidly.” directly forms sodium hydroxide and converts “Liquor droplet formation has been extensively “This is somewhat important, but since sulphite back to the metaborate. The metaborate is fully “Limited application.” studied, with good results but just small impact recovery boilers are an endangered species this soluble and will follow the recovery solutions all to the practice.” topic is slightly out-dated.” the way to the digester and further to the black Concluding Remarks liquor and back to the boiler. This way the borate “It gives, however, a nice overview of fly ash The development of the recovery boiler technol- Effort 10. Partial autocaustization is recycled and used again, and new borate is chemistry and supplements the story of Na-S ogy has been impressive. Modern boilers are only needed to cover any chemical losses from chemistry” Many ideas have been suggested to change the remarkably clean, reliable and – huge in size. recovery boiler process in such a way that the the cycle (Figure 7). Important contributors to this development are “Interesting issue from the dew-point chemistry causticizing step could be partly or completely the many long-term research efforts by a number The research included laboratory experiments point of view.” eliminated. This can be achieved with a suitable of people all over the world. This paper wanted and thermodynamic modelling of the exact added substance, which changes the smelt com- to salute this successful research by reviewing “Limited applicability to sulphite boilers” stoichiometry and yield of the smelt-borate position in such a way that no or less carbonate ten examples of efforts that have significantly reactions at different conditions (Lindberg & is formed (Jansson 1997). helped to improve the design and operation of Backman 2004). It also included studies on the Effort 9. Droplet size measurements the recovery furnace process. A common feature In the 2000s there was a great interest in the potential changes in the boiler dust composition In the “Experimental characterization of with the examples qualifying to this “top-10” use of boron compounds to produce such an and properties caused by the boron addition liquor spray droplet size distributions and ve- list was that for all of them it took many years autocausticizing effect. The research in “Partial (Hupa et al. 2002). Finally, the partial auto- locities – basis for better spraying technologies – typically a decade - of active work before they autocaustization using boron (2000)” was initi- caustization was tested and applied in full scale (2000–2010)” researchers from the IPST and finally produced results which could be used in in several boilers (Ellis & Kochesfahani, 2005) ated by the U.S. Borax Company and the first the industry. 14 15 TopAnalytica, UPM, Valmet Power, and Tekes 11. Hupa, M., Forssén, M., Backman, R., Stubbs, A. of potassium. Oxidation of Metals Vol. 64, No. 1-2, – the Finnish Funding Agency for Technology , Bolton, R., 2002, Fireside behavior of black liquors pp. 23–41. and Innovation within the consortium project containing boron. TAPPI Journal , Vol. 1, No. 1, pp. 24. Poon, W., Barham, D., Tran, H.N., 1993, FUSEC. The research partners in FUSEC, Aalto 49-52. Formation of acidic sulfates and its impact on recov- University, Tampere University of Technology 12. Hupa, M., 2004, Research highlights in recovery boiler fouling and corrosion. TAPPI Journal, Vol. and Lappeenranta University of Technology ery boiler chemistry. Proceedings: International 76, No. 7, pp. 187-193. are also acknowledged. The fruitful long-term Recovery Boiler Conference “40 Years Recovery 25. Ryham, R., 1989, High solids evaporation through collaboration with the Finnish Recovery Boiler Boiler Co-operation in Finland”, Finnish Recovery thermal polymerization of black liquor. Proceedings: Boiler Committee. Committee has been of great importance to our International Chemical Recovery Conference 1989, Figure 7. Partial autocaustization using boron research at Åbo Akademi. 13. Hupa, M., Solin, P., Hyöty, P., 1987, Combustion TAPPI Press. additions (Tran et al. 1998). behavior of black liquor droplets, Journal of Pulp and 26. Saviharju, K., Aho, K., 2006, Nitrogen bal- Paper Science, Vol. 13, No. 2, pp. J67-72. ances in recovery boilers. Proceedings: 2006 TAPPI References 14. Hyöty, P., Ojala, S., 1987, Super combustion of Engineering, Pulping, & Environmental Conference, Besides the ten examples described above, black liquor. Proceedings: 1987 TAPPI Engineering TAPPI Press. 1. Aho, K., Hupa, M., Vakkilainen, E., 1994, Fuel several other suggestions received support from Conference, TAPPI Press pp. 49-53. the evaluation panel: nitrogen release during black liquor pyrolysis, Part 27. Skrifvars, B-J., Backman, R., Hupa, M.; I: Laboratory measurements at different conditions. 15. Järvinen, M., 2002, Numerical modeling of the Salmenoja, K., Vakkilainen, E., 2008, Corrosion of • Research on char bed cool-down mechanisms TAPPI Journal Vol. 77, No. 2, pp. 121-127. drying, devolatilization and char conversion pro- superheater steel materials under alkali salt deposits, cesses of black liquor droplets, PhD Thesis, Helsinki Part 1: The effect of salt deposit composition and and means to accelerate cool-down. 2. Backman, R., Hupa, M., Hyöty, P., 1984, Corrosion University of Technology (now: Aalto University), temperature. Corrosion Science Vol., 50, No. 5, pp. relating to acidic sulfates in kraft and sodium sulfite • Removal mechanisms of soot blowing, and Finland. 1274-1282. the development of high-efficiency soot recovery boilers, TAPPI Journal, Vol. 67, No. 12. pp. 16. Jansson, J., 1979, The use of unconventional alkali 28. Strang, W., Morgan, B., McBroom, R.B., Bair, blowers. 60-64. in cooking and bleaching, Part 5: Autocausticizing C.M., Tran, H., 1999, Partial autocausticizing of kraft 3. Backman, R., Hupa, M., Uppstu, E., 1987, • Research showing the role of the smelt freez- reactions, Paperi ja Puu – Paper and Timber, Vol. 61, smelt with sodium borates, Part 1: Effects on recov- Fouling and corrosion mechanisms in the recovery ing temperature on smelt viscosity and the No. 1, pp. 20-24. ery boiler performance. Proceedings: 1999 TAPPI boiler superheater area. TAPPI Journal, Vol. 70, No. formation of the so-called jelly-roll smelt. Engineering Conference. 6. pp. 123-127. 17. Kankkunen, A., Helpiö, T., 1995, Small scale • Research on sodium and sulphur release and measurement of black liquor spraying with splash 29. Tran, H., 1997, Upper furnace deposition and 4. Ellis, M., Kochesfahani, S., 2005, Partial borate dust formation, the historical interrupted com- plate nozzles. Proceedings: 1995 TAPPI Engineering plugging. in Adams, T., Frederick, J, Grace, T., Hupa, autocausticizing technology mill trial experiences. bustion tests at Rauma and Kaskinen showing Conference, TAPPI Press, pp. 207-214, M., Iisa, K., Jones, A., Tran, H. (Eds.), Kraft Recovery Proceedings: 2005 TAPPI Engineering, Pulping and that most of the fume is formed during in-flight Boilers, TAPPI Press Environmental Conference, TAPPI Press. 18. Levesque, D., Fard, M., Morrison, S., 2004, and only a minor part is coming from the char Understanding the effect of black liquor proper- 30. Tran, H., Mao, X., Cameron, J., Bair, C.M., 5. Enestam, S., Bankiewicz, D., Tuiremo, J., Mäkelä, bed. ties and splash plate nozzle configuration on spray 1998, Autocausticizing of smelt with sodium borates. K., Hupa, M., 2013, Are NaCl and KCl equally corro- • Composite tube cracking – a major problem characteristics. Proceedings: 2004 International Proceedings: 1998 International Chemical Recovery sive on superheater materials of steam boilers? Fuel which now has been cured by the use of cor- Chemical Recovery Conference, TAPPI Press. Conference, TAPPI Press. Vol. 104, pp. 294-306. rect materials. 19. Lindberg, D., Backman, R., 2004, Effect of 31. Uppstu, E., Hyöty, P., Karvinen, R., Siiskonen, P., 6. Empie, H.J., Lien, S.J., Adams, T.N., Yang, W., temperature and boron contents on the autocausticiz- 1989, Alternative air supply system for recovery boil- 1995, Spraying characteristics of commercial black ing reactions in sodium carbonate/borate mixtures. ers. Proceedings: International Chemical Recovery Acknowledgements liquor nozzles. TAPPI Journal, Vol. 78, pp. 121-128. Industrial & Engineering Chemistry Research , Vol. Conference 1989, TAPPI Press. Special thanks go to the four-person panel for 7. Forssén, M., Kilpinen, P., Hupa M., 2000, NOx 43, No. 20, pp. 6285-6291. reduction in black liquor combustion – Reaction volunteering to do the ranking of the efforts 20. Miikkulainen, P., Kankkunen, A., Fogelholm, mechanisms reveal novel operational strategy op- presented. C-J., Heikkilä, V-P., Raukola, A., 2008, Black liquor tions. TAPPI Journal Vol. 83, No. 6. spraying six splashplate nozzles in two recovery boil- A great number of colleagues around the world 8. Frederick, W.J., Noopila, T., Hupa, M., 1991, ers. TAPPI Journal, Vol. 7, No. 10. pp. 23-27. Swelling of spent pulping liquor droplets during have been contributing to the ideas and efforts 21. Moberg, O., Ahlers, P.E., Dahl, L.,1974, Recovery combustion. Journal of Pulp and Paper Science, Vol. presented in this review. At Åbo Akademi the boiler corrosion. Proceedings: Pulp and Paper 17, No. 5, pp. J164-J170. thanks go to the black liquor research team, es- Industry Corrosion Problems, National Association pecially to Niko DeMartini, Anders Brink, Patrik 9. Grace, T., 1996, A critical review of computer of Corrosion Engineers pp. 125-136. modeling of kraft recovery boilers. TAPPI Journal, Yrjas, Markus Engblom, Oskar Karlstrom, Emil 22. Mueller, C., Eklund, K., Forssén, M., Hupa, M., Vol. 79, No. 7, pp. 182-190. Vainio and Niklas Vähä-Savo. 2004, Influence of liquor-to-liquor differences on re- 10. Grace, T., Walsh, A., Jones, A., Sumnicht, D., covery furnace processes - A CFD study. Proceedings: The recent fundamental combustion research Farrington, T., 1989, A three-dimensional mathemati- International Chemical Recovery Conference 2004, cal model of the kraft recovery furnace. Proceedings: at Åbo Akademi has been made possible by pp. 720-738, TAPPI Press. International Chemical Recovery Conference 1989, 23. Pettersson, J., Asteman, H., Svensson, J-E., the support of the companies Andritz, Clyde TAPPI Press. Bergemann, Foster Wheeler, International Paper, Johansson, L-G. 2005, KCl induced corrosion of a 304-type austenitic stainless steel at 600 °C; the role 16 17 recovery boilers - history and future

esa k. vakkilainen lappeenranta university of technology FINLAND

Introduction

Recovery boilers are built all over the world. Recovery of pulping chemicals could be based The roots of recovery technology are longer to French chemist Nicholas LeBlanc’s process than the roots of recovery boilers. But it wasn’t for producing soda at reducing furnace. A flame until the invention of recovery boilers before the oven was hand filled with black liquor, Figure Second World War that the pulping technology 1. After that, the black liquor was dried with was revolutionalized. This led to long develop- flue gases from burning wood. The dried black ment of essentially the same type of equipment, liquor was then scraped to floor, collected and culminating into units that are largest biofuel sent to separate smelt pot, Figure 2, for reduction boilers in the world. and burning the remaining organics (Rydholm 1965). Recovery of chemicals with this type Early recovery technology concentrated on of system was inefficient. Chemicals recovery chemical recovery as chemicals cost money hardly exceeded 60% (Whitney 1968). and if one could recycle these chemicals then the profitability of pulp manufacture would improve. For pulp mills the significance of electricity generation from the recovery boiler was for long secondary. The most important design criterion for the recovery boiler was a high availability. The electricity generation in recovery boiler process can be increased by elevated main steam pressure and temperature or by higher Figure 1. Early flame oven from late 1800s black liquor dry solids as well as improving its (Edling 1981). steam cycle. This has been done in the modern Scandinavian units.

Early recovery technology Early recovery technology concentrated on chemical recovery (Deeley & Kirkby 1967). Chemicals cost money and it was easy to discover that recycling these chemicals would improve the profitability of pulp manufacture Figure 2. Early smelt pot from late 1800s (Vakkilainen 2005). (Edling 1981).

18 19 Hand operated recovery grew more complicated the major operating problems. The combustion First recovery boilers boiler could produce 5000-5300 kg/ADt steam with additional heat recovery surfaces. Pre- was often conducted at very high air ratio lead- when rotating furnace could produce only 4500 The modern recovery boiler has a few strong evaporation and scrubbing in a rotary device ing to inefficient energy use. One could generate kg/ADt. The most modern boilers produce 4500 ideas that have remained unchanged until today. was invented by Adolph W. Waern (Combustion 3000-4000 kg of 3.0 MPa steam for each ton of kg/ADt (Fernandez 2001, Antônio 2011). The It was the first recovery equipment type where all Engineering 1949). The direct contact evapora- pulp (Roschier 1952). early recovery boilers had net efficiency of processes occurred in a single vessel. The drying, tor improved the heat economy of the recovery about 70% (LHV). combustion and subsequent reactions of black system. The hand feeding operation was soon The main recovery equipment itself remained liquor all occur inside a cooled furnace. This is replaced by rotative oven, Figure 3. unchanged, but details were improved on. Smelt the main idea in Tomlinson’s work. Secondly, The second early pioneer, Combustion dissolving tank was introduced, final smelting the combustion is aided by spraying the black Engineering based its recovery boiler design on was improved on and capacities grew. The use liquor into small droplets. Controlling process the pioneering work of William M. Cary, who of refractory and rotative oven tended to limit by changing spray proved easy. Stationary in 1926 designed three furnaces to operate with the recovery capacity to 70-75 tds/d (Sebbas et spraying was used in early rotary furnaces and direct liquor spraying and on work by Adolph al. 1983). Rotary part lengths were 7-10 m and with some success adapted to stationary furnace W. Waern and his recovery units. The first CE diameter about 1.5 m (Swartz & MacDonald by H. K. Moore (Toivanen 2012). Thirdly one recovery unit, looks a lot like a modern recovery 1962). can control the charbed by having primary air boiler. level at charbed surface and more levels above. Figure 3. Rotating oven from 1890, liquor in at The boilers parts were improved on. In 1930 Multiple level air system was introduced by C. Recovery boiler were soon licensed and pro- 20% ds (Edling 1981). seven LaMont type forced circulation units L. Wagner. duced in Scandinavia and Japan. These boilers were built. The use of rotary furnaces pinnacled were built by local manufacturers from drawings and with instructions from licensors. One of the in Murray-Waern type units which were suc- Recovery boiler also improved the smelt re- Use of rotary oven improved the heat economy. early Scandinavian Tomlinson units employed a cessfully built around the world. In these the moval. It is removed directly from the furnace Then it was a small step to introduce heat 8.0 m high furnace that had 2.8*4.1 m furnace rotary precombustion was combined with totally through smelt spouts into a dissolving tank. recovery equipment as was done with other bottom which expanded to 4.0*4.1 m at super- water-cooled furnace with lower part refractory Some of the first recovery units employed the types of boilers at that time. In 1912 the S-S heater entrance (Pettersson 1983). This unit lined. The Murray-Waern recovery units were use of Cottrell’s electrostatic precipitator for system (Sundblad-Sandberg) was taken online stopped production for every weekend. In the popular until the fifties. dust recovery. at Skutskär. In this system, liquor was sprayed beginning economizers had to be water washed into rotary furnace at 50% ds. The evaporation twice every day, but after installation of shot took place in a four stage evaporator. The heat An epitome of inventiveness of that age was the After several attempts and learning from Godell recovery unit at Stevens Point Wise in sootblowing in the late 1940s the economizers was recovered with vertical tube boiler. co-operation with Wagner when they built a could be cleaned at the regular weekend stop. 1940s. There the liquor was completely dried stationary recovery furnace at the soda pulp mill in a chamber after the boiler (Edling 1983). The construction utilized was very successful. Tampella was among the first manufacturers to of Howard Smith Paper Mills Ltd., in Cornwall, But that unit and most of those other alternative One of the early Scandinavian boilers 160 t/day build S-S type furnaces, Figure 4. Preventing un- Ontario, Tomlinson built and put into service systems were hard to operate and did not achieve at Korsnäs, operated still almost 50 years later necessary air flow through sealing arrangement his first experimental stationary furnace with a high availability. (Sanquist 1987). Edling states in 1937 that more between rotary drum and fixed parts was one of heat recovery boiler in 1929 (Steam 1992). This than 20 units had already been built of which 10 was soon followed by the world’s first recovery in Scandinavia. boiler unit with completely water cooled furnace at Windsor Mills which started operation on June 27, 1934 (Tomlinson 1975). After reverberatory Development of recovery boiler and rotating furnaces the recovery boiler was on technology its way (Jones 2004). Recovery boiler had two Spread of kraft recovery boilers was fast as substantial benefits. One could build more easily functioning chemical recovery gave kraft larger recovery boilers than rotating furnaces. pulping an economic edge over sulfite pulping The second was energy efficiency. (Boniface 1985). They had about 20% better energy efficiency as more than 5000 kg of In Finland Oulu Mills invested in the first 3.0 MPa steam for each ton of pulp could be five 120 tds/d recovery boilers in 1937. They generated (Roschier 1952, Alava 1955). The followed in the footsteps of Husum 1936 and first recovery boilers had horizontal evaporator Skutskär 1937 which both had invested in surface followed with superheaters and more similar recovery boilers. Kotka mills meanwhile evaporation, Figure 5-6. These boilers resem- invested in rotating furnaces. When the energy bled the state-of-the-art boilers of some 30 years efficiency of recovery boiler and rotating- fur earlier. This trend has continued until today. It is Figure 4. Early Tampella rotary furnace from about 1925 (Tampella). nace was investigated it was found that recovery easy to understand that when any stop will cost a 20 21 The first recovery boilers had severe problems B&W favored use typically a single black liquor faster startup and stop-down. The gas flow to with fouling (Deeley & Kirkby 1967, Roschier gun at front wall. In larger units additional gun the boiler bank is smoother and heating surface 1952). Tube spacing wide enough for normal was placed on back wall (Tomlinson & Richter arrangement is simple. The erection period is operation of a coal fired boiler had to be wider 1969). They preferred a significant part of the shorter because of large block construction. for recovery boilers. This gave satisfactory liquor to be sprayed to walls for drying. Boiler There is no rolled tube work. Enhanced and performance of about a week before a water bottom was in angle causing smelt to flow steady water circulation by separated and un- wash. Mechanical steam operated sootblowers quickly out. Hardly any space was reserved heated downcomers. The largest advantage is were also quickly adopted. To control chemical for smelt layer in the furnace. Thus this kind that single drum boilers can be made larger. Tube losses and lower the cost of purchased chemicals of furnace is named sloping bottom type. Final stiffness limits cross flow two drum arrangement electrostatic precipitators were added. Lowering black liquor evaporation was often carried in a to about 2300 tds/d size (Steam 1992). Vertical dust losses in flue gases has more than 60 years direct contact evaporator of venturi scrubber or flow twodrum constructions have suffered from of practice. One should also note the use of cyclone evaporator type. The highest practical plugging because of vibration stiffeners. electrostatic precipitators in an early Finnish black liquor solids was 60-65% depending on recovery boiler, Figure 6 (Roshier 1952). The air black liquor properties. Use of wall spraying was First recovery units had brick lined lower levels in recovery boilers soon standardized to promoted by B&W and its licensees Götaverken furnace with straight tubes forming cooling two. The primary air level was placed at the char and Babcock Hitachi. B&W adopted three level section behind bricks. This design persisted bed level and the secondary above the liquor air in the late 1960s. until the 1960s. Some of these units are still guns. operating today. Another design provided corro- Figure 5. First Tomlinson kraft recovery boiler Early on the CE design stressed use of multiple sion protection of furnace tubes with studs and with water cooled furnace from Babcock & In the first tens of years the furnace lining was guns in all walls (Tomlinson & Richter 1969). refractory. Some manufacturers use studs even Wilcox in 1934. Note spray tower using weak often of refractory brick or refractory on cast Boiler bottom was flat with space for smelt layer today, but the need of stud replacement has led black liquor before the ID fan. blocks. The flow of smelt on the walls causes on top of the whole floor. Thus this kind of fur- to decline in stud use. From late fifties onwards extensive replacement and soon designs that nace is named decanting floor type. Final black the membrane wall design took over, first with lot of money the adopted technology tends to be eliminated the use of refractory were developed. liquor evaporation was carried in a direct contact carbon steel walls. Tangent tube design was conservative. Conservatism meant that e.g. the The standard then became the tangent furnace evaporator of cascade evaporator type. The basic replaced with membrane design. The drawbacks new Oulu Oy, 100000 t/a sulphate mill installed wall. Membrane wall use became widespread in aims of recovery boiler design could soon be of tangent design were the difficulties in inspect- four Tomlinson boilers when it started operating the 1960s. summarized as; highest possible recovery of ing welds and doing maintenance work in 1937 (Oulu 1937). chemicals, high efficiency, high utilization of the calorific values in black liquor and highest safety First furnace walls were of carbon steel. With of operation (Hochmuth, 1953). CE sticked for a increasing design pressure there were several long time with a two level air system that had corrosion problems in lower furnace. The ad- corner fired secondary. They used similar system vantage of chrome containing alloys as wall in PCF-boilers. corrosion inhibitor was discovered as an answer to high pressure boiler sulfidation corrosion There are some early examples of single drum (Moberg 1974). In 1972 Tampella delivered first recovery boilers. Both B&W and Ahlstrom totally compound tube recovery boiler furnace delivered a single drum boiler in the late 1950s. to ASSI Lövholmen mill in Piteå, Sweden. By The first modern single drum recovery boiler 1982 there were 30 recovery boilers with AISI was delivered in 1984 by Götaverken to Leaf 304 compound tube bottoms in Scandinavia River at Hattiesburg, Mississippi. The boiler (Westerberg 1983). Use of composite tubing in size was 1966 tds/d. By 1990 all manufacturers United States started as late as in 1981. started providing single drum boilers. Excluding very small boilers, all modern boilers are now of Sanicro 38 is a widespread material that offers single drum design. improved corrosion protection for lower furnace. The first lower furnace made from Sanicro38 There are several advantages in a single drum was delivered by Valmet. They used Sanicro38 boiler. Single drum construction eliminates the in the lower furnace up to primary ports in 1994 possibility of water leakage to furnace as it is for their Rauma recovery boiler. placed outside the furnace. There are significantly less holes in a drum wall. Therefore it can be built thinner. Thinner wall of drum allows Figure 6. Early Finnish recovery boiler (Roshier 1952). 22 23 of vertical flow design the gas flows downwards run at Metsä-Botnia Kemi and Rosenlew, Pori The second generation air system targeted high and water counter currently upwards (Hyöty (Finland) recovery boilers (Hyöty & Ojala reduction. In 1954 CE moved their secondary air 1994). In a period of few years the current 1987). They noticed that above 75% dry solids from about 1 m below the liquor guns to about 2

long flow economizer design emerged, Figure the SO2 and H2S emissions were practically m above them (Llinares & Chapman 1989). The 7 (Moberg 1967). Vertical economizer design zero. Also reduction increased more than one air ratios and temperatures remained the same, spread fast in Scandinavia where by mid 1970’s percentage point. Other benefits listed were, but to increase mixing 50 m/s secondary air ve- more than a half of the recovery boilers had long that steam generation and boiler controllability locities were used. CE changed their frontwall/ flow economizers without direct contact evapo- increased. High dry solids require that ESP ash backwall secondary to tangential firing at that rator (Environmental 1976). is mixed to the black liquor with 62-65% liquor. time. In tangential air system the air nozzles are Higher retention time also improves the stability in the furnace corners. The preferred method In competition to purely vertical, the three pass of resulting black liquor. is to create a swirl of almost the total furnace design featured gas flow which was forced width. In large units the swirl caused left and crosswise the economizer tubes to improve heat Improving air systems right imbalance. This kind of air system with transfer. increased dry solids managed to increase lower Air system development continues and has been furnace temperatures and achieve reasonable continuing as long as recovery boilers existed There have been several rounds of economizer reduction. B&W had already adopted the three (Vakkilainen 1996). As soon as the target set for header designs. In a typical old design each level air by then. the air system has been met other new targets economizer tube is connected to a common large are given. Currently the new air systems have header. As maximum number of tube rows that At first the airport openings were made by bend- achieved low NOx, but are still working on with fit to this type header is about 8-10. The larger ing one tube away from the opening sideways lowering the fouling. economizers must have front and back headers. and making room for this by bending another This design has the disadvantage of having a tube back. Airport width was about tube spac- header in the gas flow. The header can corrode The first generation air system in the1940s and ing and large plate areas were needed to make and the welded joints tend to receive thermal 1950s consisted of a two level arrangement; airport gas tight. In 1978 CE began experiments stress. Modern economizers have flat horizontal primary air for maintaining reduction zone and with two level primary air. Upper primary was headers. secondary air below the liquor guns for final designed to about 20% of total air with velocity oxidation (Llinares & Chapman 1989). The up to 60 m/s. Total air split remained the same. recovery boiler size was 100-300 tds/d and The aim was to increase hearth temperatures. High dry solids black liquor concentration 45-55%. Frequently Dry solids at as fired black liquor was between to sustain combustion auxiliary fuel needed to Third generation air system was the three level 60 and 65% in Sweden in the beginning of the be fired. Primary air was 60-70% of total air with air. In Finland the use of three level air with 1960s (Jönsson 1961). In the beginning of 1950s secondary the rest. In all levels openings were primary and secondary below the liquor guns the typical as fired black liquor concentration was small and design velocities were 40-45 m/s. started about 1980. At the same time station- 50% (Vegeby 1961). The final concentration was Both air levels were operated at 150 oC. Liquor ary firing gained ground. Use of about 50% often done with cascade or cyclone evaporator. In gun or guns were oscillating. Main problems secondary seemed to give hot and stable lower practice the as fired dry solids could remain dan- were high carryover, plugging and low reduc- furnace (Westerberg 1983). Higher black liquor gerously low before refractometers started to be tion. But the main target, burning of black liquor solids 65-70% started to be in use. Hotter lower applied in late 1960s and early 1970s (Hellström could be done. furnace and improved reduction were reported. 1970). The only reasons seen for higher dry With three level air feed and higher dry solids solids were the energy economy and increase of the sulfur emissions could be kept in place. bottom loading (Vegeby 1961). One advantage Table 1. Development of air systems. noted was that partial load capability improved with higher dry solids. Increasing black liquor Air system Main target Additional targets dry solids from 60% to 68% enabled running 1st generation Stable burning of black liquor Figure 7. One of the first long-flow economizers, recovery boiler without auxiliary fuel firing at 2nd generation High reduction Burn liquor Sunila (Moberg 1967). 65% of rated MCR (Rissanen 1965). At 60% dry solids hardly any partial load could be run. 3rd generation Decrease sulfur emissions Burn black liquor, high reduction Earlier the recovery boilers had horizontal tube 4th generation Low NOx Burn black liquor, high reduction economizers. They plugged fast and had to be In 1980s the first high dry solids’ units started and low sulfur emission water washed at intervals of 1-4 weeks (Rissanen their coming on line. Extensive tests of effect 5th generation Decrease superheater and boiler Burn black liquor, high reduction, 1965). It was not until the early 1960 that install- of increasing dry solids from 72% to 84% were bank fouling low emissions ing vertical economizers started. In economizers

24 25 Fourth generation air systems are the multilevel air. By placing the planes to 2/3 or 3/4 arrange- air and the vertical air. As black liquor dry solids ment improved mixing results. Vertical air has a to the recovery boiler have increased, achieving potential to reduce NOx as staging air helps in low sulfur emissions is not anymore the target decreasing emissions (Forssén et al. 2000). of the air system. Instead low NOx and low car- C ryover are the new targets. High temperature and pressure o The three level air system was a significant recovery boiler improvement, but better results were required. Development of recovery boiler main steam Use of CFD models offered a new insight of air pressure and temperature was rapid in the begin- system workings. The first to develop air system ning, Figures 8-9. By 1955, not even 20 years with additional air levels was Valmet (Tampella) from birth of recovery boiler highest steam pres- with their 1990 multilevel secondary air in Kemi, sures were 10.0 MPa and 480 oC (Vakkilainen et Steam temperature Finland, which was later adapted to a string of al. 2004). The typical pressures and temperatures large recovery boilers (Mannola & Burel 1995). then backed downward somewhat due to safety Valmet also patented the four level air system, (McCarthy 1968). By 1980 there were about 700 where additional air level is added above the recovery boilers in the world (Westerbeg 1983). tertiary air level. This enables significant NOx In Japan, because of high electricity prices, more reduction. than ten high temperature and pressure recovery boilers are in use (Tsuchiya et al. 2002). The big- Figure 9. Main steam temperature as a function of recovery boiler main steam pressure. Meanwhile Andritz adopted vertical air where gest one is the 2700 tds/d, 10.3 MPa and 505 oC primary level is arranged conventionally. Rest of recovery boiler at Iwakuni mill (Ohtomo 2000). (Heinola 2014). The rise of main steam param- the air ports are placed on interlacing 2/3 or 3/4 Main steam temperature eters after superheaters from 90 bar/490 °C to arrangement. Vertical air was invented by Erik Maximizing electricity generation is driving New large recovery boilers seem to favor 100 bar/505 °C increases generated power 2%. Uppstu (1995). His idea is to turn traditional increases in the main steam pressures and high main steam temperatures and pressures When increasing steam parameters and increas- vertical mixing to horizontal mixing. Closely temperatures. The maximum steam temperature (Vakkilainen 2004). These increase the amount ing black liquor dry solid content from 75% to spaced jets will form a flat plane. In traditional can be limited by the ash properties. The first of backpressure electricity. 82% the generated power increases by more boilers this plane has been formed by secondary melting curve at the superheater front should be than 6%. taken into account. Increasing mill closure with high chlorine and potassium decreases the melt- Often neglected way to increase generated ing temperatures. The overall mill heat balance power is the low-pressure preheating of turbine should be used to optimize the feed water and condensate and demineralized water to 100 flue gas temperatures. °C by turbine condensing section bleeds. This increased generated power by some 3% due to The main steam temperature of recent recovery reduction of low-pressure steam extraction for boilers is shown in Figure 9 as a function of water heating in feed water tank.

C MCR capacity of that boiler. The average steam o temperature increases with size. Small boilers The rise of feed water temperature from 115 °C tend to have lower pressures to reduce specific to 140 °C was the next attempt to increase power cost. There are many boilers with main steam generation. It requires only small investments. It o parameters higher than 500 C. Most of them are did not increase the generated power due to the in Japan. higher flue gas losses. The high-pressure pre-

Steam temperature heating of feed water to 180 °C has even worse Future high energy efficiency effect on generated power due to the heat losses recovery boiler of the flue gases. The solution is to increase the economizer surface areas or to add a heat. The increase of power generation in a pulp mill The flue gas heat recovery devices have high is result of larger recovery boilers being built investment costs, but they are still highly recom- (Haaga 2014). The widely adopted industry mendable. Using flue gas heat recovery devices practice of rising the black liquor dry solid with other alternatives can increase their effect content and main steam parameters will change significantly. The combination of combustion the recovery boiler to sellable energy generator Figure 8. Main steam temperature as a function of recovery boiler capacity. air / demineralized water heating and feed water 26 27 preheating has the best result increase of gener- The Wisaforest Andritz recovery boiler started- black liquor woodhandling chips for ated power. The gain is because use of extraction up in 2004 in Pietarsaari, Finland (Nordbäck wood cooking recovery turbo volatiles G steam can be reduced. These combinations have 2004) and Kymi Valmet recovery boiler started- biomass boiler generator short payback periods in slightly more than one up in 2008 in Kouvola, Finland (Tikka 2008) heat losses year. have steam parameters of 102 bar/505 °C. They LP cooling, MP fire 82-85% DS (without ash) black liquor, use drying torrefaction biochar pelletizing MP2 The coupling using high-pressure preheating turbine bleed steam for sootblowing and heat and combustion air heating means significantly combustion air to high temperature. Both the heat for torrefaction ~330ºC increased generated power. This combina- plants have Siemens extraction back-pressure condensate tion reduces high-pressure steam extraction turbines having many bleeds at different pres- LP steam ~160ºC significantly. Adding high-pressure preheating sure levels and extraction for medium pressure increases investment about 1 M€ and has a nega- steam. Figure 10. Torrefaction process integrated to kraft recovery (Hamaguchi et al. 2013b). tive payback period. In Joutseno pulp mill the ABB extraction The combustion air heating to 190 °C increases condensing turbine started-up in 1998 has a wood wood chips Pulping line pulp generated power due to higher steam flow. The low-pressure pre-heater between the turbine Woodhandling Cooking, Washing, O2-delignification, Bleaching investment cost of an additional steam/air unit condenser and the feed water tank (Vakkilainen and modernization of power generation equip- et al. 1999). The Östrand Andritz recovery boiler fines and bark wood extracted chips chips Impacts on... ment is 1 M€. Payback period is in our case a started-up in 2006 has steam parameter of 105 black recovery bolier, white few months. The energy improvement’s ef- bar / 515 °C and uses flue gas heat recovery sys- hydrolysis liquor evaporation, chemical and liquor aqueous medium wood extraction energy balance, black liquor charactheristics ficiency can be increased through combinations tems (Westberg 2007) (Jakobsson et al. 2007). steam (pre-hydrolysis) with alternatives of feed water heating and flue The Värö Andritz recovery boiler started-up Pre-hydrolyzate Impacts on... Evaporation gas heat recovery. These combinations reduce in 2003 was retrofitted in 2009 with extra feed (containing lignin removal chemicals hemicelluloses) payback period to 11-13 months due to increase water pre-heaters, additional superheater stage, pulp properties, pulping conditions, black liquor filtrate lignin of generated power. enlarged economizer and a flue gas cooler. The characteristics, energy and chemical balance, biomass pulp mill in Fray Bentos, Uruguay involves input green The combination of all energy improvements many above mentioned energy efficiency items Recovery liquor Recausticizing can increase generated power more than 15%. (Kiuru 2008). It fires 80% dry solids eucalyptus Boiler lime kiln Biomass This requires an additional investment of more black liquor in Andritz recovery boiler with boiler Turbo than 11 M€ but gives payback period of approxi- steam parameters of 93-105 bar / 495-505 °C. (optional) Generators mately one year depending on flue gas recovery It has one 60 MW Siemens extraction back- process heat application used. pressure turbine and one 75 MW extraction condensing turbine. Now new mills will have a Figure 11. Lignin removal and pre-hydrolysis: main impacts on the kraft process (Hamaguchi et al. 2013b). Hence, the total optimization of pulp mill can high efficiency recovery boiler which can gener- increase generated power up to 23 MW with ate in the order of 260 kg/s of steam. If a large solids and adopting electricity increasing fea- Modern kraft recovery boiler has evolved from a payback period of the additional investment single cylinder extraction condensing turbine is tures known from general boiler technology chemical recycler to energy generator to provid- being 13 months. When calculating at a net used then an output of more than 216 MW of (increased feed water temperature, increased ing sellable power. With next ten years even sale price rate of 50 €/MWh it will create an electricity can be achieved. air preheating, flue gas heat recovery). more radical changes are expected. additional profit of 10 M€/year from additional • New processes such as additional firing (e.g. power sales. Conclusions bark or chips) and combining torrefaction or bio-oil production with recovery boiler. References There are some other opportunities for increas- The energy efficiency of recovery boiler process ing the power output of the mill. One is of course can be improved by Figure 10 shows recovery boiler integrated with 1. Alava, P., 1955, Kokemuksia JMW-soodakatilan designing the mill for lowest possible process • New modes of operation such as improvement torrefaction (Hamaguchi et al. 2013b). If we käytöstä. (Operation experiences of a JMW recovery heat consumption, thus freeing steam for con- of skills and motivation of personnel, energy can adopt our recovery operation to produce ad- furnace) in Finnish, Paperi ja Puu, Vol. 37, No. 5, pp. 209-215. densing power. Preheating of boiler feed water audits, enhanced process integration of exist- ditional saleable products such as biochar, then between the economizer stages is possible, thus ing processes and industrial plants (secondary changes are to be expected. 2. Antônio, A.G., 2011, Upgrade da Caldeira de increasing the mass flow of steam through the heat), collecting of reliable process informa- Recuperação da Veracel Utilizando a modelagem The recovery boiler operation will be affected CFD. (Upgrade of Veracel recovery boiler using mod- turbine is another way. Many mills already use tion (monitoring and control) and decreasing by changes in pulping process such as lignin ern CFD) In Portuguese, Latin American Congress of part of the methods mentioned above: sootblowing steam usage. • New technology such as higher steam pres- recovery and hemicellulose removal, Figure 11 Recovery Boilers, August 15th to 17th, 2011, Fray sure and temperature, higher firing liquor dry (Hamaguchi et al. 2013a). Bentos, Uruguay.

28 29 3. Boniface, A., 1985, Introduction. in Chemical 30 Years Recovery Boiler Co-operation in Finland. 30. Pettersson, B., 1983, Korsnäs sodapannor under nology. Forum on Kraft Recovery Alternatives, The recovery in alkaline pulping processes, ed. Hough, International conference, Baltic Sea, 24-26 May, pp. 40 år. (40 years of recovery boilers at Korsnäs) Institute of Paper Chemistry Appleton, Wisconsin, pp. Gerald, Tappi Press, Atlanta, pp. 1-7.. 9-22. In Swedish, Sodahuskonferensen ‘83, ÅF-IPK, 15-26. 4. Combustion Engineering A reference book on fuel 18. Hyöty, P., Ojala, S., 1987, Super combustion of Stockholm, pp. 57-68. 44. Uppstu, E., 1995, Soodakattilan ilmanjaon hal- burning and steam generation, 1949, ed. de Lorenzi, black liquor. Proceedings of 1987 Tappi Engineering 31. Raukola, A., Ruohola, T., Hakulinen, A., 2002, linta. (Control of recovery boiler air distribution) In Otto, Combustion Engineering – Superheater, Inc, Conference, pp. 49-53. Increasing power generation with black liquor recov- Finnish, Soodakattilapäivä 1995, Finish Recovery New York. 19. Ishigai, S., 1999, Historical development of strat- ery boiler. Proceedings of 2002 TAPPI Fall Technical Boiler Committee, 6 p. 5. Deeley, E. and Kirkby, A.H., 1967, The develop- egy for steam power. in Steam Power Engineering, Conference, September 8-11, San Diego, Ca, 11 p. 45. Vakkilainen, E., Sandegård, R. and Veitola, J., ment of chemical recovery boiler. Journal of the Ed. Ishigai, Seikan, Cambridge University Press, 394 32. Rissanen, E., 1965, Economic design aspects of 1999, New Ahlstrom recovery boiler at Metsä-Botnia Institute of Fuel, September 1967, pp. 417-424. p. recovery boilers. Papper och Trä, No. 10, pp. 553 Joutseno mill. Sodahus-konferenssen ’99. ÅF-IPK. 6. Edling, G., 1981, Tekniken i svenska soda- 20. Jakobsson, M., Reimer, T., Gustafsson, C., -566. pp. 73-82. hus 1870-1935. Sodahuskonferenssen, Svensk Varneslahti, P., 2007, SCA Östrands senaste mil- 33. Roschier, R.H., 1952, Soodaosasto. (Soda 46. Vakkilainen, E., 2005, Kraft recovery boilers – Papperstidning, No. 6, pp. 19-26. jardinvestering, Program Skog 2007, 22-23. March department) in Finnish, in Hiokkeen ja Selluloosan Principles and practice. Suomen Soodakattilayhdistys 7. Edling, G., 1983, Sodahuset 1930-1945. 2007, Sundsvall. valmistus. Pellinen H., Roshier, R. H., eds., Suomen r.y., Valopaino Oy, Helsinki, Finland, 246 p. Ångpannaföreningen, Stockholm, 77 p. 21. Jones, A.K., 2004, 70 years of advances in re- Paperi-Insinöörien Yhdistys, Helsinki, pp. 359 -456. 47. Vakkilainen, E., Suutela, J., Kankkonen, S., 2008, 8. Fernandéz, V., 2011, Air distribution influence on covery boiler design. 40th Anniversary International 34. Rydholm, S., 1965, Pulping processes. Advanced Efficiency options – increasing electricity reduction degree target. Latin American Congress of Recovery Boiler Conference, Finnish Recovery Boiler Interscience Publishers, a division of John Wiley & generating potential from pulp mills. Pulp and Paper Recovery Boilers, August 15th to 17th, 2011, Fray Committee, Haikko Manor, Porvoo, May 12-14, 2004, Sons, Inc., London. 1269 p. Canada, Vol. 109, No. 4, pp. 14-19. Bentos, Uruguay. pp. 15-20. 35. Sandquist, K., 1987, Operational experience 48. Vannérus, T., Svensson, G., Forslund, F., 1948, 9. Forssén, M., Kilpinen, P., Hupa, M., 2000, NOx 22. Järvinen, J., Ojala, J., Melander, A., Lamberg, of composite tubing in recovery boiler furnaces. Ångpannor. (Steam generators) In Swedish, Chapter reduction in black liquor combustion - reaction J-A., 2012, The evolution of Pulp and paper Industries Götaverken Energy Systems, technical paper TP-2- 19, in Maskinteknik, Ingenjörs handboken 2, Nordisk mechanisms reveal novel operational strategy op- in Finland, Sweden, and Norway. Chapter 2 in The 87, 15 p. Rotogravyr, Stockholm, Sweden, pp. 937-1072. tions. Tappi Journal, Vol. 83, No. 6, 13 p. evolution of the US Pulp and paper Industry, 1860- 36. Soodakattiloiden syöpymistutkimus 1965-1968, 49. Westberg, Å., 2007, Optimering av elproduc- 2000. In The Evolution of Global paper Industry 10. Grace, T.M., 1975, Perspectives on recovery tech- 1968, (Corrosion study of recovery boilers 1965- tionen vid SCA Östrands massafabrik. Värme- och 1800-2050, Lamberg, Juha-Antti, Ojala, Jari, nology. Forum on Kraft Recovery Alternatives, The 1968) In Finnish, Final report 23.10.1968, Ekono, Kraftföreningen, Panndagarna 2007, 7-8 February, Peltoniemi, Mirva, Särkkä, Timo, eds., Springer, pp. Institute of Paper Chemistry Appleton, Wisconsin, pp. 151 p. Uppsala, 14 p. (2007). 19-48. 27-57. 37. Steam its generation and use, 1992, 40th edition, 50. Westerberg, E., 1983, Kraft mill recovery units 23. Kiuru, J., 2008, Käyttökokemuksia Uruguaysta. 11. Hamaguchi, M., Vakkilainen, E., Cardoso, Stultz, S.C. and Kitto, J.B. Eds., 929 p. examined. Pulp and Paper International, March (Operating experiences from Uruguay) In Finnish, M., 2012, Alternative Technologies for Biofuels 38. Swartz, J.N., MacDonald, R.C., 1962, Alkaline 1983, pp. 53-54. Soodakattilapäivä, 29 October 2008, Finnish Production in Kraft Pulp Mills—Potential and pulping. Chapter 9 in Pulp and Paper Science and 51. Whitney, R., 1968, Introduction. Chapter 1 of Recovery Boiler Committee. Prospects. Energies, Vol. 5, No. 7, pp. 2288-2309. Technology, Vol. I, Ed. Libby, C. Earl, McGraw-Hill, Chemical recovery in alkaline pulping process. Ed. 24. V., Llinares, Jr. and P. J., Chapman (1989). 12. Haaga, K., 2014, XXL recovery boilers. In New York, pp. 160-239. Whitney, Roy P., TAPPI Monograph series No. 32, Stationary firing, three level air system retrofit Continuous development of recovery boiler technol- 39. Tikka, M, 2008, UPM Kymin uusi talteenot- Mack Printing Company, Easton, Pa., pp. 1-14. experience. Proceedings of 1989 Tappi Engineering ogy – 50 years of cooperation in Finland. Suomen tolinja. (UPM Kymi new recovery line) In Finnish, Conference, Atlanta, Georgia, September 10–13. Soodakattilayhdistys – Finnish Recovery Boiler Soodakattilapäivä 29 October 2008, Finnish Recovery Committee, Helsinki. 25. Mannola, L., Burelle, R., 1995, Operating experi- Boiler Committee. ence of a 7,270,000 lb d.s./day recovery boiler. Pulp 13. Hamaguchi, M., Kautto, J., Vakkilainen, E., 40. Toivanen, H, 2004, Learning and corporate strat- & Paper Canada, Vol. 96, No. 3, March 1995, pp. 2013a, Effects of hemicellulose extraction on the kraft egy: the dynamic evolution of the North American 70-72. pulp mill operation and energy use: Review and case pulp and paper industry, 1860-1960. Ph.D. Thesis, study with lignin removal. Chemical Engineering 26. Moberg, O., 1974, Recovery boiler corrosion. in Georgia Institute of Technology, April, 2004, 341 p. Pulp and paper industry corrosion problems, National Research and Design, Vol. 91, No. 7, pp. 1284-1291. 41. Toivanen, H, 2012, Waves of Technological Association of Corrosion Engineers, Houston, Texas, 14. Hamaguchi M., Saari J., Vakkilainen E., 2013b, Innovation: Chapter 3 in The evolution of the US Pulp pp. 195-136. Bio-oil and biochar as additional revenue streams in and paper Industry, 1860-2000. In The Evolution of South American Kraft pulp mills. BioResources, Vol. 27. Nordbäck, K., 2004, Wisaforestin uusi talteenot- Global paper Industry 1800-2050, Lamberg, Juha- 8, No. 3, pp. 3399-3413. tolinja. (Wisaforest new recovery line) In Finnish, Antti, Ojala, Jari, Peltoniemi, Mirva, Särkkä, Timo, Soodakattilapäivä, 14 October 2004, Finnish 15. Heinola, M., Salmenoja, K., 2014, Review of high eds., Springer, pp. 49-80. Recovery Boiler Committee. energy recovery boilers. In Continuous development of 42. Tomlinson, C. L. and Richter, F. H., 1969, The recovery boiler technology – 50 years of cooperation 28. Ohtomo, K., 2000, The experience of high tem- alkali recovery system. Chapter 10 in Pulp and in Finland. Suomen Soodakattilayhdistys – Finnish perature and pressure recovery boiler operation. paper manufacture, Vol. I, The Pulping of Wood, Recovery Boiler Committee, Helsinki. Japan Tappi Journal, Vol. 54, No. 1, pp. 109-114. MacDonald, Ronald G. and Franklin, John, H., eds., 16. Hochmuth, F.W., 1953, New developments in 29. Oulu Osakeyhtiö’s new sulphate mill, 1937, The McGraw-Hill, New York, pp. 576 – 627. recovery unit design. Tappi, Vol. 36, No. 8, p. 359. Finnish Paper and Timber Journal, Vol. 19, No. 7A, 43. Tomlinson, G. H. II, 1975, Black Liguor recovery pp. 276-278. 17. Hyöty, P., 1994, The history of recovery boiler. – An Historical Note. Perspectives on recovery tech-

30 31 understanding of black liquor sprays

ari kankkunen Aalto University FINLAND

Introduction

In 1936 a design of spray furnace was suggested Higher black liquor temperature was needed in which the boiler was mounted above the to decrease viscosity, and sometimes spraying smelter furnace in a tower arrangement. The de- temperature was above the boiling point of sign gave many advantages and reduced the need black liquor. The spraying temperature was for manual labor (de Lorenzi 1949). Rotating often above the boiling point, although boiler furnaces were replaced by spray furnaces in operators were not aware of flashing. Liquor future designs. Spraying enabled control of the heat treatment processes were also developed to char bed and combustion by changing the spray decrease viscosity and avoid firing temperatures direction and pressure. The mass flow rate and that were too high. droplet size could be easily changed. The air distribution to the char bed could be connected The importance of spraying studies was em- to the spraying (Vakkilainen 2007). phasized by the volume of heat and power generation from black liquor burning. In Finland, In Finland, the first Tomlinson type boilers were almost 11% of primary energy was supplied by built in the 1930s and 1940s (Vakkilainen, 2007). black liquor combustion in 2001. This was im- Obviously, a few of them were using oscillating portant in the viewpoint of the Kyoto agreement nozzles to spray evenly in all parts of the fur- under which CO2 emissions were restricted. The nace. It was normal to spray on furnace walls combustion of forest-based black liquor could to a dry black liquor for better combustion. In be calculated to be neutral concerning CO2 emis- the 1980s, stationary firing became the dominant sions. Tightening emission regulations caused way of spraying black liquor, first in Europe and an increased need for better combustion control. later in the U.S. (Llinares & Chapman 1989). Splash plate nozzles were common, but swirl Boiler sizes have been increasing year by year, cone nozzles were used in small boilers. and therefore the cost of shutdowns and other problems have also been increasing. The value of The transition to higher solids content black produced steam and electricity has been increas- liquors started during the 1980s. It had many ing all the time. The need to better understand advantages. The most important is the increased the recovery process increased; and at the same throughput capacity of a recovery boiler, which time, new research findings in other research is often the bottleneck in a pulp mill. Other fields became available. The same development advantages are better sulfur reduction, higher was seen in black liquor spraying studies starting energy production, decreased emissions, and in the late 1980s. Active research on black liquor decreased risk of smelt explosion. The nature of spray started in the U.S., Canada and Finland. spraying changed with increased solids content.

32 33 Experimental studies After 1985 a large set of experimental studies have been carried out. Figures 1 and 2 show Many studies have been carried out to simulate most of the experimental studies carried out the spraying of fuels and other liquids. Most with substitute fluids and with black liquor. The often, the purpose of spraying is to atomize a relevant operational range in modern furnaces is single component liquid to very small droplets located between the green dashed lines. Smaller with a large surface area and high reactivity. nozzles, lower spraying temperature, and lower The prediction of spray properties is at a good solids content experiments carried out in Canada level for these liquids and nozzles. In the case and the U.S. are often relevant to older furnaces. of black liquor, the liquid is very challenging. The solids content is high, and a very high Most of the published experimental studies are number of different organic and inorganic carried out at the Institute of Paper Technology species is included. The goal of spraying is to and Science (IPST) or at Aalto, formerly the produce a proper droplet size for drying and Helsinki University of Technology. The indus- devolatilization, distribute the char and solids trial scale tests have been carried out mostly evenly to the char bed, and produce good at Aalto. Small commercial nozzles have been combustion. The properties of black liquor studied at IPST. The University of British and the goals of spraying have led to different Columbia (UBC) and Tampere University of nozzles that are uncommon in other fields of Figure 1. Test series carried out, defined by nozzle size and mass flow rate. Technology (TUT) were among the first dur- spraying. Experimental studies and detailed ing the last two decades to study black liquor modeling have been necessary to increase the spraying. Substitute fluid tests with splash plate understanding of black liquor spray properties. nozzles have been carried out by the University of Toronto (UofT) during the latest decade.

* D=drop size, v=velocity, bm=breakup mechanism Table 1. Black liquor spraying studies. ** W=water, WG=water/glycerol mixture, S=syrup , BL=black liquor

Performed by Year of Objective* Nozzle type Inner Fluid** Temperature Viscosity Flow Pressure pub. diameter [°C] [mPa×s] [kg/s] [MPa] of nozzle [mm]

Bennington and Kerekes, 1985 D Pressure-swirl 0.7 W, WG, BL 20, 100, 120, 135 1, 15 0.7-3.5 0.02-0.5 UBC

Mäntyniemi TUT 1987 D Pressure-swirl 2, 3 W, 20 1 0.1-0.5 Splash plate BL

Adams et al. IPST. 1990 D,bm Splash plate 9.5-12 S, BL 20, 92-104 34-325 0.15-2.5 0.02-0.65 Fan

Bousfield/U of Maine 1990 D,bm Modulated - W, WG, BL 20, 105 1, 170 0.4 - inpinging jets Figure 2. Test series carried out, defined by fluid viscosity and difference from boiling point. Paloposki and Kankkunen 1991 D Splash plate 2, 3, 5 V, WG, ML 20 1, 45, 91 0.02-0.4 0.02-0.6 Aalto Table 1 lists the nozzles, mass flow rates, and Spielbauer and Aidun IPST 1992 D,bm Splash plate 4.8-8.5 WG 20 51-102 Droplet size, shape, and size fluids studied. Some of these studies concen- Kankkunen et al. Aalto 1993 D Splash plate 2 ,3 W, WG 20 1-60 1-10 0,1-0,6 distribution trated on the sheet breakup, some concentrated Kankkunen et al. Aalto 1994 D Splash plate 2, 3 BL 97-128 3-300 4-11 0.2-1.8 on the measurement of droplet size, and some The size and shape of a black liquor particle de- Empie et al. IPST 1995 D Splash plate 9.5-18 BL 32-125 130-215 0.9-5 fines its drying, combustion, char burning rate, Fan concentrated on the spray velocity. An overall Pressure-swirl view of the research in this field can be found in and of course also the trajectory of the droplet. Helpiö and Kankkunen Aalto 1995 D,v,bm Splash plate 15, 27 BL 65,112-125 11-177 0.7-8 0.05-0.3 Figures 1 and 2, where fluid properties, nozzle Therefore, the location for those burning stages Loebker and Empie, IPST 2001 D,bm 9, 12 S 20 100- 10000 0.5, 0.75 0.05-0.65 sizes, and mass flow rates of different research in a furnace is defined by the particle size and Kankkunen et al. Aalto 2001 D,v,bm Splash plate 27,28 BL 131-135 120-200 4.2-6.1 0.015- projects are presented. Experiments with substi- shape and the size distribution of the particles. 0.11 tute fluids are presented in the lowest part of the Additional important variables are mass flow Miikkulainen et al. Aalto 2009 v, bm Splash plate 26-40 BL 100-250 3-16 chart in Figure 2. distribution and the velocity of the spray. An Karami et al. UofT 2013 D, v, bm Splash plate 2, 27 W,S, BL 20, 130-135 1, 110-160 0.01, understanding of spray behavior and droplet 4.2-6.1

34 35 size and size distribution can be utilized to op- by spraying water. Droplet size was measured Later, the image analysis was adapted to meas- ence was achieved most often by the square-root timize boiler operation. Naturally, droplet size, with a laser diffraction based method. ure the drop size of industrial scale splash plate distribution function. The median droplet size shape, and velocity are a part of the input data nozzles in a vertical test chamber, as reported was between 4 mm and 14 mm, as can be seen to computational fluid dynamics (CFD) models At the Helsinki University of Technology, by Helpiö & Kankkunen (1996). The choice of in Figure 4. and the basis for the study of experimental single Laboratory of Energy Engineering and measuring in a vertical test chamber was moti- particle combustion. Environmental Protection, the black liquor vated by two reasons. First, there is no need for spraying research started in 1990. The basis of spray cutting baffles, and secondly, the lighting Bennington & Kerekes (1985) were the first to the starting research was on the earlier experi- system for detecting the sheet breakup mecha- systematically study black liquor spraying with ence of drop size measurement with easier fluids nism can be easily arranged at the sides of the droplet size measurements. They studied labora- like water, oils, and coal-water slurry. Many of spray. In a horizontal test chamber, the lighting tory scale grooved-core nozzles with glycerol/ these studies were confidential. They enabled of the sheet must be arranged below the spray water mixtures, superheated water, and a black the purchase of research instruments and and is liable to contamination. liquor of solids content of 56%. The spraying inspired researchers to deepen their understand- temperatures of black liquor in their experiments ing of spray studies. It was important to have were 100 °C, 120 °C, and 135 °C, and the con- Finnish boiler manufacturers, Tampella Oy sequent boiling point was estimated to be 112 and Ahlström Machinery Oy, and the Finnish °C. They suggested that the droplet size could be Recovery Boiler Committee introduce the needs best predicted based on the black liquor viscos- of industry for spraying studies. They funded Figure 4. The effect of the mass flow rate on the ity. No sudden change caused by flashing was the research in co-operation with the Ministry of mass median diameter. found. The distribution of drop sizes was found Trade and Industry. to be much broader than that of a Newtonian Loebker & Empie (2001, 2002) carried out a fluid having the same apparent viscosity. The first experiments connected to black liquor broad study with an effervescent nozzle and a spraying were carried out with small-scale splash splash plate nozzle. They studied droplet sizes Spielbauer et al. (1989) studied black liquor plate nozzles and substitute fluids in a vertical Figure 3. Droplet size of black liquor at different and sheet breakup with a very wide viscosity spraying with two sets of experiments. In the spray test chamber. Dimensional analysis was temperatures (Helpiö & Kankkunen 1996). scale from 100 mPa•s to 3400 mPa•s. The meas- first set, they used black liquor of 53% solids used to scale the results to industrial scale noz- ured mass median diameters varied from 2 mm content at room temperature, so the viscosity zles (Paloposki & Kankkunen 1991). The equa- to 7 mm, depending on the gas liquid ratio. was 150 mPa•s. The splash plate nozzle orifice tions for scaling laws and droplet size based on The next tests were carried out in a horizontal test chamber to enable the most reliable test size was 9.5 mm and the droplet size was meas- the Reynolds and Weber number were presented Karami et al. (2013) measured the droplet size ured by a laser light diffraction based droplet- for the first time for black liquor. The drop size chamber conditions. Here, the breakup mechanism in a test chamber is as similar as possible of small-scale splash plate nozzle spraying water size analyzer. The second set was carried out was measured with a laser light scattering based and an industrial-scale nozzle spraying black with black liquor of solids contents from 66% particle size analyzer. to spraying into an operating recovery boiler furnace. The comparison between the breakup liquor. They developed equations for droplet size to 72%. Temperatures varied from 104 °C to prediction for substitute fluids with small-scale The next phase was to measure the drop size of mechanism in the test chamber and furnace 127 °C. These were typical values of recovery nozzles. The Sauter mean diameter is predicted small scale splash plate nozzles with cold diluted could be fulfilled with a furnace endoscope. boilers at that time and corresponded to a viscos- by the equation black liquor. Drop size measurements could The comparison indicated similarity in most of ity range of 29 mPa•s to 66 mPa•s. The splash plate nozzles used were 9.5 and 12.7 mm of be carried out with a laser light based system the cases. This is a sign that the results of drop 0.36 -0.35 µ 0.003 (1) nozzle-orifice size. The third nozzle was a swirl with modified optics using a focal length of4 size measurements in the test chamber can be D32=677d V cone type nozzle with a 12.7 mm exit orifice. meters (Kankkunen 1993). However, it was not applied in the furnace calculations as shown by Droplet size was measured by an X-ray flash possible to reliably measure the drop size of hot Miikkulainen et al. (2002) & by Kankkunen and where d is nozzle diameter, V is velocity, and µ based imaging system. They found that a square black liquor because the light was scattered by Miikkulainen (2003). is the viscosity of the liquid. root-normal distribution provides very good the temperature gradients in the gas around the Droplet size distribution of a large set of test correlations of black liquor spray distributions. particles. An image-analysis based system was In the case of the flashing liquid, the median cases was measured in a vertical test chamber The normalized size distribution was found to therefore developed to measure the drop size of droplet size could be predicted by the equation with industrial-scale splash plate nozzles and be almost independent of nozzle geometry, flow hot diluted black liquor sprays (Kankkunen & Helpiö 1994). The diameter of the splash plate industrial spraying practices by Kankkunen 1.05 -0.35 0.003 - 0.35 conditions, or fluid parameters. Dmean=1000d V µ ( Ja+12.6) (2) nozzle in these tests was 2 mm and the measured & Miikkulainen (2003). A large fraction of In Finland, the drop size of black liquor sprays median droplet size was between 200 µm and non-spherical particles in the spray was de- where Ja is the Jacob number. The Jacob number was studied first by Mäntyniemi (1987) at 300 µm. An increase of droplet size caused tected. Non-spherical particles were modified is defined by equation Tampere University of Technology. He studied by flashing was detected for the first time, as as spherical particles with consequent volume. ρ c ΔT splash plate nozzles in a small recovery boiler to presented in Figure 3. The boiling point of di- Four droplet size distribution functions were 1 p Ja= find better spraying practices. He also measured luted black liquor was 113 °C, and the dry solids fitted to the experimental data for particle size ρ h v fg (3) droplet size distribution for a swirl-cone nozzle content was 70%. distributions. The smallest least-square differ- 36 37 The same form of equation was used by Empie et higher compared to the calculated average pipe where ρ1 is liquid density, ρv is vapor density, cp is liquid heat capacity, ΔT is excess temperature, al. (1995) for three splash plate nozzles. Helpiö flow even for the non-flashing case. & Kankkunen (1995) studied two splash plate and hfg is latent heat of evaporation. Equation (2) is based on a case of a splash plate nozzle nozzle types used by Finnish boiler manufactur- Late black liquor velocity measurements with an open exit and a diameter of 27 mm and is ers, and they were able to obtain experimental were carried out with small-scale nozzles by sensitive to changes in spraying practice. values for the constants. Values for the constants Kankkunen et al. (1998) and mill scale nozzles in based on these three test series are presented in spray test chamber by Kankkunen & Nieminen Mass flow rate Table 2. It should be taken into account that in (1997) and by Kankkunen & Miikkulainen different tests, the pressure difference Δp was (2003). The total black liquor mass flow rate to the measured at different locations. furnace is defined by the pulping process. The It was suggested that spray would behave dif- number, size, and geometry of nozzles can be Table 2. Discharge coefficient constants based Figure 6. Mass flow rate defined by black ferently in the furnace than in the spray test chosen by operators. The easiest way to control on three test series. liquor temperature for a splash plate nozzle chamber. Therefore a special furnace endoscope the mass flow rate is carried out by pressure (Kankkunen 2006). was constructed to measure spray velocities Study K K n control. Flow control is an essential variable to 1 2 in furnaces as presented by Miikkulainen et achieve better combustion and controlled char Spielbauer and 1.17 373 -0.92 Velocity of the spray al. (2004). A set of measurements in recovery Adams bed processes. The velocity of black liquor has many important boiler furnaces is presented by Miikkulainen et Empie et al. 1.24 458 -0.9 effects on spray properties. At first, it defines the al. (2000, 2002, 2006, 2009). They developed a Spielbauer & Adams (1990) studied experimen- Helpiö and -0.04 671 -0.73 sheet break-up mechanism of the black liquor new way to define spray center line velocity, and tally spraying with two splash plate nozzles (9.5 Kankkunen and essentially affects the droplet size. Next, it the concept of the dimensionless velocity was mm and 12.9 mm), v-type and u-type nozzles. defines the spray pattern, spray direction, and lo- presented. Dimensionless velocity was defined The test fluid was low solids black liquors and Nozzle pressure defines mass flow rate through cations where droplets hit the char bed or furnace as the ratio of measured spray velocity and cal- corn syrup. They concluded that the flow rate a splash plate nozzle. An example for a non- walls. In addition to these, it affects heat and culated non-flashing velocity through the nozzle could be predicted by a correlation equation flashing case is presented in Figure 5 for three mass transfer processes around particles and the outlet. A very useful way to make comparable based on the Reynolds number and the fluid nozzle sizes. The dry solids content of black time available for drying and devolatilization. measurements with different nozzle sizes and pressure for all nozzle types. The flow rate was liquor is 70% (BBR 13 °C). spraying practices was by using a flashing map, concluded to be insensitive to fluctuations Spray velocity was first based only on the cal- as presented in Figure 8. In the flashing map, in black liquor viscosity, solids content, and culated values, e.g., Empie (1995). Spray veloc- the mass flux is used to normalize different temperature. ity was first measured by Helpiö & Kankkunen nozzle diameters, and dimensionless velocity (1994). They used an image analysis based triple is presented on the vertical axis. An important The mass flow rate in a nozzle is defined by the exposure method. They found that the spray parameter is the difference from the boiling following equation velocity increased when the black liquor tem- point – the excess temperature (ΔTe). With the flashing map, the velocity could be predicted perature increased, as can be seen in Figure 7. m  = C d A p 2 ρ ∆ p (4) rather reliably for a known spraying practice and nozzle geometry.

where Ap is the area of the pipe cross-section,

Cd is the discharge coefficient, ρ is the density of black liquor, and Δp is the available pressure Figure 5. Mass flow rate defined by pressure at difference. a temperature of 113 oC, DS 70%, and boiling point of 113 oC (Kankkunen 1996). Spielbauer & Adams (1990) experimentally studied Babcock & Wilcox splash plate nozzles. They suggested an equation in the form of the At constant pressure, the mass flow is decreased discharge coefficient by flashing, as shown in Figure 6. The inner diameter splash plate nozzle is 15 mm. Pressure Figure 7. Spray center line velocity as a function difference is 0.9 bar, 1.1 bar, and 1.4 bar of temperature (15 mm nozzle, Δp=0.75 bar) 1 (Helpiö and Kankkunen 1994). Cd = (Kankkunen 1996). K + K Re n n (5) K11 + K2 2 re The boiling point of black liquor was 112.8 °C, and their spraying temperatures were above where the constants K , K , and n are obtained by 1 2 boiling point up to 125 °C. They also found that regression analysis. This type of equation can be for splash plate nozzles, the spray velocity at Figure 8. A flashing map for splash plate nozzle adapted for conditions below the boiling point. the spray center line was about 1.4 to 1.9 times type A (Miikkulainen et al. 2006). 38 39 Nieminen et al. (1996) measured the spray Mass flow rate distribution flow rate decreases towards the spray edges. At surface of the sheet and therefore affect the velocity of splash plate nozzles using an im- a higher excess temperature (18.5 °C), the mass break-up process. The distribution of mass within the spray affects age analysis system. He found that velocity flow rate to all spray angles is almost constant. the penetration of the spray to the formation of decreased almost linearly from the spray center Bousfield (1990) made an interesting test serial the char bed and to the drying and combustion of line to higher spray angles, as can be seen from producing a coherent jet with low solids content droplets. The splash plate angle and the shape of Figure 9. black liquor. The jet was introduced inside and the nozzle exit area affect the mass distribution. came out of tubular furnace at temperatures of Also, flashing is assumed to affect the mass flow 950 °C and 1110 °C. He studied the break-up rate distribution. length of the black liquor jet. He found that the length of the jet was decreased for black liquor Nieminen (1996) measured spray mass distribu- at room temperature. In the case where the black tion in spray angles by a spray collector. Figure liquor temperature was 108 °C, the jet length 11 shows the mass distribution for a full-scale was increased by the hot furnace. He concluded splash plate nozzle with non-flashing black liq- that the viscosity was increased by the drying uor and a small-scale splash plate nozzle with a of the surface of the jet when the black liquor substitute fluid. The mass distribution decreases temperature was near its boiling point. almost linearly from the spray center line to the Figure 12. Mass flow rate distribution for two highest spray angle. flashing cases (Karami et al. 2013). Kankkunen & Nieminen (1997) and Kankkunen Figure 9. Spray velocity when spraying black (1998) studied black liquor sheet break-up liquor below the boiling point (Nieminen 1996). mechanisms. They found three mechanisms. Black liquor sheet break-up Below the boiling point, the main break-up o Karami et al. (2013) measured the velocity for mechanism mechanism was a wavy break-up. Around 5 C above the boiling point, the break-up was caused flashing black liquor, as presented in Figure Black liquor forms a sheet when sprayed at by sheet perforations. Around 15 °C above the 10. Experiments were carried out in a spray temperatures slightly above (<5 °C) or below boiling point, the break-up was caused by high test chamber, and velocity was measured by boiling point. The sheet thickness decreases velocity and boiling. The configuration of splash an image analysis system. The velocity was when the distance to the nozzle increases as the plate nozzles seemed to affect the break-up almost constant in the spray center line and at sheet cross-section gets wider. Disturbances in mechanism, and a combination of break-up other angles of the spray. The velocity was ap- the liquid and aerodynamics drag around the mechanisms was detected. proximately 8 m/s for the excess temperature of thinning sheet breaks the sheet into ligaments 14.2 °C and approximately 15 m/s for the excess and then to droplets. The thickness of the Miikkulainen (2006) systematically studied temperature of 18.5 °C. Figure 11. Mass flow rate distribution of a splash sheet and corresponding wavelength (i.e., the the length and break-up mechanisms of black plate nozzle (Nieminen 1996). length between two fracture points) defines the liquor sheets. He found decreasing length at ligament diameter and consequent droplet size. increasing temperature when the mass flow rate Spray mass flow rate distribution was also stud- When the spraying temperature is sufficiently was constant. He also measured the wavelength ied by Empie et al. (1997) for a splash plate and above the boiling point (typically >10 °C), the of a black liquor sheet by image analysis based V-jet nozzle. They sprayed 55% to 65% solids sheet is not necessarily formed and at least the methods. Experiments were carried out at tem- content black liquor at temperatures below sheet break-up is affected by bubble formation. peratures from 14 °C to 18 °C above the boiling boiling point and measured spray distribution in A high temperature in the furnace may dry the point. spray angles by spray patternators. They found that the mass distribution could be estimated by a parabolic function of the angular position, so that the highest mass flow was in the spray center line.

Figure 10. Spray velocity at spray angles for a Karami et al. (2013) presented a new method splash plate nozzle (Karami et al. 2013). to estimate mass distribution in the spray. The estimation was based on the light intensity variation of the spray images taken in a horizontal o o o spray test chamber. Figure 12 shows the mass ΔTe = -4 C ΔTe = 5 C ΔTe = 15 C distribution in the spray for two conditions. At Figure 13. Sheet break-up mechanisms, wave formation, perforation, and flash break-up (Kankkunen a lower excess temperature (14.2 °C), the mass 1998).

40 41 Karami et al. (2013) found four different break- spray dimensions and velocities at the distance the perpendicular splash plate and to a set Mass-momentum coupling was solved by the up regions for splash plate nozzles. The closed of 2.3 meters from the nozzle by spraying along where corn syrup was sprayed with a typical SIMPLE method. Scalar equations were solved rim break-up mechanism is applicable for small the furnace wall. Imaging took place by empty 11.9 mm splash plate nozzle. The results were by the fully implicit control volume method. flow rates. In the open rim break-up mechanism, nozzle gun ports. A decrease in the droplet ve- highly similar in predicting spay velocity, sheet Using experimentally measured spray veloci- the liquid sheet is continuous until it starts to locities was detected. A part of the images was thickness, and break-up mechanisms, and even ties, the model could predict the spray velocity oscillate and break up into ligaments and drop- showing hollow or exploded droplets, which droplet size. The code was used to simulate a of flashing liquid for a flashing black liquor gun. lets. The perforated sheet break-up mechanism was explained as being caused by excess steam new nozzle type that was developed. The initial bubble nuclei concentration in bulk produces a sheet that is perforated by flashing vaporizing from the black liquor. An example black liquor was found to be an important vari- vapor bubbles. In fully shattered spray, there is presented in Figure 15. Later, Engblom et Sarchami et al. (2010) at Toronto University used able without reliable estimation. The estimation is no sheet, but a large number of bubbles and al. (2013) showed by CFD simulations that KIVA code to simulate the sheet break-up of a could be obtained from experimental spray a high velocity at the nozzle exit disintegrates the velocity decrease could be predicted by the splash plate nozzle. They predicted the droplet velocity measurements. black liquor directly into droplets. Figure 14 decreased density of droplets. size for water sprays produced by 1 mm and 2 shows different break-up mechanisms defined mm splash plate nozzles. The predicted droplet Computational combustion by the Reynolds number and the Jacob number. Overall spray studies in the furnace was carried size was found to be similar to the experiments. modeling The secondary break-up of droplets does not out by Brink et al. (2009). They could see spray seem to be important. Velocities are low – below directions and opening angles by IR camera. Järvinen et al. (2006) started a computational Adapting the data in CFD takes the most advan- 15 m/s. Although droplet size is high, second- analysis of flashing nozzle flow based on the tage of the initial data produced by experimental ary atomization is unlikely because of the low earlier black liquor droplet combustion model sprays studies and spray modeling. Most calcu- velocity and high viscosity of black liquor, as (2002). They developed a new model for the lations are based on the Fluent program. found by Kankkunen & Miikkulainen (2003). growth of a single vapor bubble during flash evaporation of highly viscous liquids. The model CFD is an important tool for furnace design consisted of the simultaneous solution of the heat and engineering. The accuracy of simulations transfer equation combined with the generalized has improved with increased computational Figure 15. Droplet expansion in the furnace, Rayleigh equation for bubble growth dynamics. capabilities and especially with the increased triple exposure and an image with a hollow A simplified thermodynamic, Raoult law based understanding of flow, heat transfer, and chemi- droplet. (Kankkunen et al. 2007). equilibrium model described the vapor pres- cal reactions taking place in the furnace. Special sure of water in a multi-component solution. emphasis is on the increased understanding By using the new model, unit processes during achieved by the single droplet experiments by Computational spray modeling flashing could be studied and understood. It was Hupa et al. (1987) and the single droplet combustion model by Järvinen (2002). Nozzles can be simulated by computational found that when viscosity increased to values close to black liquor, the initial stages of bubble methods. Foust et al. (2002) used the volume of Working in co-operation with groups using fluid (VOF) model to simulate flow through the growth were controlled by viscous and surface tension forces. As the bubble continued to grow CFD for furnace simulations is important for Figure 14. Spray patterns at different Reynolds splash plate nozzle up to initial sheet breakup experimental spray studies. This has enabled an and droplet formation. The VOF model solves further, heat transfer to the bubble surface be- and Jacob numbers (Karami et al. 2013). came the dominating mechanism. The role of the increased understanding of phenomena affecting the conservation equations of volume fraction furnace operations by spraying. The co-oper- and momentum utilizing the finite volume tech- mass transfer of water in black liquor was not studied in detail. When water is removed from ation with Aalto University and Åbo Academy Sprays in the furnace nique. The results obtained by the model were University has been an excellent example of this. different compared to the experimental part of the surface, new water should be transported The purpose of all experiments and theoretical their work. They concluded that the coarse mesh to the evaporation surface to maintain the rate. calculations is to predict phenomena and pro- and insufficient computational power hindered Järvinen et al. (2006) assumed that as a new Conclusions and future directions surface is constantly created, water is also trans- cesses taking place in the furnace. Spray proper- reliable results. The droplet size in a non-flashing case is rather ported by the radial convection mechanism. The ties in the furnace are not necessarily similar as well understood. In the case of flashing spray, role of water diffusion to the drying front should in a spray test chamber. Gas flows in the furnace Fard et al. (2007) developed a 3D numerical the droplet size is difficult to define. The spray be better understood and further studied. and heat conduction and radiation may change model of free surface flows with an Eulerian velocity is first increased by flashing inside the spray properties. Therefore, experimental stud- fixed-grid algorithm that utilized a volume nozzle. After that, nozzle droplets are exploded Later, Järvinen et al. (2011) used this un- ies in the furnace are needed. tracking approach to track fluid deformation. by the excess steam coming out from the drop- derstanding of the bubble growth for a new The governing flow equations were assuming an lets. Experimental studies carried out for sprays computational flow model for a flashing black Sprays in the furnace have been studied by incompressible, Newtonian, and laminar flow. in the normally operating furnace should give liquor gun. The model was based on the numeri- Miikkulainen et al. (2009) with a furnace endo- A 3D fixed-grid Eulerian method was written a better understanding of a droplet’s fate in the cal solution of one-dimensional conservation scope. He could study the spray velocity, sheet specifically for free surface flows with surface furnace. length, and breakup mechanism. With another tension. Calculation results of the code, BlSpray, equations of mass, momentum, energy, steam endoscope, Kankkunen et al. (2011) studied were compared to a set with water sprayed to mass fraction, and bubble number density. 42 43 The work for a better nozzle flow model for Spraying characteristics of commercial black liquor Technologies for a Clean Environment, Lisbon, 19, 22 29. Miikkulainen, P., Järvinen, M., Kankkunen, A., flashing black liquors is actively going on. A nozzles, Tappi J., Vol. 78, pp. 121-128. July 1993, Lisbon. 2002, The effect of a furnace environment on black one-dimensional plug-flow model was presented 6. Empie, H.J., Lien, S., Samuels, D.B., 1997, 17. Kankkunen, A., Helpiö, T. 1994, Small scale liquor spray properties. Journal of Pulp and Paper by Järvinen et al. (2011). This new model uses Distribution of Mass Flows in Black Liquor Sprays. measurement of black liquor spraying with splash Canada, Vol. 103, No. 9, pp. 238-242. measured non-Newtonian properties of black In Brogdon, B. N., Innovative advances in the for- plate nozzles. The TAPPI Engineering Conference, 30. Miikkulainen, P., Kankkunen, A., Järvinen, M. liquor. The validation results of firing pressure, est products industries: the 1997 Forest Products San Francisco, 19-22 September 1994. pp. 207-214. 2004, Furnace endoscope - measuring fuel spray flow rate, and velocity predictions are promis- Symposium, 1998, p. 171. 18. Kankkunen, A., Nieminen, K., 1997, Black Liquor properties in hot and corrosive environments. ing. In the future, the model can be used to 7. Foust T.D, Hamman, K.D., Detering, S.A., the Sheet Breakup Mechanisms Effect on the Droplet Experiments in Fluids, No 37, pp. 910-916. develop better spraying practices and nozzles 2002, Numerical Simulation of Black Liquor Size. International Journal of Fluid Mechanics and 31. Miikkulainen, P., Järvinen, M., Kankkunen, A., in co-operation with CFD simulations. An im- Spray Characteristics, ASME 2002 International Research, Vol. 24, No. 1-3, pp. 350-357. 2002, The effect of a furnace environment on black liquor spray properties. Journal of Pulp and Paper portant part of improvement in the combustion Mechanical Engineering Congress and Exposition 19. Kankkunen, A., 1989, Flow Regimes inside a Heat Transfer, Volume 4. New Orleans, Louisiana, Canada, Vol. 103, No 9, pp. 238-242. and recovery process is a better understanding of Black Liquor Splash Plate Nozzle. ILASS-Europe ‘98, USA, November 17–22, 2002, pp. 155-162. spray properties, especially droplet size affected Manchester 6-8 July 1998. pp. 412-417. 32. Miikkulainen, P., 2006, Spray Formation of High 8. Helpiö, T., Kankkunen, A., 1996, The Effect of Dry Solids Black Liquor In Recovery Boiler Furnaces. by flashing. 20. Kankkunen, A., Miikkulainen, P. 2003, Particle Black Liquor Firing Temperature on Atomization Size Distributions of Black Liquor Sprays with A High PhD Thesis, Department of Mechanical Engineering, Performance, Tappi Journal Vol. 79, No 9, 1996. pp. Helsinki University of Technology. New measurement technologies connected to Solids Content in Recovery Boilers. Article Number 158-163. 200308, IFRF Online Combustion Journal. 33. Miikkulainen, P., Kankkunen, A., Järvinen, M., a better understanding of spray behavior and 9. Helpiö, T., Kankkunen, A., 1994, Measurements 21. Kankkunen, A., Miikkulainen, P., and Fogelholm, Fogelholm, J.-J., 2009, The Significance of Velocity drying and combustion process enable better of the Effect of Black Liquor Firing Temperature in Black Liquor Spraying, TAPPI Journal, Vol. 8, No. on-line process control. The lifetime of a nozzle C.J., 2007, Black Liquor Spray and Droplet Properties on Atomization Performance, Part 1: Nozzle Flow in the Furnace, International Chemical Recovery 1, pp. 36-40. in the harsh furnace environment is normally Properties and Sheet Breakup. Helsinki University of Conference, May 29 June 1, 2007, Quebec City, QC, 34. Mohammad, P., Fard, M.P., Levesque, D., not an issue after proper material selections and Technology, Laboratory of Energy Engineering and Canada, pp. 115-118. Morrison, S., Ashgriz, N., Mostaghimi, J., 2007, spraying practice have been carried out. Environmental Protection, Research Report, p. 30. 22. Kankkunen A., Jarvinen M., Miikkulainen P., Characterization of Splash-Plate Atomizers Using 10. Helpiö, T., Kankkunen, A., 1994, Measurements 2011, Initial Stages and Overlapping Processes of Numerical Simulations. Atomization and Sprays, Vol. Black liquors are changing in the future, when of the Effect of Black Liquor Firing Temperature a Black Liquor Droplet in a Kraft Recovery Boiler. 17, pp. 347-380. side flows like lignin, tall oil, and others are on Atomization Performance, Part 2: Drop Size TAPPI Peers. 2.-5.10.2011, Portland, Oregon, pp. 35. Mäntyniemi, J. 1987, Atomization of liquor in utilized. The changes will affect the behavior Measurements. Helsinki University of Technology, 532-543. swirl and splash plate nozzles (in Finnish), Tampere Laboratory of Energy Engineering and Environmental of black liquors in the nozzle and during the 23. Karami, R., Kankkunen, A., Ashrgriz, N., Tran, University of Technology, Department of Mechanical Protection, Research Report, p. 23. drying and combustion phases. The effect of H.; Effects of Flashing on Spray Characteristics of Engineering. these changes must be measured and modeled to 11. Hupa, M., Solin, P., Hyoty, R., 1987, Combustion Splashplate Nozzles, TAPPI Journal, 2013, Vol. 12, 36. Nieminen, K., 1996, Experiments on Black achieve better furnace control strategies. behavior of black liquor droplets. Journal of Pulp and No. 5, pp. 21-27. Liquor Splashplate Nozzle Performance, 12th Annual Paper Science, Vol. 13, pp. 567-572. 24. Llinares, V., Chapman, P.J., 1989, Combustion Conference of ILASS-Europe, Lund, Sweden, June 12. Järvinen, M.P., 2002, Numerical modeling of Engineering Update Stationary firing, Three Level Air 19-21. References the drying, devolatilization and char conversion System Retrofit Experience. Proc. TAPPI Eng. Conf., 37. Paloposki, T., Kankkunen, A., 1991, Dimensional processes of black liquor droplets. Acta Polytechnica p. 629. analysis and spray modeling. Sprays and Aerosols Scandinavica, Mechanical Engineering Series, No. 1. Bennington, C.P.J., Kerekes, R.J., 1985, Effect of 25. Loebker, D.W., Empie, H.J., 2001, Independently ‘91, 30.9–2.10.1991, Guildford, pp. 144-149. 163, p. 77. Temperature on drop size of black liquor sprays. 1985 controlled drop size in black liquor sprays into the 38. Paloposki, T., Kankkunen, A., 1991, Droplet Size International Chemical Recovery Conference, TAPPI 13. Järvinen, M. P., Kankkunen, A. P., Miikkulainen, kraft recovery boiler using effervescent atomization. of Black Liquor Spraying (in Finnish), Research Press, Atlanta, GA, USA, New Orleans, LA, USA, p. P. H., 2006, Bubble Growth in Flash Evaporation Journal of Pulp and Paper Science, Vol. 27, No. 1, Report 44, Helsinki University of Technology, Energy 345. of Viscous, High Solid Content Black Liquor. ECI pp. 18-25. Technology, p. 71. International Conference on Boiling Heat Transfer, 2. Bousfield, D.W., 1990, Research on Droplet 26. Loebker, D.W., Empie, H.J., 2002, Independently 39. Sarchami, A., Ashgriz, N., Tran, H., 2010, An Spoleto, Italy, 7-12 May 2006. Formation for Application to Kraft Black Liquors,. controlled drop size in black liquor sprays using ef- atomization model for splash plate nozzles. AIChE J., University of Maine, Orono, Maine. 14. Järvinen, M., Miikkulainen, P., Kankkunen, A., fervescent atomization. Part II: Splashplate nozzles. Vol. 56, No. 4, pp. 849-857. 2010, One-Dimensional Flow Model of a Flashing 3. Brink, A., Lauren, T., Hupa, M., Koschack, R., Journal of Pulp and Paper Science, Vol. 28, No. 8, 40. Spielbauer, T. M., Adams, T. N., Monacelli, J. E., Black Liquor Gun, Mar 29–April 1, Williamsburg, and Mueller, C., 2010, In-Furnace Temperature and pp. 251-258. Bailey, R.T., 1989 International Chemical Recovery Virginia, USA, 10 p. Heat Flux Mapping in a Kraft Recovery Boiler. TAPPI 27. de Lorenzi, O., 1949, Combustion Engineering. Conference, Droplet size distribution of black liquor Journal, Vol. 9, No. 9, pp. 7-11. 15. Järvinen, M. P., Kankkunen, A. P., Miikkulainen, A reference book on Fuels Burning and Steam sprays, Ottawa. (1989). P.H., Heikkilä, V.P., 2011, A One-Dimensional Flow 4. Engblom, M., Brink, A., Hupa, M., Kankkunen, Generation, Combustion Engineering – Superheater 41. Spielbauer, T.M., Adams, T.N., 1990, Flow and Model of a Flashing Black Liquor Gun: Study of A., Järvinen, M., 2013, CFD Droplet Boundary Inc., New York. pressure drop characteristics of black liquor nozzles. Vaporization Sub-Models. Swedish-Finnish Flame Conditions from Black Liquor Spray Experiments. 8th 28. Miikkulainen, P., Järvinen, M., Kankkunen A., TAPPI Journal, pp. 169-174. International Black Liquor Colloquium / Black Liquor Days 2011, Piteå. 2000, Black Liquor Spray Properties in Operating 42. Vakkilainen, E., 2007, Kraft Recovery Boilers, and Biomass to Bioenergy and Biofuels, UFMG, Belo 16. Kankkunen, A., 1993, Drop Size Measurements Recovery Boiler Furnaces. INFUB 5th European Principles and Practice. Helsinki University of Horizonte, Brazil. of Scale Model Black Liquor Splash Plate Nozzles. Conference on Industrial Furnaces and Boilers, Technology, Espoo, Finland. 5. Empie, H., Lien, S., Adams, T., Yang, W., 1995, Second International Conference on Combustion Porto, Portugal, 11-14 April 2000. 44 45 RECOVERY BOILER SOOTBLOWERS - HISTORY AND TECHNOLOGICAL ADVANCES

Honghi Tran and danny tandra University of Toronto CANADA

Introduction

The accumulation of fireside deposits on heat faces with high pressure steam (Figure 1), while transfer tube surfaces is a persistent problem in during boiler outages, they use hot water instead many kraft recovery boilers. The severity of the of steam to remove the deposits and clean the problem depends mainly on the amount and the boiler tubes. Maintaining a high deposit removal melting behaviour of fly ash particles (carryover efficiency and availability of sootblowers is of and fume) entrained in the flue gas, and the vital importance in recovery boiler operation efficacy of the deposit cleaning system that the (Chapel 1987, Barsin 1992). individual boiler employs (Tran 1997, 2004). If left unattended, deposits may form a thick insu- Over the last three decades, much research and lation coating on the tube surface, significantly development has been done on deposit removal reducing the boiler thermal performance and and sootblowing technology by researchers at steam production capacity. In severe cases, de- the University of Toronto, and by researchers posits may completely plug the flue gas passes, and engineers at recovery boiler and sootblower leading to costly unscheduled boiler shutdowns manufacturers. This includes jet hydrodynamics for cleaning. They might also fall, and if large and nozzle design (Jameel et al. 1994, 1996, enough, damage screen tubes and floor tubes Moskal et al. 1993), interactions between a jet and directly below (Villarroel et al. 2005). tubes and between a jet and deposits (Kaliazine et al. 1997, 1999, Tandra et al. 2007), and the Sootblowers are the principal means of cleaning mechanisms of deposit breakup and removal deposits in recovery boilers. During boiler op- by a jet (Pophali et al. 2009, 2013). The work eration, sootblowers blast deposits off tube sur- was carried out through laboratory experiments, mathematical modeling and computational fluid dynamic (CFD) modeling (Emami et al. 2010, 2012), as well as through mill trials (Moskal et al. 1993, Saviharju et al. 2011 , Tran et al. 2012). The results provided significant insight into how a sootblower jet strikes and removes deposits from the tube surface, and factors influencing the removal efficiency. These research and development efforts resulted in two major innovations. One is the fully-expanded sootblower jet concept (Jameel et al. 1994, 1996) which served as the basis for the development of the new Figure 1. Sootblower in action in a recovery generation of high intensity sootblower nozzles boiler. on the market today (Clyde Bergemann 2012, 46 47 Diamond Power 2007, Habib et al. 2006, Jameel sootblowers that can deliver stronger and better a long steel pipe with a handle on one end and air). The nozzle at the end of each lance tube was 1996). The other is low-pressure sootblowing controlled jets. small nozzles on the other end. Compressed air designed so that the steam jet could be directed technology (Kaliazine et al. 2006, Tran et al. was manually regulated through the pipe to pro- to a desired area. These stationary multi-nozzle 2008), that has been tested (Tandra et al. 2005, This paper reviews the history of recovery boiler duce an air jet of various strengths. Depending cleaning systems were later replaced by station- Tavares & Youssef 2006) and implemented in sootblowers, and several important techno- on the nozzle design and position, operators ary rotary cleaning systems to increase the several recovery boilers in the United States and logical advances that have been made in the past could direct the air jet straight into the boiler or coverage area. This was achieved by mounting Brazil (Tandra et al. 2008). three decades. at an angle to clean deposits. Since the passing multiple nozzles along the length of the lance air helped cool the pipe lance, hand blowing tube, which was manually rotated by pulling the The availability of high-computing power com- The history could be carried out during boiler operation. chain around the lance tube (Figure 3). puters, superfast microprocessors and data acquisition systems, sophisticated process control The word “soot” refers to black, carbonaceous Hand lancing and hand blowing were labori- The large number of nozzles mounted on each logics, and versatile CFD modeling capabilities fine particles that are formed during the combus- ous and useful only for small boilers. As the lance tube reduces the steam flow rate per noz- in the past two decades has also contributed tion of coal, wood, oil, etc. Soot forms deposits boiler size increased, these manual cleaning zle, and hence, reduces the cleaning power of the greatly to the advancement of sootblowing tech- on the walls of the combustors, chimneys and jet. Furthermore, the portion of the lance tube nology. Several deposit monitoring methods pipes that convey the flue gas. In the context of that stays inside the boiler has a short service based on boiler and sootblower operating recovery boilers, soot is synonymous with car- life since the lance tube corrodes quickly if it is parameters were developed to estimate the state ryover/fume deposits. A sootblower is simply a exposed uncooled to the corrosive environment, of fouling (or cleanliness) of the heat transfer device which is used for blowing and removing particularly in the superheater region of the boil- surfaces in different regions of the boiler, and to deposits from tube surfaces. er. Even in the cooler and less corrosive regions incorporate the information to optimize the soot- of the boiler, the lance tube temperature is still blowing operation (Adams 2010, Jones 2001, The first sootblower for on-line deposit cleaning high, close to the flue gas temperature. When the Tandra et al. 2010, Thabot et al. 2010, Uloth et was patented and first used in coal-fired boilers cleaning is initiated, the cooler steam supplied al. 1996, Versteeg & Tran 2009). by Diamond Power in 1903 (Barsin 1992). The to the lance tube causes the tube temperature to technology evolved slowly but eventually found drop abruptly, resulting in thermal cyclic stresses its way to other applications including recovery There has been enormous change in recovery that can shorten the lance tube life. As a result of boilers. boiler technology since the first Tomlinson these drawbacks, stationary rotary sootblowers recovery boiler was built in 1934 (STEAM were soon replaced by retractable sootblowers In the early days, recovery boilers were designed 2005). The black liquor firing capacity of the with close tube spacings that were difficult to boiler has increased substantially from under Retractable sootblowers keep open. Hand lancing was carried out fre- 200 tds/d (tons of black liquor dry solids per quently to remove deposits but was only able to Retractable sootblowers were developed to day) in the 1930s to slightly over 2000 tds/d Figure 2. Stationary cleaning system (Vulcan keep the boiler in operation for a short period Retractable sootblowers were developed to ad- in the 1980s, over 3000 tds/d in the 1990s, and 1921). of time (Chapel 1987). As the boiler became dress the shortcomings of stationary rotary soot- to 5000 tds/d in the 2000s (Vakkilainen 2009). plugged, the operators stopped the liquor fir- blowers. When the sootblower is not in use, the Several recently built recovery boilers are in the ing, opened the boiler mandoors to locate the lance tube is fully retracted and parked outside 6000-7000 tds/d range (Haaga 2013), and pres- deposits, and used a long steel rod (or lance) to the boiler, and thus is not exposed to the hostile ently the world’s largest recovery boiler is being manually knock them off the tubes. Hand lanc- environment in the boiler. In order to maximize built at the new greenfield pulp mill in OKI, ing could only handle deposits that were close to the jet cleaning power as well as to balance the Indonesia, with a staggering designed firing the boiler wall. For deposits that were far from lance tube, retractable sootblowers are equipped capacity of 11600 tds/d. The superheated steam the boiler wall or were obscured by tubes/other with only two opposing nozzles at the tip of the parameters have also increased, from a maxi- deposits, hand lancing was difficult as the steel lance tube. mum temperature/pressure of 350 °C/30 bar in rod needed to be extended deeper into the boiler. the 1930s, to 480 °C/60 bar in the 1990s and For deposits that were behind tubes, it was even Retractable sootblowers were introduced to 2000s (Vakkilainen 2009), with a few recovery more difficult since the steel rod needed to be recovery boilers in the 1940s; this significant boilers in Scandinavia and Japan operating at bent in order to reach the deposits. Hand lanc- Figure 3. Stationary rotary cleaning system development made large recovery boiler design 515 °C/110 bar to maximize the electricity gen- ing was dangerous to carry out during boiler (Vulcan 1921). practical (STEAM 2005). erating efficiency (Arakawa et al. 2014, Haaga operation due to the constant puffing of hot flue 2013, Tran & Arakawa 2011). As the firing ca- devices became obsolete and were replaced by gas from the boiler, particularly when the boiler Early retractable sootblowers had a chain drive pacity and steam quality increase, the boiler size fixed-position (stationary) multi-nozzle cleaning became plugged. system that moved and rotated the lance tube and heat transfer surface area also increase. This systems, as shown in Figure 2. In this system, in and out of the boiler. The chain was driven inevitably requires not only a larger number of several lance tubes were connected to a common Hand blowing became a preferred means of by a stationary motor mounted on top of the sootblowers per boiler, but also longer and larger steam header (steam is mostly used since it is deposit removal. A hand blower was basically sootblower canopy (Figure 4). Safety and reli- readily available and less costly than compressed 48 49 ability were the main concerns for this version In 1991, with the support of 11 pulp and paper available in the steam jet is dissipated as the jet of retractable sootblowers. The chain system related companies and the Ontario University passes through the shock waves, and only a frac- required constant maintenance. If it failed and Research Incentive Fund, an industry-university tion of the initial jet energy is delivered to heat broke during sootblower operation, the high collaborative research program on “Recovery transfer surfaces for cleaning. pressure steam in the lance tube would cause Boiler Fireside Deposits” was initiated at the the tube to torpedo forward and strike the boiler University of Toronto. The research program Jameel et al. (1994, 1996) also showed that wall on the opposite side at a high speed. This included, for the first time, projects on soot- with under-expanded nozzles, increasing the up- could cause severe damage to the wall tubes and blower jet fundamentals and issues related to stream steam pressure above the nozzle design injuries to mill personnel who happened to be deposit removal in recovery boilers. The results condition only results in a small increase in jet nearby. of this research program, and of programs that PIP, due to the formation of a larger shock wave. followed over the subsequent two decades, and By extending the nozzle length and redesign- Modern retractable sootblowers replaced the the work of recovery boiler and sootblower ing the nozzle contour (Figure 8b) to allow the stationary motor and chain system in the early Figure 6. Diamond Power’s IK-555 sootblower manufacturers, have significantly advanced version to a traveling motor with dual-rack and (Diamond Power). the basic knowledge of sootblowing: how a a) pinion drives. With such an arrangement, the sootblower jet behaves, how it interacts with motor travels together with the carriage as the tubes and deposits, and identification of the key lance tube is pushed in and retracted from the parameters that influence the deposit removal boiler. The dual-rack and pinion drives have efficiency. Two important innovations that greatly improved the reliability of the lance tube resulted are fully-expanded nozzles, and low movement. This retractable sootblower design pressure sootblowing technology.

Fully-expanded nozzles The design of sootblower nozzles used in indus- b) trial boilers until the mid 1990s ranged from two simple holes drilled into the lance tube to a more sophisticated venturi type, known conventionally as High performance Peak Impact Pressure Figure 7. Schematic of retractable sootblower (Hi-PIP) nozzles (Figure 8a). Hi-PIP nozzles installation in a modern recovery boiler. were the most commonly used nozzles in recov- Figure 8. Nozzle design comparison. ery boilers from the mid 1980s to the mid 1990s. is currently used by the two major sootblower manufacturers, Clyde Bergemann (Figure 5) and The ability of a sootblower to remove deposits Diamond Power (Figure 6). is directly proportional to the jet energy, which is closely related to the peak impact pressure A schematic of multiple retractable sootblowers (PIP) of the jet. PIP depends greatly on the steam installed in a modern recovery boiler is shown in pressure in the lance tube and the design of the Figure 4. Early retractable sootblower (Clyde Figure 7. The number of such sootblowers may nozzle through which the steam jet passes. By Bergemann). range from 60 in a small unit to as many as 150 examining the hydrodynamics of a jet pass- in a large unit. ing through a convergent-divergent section of a Hi-PIP nozzle, Jameel et al. (1994, 1996) showed that the steam jet at the nozzle throat is Technological advances supersonic. As the jet passes beyond the nozzle Figure 9. Peak impact pressure profile comparison (Jameel et al. 1994). Up to the early 1990s, most improvements made throat, it expands and accelerates to a velocity well above the speed of sound, while its pressure on sootblowing technology were limited to the steam to expand fully before it leaves the nozzle, decreases in a manner controlled by the shape design and mechanical aspects of the equip- the shock waves can be substantially minimized, of the divergent section of the nozzle. Because ment; little attention was given to optimizing and this can greatly improve the sootblower of the short distance between the throat and the the sootblower cleaning power, the steam usage performance. As shown in Figure 9, for the nozzle exit, the jet pressure cannot adjust to the and deposit control strategies. The sootblowing same steam consumption, fully expanded noz- ambient pressure; it is then “under-expanded”, technology used in recovery boilers was literally zles provide much greater cleaning power than and must complete its expansion outside the the same as that used in coal-fired boilers. under-expanded nozzles (Jameel et al. 1996, nozzle through a series of shock waves. This Figure 5. Clyde Bergemann’s sootblower (Clyde Kaliazine et al. 1997, Moskal et al. 1993). Bergemann). means that a substantial fraction of the energy 50 51 This discovery led to the development of a new Sootblowing steam is typically taken directly The concept of low pressure sootblowing was generation of sootblower nozzles that allow from the final superheater steam outlet and is first introduced to us by a mill representative the jet to achieve full (or close-to-full) expan- passed through a poppet valve to reduce the pres- during our annual research review meeting at sion before it exits the nozzle. With the help of sure 7 to 24 bar before entering the sootblower the University of Toronto in 2001. Theoretical CFD modeling and efforts made by sootblower feed tube. Since sootblowers consume 3 to 12% analysis, laboratory studies and mill trials manufacturers, the dimension, shape, and posi- of the total superheated steam produced by the (Kaliazine et al. 2006, Tandra et al. 2005, 2008, tion of the nozzles around the tip of the lance boiler, sootblowing with high pressure steam can Tavares & Youssef 2006, Tran et al. 2008) were were refined to make it easier for the stream to be a costly operation. If sootblowers were to op- subsequently performed to evaluate the feasibil- pass through the nozzles, thereby increasing the erate at a lower pressure, for instance 9 to 14 bar, ity of the concept. The results clearly indicated cleaning power of the steam jets even further. there would be a significant economic advantage that low pressure sootblowing is technically Figure 10 shows the most advanced nozzles to pulp mills. This is because low pressure steam and practically feasible. However, in order for presently offered by the two major sootblower is less valuable than high pressure steam, as it low pressure sootblowers to provide a deposit manufacturers: the Diamond Power Gemini noz- can be taken from the steam turbine exit after cleaning power that is comparable to that of Figure 13. Jet force vs. Lance pressure at zles (Diamond Power 2007, Habib et al. 2006) the steam has been used to generate electricity high pressure sootblowers, they require a larger different distances (Tran et al. 2012). and the Clyde Bergemann CFE (Contoured (Figure 12). jet (a higher steam flow rate) to compensate for Fully Expanded) nozzles (Clyde Bergemann the adverse effect of reduced pressure. This can 2012, Jameel 1996). be accomplished by using larger fully-expanded nozzles with a modified design to achieve optimum performance at the lower pressure.

The economic benefits of implementing low- pressure sootblowing technology depend greatly on the amount of additional low pressure steam required to make up for the low pressure, in order to attain the same deposit cleaning power as high pressure sootblowers, and the differential cost between high pressure steam and low pressure steam (Tran et al. 2008). Furthermore, the Figure 14. Average Jet force vs. distance at technology is difficult to implement on existing Lance pressure of 9-11 bar (Saviharju et al. recovery boilers due to the need for re-piping 2011, Tran et al. 2012). the sootblowing steam lines to accommodate the higher steam flow rate. For new recovery boil- Figure 11. Conventional high-pressure soot- At a given lance pressure, the jet force dimin- ers, this is not an issue, as sootblowing steam Figure 10. New generation of sootblower blowing operation (Tran et al. 2008). ishes markedly with an increase in distance from lines can be properly designed and installed nozzles (Habib et al. 2006, Jameel 1996). the nozzle. At a distance farther than 1 m from from scratch. Presently there are three recovery the nozzle, the jet exerts only 10% of its original boilers in the United States and two in Brazil us- force on the same target (Figure 14). At greater ing low pressure sootblowers. A few other new Due to their superior performance compared to distances, the jet strikes the target not only with recovery boilers are expected to follow suit. Hi-PIP nozzles, these fully expanded nozzles are a weaker force but also for a shorter period of now used in virtually all new recovery boilers time. and utility boilers today. Sootblower jet strength In 2007, a collaborative project was initi- The angle at which the jet hits the target also has Low pressure sootblowing technology ated by Andritz Inc., in collaboration with the an impact on jet force (Tran et al. 2012). For a Recovery boilers operate at a superheated steam University of Toronto, Diamond Power and given projected area of the target, the larger the pressure ranging from 41 bar on older units to Clyde Bergemann, to systematically examine impact angle (closer to the normal direction),

as high as 103 bar on modern units (Vakkilainen the strength of sootblower jets in-situ, using a the greater is the jet force, i.e. Fɵ = F90 x ɵ/90,

2009). The high pressure steam from the final jet force measurement system (Saviharju et al. where Fɵ is the jet force exerted on the target at o superheater outlet is passed through a steam 2011, Tran et al. 2012). Four trials were con- ɵ angle and F90 is the jet force at 90 (head-on). turbine to generate electricity. The exhaust ducted under various sootblowing conditions in steam from the turbine, typically at 9 to 17 bar, recovery boilers at two different kraft pulp mills. is subsequently used in various processes in the The main conclusions were that the jet force pulp mill (Figure 11). Figure 12. Low-pressure sootblowing operation increases linearly with an increase in lance pres- (Tran et al. 2008). sure (Figure 13). 52 53 Sootblower Dimension Sootblowing Efficiency the rods, which is directly proportional to the weight of the deposit buildup on the platen. Presently, there are only a few old and small re- Flue gas pressure drop (draft loss) across indi- • Local deposition rate: This system analyzes covery boilers in operation that have sootblow- vidual heat transfer banks, attemperator flow the instantaneous fluctuations of flue gas tem- ers mounted on one sidewall, with lance tubes rate (% valve opening), steam production rate, peratures at the boiler bank exit during single that extend across the boiler width, 8 to 10 m, boiler bank exit gas temperature, ID fan speed, strokes of all sootblowers in the boiler, and to the other side wall, in order to achieve full etc. have traditionally been used to indicate converts them into a time series of raw data coverage. In most recovery boilers, sootblowers the state of cleanliness (or fouling) of recovery that indicates fouling of the tube surfaces are mounted on both sidewalls so that they need boilers, and to decide when to operate the soot- (Adams 2010). This, in turn, can be used to to reach only half of the boiler width. Since the blowers more intensively in the heavy buildup indicate the short-term carryover deposition lance tube rests at about 6” outside the boiler area. While these operating variables provide rate and the efficiency of deposit removal by when it is fully retracted, the travel length (or some indications of boiler fouling, they are not the sootblowers. This approach yields a 3-D distance) is slightly greater than 1/2 of the boiler accurate, and often are influenced by other boiler map of the local fouling rate in the vicinity of width. operating variables. each sootblower that is useful in establishing the sootblowing sequence for minimum steam Figure 16. Images recorded by an infrared Prior to 1990, recovery boilers were relative In the early days, sootblowers were operated use or for plugging prevention. inspection camera showing a deposit being small with a firing capacity under 2500 tds/d. manually by pushing a button beside each soot- • Principle component analysis (PCA): This removed by a sootblower from behind. Sootblowers were equipped with an 3.5” OD blower or in the recovery boiler control room. system uses a multivariate statistical program lance tube and had a travel distance of 5 m. As Based on the experience and logic developed at to monitor the gradual and sudden shifts in boilers become larger and larger in recent years, the mill over the years, the blowing sequence boiler thermal efficiency for individual runs, larger 4” OD sootblowers have been used to and frequency were programmed into the mill allowing day-to-day analysis of changes in support a longer lance tube, as well as to accom- programmable logic controller (PLC) and dis- boiler fouling conditions (Versteeg & Tran modate longer fully-expanded nozzles better. tributed control system (DCS), which automati- 2009). Figure 15 shows the sootblower travel length cally operate the sootblowers “blindly” without (L) versus boiler firing capacity (F) for a few any feedback on the state of fouling or boiler selected recovery boilers (actual data points). cleanliness. In recent years, thanks to the avail- Inspection cameras Since most boilers are designed such that the ability of high speed microprocessors and data The availability of compact, high quality and total heat input to the boiler per unit furnace acquisition systems, several intelligent/smart inexpensive infrared inspection cameras in floor area is constant, roughly under 3.3 MW/m2, sootblowing systems have been developed. recent years has enabled mills to inspect the L is proportional to the square root of F (broken These systems use more reliable fouling indica- state of deposit buildup at different locations in curve). tors to advise boiler operators when and where the boiler during operation, as well as to assess to activate the sootblowers, and to optimize the deposit removal efficiency of sootblowers. Figure 17. Ineffectiveness of a sootblower in The world’s largest recovery boiler to be built at sootblowing operation. These include: Figure. 16 shows images of a piece of deposit removing massive superheater deposits during a the pulp mill in OKI, Indonesia will be equipped being knocked off by a sootblower blowing from blowing cycle (time=mm:ss). with 5” OD and 11.6 m long sootblowers. While • Heat balance: This involves the use of a foul- behind the tube, while Figure. 17 shows little they will be the largest and longest sootblowers ing index based on the steam side heat balance evidence of deposit removal by a sootblower ever installed in recovery boilers, sootblowers of performed on each individual superheater in a different recovery boiler as the sootblower Future outlook this size are rather common in utility boilers. section (i.e. primary, secondary and tertiary moves in and out of the boiler. Sootblowing is a fairly mature technology, superheaters), boiler bank and economizer, and the gas side temperatures (Tandra et al. particularly with respect to hardware. With the Inspection cameras have also helped identify i) trend toward operating recovery boilers with 2010, Uloth et al. 1996). The fouling index the swinging of sootblower lances caused by the is essentially a measure of the efficiency of a fewer and fewer operators in years ahead, it will imbalance of jet forces exiting the two opposing be a challenge for sootblower manufacturers to tube bank to transfer heat from the flue gas to nozzles at the tip of the lance tube, ii) locations the steam flowing inside the tubes. Thus, it can provide equipment that is easy to operate and to where massive deposits buildup in the boiler, iii) maintain, e.g. gear boxes and other moving parts also serve as a key indicator of the sootblow- the cleanliness of the tube surface after thermal ing effectiveness. can be quickly greased; poppet valves and lance shedding events, iv) the swinging of superheater tubes can be easily replaced, etc. • Strain gauge system: This system was pat- platens and the vibration of boiler bank tubes ented by International Paper (IP) (Jones 2001) caused by sootblowing. and has been installed exclusively in recovery Sootblowing efficiency depends on two factors: boilers at IP mills. It uses strange gauges i) the force exerted by the jet on the deposit, and installed on the hanger rods that support su- ii) the strength/tenacity of the deposit on the tube. The peak impact pressure of the sootblower Figure 15. Sootblower travel lengths of selected perheater platens to measure the elongation of jet and the resulting force can now be predicted recovery boilers. 54 55 accurately based on the nozzle design, stream important innovations: fully-expanded soot- References 15. Jameel, M.I., Cormack, D.E. and Tran, H.N., conditions inside the lance tube, and the distance blower nozzles are used in virtually all recovery “Advanced Sootblower Nozzle Design”, U.S. Patent between the nozzle and the deposits. The de- boilers today, and the low pressure sootblowing 1. Adams, TN, “Sootblowing Control Based on No. 5,375,771 (1996). posit strength, however, is not as straightforward technology that has been implemented in several Measured Local Fouling Rate”, TAPPI PEERS 16. Jameel, M.I, “Sootblower Nozzles”, United States Conference, October 18-20, Norfork, VA, 2010 since it varies widely with location in the boiler, new recovery boilers. Patent, No. 5,505,163, April 9, 1996. depending on gas temperature, deposit composi- 2. Arakawa, Y., Ukeguchi, Y., Maeda, T., Sakai, 17. Jones, A.K, “System and Method for Measuring tion, and tube surface condition. These, in turn, The availability of powerful computing systems, T., and Tran, H.N., “Maximizing the Electricity Weight of Deposits on Boiler Superheaters”, United Generation Capacity of Recovery Boilers and are a strong function of black liquor composition superfast microprocessors and data acquisition States Patent, No. 6,323,442 B1, November 27, 2001. Superheater Tube Materials in Japanese Pulp and 18. Kaliazine, A., Piroozmand, F., Cormack, D.E., and boiler operating conditions. Without the systems, and versatile CFD modeling capabil- Paper Mills”, International Chemical Recovery and Tran, H.N., “Sootblower Optimization - Part 2: knowledge of deposit strength, it would be dif- ity in the past two decades has also contributed Conference, Tampere, Finland, June 9-12, 2014. ficult to optimize the deposit removal process, greatly to the advancement of sootblowing tech- Deposit Sootblower Interaction”, TAPPI Journal, 80 3. Barsin, J., “Recovery Boiler Sootblowers”, [11] 201-207 (1997). unless reliable fouling indicators can be devel- nology. High quality infrared inspection cameras Proceedings of Kraft Recovery Short Course, p.219- 19. Kaliazine, A., Cormack, D.E., Ebrahimi-Sabet, A. oped to advise boiler operators when and where have enabled mills to inspect the deposit buildup 227, TAPPI Press (1992). to activate the sootblowers and to optimize conditions in the boiler during operation, and and Tran, H.N., “The Mechanics of Deposit Removal 4. Chapel, R.E, “A Basic Understanding sootblowing operation. helped identify problems with sootblower lance in Kraft Recovery Boilers”, Journal of Pulp & Paper of Sootblowing for Recovery Boiler Cleaning”, Science, Vol. 25, No. 12, p. 418-424 (1999). swinging, and superheater platens and boiler Proceedings of Kraft Recovery Short Course, p. 159- 20. Kaliazine, A., Cormack, D. E., Tran, H.N., Fully-expanded nozzles will remain the pre- bank tube vibrations. 163, TAPPI Press (1987). “Feasibility of Using Low Pressure Steam for ferred choice. Since the present nozzle designs 5. Clyde Bergemann, Contoured Fully Expanded have been optimized for best performance, these As the recovery boiler firing capacity and steam Sootblowing”, Pulp & Paper Canada, 107(4): T80- (CFE) III nozzle brochure, 2012. 84 (2006). nozzles will be used for years to come. New re- parameters have increased markedly in recent 6. Clyde Bergemann’s archive photos. covery boilers will likely adopt the low pressure years, sootblowers have become larger and 21. Moskal, T.E., Burton, M.A. and Jordan, C.A., 7. Clyde Bergemann’s sootblower brochure, http:// sootblowing technology due to its economic longer, and this can present a challenge in terms “Results of Laboratory Testing and Field Trials www.boilercleaning.org/Literature/brochure/SBD/ of Improved Sootblower Nozzles”, Proceedings of advantage. of both sootblower design and operation RS%20Recovery%20Service%20Sootblower.pdf TAPPI Engineering Conference, p. 963-967, TAPPI 8. Diamond Power International Inc., “Gemini Press (1993). We will probably see more research and de- Acknowledgements Nozzle for Improved recovery boiler cleaning”, 22. Pophali, A., Eslamian, M., Bussmann, M., velopment work on CFD modeling of jet-tube Brochure, DPII-9644-1206-00, 2007. Kaliazine, A., Tran, H.N., “Breakup Mechanisms and jet-deposit interactions in the near future. This work was conducted as part of the research 9. Diamond Power’s IK-555 Sootblower Brochure of Brittle Deposits In Kraft Recovery Boilers – A Application of CFD results will help improve program on “Increasing Energy and Chemical Recovery Efficiency in the Kraft Process’, http://www.diamondpower.com/Resources/IK555_ Fundamental Study”, TAPPI Journal, Vol. 8, No. 9, understanding of sootblowing jet behaviors. p. 4-9 (2009). jointly supported by the Natural Sciences and Brochure.pdf Engineering Research Council of Canada 10. Emami, B., Bussmann, M., Tran, H.N., and Tandra, 23. Pophali, A., Emami, B., Bussmann, M., and Tran Summary (NSERC) and a consortium of the following D., “Advanced CFD Simulations of Sootblower Jets”, H.N., “Studies on Sootblower Jet Dynamics and International Chemical Recovery Conference, TAPPI/ Ash Deposit Removal in Industrial Boilers”, Fuel Sootblowers are required to keep heat transfer companies: Andritz, AV Nackawic, Babcock & PAPTAC, Williamsburg, VA, March 29-April 1, 2010. Processing Technology, 105, 69-76 (2013). surfaces clean in order to ensure high steam Wilcox, Boise, Carter Holt Harvey, Cellulose Nipo-Brasileira, Clyde Bergemann, DMI Peace 11. Emami, B., Bussmann, M., and Tran, H.N., 24. Saviharju, K., Kaliazine, A., Tran, H.N., and production efficiency and high runnability of Habib, T., “In-Situ Measurements of Sootblower Jet River Pulp, Eldorado, ERCO Worldwide, Fibria, “A Semi-Analytical Solution for a Compressible recovery boilers. Sootblowing technology used Strengths”, TAPPI Journal, Vol. 10, No. 2, p. 27-32 FP Innovations, International Paper, Irving Turbulent Axisymmetric Jet”, SIAM Journal on in recovery boilers originated from that used in Applied Mathematics 72(1), 85-98 (2012) (2011). Pulp and Paper, Kiln Flame Systems, Klabin, coal-fired boilers. It started with manual clean- 25. “STEAM: Its Generation and Use”, 41st addi- MeadWestvaco, Stora Enso Research, Suzano, 12. Haaga, K., “Development Path of Recovery ing with hand lancing and hand blowing, and tion, Edtied by Kitto, J.B., and Stutz. S.C., Babcock & Tembec, Tolko Industries, and Valmet. The Boilers From Small Ones to Big Ones and How evolved slowly into on-line sootblowing using We Made it Happened?”, the 8th International Wilcox Company, (2005). authors also wish to acknowledge Diamond retractable sootblowers. Colloquium on Black Liquor and Biomass to 26. Tandra, D., Kaliazine, A., Cormack, D. E., Tran, Power International for its past support of the Bioenergy and Biofuel, Belo Horizonte, Brazil, March H.N., Jameel, I., McCable, D., Mott, D., “Mill Trial Prior to 1991, most improvements to sootblow- sootblowing research program at the University 19-23, 2013. on Low Pressure Sootblower Performance in a of Toronto. ing technology were limited to the design and 13. Habib, T, F., Keller, D.L., Fortner, S.R., Recovery Boiler”, TAPPI Engineering Conference, mechanical aspects of the equipment. Since “Sootblower Nozzle Assembly with Nozzles Having Philadelphia, August 2005. then, intensive research and development has Different Geometries”, US Patent No. 7028926 B2 27. Tandra, D., Kaliazine, A., Cormack, D.E., and focused on sootblowing jet fundamentals and (2006). Tran, H.N., “Interaction between Sootblower Jet deposit removal in recovery boilers. The results 14. Jameel, M.I., Cormack, D.E., Tran,H.N., and and Superheater Tubes in Recovery Boilers”, Pulp & have provided much insight into sootblower jet Moskal, T.E., “Sootblower Optimization - Part 1: Paper Canada, Vol. 108, No. 5, p. 43-46 (2007). hydrodynamics, how a sootblower jet interacts Fundamental Hydrodynamics of a Sootblower Nozzle 28. Tandra, D.S, Hinman, J., Olson, M., Breaux, S., with tubes and deposits, and factors influencing and Jet”, Proceedings of 1993 TAPPI Engineering “Energy Efficiency and Cost Reduction Through its deposit removal efficiency, and led to two Conference, and TAPPI Journal, 77 [5], 135-142 the Implementation of Low Pressure Sootblowing (1994). 56 57 System”, TAPPI PEERS Conference, August 24 – 27, 41. Vulcan Soot Cleaner Handbook, 1921. Portland, OR, 2008. 29. Tandra, D.S., Manay, A., Edenfield, J.A., “The Use of Energy Balance Around Recovery Boiler Heat Exchangers To Intelligently Manage Sootblower Operations: A Case Study”, TAPPI PEERS Conference, October 18-20, Norfork, VA, 2010. 30. Tavares, A. and Youssef, S., “Retrofitting Existing Sootblower System to use a lower Pressure Steam Source”, TAPPI Engineering Conference, Atlanta, GA, November 6-8, 2006. 31. Thabot, A., Tandra, D.S., Oehrig, B., “The Use of Intelligent Sootblowing System to Deal with Various Operational Issues Related to Heavy Backpass Fouling”, Power Gen International Conference, December 14 – 16, Orlando, FL, 2010. 32. Tran, H.N., “Kraft Recovery Boilers - Chapter 9: Upper Furnace Deposition and Plugging”, edited by T.N. Adams et al, p.247-282, TAPPI Press (1997). 33. Tran, H.N., “Fouling of Tube Surfaces in Kraft Recovery Boilers”, Proceedings of the 40th Anniversary International Recovery Boiler Conference, Porvoo, Finland, p. 91-102, Finnish Recovery Boiler Committee, May 12-14, 2004. 34. Tran, H.N., Tandra, D. and Jones, A.K., “Development of Low Pressure Sootblowing Technologies”, Pulp & Paper Canada, 109:12, T129-T134 (2008). 35. Tran, H.N., and Arakawa, Y., “Recovery Boiler Technology in Japan”, Proceedings of TAPPI Engineering Conference, TAPPI Press, 2011. 36. Tran H.N., Pophali, A., Emami, B., Bussmann, M., and Miikkulainen, M., “Sootblower Jet Strength In Kraft Recovery Boilers-Part 2”, TAPPI Journal, September, Vol. 11, No. 9, p. 31-35 (2012). 37. Uloth, V.C., Markovic, C.M., Wearing, J.T., and Walsh, A., “Observations on the Dynamics and Efficiency of Sootblowing in Kraft Recovery Furnaces: Part II – Efficiency”, Pulp & Paper Canada, 97:7, T223-T226 (1996). 38. Vakkilainen, E., “Boundaries of Recovery Boiler Development”, Proceedings of the 45th International Recovery Boiler Conference, p.7-15, Finnish Recovery Boiler Committee, Lahti, Finland, June 3-5, 2009. 39. Versteeg, P., and Tran , H.N., “Using Principal Component Analysis to Monitor Kraft Recovery Boiler Fouling”, TAPPI Journal, November, p.22-28 (2009). 40. Villarroel, R., Gonçalves, C., and Tran, H.N., “Experience of Screen Tube Damage Caused By Falling Deposits in Kraft Recovery Boilers”, Pulp & Paper Canada, 106 (12): T273 - T276 (2005).

58 59 review of recovery boiler superheater material studies

Martti Mäkipää

Pekka Pohjanne VTT Technical Research Centre of Finland

Introduction

Higher steam values, higher dry solid contents in conservative. Economic reasons and environ- black liquor and mill closure have led to major mental constraints have put pressure to increase changes in the flue gas and superheater deposi- the energy efficiency of the recovery boilers tion chemistry in recovery boilers. The effects and traditionally this is achieved by increasing of these trends on superheater corrosion and the steam values i.e. temperature and pressure materials selection are considered with refer- Figure 1). The trend to increase electricity ence to literature data. There is a wide variety of generation is forcing to increase steam values available superheater materials to choose from. further. It is calculated that if the steam pressure There are even few newer materials that should and temperature can be increased to 104 bar and be considered. The operating conditions and 520 °C from typical current values of 80 bar deposit properties must be looked at when mate- and 480 °C the electricity generation capacity rial selection is done to ensure operational safety of the recovery boiler plant increases about 7% and high availability. (Vakkilainen 2004).

The black liquor recovery boiler has multiple A key issue for the recovery boiler, as mentioned roles in a pulp and paper mill which makes it earlier, is the high availability need. And when a critical component of a mill. The black liquor recovery boilers are designed one of the most recovery boiler is fired with evaporated black difficult questions that arise is; what kind of ma- liquor that is the spent cooking liquor from the terials should one use for different parts of the kraft process. It burns the organic material that recovery boiler. In future this is expected to be- is contained in the black liquor to generate high come even more complicated since the demand pressure steam and nowadays also increasing to increase steam value increase the corrosion amounts of electricity; it regenerates the used risk especially in the superheaters. When select- chemicals in black liquor, and reduces some ing superheater materials one needs to consider waste streams in an environmentally friendly several factors (Bruno 1997, 1999). Set design way (Vakkilainen 2005). conditions (pressure, temperature, corrosion, erosion) determine the boundary conditions. The modern black liquor recovery boiler tech- Tube material mechanical properties (strength) nology originates from 1930s and after that determine the tube thickness. Manufacturing a lot of changes have taken place. The most properties (weldability, need for heat treatment) prominent fact characterizing the development determine the required time to manufacture a of recovery boilers is that their features have piece of heat transfer surface. Cost of material changed rather slowly; when any outage will and relevant manufacturing cost give the total cost a lot of money, the technology tends to be cost for heat transfer surface. Often also material

60 61 availability (time the material can be delivered position. Oxidation behaviour is affected by Sulfidation tubes. The corrosion conditions in question may for manufacturing) plays a significant role. combustion products (e.g. carbon monoxide, be modelled in laboratory exposing materials to Typical deposition conditions prevailing in the One should not forget that variation of process carbon dioxide) and generally the oxidation molten mixtures alkali sulphates, carbonates, superheater area of recovery boilers until late values during operation (e.g. temperature rate in flue gases is higher than that in the air. sulphides and chlorides. Klöwer and White nineties, and still common in most of the op- spikes, overheating and liquor chemistry) can Respectively studies have shown that oxidation (1995) found that under deposit of 45% Na S, erating boilers, have been discussed by various 2 significantly affect corrosion rate and strength of rates in moist environments are higher than in 21% Na CO , 21% of Na SO and 13% NaCl workers in detail (Tran 1997, Hupa et al. 1990, 2 3 2 4 some materials but not others. And in the case of dry environments (Figure 2). Temperature limits and at temperatures ranging from 550 to 650 Bruno 1997, see also Mäkipää et al. 2001b). The new materials high temperature aging may cause for different type boiler materials above which °C austenitic stainless steels and high nickel- ash depositing in the superheater area consists unexpected failures. oxidation is excessive for 25-30 service years high chromium steels suffer corrosion. Silicon of both fume particles and carryover particles. are presented in Table 1. Oxidation resistance is containing alloy 45-TM performed best, as may Fume particles are composed largely of alkali improved by alloying e.g. with chromium, sili- be expected considering the solubility behaviour sulphates, with lesser amounts of alkali chlo- con and aluminium. High temperature oxidation of silica in the fluxing conditions in question. rides and alkali carbonates. Carryover particles is not normally considered to be a problem in Proper boiler design and high-solids operation, are, in the main, black-liquor residues. In terms recovery boilers due to moderate steam values. however, effectively eliminate carry-over corro- of chemical composition, the greatest difference sion. High solids firing enables good mixing and between carryover particles and fume particles eventually more complete combustion of black is the higher content of alkali carbonates in car- Table 1. Temperature limits for different type liquor droplets entrained in the flue gas flow. In ryover particles. Superheater tube deposits ide- boiler materials above which oxidation rate high solids boilers there seems to be low sulfide ally consisting of sodium sulphate and sodium exceeds 25-30 year service life (French 2014). content in deposits (Raukola & Haaga 2004). carbonate that are mutually soluble in the solid Figure 1. Maximum steam temperature and Material Temperature state, may form melts but in the temperature As a concluding note, high chlorine pressures pressure of recovery boilers in the world vs. °C range about from 826 to 884 °C and higher. advocating active oxidation may be generated in their start-up year (Bruno 1997). Sodium sulphide in the deposit composition Carbon steel (SA210, 450 the presence of a thin molten salt film of complex may decrease the first melting point of down to SA106) composition (Mäkipää & Koivisto 1999). This 715 °C. Recovery boiler ash in contact with the Superheater corrosion Carbon steel -1/2Mo 480 corrosion morphology might relate to operation tubes is thus normally solid. Superheater tube conditions of a peculiar kind, where alkaline The superheater corrosion behaviour depends on (SA209) material wastage rates appreciably less than 0.1 deposit conditions prevail but SO3-pressure in three factors (Salmenoja 2004): 1) The tube sur- Low alloy steel 1-1/4Cr- 550 mm/a are typical under optimal conditions. flue gases is at least temporarily high. Then, face temperature, 2) the deposit chemistry and 1/2Mo (T11, P11) molten alkali chloride – alkali sulphate film Significant corrosion risks are associated with the 3) the local atmosphere at the vicinity of tubes. Low alloy steel 2-1/4-1Mo 580 will react vigorously with flue gas atmosphere high sulphidity of the black liquor (high value of (T22, P22) and protective oxide scales to release chlorine, (S/Na+K)), which might lead, especially when Typical steam temperatures in current recovery as previously shown in (Karlsson et al. 1990). Austenitic stainless steel 700 the temperature in the lower furnace is low, to boilers are ranging from 480 to 500 °C, i.e. sub- As far as known, no metallic material including 18Cr-8Ni (304, 321, 347) high partial pressures of sulphur oxides (up to stantially lower than that of the coal fired boilers high molybdenum nickel alloys is completely 1000–5000 vppm of SO , 100 vppm of SO ). where superheating temperatures in top-of-the- 2 3 resistant under such conditions. line plants exceed 600 °C. In comparison to coal If the material temperature is high, high partial fired boilers kraft recovery boilers have higher pressures of sulphur oxides cause sulphidation- As discussed by Bruno (1997) and Ahlers (1987), rates of alkali metals, chloride in gaseous form oxidation of superheater tubes. The mechanisms among others, more favourable corrosion condi- and often highly reducing conditions caused by of sulphidation-oxidation in recovery boilers are tions occur in the superheater in boilers operated carryover particles, which alter superheaters to basically the same as in power boilers (Bruno in the high-solids firing mode. In modern boilers deposit induced corrosion. On the other hand 1997). This mode of corrosion is successfully firing high-solids black liquor, temperatures in levels of heavy metal chlorides causing cor- counteracted by using austenitic stainless steel the lower furnace are high, from which it follows rosion problems in waste to energy plants are or composite tubing (Oldestam 1982, Ahlers that, under steady-state conditions, relatively 1987), and/or by shielding the lower parts of typically low. low concentrations of SO2 are encountered in the superheater from direct heat radiation by the upper furnace. Maximum steam conditions Oxidation applying screen tubes in conjunction with the that were reached during seventies and eighties “bullnose”, which is the protruding part of the were in the range from 480 to 490 °C. Low al- High temperature oxidation occurs when metals water wall. loy steel tubing (10CrMo910) with lower bends are exposed to temperatures above 300 °C in made of austenitic stainless steel of type 304 Carry-over deposits, composition of which gases containing more than 1 vol.-% oxygen. were used with success (Jaakola & Rose 1983). Oxidation resistance depends on operational Figure 2. Effect of moisture on isothermal are close to black liquor composition may, if oxidation of 9% Cr P91 steel in N –1 vol% O – x More lately, gradually increasing mill closure conditions like temperature, gas composition 2 2 molten, cause excessive material wastage on vol%H O mixtures at 650 °C (Ehlers et al. 2006). leading to enrichment of chloride and potassium and moisture level as well as on material com- 2 the windward side of the leading superheater 62 63 in the flue gases did put an end to this develop- cipitation reactions may take place in thin salt mersed into a fully molten alkali chloride - alkali a) ment. The steam conditions from the high level films and, depending on the melt and gas phase sulphate salt mixture of eutectic composition in achieved e.g. in Scandinavia backed down to the composition, may therefore radically change the 520–530 °C. Materials used in the test were range of 430-450 °C. corrosion potentials and the ranking of various low-alloy steels, stainless steels and nickel al- alloys and steels. loys: 13CrMo44, 10CrMo910, AISI 304, HR3C, Deposits AC66, 45TM, YUS 170 TBM S-C, Sanicro High cost of 25% Cr -grade austenitic materi- 28 and Inconel 625. For low-alloy steels and Recent trends in kraft recovery process as well als and their tendency to localised corrosion in austenitic stainless steels the characteristic cor- as trends in fuel technology related to produce the presence of alkali chlorides has motivated rosion attack mode was intergranular attack and more electricity using renewable fuels has led a number of materials development and test- growth of oxide scales. More corrosion resistant to major changes in operation conditions. In ing programmes in North America, Japan and steels and alloys suffered from pitting attack. specific, there is an increased risk of significant Europe. However, methodologies used for Corrosion morphologies observed had a close alkali chloride deposition in the superheater area corrosion testing in laboratory or for material resemblance to practical corrosion morpholo- of kraft recovery and biomass fuelled boilers exposures in full scale using cooled probes are gies in cases involving alkali sulphate - alkali b) (Jokiniemi et al. 1996). Depending on the alkali not uniform. The only standardised hot corro- chloride melt attack. Corrosion morphologies, metal volatilisation in the furnace, and sulphur sion test method, which applies thin salt layer which were obtained in 44 or 65 h testing for and chlorine contents of the fuel, condensation covering replenished with relatively regular low-alloy steels, correspond reasonably well of vaporised alkali chlorides on flue gas particles intervals, does not reproduce corrosion rates with corrosion morphologies observed for low- or directly on heat transfer surfaces may cause observed in practice in cases, where salt cover alloy steels in various corrosion failure cases, accelerated corrosion in the superheater area. and atmosphere goes to reactions, which im- where excessive material wastage has occurred. Known deposit corrosion mechanisms include poverish corrosive species or increase melting Short-term exposures give some additional in- corrosion by solid alkali sulphates or alkali point. Corrosion rates more close to practical formation about the initiation stage of the rapid chlorides and molten salt attack by low melting ones have been reported to be achievable us- general material wastage of low-alloy steels and mixtures of alkali sulphates, alkali chlorides ing so-called crucible method. This method is austenitic stainless steels as well. As shown in and alkali carbonates. Contact of tube surface rather simple to apply. In a recent variation of Figure 3a for 10CrMo910, and in Figure 3b for with molten alkali salts causes rapid corrosion. the crucible method, metallic specimens have austenitic stainless steel type AISI 304, deep Active oxidation can affect superheater tubes been exposed to flowing flue gas atmosphere as intergranular attack occurs causing individual even at temperatures lower than the first melt- buried in synthetic or real deposit, which may grains to detach from the substrate to form in- Figure 3. a) Cross-section of 10CrMo910 and ing temperature (FMT). Therefore operating be partially molten at the experimental tempera- dividual islands of metal. These results suggest b) cross-section of AISI 304 stainless steel after the recovery boiler without gaseous chlorine at ture (Kawahara et al. 1997). Test temperature that the ranking of austenitic stainless steels is 20 h exposure in eutectic alkali chloride melt superheaters is essential. As sulfidity increases (6.5 wt-% NaCl , 59.0 wt-% Na SO and 34.5 should be chosen so that the content of melt in largely dependent on their relative susceptibility 2 4 the polysulfide content of deposits on the heat the bed is less than about 30% allowing for the to grain boundary attack. YUS 170 material had wt-% KCl) at 525 °C (Oksa & Mäkipää 2001). transfer surfaces can increase which increases gaseous atmosphere an access to the corrosion endured test well below the melt line. Material corrosion (e.g. Yli-Olli et al. 2007). front, typically leading to test temperatures in surface, which had been above the melt, was bility of silica is low in sulphate-chloride melts the range of 550–560 °C. Specimen cooling may covered with small corrosion pits. Thus, steels independently of the melt acidity. Furthermore, Molten salt attack be applied to simulate temperature variation like YUS 170 without any signs of grain bound- it has been noted that it is rather basic fluxing by ary attack in the half-immersion test, should Na O than acid fluxing by SO that is respon- Sodium and potassium may enhance formation and temperature gradient. Materials evaluation 2 3 perform better than some other steels of similar of ash deposits with low melting points 510–600 is based on post-experiment examination of sible of protective scale dissolution in sodium composition but prone to grain boundary attack °C. Minimum melting points in deposits consist- specimens and their cross-sections. However, carbonate- sodium sulphide - sodium sulphate in the test. ing of complex mixtures of alkali and alkaline this method does not allow with ease to separate - sodium chloride salt mixtures found on BLRB earth metal salts are in the range of 490–510 between reactions of solid and liquid phases superheater tubes under carryover conditions Initiation of grain boundary attack may be °C, which is much lower than 550–575 °C, i.e., in the deposits, or to analyse the effect of gas (Klöwer & White 1995). Accordingly, various suppressed under various conditions, basic or material temperatures related to steam values phase composition on the corrosion reactions. silica forming alloys have been developed for acidic, by alloying silicon. Silicon is known to 500–525 °C. In various practical corrosion Electrochemical measurements in combination use in superheaters of waste incinerators, and be useful in two ways. Silicon enhances chro- failure cases one finds that corrosion environ- with crucible tests are in principle very helpful, recommended to be used also in high-pressure mium diffusion in the alloy accelerating protec- ment has actually been a low melting salt film but these measurements are too intricate to be recovery boilers. Half-immersion tests made tive chromium oxide scale formation. On the spreading on the superheater tube surfaces applied widely. put this proposition in some doubt: silicon-con- other hand, silica may form a thin amorphous leading to excessive material wastage in spite taining nickel alloy was found to be susceptible silica subscale, which is not soluble into alkali of reasonable material temperatures (Mäkipää In the study of Oksa and Mäkipää (2001) modi- to grain boundary attack on specimen surface sulphate melts, and is practically impermeable & Koivisto 1999, Crowe & Youngblood 1998, fied crucible test method was introduced that areas located above the immersion line (Oksa to chromium, nickel, chlorine etc. Furthermore, Fujikawa et al. 1999). Complex dissolution pre- simulates this corrosion condition. Metallic & Mäkipää 2001). In opposition to the case of specimens exposed in air atmosphere as half-im- it has been claimed by various workers that solu- alkali-sulphate-chloride-carbonate-sulphide 64 65 melt, see (Klöwer & White 1995), dissolution a) 1997, Salmenoja & Mäkelä 1999, Mäkipää et al. iron oxide to from stoichiometric alkali ferrite

of chromium and iron into alkali sulphate-alkali 2001a, 2001b). However, the HCl content in the (Na,K)FeO2 and gaseous chlorine (Mäkipää chloride melt seems to create extremely basic flue gases of such boilers is rather low. Hence, & Fordham 2000). Instead, initial formation

conditions in present test conditions, counteract- in exposure experiments aiming to simulate of various ferrates (of type (Na,K)2FeO4) and ing beneficial effects of silicon. such boiler conditions another source of chlorine various nonstoichiometric gamma-ferrites and

(e.g. chlorides) rather than gaseous HCl should polyferrites (i.e. of type (Na,K)Fe11O17), ac- The corrosion behaviour of pure metals and be used. Alkali chloride sulphation reactions companied by adsorption of liberated chloride commercial high-temperature alloys exposed in recovery boilers are slow or non-existent at ions on oxide surfaces would be more plausible to recovery boiler superheater tube deposits has temperatures below 800 °C. Accordingly, some explanation for experimental findings. Contrary been studied recently (Mayer 2012). Theoretical workers exclude active oxidation as an operat- to the case of alkali ferrites, thermodynamics principles developed previously by Rapp (2002), ing corrosion mechanism in recovery boilers, strongly favour formation of alkali chromates, Rapp and Zhang (1986) seem to apply. Alloys whereas rapid chloride induced low-temperature molybdates and manganates as corrosion prod- and steels showing negative solubility gradient oxidation corrosion, active oxidation (Spiegel et ucts under oxidative conditions. Using X-ray in the molten salt film showed poor corrosion b) al. 1997, Reese & Grabke 1993), may occur in diffraction, sodium chromate (Na2CrO4) was performance. All of the stainless steels studied fluidised bed at material temperatures which are found in the phase composition of corrosion had negative solubility gradients but less alloy- well below the first melting point of the depos- products detached from cooled probe exposed ing reduced the effects of mutual synergistic its, then being due to alkali chloride sulphation for 2000 h in recovery boiler superheater area reactions between iron oxides and refractory reactions (Salmenoja & Mäkelä 1999, 2000). It (Mäkipää et al. 2001). The presence of chromate oxides. Overall, the best alloys were high nickel, has been found, however, that alkali chlorides has been reported several times by Japanese high chromium, low iron alloys where nickel induce active oxidation even under SO2-free workers. provides molten salt resistance and chromium atmospheres (Mäkipää et al. 2001b, Hilden et reduces sulphidation. The experimental results al. 1999, Mäkipää & Fordham 2000, Mäkipää When instead of oxides of the alloying elements are in good accord with previous work (Mäkipää & et al. 2001c, Skrifars et al. 2008, Lehmusto their metallic form is considered, affinities of & Koivisto 1999, Fujukawa 1999, Oksa & et al. 2013). Material wastage rates of low alloy the gross corrosion reactions referred to above Mäkipää 2001). steel specimens exposed to alkali chlorides un- are greatly enhanced. Accordingly, metal and der oxidative atmospheres are excessively high alloy specimens prepared using standardised One may consider closed volumes of molten salt Figure 4. a) Large corrosion pit on the surface (> 0.1-10 mm/a) starting from temperatures of methods, respectively the formers of a very thin film and metal surface in the case of higher al- and b) cross-section from the bottom of a large about 450 °C and depending on the alkali chlo- oxide (Cr O ) scale only, may be suspected to loys showing no intergranular attack. One finds 2 3 corrosion pit (BE-image) on UNS N08028 after ride and actual gaseous atmosphere composition underperform in laboratory testing or boiler that chromium and iron will be depleted from 20 h exposure in eutectic alkali chloride melt and flowing conditions in different corrosion exposures using ground specimens attached in the alloy (Figure 4). Molybdenum is considered (6.5 wt-% NaCl , 59.0 wt-% Na SO and 34.5 morphologies. cooled probes, see Figure 5 and Figure 6. beneficial alloying element because it stabi- 2 4 wt-% KCl) at 525 °C (Oksa & Mäkipää 2001). lizes chromium oxide instead of alkali chromate Note: Selective corrosion of chromium (and As noted by Keiser et al. (2009) there is, how- ions as stable corrosion product in oxidative iron) and enrichment of nickel (and iron) at the ever, seemingly no consensus between various conditions. However, it is notable that two high- pit bottom. groups on the relative corrosivity of various chromium alloys with moderate molybdenum alkali chlorides and on the critical temperature content suffered from serious pitting corrosion at molten salt corrosion are reasonably consistent interval of corrosion reactions. One may write and above of the immersion line as studied Oksa with each other. The prevailing corrosion mode a number of hypothetical corrosion reactions & Mäkipää 2001. Again, this may be attributed is mostly localised intergranular, for low alloy suitable, in principle, also for their experimental to the specific test conditions, where extremely and austenitic stainless steels (Figure 3), and verification in laboratory, see Rapp & Grabke et basic conditions may be locally created. In a pitting for highly with chromium and nickel al- al. (1993), Spiegel et al. (1997), Skrifvars et al. modern recovery boiler operated in high solids loyed materials (Figure 4). For the latter the iron (2008), Lehmusto et al. (2013), among others: firing mode, carry-over of unburnt black liquor content is critical. is minimal and basic deposit conditions may Na2CO3/2NaCl(s,g) + Fe2O3(s) +O2(g) → develop. However, considering ranking tests for Active oxidation / Low temperature oxidation Na2Fe2O4(s) + CO2(g)/Cl2(g) (1) Figure 5. TGA-curves showing weight gain in superheater parts exposed to carryover deposi- under thermal cycling conditions K2CO3/2KCl(s,g) + Fe2O3(s) + O2(g) → various cases. Specimens embedded in NaC1 or tions, test conditions should be modified to be K Fe O (s) + CO (g) (2) A number of corrosion failures concerning recov- 2 2 4 2 KC1 beds in a Pt crucible. Total exposure time less basic. 4K CO + 2Cr O + 3O2(g) → 4K Cr O + ery boilers and biomass-fuelled power boilers of 2 3 2 3 2 2 4 35 h (Mäkipää et al. 2001a). 4CO2(g) (3) In summary and with reference to the excellent the pulp and paper industry are reported where the morphology of the corroded tubes exhibits One must however, be aware that the affinities of review prepared by Keiser, Kish and Singbeil Validated thermodynamic data do not sup- Cl-enrichment at their metal surfaces (Bruno the hypothetical corrosion reactions considered (2009) major findings of the studies related to port the view that alkali chlorides react with 66 67 gressiveness (Malkow et al. 2001). Corrosion corrosion attack as exposed to both sodium and reaction schemes in simplified theoretic models potassium chlorides, (Mäkipää et al. 2001b, may fail in practical work. Seemingly contradic- Mäkipää & Fordham 2000, Malkow et al. 2001), tory results concerning corrosiveness of sodium whereas various high chromium alloys corroded and potassium salts as pure or in their mixtures only if exposed to potassium chloride. Most may be obtained by various laboratories as resistant in both sodium and potassium chlorides noted before. Experiments using thermal ramps are high-nickel, high-molybdenum alloys of or thermal cycling may result to very different type Alloy 625. This finding is consistent with corrosion morphologies as compared to experiments at constant temperature, Figure 7.

In particular, due to the higher thermodynamic stability and to the higher vapour pressure of sodium iron oxide and complex sodium iron chloride, respectively, as compared to the corresponding potassium compounds, metallic specimens of low alloy steels (10CrMo910 and 15Mo3) tend to lose some weight when exposed to sodium chloride whereas they show weight gain and thick multi-layered oxide scales Figure 9. Overall thermogravimetric weight when exposed to potassium chloride, Figure changes of various high alloyed steels during 8. Thermodynamic background of the very the 35 h exposure in alkali chlorides measured complicated corrosion reactions in question is after the initial weight losses happened until Figure 6. SEM imaging of Sanicro 28 specimen surfaces exposed to different alkali chlorides for 35 h known inadequately, and should be carefully cooling starts (Malkow et al. 2001). at 530 °C. Left column: NaCl, Right column: KCl. Atmospheric conditions; Top row: Air, Bottom row: scrutinized. As an example of the usefulness of Moist air. On the basis of EDS-analyses of the corroded surfaces, evaporation of complex alkali metal such exercises for practical corrosion science, the view that corrosion resistance of high- -containing species was obvious under the moist air atmosphere (Mäkipää et al. 2001b). the works on vaporised species in the system performance alloys under severe alkali chloride Cr-O-Cl-H may be referred to (Liu et al. 2009, deposition conditions largely correlates to the above are themselves critically dependent of gamma-iron oxide (maghemite), thus having the Jiao et al. 2011, Asteman et al. 2002). See also sum of nickel, chromium and molybdenum material temperature and gas phase composition potential to affect stress and growth properties Figure 5 and Figure 6. contents in the alloy, see e.g. Figure 9. both. Actual, or local, values of pO2/pCO2 and of the oxide scale (Aleksander 1963, Davies

pCO/pCO2 are highly important parameters in et al. 1973, Boettger & Umland 1974). Alkali theory, but experimentally hardly controllable. chlorides possess this ability even when present Material selection for superheaters Thus, reaction products formed in corrosion at ppm-level. Exposure to alkali chlorides even Common understanding is that corrosion of scales may dissociate and form again locally and temporarily may enhance oxide scale growth superheaters is not a problem in boilers where timely, as well in full-scale as in experimental (Boettger & Umland 1974, Ross & Umland steam temperature is below 450 °C. According + conditions in laboratory. Alkali metal-, H3O , 1984). Water vapour enhances, and carbon to (Vakkilainen & Pohjanne 2010) typical + NH4 and chloride ions are known to stabilize dioxide somewhat reduces alkali chloride ag- recovery boilers fire black liquor at a dry solid of 80% and operate at a steam pressure of 80 a) b) bar and a temperature of 480 °C to reduce superheater corrosion and ensure boiler availability. In the future the situation is expected to change. Figure 8. Cyclic oxidation of 15Mo3 in synthetic Reason to this is the aim and need to increase air at 440–575 °C when embedded in solid alkali CO2 free, green electricity. This requires that chlorides (Malkow et al. 2001). the steam values are increased maybe even to a level of 160 bar and 550 °C. At present a number

of modern boilers exist in Sweden, Finland and The experimental findings referred to above Japan that operate at more than 100 bar and 500 experimental are in line with earlier results, e.g. °C. No current recovery boiler operates at higher Figure 7. SEM images of the scale morphologies of 10 CrMo 9 10 exposed to KCl a) in synthetic air with those of Alexander (1963), Davies et al. pressure than 120 bar and higher temperature cyclically at 480–530 °C and b) isothermally at 480 °C (Malkow et al. 2001). (1973, Boettger and Umland (1974), and also than 520 °C. of Shinata and Nishi (1986). Austenitic stainless steels have been found to suffer from general 68 69 The main steam temperature is the main param- Carbon steel and low alloyed steels either as mono or compound tube. Compound changes in the flue gas and superheater deposi- eter that affects the choice of superheater materi- tubes and the grade X1NiCrMoCu31-27 tion chemistry in recovery boilers and increased als, this because the corrosion rate increases as Carbon steel is only used in the coldest part of the (27Cr31Ni3.5Mo) is targeted to European mar- super heater corrosion failures. Following the superheater steam temperature increases, superheaters, their corrosion rates are acceptable ket whereas mono tubes are supplied to domes- corrosion modes may be present either alone Figure 10. However it should be noted that the if steam temperatures are below 400 °C (corre- tic customers (Nippon 2014). Composite tubes or in combination; 1) molten salt attack, 2) low superheater corrosion rates are not uniform, sponding metal temperature 430-450 °C) (Sharp are attractive, because they are not susceptible temperature oxidation / active oxidation and 3) but depend strongly on the local conditions. 2007). According to Vakkilainen & Pohjanne, to steam/water side stress corrosion cracking. sulfidation under reducing deposit conditions. Superheater surface temperature can be tens of (2010) typical primary superheater materials, The experiences have been mainly positive. In High sulphidity of the black liquor (high value degrees higher than the bulk steam temperature, when they are protected from direct furnace example cases Unifuse 310 welded coating is of (S/Na+K)), which might lead, especially Figure 11. A rough estimate is that in areas were radiation are carbon steel (SA/A-210 Gr A-1, reported to last more than two years without when the temperature in the lower furnace is radiation is present the metal temperatures are S235JRG2) which can be used up to 350–400 visible damage. In the same conditions the old low, to a high partial pressure of sulphur. If the about 50 °C higher than the steam temperature °C. Secondary and tertiary superheater materi- steel (T11) corroded 3.9 mm/year (Lai & Blogg material temperature is high or even moderate, (LaFond et al. 1992). This is taken into account als contain often 1-3% Cr to improve corrosion 2003). high partial pressures of sulphur oxides may when positioning superheaters. The corrosion resistance. Examples from these are 13CrMo44, cause sulphidation-oxidation of superheater T11/10CrMo5-5 and T22/10CrMo910 which are risk of hottest superheaters can be reduced High Cr – high Ni – Fe base alloys tubes. The mechanisms of sulphidation- oxida- by placing behind the nose or protecting it by used up to 400–480 °C. These are characterized tion in recovery boilers are basically the same screen tubes. Corrosion risk can also be reduced by significantly increased strength at elevated Sanicro 28 (UNS N08028) is currently used in as in power boilers. This mode of corrosion is by controlling deposit chemistry and avoid- temperatures and elevated scaling temperature. recovery boiler superheaters as compound tube. successfully counteracted by using austenitic ing condensation of potassium chloride i.e. by However, their corrosion resistance is quite It has a good corrosion resistance because of high stainless steel or composite tubing. Excessive removing non-process elements (NPE) (potas- limited and therefore they are useful at tempera- Cr- (27%) and Ni- (31%) content, other alloying carryover of unburned black liquor droplets sium and chlorine) from the chemical recovery tures significantly below the FMT (Sharp 2010). elements being 3.5% Mo and 1% Cu. Sanicro 28 may cause conditions in superheater deposits process (Tuiremo & Salmenoja 2001, Kaneko et Similar guideline is presented in Clement & has low carbon content to which also Japanese which are locally reducing, and sifts the deposit al. 1998, Jaakkola 2009). Grace (2009) where it is mentioned that T22 can tube manufacturers are targeting. AC66 (1.4877) composition to the one where the melt is rich be operated at steam temperature of 496 °C if is like Sanicro 28, but it doesn’t contain Cu and in alkali sulphides. The melting point might chloride levels are kept low. it is Nb-stabilised because of the higher carbon be suppressed to the extent that rapid material content. This material has not been widely used wastage occurs. Even polysulphide rich melts Stainless steels in recovery boilers, but in comparable corro- may become stable in the salt system in ques- sion conditions at biomass boilers it has been Corrosion resistance is increased with chromium tion in presence of elemental carbon, leading to proven as one of the best materials. Alternative alloying and martensitic (X10CrMoVNb9-1, extremely rapid material wastage. grades from Nippon Steel & Sumitomo Metal X20CrMoV11-1) and austenitic (X1NiCr-Mo- Corporation (NSSMC) are UNS N08825 and Cu31-27-4, NiCr22Mo9Nb) steels are used in Another typical corrosion problem is molten salt HR11N (UNS N06811) (Nippon 2014). hottest and most corrosive superheaters up to attack, which is most likely to occur when the 500–520 °C (Vakkilainen & Pohjanne, 2010). boiler is operated outside the designed capac- Figure 10. Effect of temperature on superheater AISI 347 has been widely used in current su- High Cr – high Mo – Ni base alloys ity. In reducing flue gas conditions the partial corrosion rates (Vakkilainen & Pohjanne 2010). perheaters typically as short pieces in most cor- Materials belonging to this category are e.g. pressures of oxygen and sulphur oxides may rosive places. Parts manufactured from AISI 347 Sanicro 63, Sanicro 67, Super 625. These high be too low, sulphation of alkali chlorides in the have no record of corrosion failure in Finland. molybdenum grades have excellent service furnace is much less, and chloride rich deposits In high temperature boilers, where improved records from waste incineration. When selection with low melting points may form. Molten intergranular corrosion resistance is needed these materials special attention should be paid phase corrosion can be mitigated by keeping alloys like 309 and 310, with higher amounts on manufacturing as well as on long-term stabil- the superheater metal temperatures below the of chromium should be used. According to ity and ageing at high temperatures (Kawahara FMT of the deposits. This can be achieved either Tuthill (2002) austenitic type 310 stainless steel & Kaihara 2001). WSI has also developed and with correct steam values or by removing NPE or more highly alloyed material either as solid used Unifuse 52 type welded coating (28% Cr, (potassium and chlorine) from the chemical or compound tubes is recommended for super- 14% Fe with minor amounts of Al and Ti) for recovery process. As known from other fields heater tubing when the outlet steam temperature use in coal fired boilers (Lai & Blogg 2003). of boiler technology, alkali chloride rich melts exceeds 480 °C. Coating can be welded to carbon or austenitic have the potential to cause rapid corrosion even Figure 11. Effect of steam velocity on super- steel tube. at material temperatures of 400 °C and lower. heater surface temperature (Vakkilainen & Nippon Steel & Sumitomo Metal Corporation Pohjanne 2010). (NSSMC) is supplying grades HR2M Conclusions Low temperature oxidation / active oxidation (22Cr14Ni1.5MoN), MN25R (25Cr14Ni0.8 mechanisms which occur below FMT of the Mo0.3N), HR3C (25Cr20NiNbN) and UNS Higher steam values, higher dry solid contents in deposits, involving complex volatile corrosion N08028/X1NiCrMoCu31-27 (27Cr31Ni3.5Mo) black liquor and mill closure have led to major products and low melting mixtures, are not well

70 71 enough understood, when applied in recovery Part 1. Tappi Engineering, Pulping & Environmental 20. Jones, A.K., 2004, 70 years of advances in 215. 10th International Symposium on Corrosion in boiler deposit conditions and under thermal Conference, October 11-14, 2009, Memphis, TN, recovery boiler design. In: 40 years recovery boiler the Pulp and Paper Industry (10th ISCPPI). Helsinki, cycling. USA, 27p. co-operation in Finland. International Recovery 21 - 24 August 2001. Volume 2. VTT Manufacturing 9. Crowe, D.C., Youngblood, W.C., 1998, Recovery Boiler Conference, Haikko Manor, Porvoo, May 12- Technology, pp. 509-521. Recovery boiler operation in a pulp mill is a Boiler Superheater Corrosion. 9th Int. Symp. on 14, 2004, pp. 15 - 20. 32. Mayer, J., 2012, Recovery boiler superheater complex process with several interrelated pa- Corrosion in the Pulp & Paper Industry, CPPA, 21. Karlsson, P.J. Möller, V. G. Johansen, K., 1990, corrosion – solubility of metal oxides in molten Toronto, Canada, pp. 225–230. Iron and steel corrosion in a system of SO , SO and salt. M.Sc. Thesis, Georgia Institute of Technology, rameters and almost infinite variables. Therefore 3 2 the selection of materials for long time trouble 10. Davies, M., Hirst, R.C., Hancock, P., 1973, alkali chloride. The formation of low melting point December 2012, USA, 133p salt mixtures, Corrosion Science, Vol. 30, p. 153. free service is not an easy task. Optimal boiler The Influence of Oxide Adherence on the Hot-Salt 33. Mäkipää, M., Koivisto, L., 1999, Corrosion and operation, together with application of suit- Corrosion of Mild Steels, Metallurgical Transactions, 22. Kaneko, S., Takatuka, H., Arakawa, Y., Tanaka Material Problems in Black Liquor Recovery Boilers. able tube materials, allows increasing the steam Vol. 4, p. 2196. R., 1998, Mitsubishi-CBC High Efficiency Chemical In: Eurocorr´99, 30 Aug-2 Sept 1999, Aachen, Recovery Boiler Power Generation System. 31st Germany, Dechema, 1999. values from the present ones. Low alloy steels 11. Ehlers, J., Young, D.J., Smaardijk, E.J., Tyagi, Pulp and Paper Annual Meeting, ABTCP, Sao Paolo, like T22/10CrMo910 can usually be used up to A.K., Penkalla, H.J., Signheiser, L., Quadakkers, W.J., 34. Mäkipää M, Oksa M, Koivisto L., 2001, 2006, Enhanced oxidation of the 9%Cr steel P91 in Brazil, October 19-23, pp. 693-704. Superheater Tube Corrosion in Changing Operation 480 °C. Austenitic stainless steels like grades water vapour containing environments. Corrosion 23. Kawahara, Y., Kira, M., 1997, Effect of Physical Conditions of Recovery Boilers. Corrosion 2001, 347, 310 are recommended for superheaters Science, Vol. 48, p. 3428. Properties of Molten Deposits on High-Temperature March 11–16, 2001, Houston, Paper 1424, 14 p. when steam temperature exceeds 480 °C. For 12. French, D.N., 2014, Microstructural degrada- Corrosion of Alloys in a Waste Incinerator 35. Mäkipää, M., Kauppinen, E., Lind, T., Pyykönen, the future recovery boilers, where the targeted tion. The National Board of Boiler and Pressure Environment. Corrosion Engineering, Vol. 46, p. 13. J., Jokiniemi, J., McKeough, P., Oksa, M., Malkow, steam temperatures are clearly beyond 500 °C, Vessel Inspectors. http://www.nationalboard.org/. 24. Kawahara, Y., Kaihara, Y., 2001, Recent trends in Th., Fordham, R.J., Baxter, D., Koivisto, L., materials solutions are looked from the highly Downloaded 2014-04-23. corrosion resistant tube materials and improvements Saviharju, K., Vakkilainen, E., 2001, Superheater alloyed austenitic iron and nickel based alloys. 13. Fujukawa, H., Makiura, H., Nishiyama, Y., of corrosion environments in WTE plants. Corrosion tube corrosion in recovery boilers, VTT Symposium 1999, Corrosion behavior of various steels in black 2001, NACE, Paper 01173. 215. 10th International Symposium on Corrosion in References liquor recovery boiler environment. Materials and 25. Keiser, J.R., Kish, J.R., Singbeil, D.L., 2009, the Pulp and Paper Industry (10th ISCPPI). Helsinki, Corrosion, Vol. 50, p. 154. Alternate materials for recovery boiler superheater 21 - 24 August 2001, Volume 2, VTT Manufacturing 1. Ahlers, P.-E., 1987, Corrosion problems in re- tubes. 45th Anniversary International Recovery Technology, pp. 157-180. covery boilers and their causes. (In Finnish) Paper 14. Hilden, J., Mäkipää, M., Oksa, M., 1999, Degradation of Protective Oxide Scales on the Boiler Conference. Sibelius Hall, Lahti, June 3 - 5, 36. Mäkipää, M., Fordham, R.J., 2000, The stability presented at Savon Korroosiopaivat 1987, 6–7.8. 2009, pp. 117-130. of some alkali ferrites in relation to chloride-enhanced 1987, Lomakoti Heimari, Mikkeli, Finland. 5p. Low-alloy Steel 10CrMo910 under HCl-bearing atmospheres: Initial Stage Monitoring with the 26. Klöwer, J., White, F. E., 1995, High-temperature oxidation. Schriften des Forschungscentrums Jülich: 2. Alexander, P.A., 1963, Laboratory Studies of the CER-technique. In: Full Papers on CD-ROM of corrosion of commercial heat resistant alloys under Reihe Energitechnik/Energy Technology, Volume 15, Effects of Sulphates and Chlorides on the Oxidation Eurocorr’99, 30 Aug–2 Sept 1999, Eurogress, deposits of alkali salts: recent laboratory data. 8th Part II, pp. 553–576. of Superheter Alloys, Dummy Tijdschrift, 1963, pp. Aachen, Germany. Dechema, e.V.. Int. Symp. on Corrosion in the Pulp & Paper Industry, 37. Mäkipää, M., Malkow, Th., Baxter, D.J., 1997, The 571–582. 15. Hupa, M., Backman, R., Skrifvars, B-J., Hyöty, Swedish Corrosion Institute, Stockholm, pp. 179–188. Effect of Salt Composition on the Chlorine Corrosion 3. Asteman, H., Svensson, J.-E., Johansson, L.-G., P., 1990, The influence of chlorides on the fireside 27. LaFond, J. F., Verloop, A., Walsh, A. R., 1992, of Low Alloy Steels. Corrosion 2001, Paper 01424. 2002, Oxidation of 310 steel in H O/O mixtures at 2 2 behavior in the recovery boiler, Tappi Journal, Vol. Engineering analysis of recovery boiler superheater NACE, Houston, Texas, USA, 2001. 14 p. Spiegel o 600 C: The effect of water-vapour-enhanced chro- 73 p. 153. corrosion. Tappi Journal, Vol. 75, p. 101. M, Schroer C, Grabke HJ., Corrosion of high alloy mium evaporation. Corrosion Science Vol. 44 p. 2635. steels and Fe-Cr-alloys beneath deposits from waste 16. Jaakkola, H. 2009, Getting rid of potassium and 28. Lai, G., Blogg, N. 2003, Unifuse cladding for incineration plants. Materials Science Forum, Vol. 4. Boettger, H., Umland, F., 1974, Untersuchungen chloride. 45th Anniversary International Recovery surface protection against corrosion and erosion/ 251–254, p. 527. zur Hochtemperaturkorrosion durch Chloride, Boiler Conference. Sibelius Hall, Lahti, June 3 - 5, corrosion in power boilers and waste heat recovery Werkstoffe und Korrosion, Vol. 25, p. 805. 2009, pp. 109-115. systems. 2003 Tappi Fall Meeting, 19 p. 38. Odelstam, T., 1982, Material Evaluation for BLRB Superheter Tubes. Black Liquor Recovery Boiler 5. Bruno, F., 1997a, Superheater corrosion in kraft 17. Jaakola, A., Rose. T., 1983, Operational experi- 29. Lehmusto, J., Skrifvars, B.-J., Yrjas, P., Hupa, M., Symposium 1982, Helsinki, August 31–September 1, boilers. (in Swedish) Rapport M3-613. Stiftelsen for ence of austenitic superheater tube bends in recovery 2013, Comparison of potassium chloride and potas- 1982, pp. D7-1-D7-9. Värmeteknisk Forskning, Stockholm, 129 p. boilers. Pulp and Paper Industry Corrosion problems, sium carbonate with respect to cause high tempera- 6. Bruno, F., 1997b, Superheater corrosion in Vol. 4. Swedish Corrosion Institute, Stockholm, ture corrosion of stainless 304L steel, Fuel Processing 39. Oksa, M., Mäkipää, M., 2001, Comparative labo- kraft recovery boilers (Korrosion hos överhettare i Sweden, p. 82. Technology, Vol. 105, p. 98. ratory testing of molten salt attack of superheater tube materials. VTT Symposium 215. 10th International sodapannor. En översikt och diskussion). Stockholm, 18. Jiao, F., Wijaya, N., Zhang, L., Ninomiya, Y., 30. Liu, F., Tang, J.E., Asteman, H., Svensson, Symposium on Corrosion in the Pulp and Paper Sweden: Stiftelsen för Värmeteknisk Forskning, Hocking, R., 2011, Synchrotron-based XANES specia- J.-E., Johansson, L.-G., Halvarsson, M., 2009, Industry (10th ISCPPI). Helsinki, 21 - 24 August Technical Report SVF/602, 134 p. tion of chromium in the oxy-fuel fly ash collected from Investigation of the Evolution of the Oxide Scale 2001. Volume 2. VTT Manufacturing Technology, pp. 7. Bruno, F., 1999, On the influence of chlorides lab-scale drop-tube furnace. Environmental Science Formed on 310 Stainless Steel Oxidized at 600oC in 523-538. and sulphurous compounds on the corrosion of su- and Technology, Vol. 45, pp. 6640. Oxygen with 40% Water Vapour Using FIB and TEM, 40. Rapp, R. A., 2002, Hot Corrosion of Materials: a perheater tubes in boilers with special consideration 19. Jokiniemi, J., Pyykönen, J., Mikkanen, P., Oxid Met, Vol. 71, p. 77. Fluxing Mechanism? Corrosion, Vol. 44, p. 209. on Kraft recovery boilers. Värmeforsk Service AB, Kauppinen, E., 1996, Modeling fume formation and 31. Malkow, Th., Mäkipää, M.I., Ungeheuer, J., Technical Report VF-664, 36 p. deposition in kraft recovery boilers. Tappi Journal, Fordham, R.J., Baxter, D.J., 2011, Cyclic TGA 41. Rapp, R. A., Zhang, Z. S., 1986, Journal of the 8. Clement, J.L., Grace, T.M., 2009. Investigation of Vol. 79, p. 172. testing to evaluate chloride enhanced oxidation Electrochemical Society, Vol. 133, No. 655. recovery boiler superheater design and performance - resistance of superheater steels, VTT Symposium 42. Raukola, A., Haaga, K., 2004, Expandable

72 73 recovery boiler, a novel method for future capac- Toronto, Ontario, Canada, pp. 48. ity increase. Proceedings of International Chemical 55. Nippon Steel & Sumitomo Metal, 2014, Seamless Recovery Conference, June 6-10 2004, Charlston, SC, steel tubes and pipes for boilers, http://www.nssmc. USA, pp. 241-246. com, Downloaded 2014-04-23. 43. Reese E, Grabke H.J., 1983, Einfluß von 56. Vakkilainen, E.K., 2005, Kraft recovery boilers Natriumchlorid auf die Oxidation von hochlegierten – Principles and practice. Suomen Soodakattila- Chrom- und Chrom-Nickel-Stählen - Effects of so- yhdistys ry. Helsinki, Finland, Valopaino Oy, 246 p. dium chloride on the oxidation of high alloy Cr- and ISBN 952-91-8603-7. Cr-Ni-steels. Werkstoffe und Korrosion, Vol. 44, p. 41. 57. Vakkilainen, E.K., 2004, Future of recovery boiler 44. Ross, W., Umland, F., 1984, Untersuchung technology. In: 40 years recovery boiler co-operation chlorid-induzierter Hoch-Temperaturkorrosion an in Finland. International Recovery Boiler Conference, Eisen-, Nickel-, Cobalt-Basislegierungen, Werkstoffe Haikko Manor, Porvoo, May 12-14, 2004, pp. 51–58. und Korrosion, Vol 35, p. 47. 58. Vakkilainen, E.K., Pohjanne, P., 2010, Selecting 45. Salmenoja, K., 2004, Superheater corrosion in the right material for recovery boiler superheaters. modern recovery boilers. 70 years of advances in Baltica VIII. Life Management and Maintenance for recovery boiler design. In: 40 years recovery boiler Power Plants. Vol. 2, VTT Symposium 265, VTT, pp. co-operation in Finland. International Recovery 37-51. Boiler Conference, Haikko Manor, Porvoo, May 12- 59. Yli-Olli, S., Pohjanne, P., Heikinheimo, L., 14, 2004, pp. 103 - 113. Kinnunen, T., Pankakoski, P. H., 2007, Sulphidation 46. Salmenoja, K., Mäkelä, K., 2000, Relations of advanced kraft recovery boiler materials. between superheater corrosion conditions in a black International Chemical Recovery Conference - liquor recovery boiler and a power fired with biofuels, Efficiency and Energy Management. Quebec City, Nordic Pulp and Paper Research Journal, Vol. 15, p. Canada, May 29 - June 1, 2007, pp. 273-278. 30. 47. Salmenoja, K., Mäkelä, K., 1999, Corrosion in fluidized bed boilers burning potassium and chlorine containing fuels. TAPPI Journal, Vol. 82, p. 161. 48. Sharp, W.B.A., 2007, Corrosion and cracking in recovery boilers. Tappi Kraft Recovery Short Course St. Petersburg, FL, USA, January 7-10, 2007. 49. Sharp, W.B.A., 2010, Superheater corrosion in biomass boilers - Today’s Science and Technology. ORNL/TM-2011/399. Oak Ridge National Laboratory, Oak Ridge, TN, USA, 105p. 50. Shinata, Y., Nishi, Y., 1986, NaCl-induced Acclerated Oxidation of Chromium, Oxidation of Metals, Vol. 26, p. 201. 51. Skrifvars, B.-J., Backman, R., Hupa, M., Salmenoja, K., Vakkilainen, E., 2008, Corrosion of superheater steel matrerials under alkali salt deposits, Part 1: the effect of salt deposit composition and temperature, Corrosion Science, Vol. 50, pp. 1274. 52. Tran, H., 1997, Upper Furnace Deposition and Plugging. In: Terry Adams (Ed.) Kraft Recovery Boilers. TAPPI Press, Atlanta, GA, USA, pp. 247–282. 53. Tuiremo, J., Salmenoja, K., 2001, Control of superheater corrosion in black liquor recovery boilers. VTT Symposium 215, 10th International Symposium on Corrosion in the Pulp and Paper Industry (10th ISCPPI), Helsinki, 21 - 24 August 2001, Volume 1, VTT Manufacturing Technology, pp. 181-188. 54. Tuthill, A.H. (Ed.), 2002, Stainless Steels and Specialty Alloys for Pulp and Paper. NiDi Reference Book Series No. 11025, Nickel Development Institute, 74 75 UTILIZATION OF process HISTORICAL DATA IN RECOVERY BOILERS

SONG WON PARK UNIVERSITY OF SAO PAULO

GUSTAVO MATHEUS DE ALMEIDA FEDERAL UNIVERSITY OF SAO JOAO DEL-REI BRAZIL

Introduction

Continuous chemical process industries seek is how to transform these massive amounts of to reduce economic, safety, and environmental raw data into relevant information to be applied losses in order to primarily meet societal in the benefit of the company. In other words, demands, governmental regulations, and the the common lack of data in the past decades has competitive market. All of them are increas- been replaced by the challenge of mining rele- ingly tight. Any information that contributes to vant information from massive amounts of data. at least mitigate such potential losses is desir- This new paradigm imposes new questions with able. However, large-scale chemical systems regard to data analysis (Venkatasubramanian, are complex by nature. They are multivariable, 2009). Due to the limitations of both quantita- with non linear relationships and noisy measure- tive and qualitative models, in conjunction to ments, and are of partial knowledge. In addition, the growing availability of process historical continuous chemical systems are characterized data, data-driven models, which are developed by presenting multiple normal operating modes. directly from the data, have increasingly been This scenario makes its description difficult in employed to transform raw data into useful practice. Quantitative (purely mathematical) and information. This approach opened a new pos- qualitative (such as expert systems) models are sibility of directly extracting relevant informa- limited in this task, either because of a lack of tion from very large data sets. Its use in place of knowledge or an arduous mathematical descrip- the process-driven concept gains strength every tion (Venkatasubramanian et al. 2003a,b,c). An day. Handling and interpreting massive amount alternative approach is getting information di- of data manually is infeasible in practice, and rectly from data, which is more and more avail- supporting computational tools are absolutely able in chemical industries. Nowadays, with essential. These tasks become more difficult as the advance of the informatics, database and the number of process variables increases and industrial instrumentation areas, hundreds and the sample time gets lower. even thousands of process variables, among flow rates, temperatures, pressures, concentrations, Next section describes the way collected data and so on, are continuously collected from a are usually employed in recovery boilers. In multitude of plant sensors and stored until every sequence, recent studies making use of new second (Chiang et al. 2001, Wang, 1999). The techniques or approaches to transform process resulting databases are called process historical historical data into relevant information are data (PHD) since they carry the history of the illustrated. operations. Thus, the availability of data is no longer a problem. From a past poor-data world to a current rich-data world, the new challenge

76 77 Use of process historical data in chemical process industries in general, process The structure of each case study is given by the Each one relates to a distinct heat source in the recovery boilers historical data are still little explored in recovery practical challenge in question, the goal of the boiler. This segmentation makes the interpreta- boilers. application, and the employed data-driven tech- tion of the contribution of each group (or each The availability of process data collected in nique, the main obtained results, followed by a variable in a particular group) on the steam pro- recovery boilers increased from the nineties. The following studies give examples of uti- brief closing. The title of each section is related duction easier. The map shows arrows from all Nowadays, massive databases are available in lization of process historical data in recovery to the goal that motivated the study. input variables towards the output steam. Their every medium or big mill. Like chemical indus- boilers. Almeida et al. (2004a) applied variable thicknesses are proportional to the previous tries, this data is very little explored. Its on-line selection techniques and artificial neural net- Communication facilities results obtained for the MSE values (Table 1), use is in general limited by monitoring key works to select best variable subsets with respect and hence they represent the contributions of the process variables through univariate trend plots. The understanding of massive numerical da- to steam generation. Almeida et al. (2004b) variables over the steam produced by the boiler. One drawback of this independent procedure is tabases is not a natural task for human beings. used data visualization techniques that allowed The thicker the arrow, the stronger the effect. that the spatial correlation among the variables This fact explains in part the growing tendency a visual representation of a set of process vari- Block 1 gives the main source of heat (76.9%) is not taken into account. Then, a measurement of representing them in a graphical way. The ables simultaneously. Versteeg and Tran (2008) in the boiler. In modern recovery boilers, which seen as normal may be in fact the onset of an motivation behind it is the human capacity to and Versteeg et al. (2007) applied principal it was not the case here, it may account for more abnormal condition. In addition, this disregard process visual information (Marakas, 2002). component analysis and partial least squares to than 85% of the total heat available. Block 2 an- causes distrust of the alarm levels, and in conse- Graphical representations can be employed to analyze the fouling problem in recovery boilers. swers for the second greater influence on steam quence an unsafe operating condition may arise. attend explanatory, exploratory, confirmatory, Major operating variables that caused fouling generation (23.0%). The contribution of Block 3 Another characteristic of process data is the seri- and communication goals. and plugging were identified with monitoring is negligible. Figure 2 shows a submap derived al correlation. When constructing control charts purpose. Martinelli et al. (2008) employed arti- from Block 2. It allowed a cleaner visualization it is also needful to consider this point mainly ficial neural networks for detection of incipient The following results were presented in Almeida of the contribution given by each level of com- in case of low sample times. The off-line use of defects in control systems. The case study was et al. (2008a, 2010). The collected database re- bustion air on the steam generation. It showed the stored data is mostly in diagnosis tasks. In the secondary air system of a recovery boiler. fers to four months of a recovery boiler operation the greater effects of the secondary (37.8%) and general, it is carried out with the aid of scatter Examples of graphical analysis using scatter that belongs to a pulp mill in Brazil. Its operation tertiary (51.7%) levels of air compared to the plots, which show the relationship between two plots and trend plots were given by Porter et al. was considered normal during the whole period. primary level of air (10.5%), whose focus is on process variables. However, as a multivariate (2010) and Vakkilainen et al. (2010). The former There are eight process variables with about the char bed. The obtained cause-effect strengths process with high correlations among variables, carried out a data analysis on the black liquor three thousand measurements for each one. The are in accordance with the literature (Adams et diagnosis tasks in recovery boilers would benefit being fed to a recovery boiler, and the latter goal was to construct a cause-effect map show- al. 1997, Vakkilainen, 2005) and common oper- from a simultaneous visualization of a group of examined the emissions from a recovery boiler ing the contribution of this set of process vari- ating practices. The relative magnitude of each process variables. Visualizations based on the using different time point views. Vakkilainen ables to the output steam. This was achieved by effect became more intuitive by using both maps orthogonal space (as scatter plots) are limited (2010) also made use of such plots to study the performing a sensitivity analysis study. Firstly, compared to the numerical results (fifth column to a maximum of three variables. By applying relationship between dust properties and a set of a predictive artificial neural network model for of Table 1). The higher the number of process attributes such as color, shape, and size, it is operating variables. the output steam was obtained. This data-driven variables, the clearer the role visual representa- feasible to represent about eight dimensions technique belongs to the computational intelli- tions play. The grounded knowledge concerning (Cleveland, 1993). Other solutions include sub- New approaches for use of process gence field. It was used due to its capacity of ac- steam generation by recovery boilers served as space selection (i.e. multidimensional reduction) complishing direct input-output mappings once a benchmark for validation of the methodology. or axis translation by using for example princi- historical data in recovery boilers only process variables were available, as well Having process historical data available, at first pal component analysis. Both approaches cause This section presents a collection of studies as of dealing with noisy data. Next, the input it can be applied to any other boiler's subsystem. loss of information. Then, multidimensional conducted by the authors with respect to the use process variables were disturbed, one at a time, visualizations would be valuable in helping the of process historical data in recovery boilers. and the output steam measurements collected at Graphical representations provide a more understanding of recovery boilers's operations. The data-driven techniques employed belong the mill compared to the corresponding model intuitive understanding of system's structures. Another point is related to the definition of a key to the multivariate statistics, computational outputs. The Mean Squared Error (MSE) metric Here numerical results (Table 1) involving nine variable subset with the purpose of monitoring, intelligence, and the signal processing fields. An was used as a measure of this discrepancy (Table process variables (eight inputs and one output) whose selection is based on process expertise example of simultaneous multidimensional visu- 1). The larger it is, the higher the contribution were transformed into 2D visual representations in general. However, in addition to this chosen alization is also given. The case studies include (fourth column of Table 1) of the input variable (Figures 1 and 2). For instance, they can be set of process variables, it may be possible that sensitivity analysis, visual representations, and on the steam generation. useful for knowledge dissemination, decision- other relevant variables exist. These variables communication facilities; abnormal situation making processes such as variable selection, and may be uncovered by investigating the stored management; fault detection and diagnosis, and In sequence, a graphical representation of the for leveling when putting together analysts from process historical data. Subsets of variables may visual analytics; abnormal operating variability previous (numerical) results was constructed distinct areas. Visual representations may also also contain redundant information. By using detection; and handling of multiple normal op- (Figure 1). The eight input variables were al- give unexpected or new insights into the system. them to construct data-driven models such char- erating states, and minimization of false alarm located into three groups, namely combustion acteristic may cause the deterioration of their rates. All case studies employ real data collected process heat (QSS), preheated combustion air performances. In brief, as occur in continuous in recovery boilers from pulp mills in Brazil. (QCA), and sensible heat (in the fuel) (QBL).

78 79 2 Table 1. Deviations (Dev) from the Reference Mean Squared Error (MSER = 74.2 (ton/h) ).

Disturbed Variable Symbol MSE Dev1 Dev2 (ton/h)2 (ton/h)2 (%)

Black liuor flow rate FBL 551.6 477.4 51.2

Black liuor temperature TBL 75.2 1.0 0.1

Dry solids content %SS 313.8 239.6 25.7

Primary air flow rate FPA 89.2 15.0 1.6

Primary air temperature TPA 81.7 7.5 0.8

Secondary air flow rate FSA 113.7 39.5 4.2

Secondary air temperature TSA 115.8 41.6 4.5

Tertiary air flow rate FTA 185.2 111.0 11.9 Total - 932.6 100.0

2 2 † Dev1 (ton/h) = MSE – MSER, where MSER = 74.2 (ton/h) . 2 ‡ Dev2 (%) = (Dev1/Total Dev)•100, where Total Dev = 932.6 (ton/h) . Figure 2. Graphical representation showing the individual contributions in the preheated combustion air block (see Figure 1) for all three levels of air, namely primary, secondary, and tertiary, on the output steam.

Abnormal situation management The following study (Almeida and Park, 2009) constructed a system to monitor the state of the Abnormal situations can be defined as distur- ash deposits over the heat section. The database bances that deviate the operations from the covers ten months of operation of a recovery normal states. Their causes are the human boiler at a pulp mill in Brazil, with a sample time factor, the process itself, and the equipments. of one minute. Five measurements of fuel gas Human factors relate to the lack of procedures, pressure drops (∆P) across the heat exchangers incorrect actions, and noncompliance with of the heat section and also the electrostatic pre- instructions. Overloads and mechanical failures cipitator were employed as monitoring variables are the main sources of abnormal situations with (Figure 3). The monitoring system was given by repest to process and equipment disturbances, a data-driven technique belonging to the signal respectively (ASM, 2005). To prevent or at least processing field called hidden Markov model. mitigate potential losses, automatic and reliable monitoring systems are more and more needed in mill control rooms (PAS, 2000). Every chemical process is under random influences due to an inherent variability. Then, measurements of process variables may be A risk situation in recovery boilers refers to the seen as realizations of an underlying stochastic accumulation of ash deposits, mainly composed process. This way, specific operating condi- for sulfur- and sodium-based salts, along its tions can be described by particular probability convective heat transfer section. Potential losses distributions (Venkatasubramanian et al. 2003c). in this case refer to the reduction of the steam This is the motivation for applying the hidden generation efficiency, the lower disponibility of Markov modeling to chemical process monitor- the boiler due to blockage of the fuel gas path, ing, once it is capable to identify changes of and the greater maintenance costs. Pulp mills statistical nature in signals (in this case, given manage it in day-by-day through the use of a set by process variables measurements) over time. of sootblowers as well as the continuous moni- Its output is a likelihood value (–log[P(O|λ)]), toring of key process variables in the control Figure 1. Graphical representation of the contributions of the input process variables (Table 1) over the which is a measure of the capacity of the model rooms (Vakkilainen, 2005, Adams et al. 1997). steam generated by the boiler. (with parameter λ) in generating a temporal

80 81 sequence of data (O) continuously made avail- second economizer higher than 850 Pa. Figure Table 2. Lower control limits. able over time (Rabiner, 1989). For instance, a 4a shows the significant correlation (ρ = 0.985) †∆P = Fuel gas pressure drop across the second economizer. hidden Markov model characteristic of normal between the model outputs (likelihood values) EcoII operating conditions can be used to detect de- and the measurements of fuel gas pressure Meaning: An output value, given by the .... means a possibility of already having a viations from it (caused by disturbances) over drops across the second economizer (∆PEcoII). monitoring system, equal to … value for ∆P † equal to ... time. This new situation would be perceived in Higher the monitoring system outputs, greater EcoII –3.2 950 a control room through lower likelihood values the fuel gas pressure drop (towards to the higher in a trend plot in the course of time. The suc- values above 1000 Pa this boiler can undergo in –19.6 900 cessful applications are in the speech processing practice). This result illustrates the use of hidden –36.1 850 field, including speech recognition and speaker Markov models as decision support systems in –51.8 800 verification, since the seventies (Rabiner, 1989). monitoring tasks. Figure 4b shows the resulting –110.3 700 More recently, others fields have experienced its monitoring chart. potential, namely telecommunications (Cappe –160.3 600 et al. 2010), bioinformatics (Kosky, 2002), Hidden Markov models are able to handle noisy and financial engineering (Mamom and Elliott, data, spatial and serial correlations, and non 2007). Gaussian distributions. These features are valuable for developing reliable systems regarding Then, a hidden Markov model-based monitor- abnormal situation management in continuous ing system, characteristic of fuel pressure chemical processes. Its explicit modeling, based drops higher than 1000 Pa across the second on the probability theory and statistics, also al- economizer (a critical condition in the present lows both getting insights learned by the model boiler), was constructed. In sequence, lower about the process, and in the opposite direction control limits were calculated for a set of fuel adjusting the model parameters to capture the gas pressure drops across it. (The correlation of phenomena more efficiently. Other case study fuel gas pressure drop measurements across this in recovery boiler applying this technique with heat exchanger with those across the convector purpose of monitoring is given in Almeida et al. and the first economizer is close to 1. These (2008b). Figure. 4 a). Significant correlation between model outputs (likelihood values) and measurements of fuel heat exchangers are the most critical regard- gas pressure drops across the second economizer (∆PEcoII), and b) resulting monitoring chart. ing ash deposits accumulation (Vakkilainen, 2005)) Table 2 summarizes the results. For instance, a model output above –36.1 given by Fault diagnosis artificial neural network model for chemical the monitoring system means that it is already process description since the nineties (Bulsari, possible to have a pressure drop across the In on-line process monitoring activities diagno- 1995). The monitoring task made use of the sis task arises after the detection of a fault. The Shewhart control chart for individual measure- goal is to uncover the root causes in order to ments (Montgomery, 2004). On September 3rd, intervene and normalize the operations. On one a fault condition was detected by such monitor- side the pressure for fast decision makings, and ing system (Figure 5; register 2204). on the other hand interactions among process variables and dynamic operations, make it a Due to the human being capacity for processing hard task in practice. Usually fault diagnosis is visual information, data visualization techniques carried out off-line with the aim of learning and are the most powerful tools that can help ana- then avoiding the same occurrence in the future. lysts to deal with very large data sets (Marakas, 2002). Here, parallel coordinate plots were used The following study is given in Almeida et al. for accomplishing fault diagnosis, with the aim (2012). The focus is also a recovery boiler from a of uncovering the root causes of the event previ- pulp mill in Brazil. The collected database refers ously classified as faulty. This data visualization to four months of operation. It has fifteen process technique provides a simultaneous projection variables with about three thousand measure- of multidimensional data onto a single plane ments for each one. Initially, a fault detection (Inselberg, 2009, 1985). As an example, con- system for monitoring the output steam flow sider that in a given sample time, the flow rate rate was obtained. It was based on a multi-layer (F), the temperature (T) and the pressure (P) Figure 3. Scheme for the convective heat transfer section and the electrostatic precipitator of the boiler, perceptron (MLP), which is the most employed with the five measurements of fuel gas pressure drops along it. of the output steam are equal to 297.8 ton/h,

82 83 Figure 5. Fault detection (at register 2204) by a reference monitoring system.

740.2 oC and 6.4 MPa, respectively. Figure 6a one associated to a specific sample time. The shows the representation of this 3-dimensional simultaneous visualization of several variables, point (297.8, 740.2, 6.4) in the commonly used together with this time information, made the Figure. 7. Final parallel coordinate plot. (group of black lines: 1st normal operating mode, with low orthogonal Euclidean space. This point in such understanding of the interactions among them as black liquor flow rates (F-BL) and high dry solids contents (C-DS); group of red lines: 2nd normal op- space becomes a polygonal line in the parallel well as the dynamic behavior of the boiler easier. erating mode, with an opposite condition; and green, blue, red, and group of light green lines: transition space, as shown in Figure 6b, where each verti- period between both normal operating modes, where P-SD: steam drum pressure and F-HPS: output cal axis (coordinate) represents a process vari- Figure 7 shows the final parallel coordinate plot steam flow rate). able. This approach provides an extension of the presented in the study. It was possible to visual- Euclidean space, since theoretically an unlimited ize two distinct but normal modes of operation site condition. From register 2202 (green line), Figure 8 presents univariate trend plots, which number of variables can be represented. Due to in the boiler during the period encompassing register 2203 (blue line), register 2204 (thick red are commonly used in mill control rooms interactions, simultaneous visualization of sev- register 2204 (Figure 5). The former, between line), until register 2215 starting from register worldwide, for the same operating interval as in eral process variables is valuable. Now, consider registers 2192 and 2201 (group of black lines), 2205 (group of light green lines), there was in Figure 7. The easier way to extract information a set of successively collected multivariate data is characterized by low black liquor flow rates fact a period of transition between both normal from parallel coordinate plots in comparison to points as is the case regarding recovery boilers. (F-BL) and high dry solids contents (C-DS), modes of operation, rather than a faulty condi- a simultaneous analysis of a set of trend plots Plotting them in a parallel coordinate plot results whereas the latter, from register 2216 up to tion as stated by the fault detect system (Figure is clear. Such information refer to dynamic in a corresponding set of polygonal lines, each register 2224 (group of red lines), had the oppo- 5). The stabilization of the boiler could also be behavior, interactions among process variables, visualized. The steam drum pressure (P-SD) modes of operation, and anomalous data. It can returned to the same operating range as before also support the setting of more reliable alarm the transition period. Thereafter, there was also levels (not shown). In brief, visual data analytics the stabilization of the output steam flow rate in recovery boilers may benefit by the use of (F-HPS). (Both occurrences can be noted by multidimensional data visualization tools, as are means of the overlapping of the groups of black parallel coordinate plots. and red lines.) Perceptions like these are valuable in order to achieve more efficient operations.

Figure 6. Representation of a multivariate measure (F = 297.8 ton/h, T = 740.2 K, P = 6.4 MPa): a) a point in the common 3-D orthogonal space, and b) an equivalent polygonal line in the parallel space. 84 85 efficiencies (concerning variability reduction), dry solids content of 74%. Figure 10b shows given a fixed dry solids flow (in3 m /h). the optimum operating ranges (concerning higher steam production) for the eight original Principal component analysis is a multivariate process variables (which were used in the model statistics technique. Its main purpose is dimen- construction) considering the operating region sionality reduction contributing to an easier han- highlighted in Figure 10a. For instance, the dling and understanding of multivariate systems secondary air flow rate (F-SA) should operate such as recovery boilers. The key point is its ca- above whereas its temperature (T-SA) should pacity to explain most of the variability present remain below the respective mean operating in a data set, composed by p process variables, values. through only k linear combinations of them, being k << p. Such combinations, called principal In practice, multivariate control charts using components, are the axes of a new coordinate operating regions can be employed to keep system, derived by rotation of the original one. the process condition inside desired operating Their directions are those of maximum variance ranges (not shown), which are defined as opti- (Joelliffe, 2002). Most applications of principal mum according to pre-established criteria. This component analysis in the chemical engineering control chart may provide information about the field is related to process monitoring since the current state as well as the trend of the opera- nineties (Piovoso and Kosanovich, 1996; Kresta tions. Concluding, unexpected variations of key et al. 1991). The multivariate control charts process variables are responsible for abnormal usually employ both monitoring metrics Q and process variability. Its identification and diagno- T2. Another possibility (illustrated in this study) sis contribute to more effective actions towards is its use based on operating regions (windows) (a) (Wang, 1999). This approach is still little explored in chemical industries worldwide.

Initially, a filtered database including only samples associated to black liquor flow rates of 50 m3/h and dry solids contents of 74% was prepared. This combination refers to the most Figure. 8. Trend plots for the same operating interval as in Figure 7. frequent operating condition in the boiler. In sequence, a model based on principal component Detection of abnormal operating the mill.) More than one operating condition can analysis was obtained. Its first three components variability also be noted in Figure 9a. This short example were able to explain 80% of the total variance shows the relevance of the subject. Undesired in the original data, composed by eight process (b) Continuous chemical processes suffer from process variable dispersion is what we are call- variables related to combustion air. Hence, it was an inherent variability even under fixed condi- ing abnormal operating variability. The resulting possible to describe the system under analysis tions. The problem arises when the operations operating inefficiency is responsible for losses using a smaller number of dimensions (since k significantly exceed it. For example, consider a mainly economic. Its detection is not an easy = 3 and p = 8), which makes its handling, visu- fixed condition in a recovery boiler, given by a task. alization, and understanding easier. After a vali- black liquor flow rate of 50 3m /h and a dry solids dation step, operating regions related to higher content of 74%. Figure 9a shows the dispersion The following study (Almeida and Park, 2012a) steam production were determined. Figure 10a in the flow rate of the primary combustion air. employed principal component analysis to shows scatter plots for all possible combinations The corresponding coefficient of variation identify and determine more efficient operating among the three principal components (cp1, (coefficient of variation = sample standard regions regarding steam production. The case cp2 and cp3), where samples related to steam deviation / sample mean) is equal to 5.4%. The study also employs a recovery boiler from a pulp flow rates higher than 270 ton/h are highlighted variability in input process variables cause vari- mill in Brazil. The collected database containing (blue points). It can be verified that higher steam ations in the outputs. Figure 9b shows the case eleven process variables refers to one month of generation efficiencies are associated to specific Figure 9. Operating variability in the a) primary for the output steam flow rate having a variation operation, with a sample time of five minutes. regions. Remember that this result is derived combustion air and b) output steam flow rate, for around 100 ton/h. The corresponding coefficient The goal was to determine combustion air oper- from a fixed operating condition in the boiler, a fixed black liquor flow rate (50 ton/h) and dry of variation is equal to 7.9%. (Such values can 3 ating regions related to higher steam generation given by black liquor flow rates of 50 m /h and solids content (74%). be considered satisfactory or not depending on 86 87 consist of transition periods between normal Final considerations states in the boiler. To the first, the false alarm From a set of real case studies, examples of rate given by the hidden Markov model-based utilization of process historical data in recovery system was slightly worse than the others (9.1% boilers were illustrated. Its complex operation, against 8.8% and 8.1%) (second row of Table by being multivariate, non linear, noisy and of 3), whereas to the second (third row of Table partial knowledge, as chemical processes usually 3), the false alarm rate compared to the others are, makes hard a pure mathematical description was significantly lower (8.3% against 15.1% of it. Despite the knowledge accumulated over and 11.6%). This result suggests its potential in the years, naturally there are knowledge, insights handling multiple normal modes of operation, and perceptions to be still uncovered. This sce- an inherent and challenging characteristic of nario, in conjunction to a plenty of process vari- continuous industrial chemical processes as are (a) (b) ables continuously collected through a multitude recovery boilers, contributing to more satisfac- of plant sensors nowadays, has encouraged the Figure 10. a) Optimum operating regions given by steam flow rates higher than 270 ton/h (blue points), tory false alarm rates. Also, the proximity of employment of data-driven approaches since the where cp1, cp2 and cp3 are the resulting scores of the principal component analysis, and b) operat- both rates (9.1% and 8.3%) suggests its ability nineties. Current applications involve mainly ing ranges (blue points) of the combustion air variables associated to the optimum operating region to deal with normal transitions in a more stable the use of both techniques principal component highlighted in a). way. The reason for this relates to the speciali- analysis and (feedforward) artificial neural net- zation of the model in describing the multiple its reduction. The result is the achievement of positive correlation. The goal was to verify the works. This chapter illustrated the use of mul- normal modes of operation (Figure 11b). more stable operations and lower maintenance capacity of hidden Markov models (described in tivariate statistics, computational intelligence, costs. a previous section) to manage false alarm rates. and signal processing tools. The applications Its ergodic topology was adopted. Theoretically, Hence, the hidden Markov model approach may covered sensitivity analysis, visual representa- contribute to reach reasonable false alarm rates Management of false alarm rates such model structure allows the description of tions, communication facilities, abnormal situa- any transition that occurs in practice between the once it may handle multiple states of normal tion management, fault detection and diagnosis, An inherent characteristic of continuous chemi- normal states. Figure 11b shows three Gaussian operation. This ability is also welcomed to visual analytics, abnormal variability detection, cal processes is the presence of multiple states probability distributions each one associated to achieve early fault detection since it competes handling of multiple normal operating states, of normal operation with frequent transitions a particular state of the model constructed. An with satisfactory false alarm rates. Both issues and minimization of false alarm rates. between them. For instance, black liquor flow overlapping of the whole region given by the are crucial for developing automatic and reliable rates fed to recovery boilers change from time fault detection systems, yet an open question in temperature ranges (TSH and TBB) can be ob- Concluding, the growing availability of very to time inside a considerable operating range practice. served with each distribution covering a portion large data sets in chemical industries worldwide (Figure 11a). Higher number of normal states of the process data in particular. This specializa- in the last years established another paradigm. pose another difficult regarding the development tion means that the model is able to describe The poor-data world, where getting data to finish of reliable monitoring systems: How to manage specific boilers' operating conditions (given by them in order to achieve reasonable false alarm distinct black liquor flow rates; shown for 25, 40 rates? and 50 m3/h), as desired. This point motivates the study in Almeida and Park (2012b). The data comprehends three Table 3 summarizes the quantitative results to months of operation of a recovery boiler from a false alarm rates. The performance of the hid- mill in Brazil, with a five minutes sample time. den Markov model-based monitoring system The operations were considered normal during was compared to monitoring systems based on the whole period. The case study refers to the principal component analysis, a well-established critical situation concerning thermal efficiency multivariate statistical process control technique reduction that may happen in the heat section (Piovoso and Kosanovich, 1996; Kresta et al. due to ash deposits accumulation. Common 1991), and on self-organizing maps, a compu- practices worldwide are routine patrols around tational intelligence technique (Kohonen, 1995). the boiler and monitoring of key process vari- Both of them were employed in a dynamic way. ables mainly fuel gas temperature (Vakkilainen, For a fair basis comparison, they were previ- 2005, 2000; Adams et al. 1997). The case study ously adjusted to give the same false alarm rate made use of two of these temperatures, namely as the hidden Markov model-based monitoring Figure. 11. a) Multiple normal operating states, and b) overlapping of the whole operating range, those measured after both the super-heater system, equal to 4.5% (first row of Table 3). defined by TSH and TBB (temperature normalized values), by Gaussian probability distributions (each (TSH, in oC) and the boiler bank (TBB, in oC). Two independent test data sets were then used. one belonging to a particular model state), and association of them with boilers’ operating conditions, Under normal conditions, they present a high Unlike the former, all data samples in the latter given by distinct black liquor flow rates (shown for 25, 40 and 50 3m /h).

88 89 Table 3. Comparison of false alarm rates (in %) among monitoring systems. (IDEAL), Lecture Notes in Computer Science (LNCS), 22. Mamom, R.S., Elliott, R.J., 2007, Hidden Markov Natal, Brazil, 7435, 743-752, Springer. models in finance. Springer, New York, 188 p. Data set Principal Self-organizing Hidden Markov 8. Almeida, G.M., Park, S.W., 2012a, Avaliação de 23. Marakas, G.M., 2002, Modern data warehouse, component map† model desempenho de operações em processos químicos mining and visualization: Core concepts, Prentice analysis industriais via análise por componentes principais: Hall, New Jersey, 274 p. Validation data set (reference basis) 4.5 4.5 4.5 Exemplificação a partir de um estudo de caso real 24. Martinelli, S.H.S., Neitzel, I., Vieira, O., 2008, (Performance evaluation of operations in indus- Test data set: Without transitions 8.8 8.1 9.1 Defect detection by neural networks. Application: trial chemical processes through principal component (n‡ = 7382) Secondary air system in a chemical recovery boiler. analysis: An example from a real case study). In: In: Proceedings of the IberoAmerican Congress on Test data set: With transitions 15.1 11.6 8.3 Proceedings of the 20th National Symposium on Pulp and Paper Research (CIADICYP), Guadalajara, (n = 965) Probability and Statistics (SINAPE). Joao Pessoa, CD-ROM. Brazil, CD-ROM. † An extension of Barreto et al. (2005). 25. Montgomery, D.C., 2008, Introduction to statisti- 9. Almeida, G.M., Park, S.W., Cardoso, M., 2004a, ‡ Sample size. cal quality control. Wiley, New York, 6th ed., 734 p. Variables selection for neural networks identification 26. PAS (Plant Automation Services, Inc.), 2000, The for kraft recovery boilers. In: Proceedings of the 2nd a report was a hard task, gave way to a rich-data References cost/benefit of alarm management - An economic jus- IFAC Workshop on Advanced Fuzzy/Neural Control world. The current challenge is how to transform tification for alarm system re-engineering. Texas, 8 p. 1. Adams, T.N., Frederick, W.J., Grace, T.M., Hupa, (AFNC), Oulu, Finland, pp. 91-96. massive raw data into useful information for 27. Piovoso, M.J., Kosanovich, J.P., 1996, The use of M., Iisa, K., Jones, A.K., Tran, H. 1997, Kraft recov- 10. Almeida, G.M., Park, S.W., Cardoso, M., 2004b, cleaner, safer and more economical operations. multivariate statistics in process control. The Control ery boilers, Tappi Press: Atlanta, 380 p. Predicting recovery boiler performance and exploit- New ways of data analysis are needed to reach Handbook, CRC Press: Boca Raton, pp. 561-573. 2. Almeida, G.M., Cardoso, M., Park, S.W. 2012, ing data visualization for the critical variables. In: such goals, however there are still no established 28. Porter, J., Trung, T., Sands, T., 2010, Improving Detecting an abnormality in a recovery boiler using Proceedings of the International Chemical Recovery recipes for this new required intelligent data Recovery boiler performance by controlling variabil- dynamic multivariate data analysis with parallel Conference (ICRC), CD-ROM. analysis. ity (New tools for an old problem). In: Proceedings coordinate plots. Appita Journal, Vol. 65, No. 1, pp. 11. ASM, 2014, (Abnormal Situation Management) of the International Chemical Recovery Conference 78-86. - A Joint research and development consortium. (ICRC), Williamsburg. Acknowledgements 3. Almeida, G.M., Cardoso, M., Rena, D.C., Park, . 24 March, 2014. 29. Rabiner, L.R., 1989, A tutorial on hidden Markov S.W., 2008a, Graphical representation of cause-effect 12. Barreto, G.A., Mota, J.C.M., Souza, L.G.M., The authors thank the financial support of models and selected applications in speech recogni- relationships among chemical process variables us- Frota, R.A., Aguayo, L., 2005, Condition monitoring Brazilian Research Foundations, namely tion, Proc. of the IEEE, Vol. 77, No. 2, pp. 257-287. The State of Sao Paulo Research Foundation ing a neural network approach. In: Proceedings of the of 3G cellular networks through competitive neural IEEE World Congress on Computational Intelligence 30. Vakkilainen, E., 2000, Chemical Recovery. In: (FAPESP Project 03/00405-9), The State of models. IEEE Transactions on Neural Networks, Vol. (WCCI)/International Joint Conference on Neural 16, No. 5, pp. 1064-1075. Gullichsen, J., Fogelholm, C.J. (Eds.), Chemical pulp- Minas Gerais Research Foundation (FAPEMIG ing, Book 6B, Papermaking and Science Technology Networks (IJCNN), Hong Kong, pp. 2669-2674, . 13. Cappe, O., Moulines, E., Ryden, T., 2010, Projects APQ-02544-09 and TEC 00142/06), Series, Chapter 11, pp. B6-B34. 4. Almeida, G.M., Park, S.W., Cardoso, M., 2008b, and The National Council for Scientific and Inference in hidden Markov models. Springer: New Detection of abnormal situations in a heat transfer York, 654 p. 31. Vakkilainen, E., 2010, Predicting ash properties in Technological Development (CNPq Project recovery boilers. In: Proceedings of the International section of a boiler using a hidden Markov model 14. Chiang, L.H., Russel, E.L., Braatz, R.D., 2001, 482455/2010-1), as well as the technical support approach. In: Proceedings of the 3rd International Chemical Recovery Conference, Williamsburg. of kraft pulp mills in Brazil, through techni- Fault detection and diagnosis in industrial systems. Symposium on Advanced Control of Industrial Springer, London, 279 p. 32. Vakkilainen, E., Hamaguchi, M., Laux, D.C. 2010, cal visits, short time internships, experience Processes (AdCONIP), Jasper, pp. 493-497. Statisticality of emissions from a recovery boiler. In: 15. Cleveland, W.S., 1993, Visualizing data. Hobart exchanges, and cession of process historical Proceedings of the International Chemical Recovery 5. Almeida, G.M., Park, S.W., 2009, Ash Deposits Press, New Jersey, 360 p. databases. monitoring in a convective heat transfer section Conference (ICRC), Williamsburg. 16. Inselberg, A., 2009, Parallel coordinates: Visual of a kraft recovery boiler. In: Proceedings of the 33. Vakkilainen, E.K., 2005, Kraft recovery boilers: multidimensional geometry and its applications, 10th International Symposium on Process Systems Principles and practice, Valopaino Oy, Helsinki, 246 Springer, New York, 554 p. Engineering (PSE), Salvador, Brazil, pp. 1467-1472. p. 17. Inselberg, A., 1985, The plane with parallel coor- 6. Almeida, G.M., Cardoso, M., Rena, D.C., Park, 34. Venkatasubramanian, V., 2009, Drowning in data: dinates. The Visual Computer, Vol. 1, No. 2, pp. 69-91. S.W. 2010, Graphical representation of cause-effect Informatics and modeling challenges in a data-rich relationships among chemical process variables us- 18. Jolliffe, I.T., 2002, Principal Component analy- networked world. AIChe Journal, Vol. 55, No. 1, pp. ing a neural network approach. International Journal sis. 2nd ed., Springer, New York, 489 p. 2-8. of Computational Intelligence and Applications, Vol. 19. Kohonen, T., 1995, Self-Organizing maps, 35. Venkatasubramanian, V., Rengaswamy, R., Yin, 9, pp. 69-86. Springer, Helsinki, 521 p.. K., Kavuri, S.N., 2003a, A review of process fault di- 7. Almeida, G.M., Park, S.W., 2012b, Fault detection 20. Kosky, T., 2002, Hidden Markov models of bioin- agnosis - Part I: Quantitative model-based methods. in continuous industrial chemical processes: A new formatics. Springer, New York, 391 p. Computers and Chemical Engineering, Vol. 27, No. approach using the hidden Markov modeling. Case 21. Kresta, J.V., Macgregor, J.F., Marlin, T.E. , 1991, 3, pp. 293-311. study: A boiler from a Brazilian cellulose pulp mill. In: Multivariate statistical monitoring of process operat- 36. Venkatasubramanian, V., Rengaswamy, R., Proceedings of the 13th International Conference on ing performance. The Canadian Journal of Chemical Kavuri, S.N., 2003b, A review of process fault diagno- Intelligent Data Engineering and Automated Learning Engineering, Vol. 69, No. 1, pp. 35-47. sis - Part II: Qualitative models and search strategies.

90 91 Computers and Chemical Engineering, Vol. 27, No. 3, pp. 313-326. 37. Venkatasubramanian, V., Rengaswamy, R., Kavuri, S.N., Yin, K., 2003c, A review of process fault diagnosis - Part III: Process history based methods. Computers and Chemical Engineering, Vol. 27, No. 3, pp. 327-346. 38. Versteeg, P., Tran, H., 2008, Monitoring Kraft recovery boiler fouling using Principal Component Analysis. Engineering, Pulping and Environmental Conference, TAPPI, Portland. 39. Versteeg, P., Tran, H., Sayad, S., de Leo, V., 2007, A multivariate analysis of recovery boiler fouling. Proceedings of the International Chemical Recovery Conference Proceedings. 40. Wang, X.Z., 1999, Data mining and knowledge discovery for process monitoring and control: Advances in industrial control. Springer, London, 251 p.

92 93 XXL recovery boilers - how we developed them

Kari haaga VALMET POWER OY FINLAND

Introduction

The chemical wood pulp production has during Other interests are in reductions in usage of fos- the last decades grown with about 2-3 Mt/a. This sil fuels and with high power recovery boilers it capacity increase consists of new installations, is possible use less fossil fuels like coal, oil and including green field mills and capacity increases natural gas in power boilers if located at pulp on existing recovery boilers. The main part of mill site. This trend means that new technical the new projects is hardwood lines operating on features used now in recovery boilers have been planted eucalyptus and acacia. Those mills are adapted from power boilers like high steam typically located in South America and Far East. parameters, high pressure feed water heaters, high combustion air temperatures, flue gas cool- New developments in technology have made it ers etc. These new solutions have caused certain possible to take advantage of the economy of changes to present designs of recovery boilers scale when building those new big pulping lines. compared to units built 10-20 years ago. Obviously the recovery boilers had to grow in size in order to meet the production capacity in World’s biggest boilers consist of the latest the large new single line pulp mills. Excellent design from environmental point of view; all process and availability experiences from big NCG and vent gases are burnt in recovery boiler. and XL size recovery boilers have smoothened Diluted non condensable gas, weak gas (DNCG) the way to XXL size recovery boilers. dissolving tank and mixing tank vent gases are typically introduced to the high secondary air The new XXL size recovery boilers are charac- system, concentrated non condensable gas, terised by high black liquor firing capacity, high strong gas (CNCG) burner is located at second- dry solids content, high power features, low air ary air level. Boilers are also designed for high emissions and environmenttaly friendly process dry solids firing of 80% or higher black liquor solutions like burning non condensable gases dry solids content in virgin black liquor. (NCG) and other side streams coming from other parts of the mill. The plant designs have been made in PDMS which has given a complete 3D model of re- High power XXL size recovery boilers are covery boiler including all details. The PDMS becoming more and more common in pulp model is also of great help in the erection phase. mill industry as electricity price has increased. Other sophisticated engineering tools like Electricity produced by recovery boilers is seen computational fluid dynamics (CFD), fouling as green energy in many countries meaning advisor, recovery boiler designer (RBD) have incentives to electricity price which make high been used when big recovery boilers have been power feature investments more profitable. designed.

94 95 Very important nowadays when engineering Two 7000 tds/d boilers started up in 2010 in XXL size recovery boilers is the knowledge of Indonesia and in China. Those boilers have floor corrosion theory because huge boilers have a area of 293 m2 and furnace total height is 64.5 m. lot of heating surfaces and it is not possible that Boilers have quite traditional steam parameters; those units will encounter sudden corrosion be- temperature is 480 ºC and pressure 84 bar and cause production losses of big pulp mills would their gross steam generation is 285 kg/s. They be huge. do not have any high power features except 80% black liquor dry solids content. Mainly they are Increase of recovery boiler size burning Eucalyptus and Acacia black liquors. during last decades Boilers are sister boilers except that the boiler in Pulp mill and so also recovery boiler capacity China has Quaternary air level for reducing NOx has increased a lot during last decades. In the emissions. This also means that composite tube beginning of 1980s there was a common opinion level on furnace walls is extending higher than that recovery boilers had reached their maximum in Indonesia boiler. physical dimensions. During that time the floor cross section was about 100-120 m2. However, Present biggest recovery boiler in the world is one recovery boiler in Finland started up in 1990 located in Brazil and its floor area is 323 2m and which had black liquor firing capacity of 2600 furnace total height is 68.2 m. Its main steam tds/d. Boiler can be called as first so called XL temperature is 492 ºC and pressure 94 bar. Steam size recovery boiler. This big boiler was some is generated 325 kg/s and boiler has certain high Figure 2. Boiler sizes higher than 5500 tds/d. kind of threshold for other coming big boilers. power features like high dry solids firing (80%), This boiler was also a forerunner for black liquor Figure 3 shows what are certain physical differ- Most interesting from above values is furnace ences between small and big recovery boilers. height. As can be seen it is only about two On the left hand side of the figure there is shown times taller in big unit than in small one. Other one new small boiler delivered to India in 2013. parameters are varying much more between This boiler size was quite typical about 30-40 those two compared boilers and differences years ago. The middle one is built in the begin- are more logical to understand. Some design ning of 1990 and the boiler on the right is world’s criteria in recovery boilers are furnace volume biggest recovery boiler. If we compare certain load and flue gas residence time in the furnace. design parameters between smallest and big- In small boilers volume load can be 160 MW/ gest ones we can see the following differences: m3 and corresponding value in big boilers has typically been 80-100 MW/m3. Volume load is • Capacity is 11.7 times higher indicating also what flue gas residence time in • Floor area is 7.9 times bigger the furnace is and in small boilers it is 3.5-4.5 • Furnace total height is 2.1 times taller s and in big units 7-8 s (calculated from floor to • Furnace volume is 16.9 time bigger nose). Theoretically the furnace of big recovery Figure 1. Capacity increase of recovery boilers. high dry solids firing, CNCG combustion and all combustion airs are heated to 185 ºC, feed bio sludge combustion. Six years later the first water tank is so called full pressure tank (140 3800 tds/d boiler was started up in Indonesia. ºC), feed water before economizer I is heated to At that mill there are three similar boilers which 160 ºC in preheater. As can be noticed the boiler have been upgraded to the load of 4000 tds/d. does not have very high steam parameters and Development continued. Again after eight years that has been typical for big boilers. Boiler was first so called XXL size 5500 tds/d recovery started up successfully in January 2014. boiler was delivered to China. With that boiler 6000 tds/d peak load was reached in 2005 and What will happen in the future? Even bigger it was world record during that time. Figures 1 and bigger boilers have been discussed and most and 2 show how the size of recovery boilers has probably 10000 tds/d load will be exceeded in a increased during last decades. The pictures also couple of years, Figure 2. shown the floor area and start-up year of those boilers. Figure 3. Comparison of boilers’ physical sizes in real scale.

96 97 boiler is not needed to be as tall as it is if only 2000 tds/d load. Real big step happened in 2004 volume load and residence time had taken into when boiler load increased from 3800 tds/d to account because many small boilers are operated 5500 tds/d. well with above mentioned design parameters. However, when furnace height is determined One of the most important design criteria in then also flue gas temperature after furnace and recovery boilers is a floor loading. Floor loading superheaters, sticky area location etc. must be is calculated by using black liquor dry solids considered. higher heating value (HHV), black liquor load and furnace floor area. Nowadays typical floor Figure 4 shows world’s biggest recovery boiler load design in big boilers is about 3.5 MW/m2 or height vs. hotel Ilves which is a famous hotel even higher when it was earlier 2.0-3.0 MW/m2. in Tampere. As can be seen the recovery boiler In some new upgrade cases even 3.8-4.0 MW/m2 located in Brazil is about 23 m taller than hotel floor loads have been operated. Ilves. Why is it possible to have so high floor loading? Nowadays black liquor dry solids content starts to be 80 % (earlier it was 60-70%), boilers have Figure 6. Development of floor loadings. modern control systems, air systems and black liquor firing practices have been developed a Figure 7 shows how floor area has been in- Pre-engineering of first xxl size lot during last decades, etc. Figure 6 shows how creased during the last decades. From mid 1970s recovery boiler floor loadings have developed during the last to mid 1980s in eight years period the floor area 35 years. Blue dots are indicating floor load- increased by 45% and then in the following 8 In the beginning of 2000 special development 2 ing with unit of tds/d/m and red ones are for years by 26%. From 1992 to 2004 floor cross project was carried out in order to study pos- 2 MW/m . As can be seen somewhere in the mid- section increased only by 3% meaning about sibilities to build and deliver a real big recovery dle of 1990’s floor loading exceeded 3.0 MW/ 0.25% annually. Then suddenly the area jumped boiler. Main emphasize in that project was to m2. As an example from the floor loadings we identify possible items/designs which may af- Figure 4. World’s biggest recovery boiler vs. upward by 43% when big recovery boiler in could analyze world’s biggest recovery boiler; if China started up. After this project the furnace fect boiler availability, furnace processes and a famous hotel Ilves in Tampere. 2 it had dimensioned by using 2.5 MW/m loading size has increased quite steadily meaning about mechanical design. All boiler parts and equip- 2 then floor area would be 452 m instead of 323 20% in ten years. As can be noticed from the ment were analyzed and when certain risk item 2 Figure 5 shows how recovery boilers’ black m . This would mean 21 m furnace width and Figure 7 the highest floor area increase and also was detected then it received a code number for liquor burning capacity has increased since 1970 depth compared to the design of 18.3 m x 17.7 boiler load increase happened in 2004. Before further analysis, see Figure 8. based on our reference list. As can be seen about m furnace. and after that average boiler floor area has in- in 1990 black liquor burning capacity exceeds creased by 3% and 1.3% annually.

Figure 5. Recovery boiler capacity increase since 1970. Figure 7. Increase of floor areas in recovery boilers.

98 99 Table 2 Typical tube materials in recovery boilers.

EN ASME COMPONENTS P265GH SA-210 A1 Furnace, primary superheaters, screen tubes, boiler bank, economizers 16Mo3 SA-209 T1 Furnace, primary superheaters Composite 304L Composite 304L Lower part of furnace 13CrMo4-5 SA-213 T11 / T12 Secondary superheaters 10CrMo9-10 SA-213 T22 Secondary, tertiary and quaternary superheaters 7CrMoVTiB10-10 SA-213 T24 Tertiary and quaternary superheaters X10CrMoVNb9-1 SA-213 T91 Tertiary and quaternary superheaters Figure 8. Indentified possible risk areas. X7CrNiNb18-10 (1.4912) SA-213 TP347H Hottest superheaters in corrosive atmosphere Every item was studied deeper by using for like downcomers, headers, main steam pipe, Composite Sanicro 28 Sanicro 28, overlay welded Hottest superheaters in corrosive example risk analysis. When certain risks were drain and vent pipes, process pipes etc. If those and 310, Overlay welded materials atmosphere detected then action plans were developed in are added then total tube/pipe length is close to materials order to find corrective actions. As an example 600 km. code 19 which is located just beside secondary superheater. When this item was analyzed then Table 1. Heating surface tubes in world’s biggest main targets were concentrating on manufactur- recovery boiler. ing of very long elements (22-25 m), how forth Heating surface Tube amount (km) and back movement of elements is minimized especially during boiler start ups when heating Economizers 192 surfaces are relatively clean but sootblowers are Boiler Bank 55 used etc. In economizer and boiler bank areas Screen tubes 43 long 24-27 m elements are very flexible and Primary I SH 22 therefore it was important to find solution for minimizing sideward vibration/movement of Primary II SH 22 elements. Secondary SH 29 Tertiary SH 29 After this study we principally had a readymade Figure 9. Boiler bank tubes and readymade boiler bank elements/package. big recovery boiler concept for upcoming big Quaternary SH 25 pulp mills. Soon after this work first inquiry of SH Total 149 mentioned in Table 1 are typically the following truck when transporting it from the manufactur- XXL size recovery boiler came and it was very Furnace walls 83 (EN and ASME materials) Table 2. ing facility to harbor for sea transportation. In easy to start to prepare technical specification TOTAL ~ 500 when knowing how this kind of big unit should Figure 10b there is erection of steam drum going Figure 9 shows tubes/elements for manufactur- on in one XXL size boiler in China. Figure 10c be built. That project was awarded to us in 2002. As an example in world’s biggest recovery boiler ing of boiler bank in one big recovery boiler. shows drum inside of world’s biggest recovery there is about 190 m furnace downcomer pipes On the left hand side there are bundles of tubes boiler ready for installation of steam drum in- Some physical facts from xxl size with size of 711 x 55 mm (Outer diameter OD x which have come from the tube supplier and on ternals. In Figure 10d there is welding of risers wall thickness). Weight of this pipe is 890 kg/m recovery boilers the right hand side there is the output from the going on. and so total weight of downcomers is about 170 Steam boilers are built by using a lot of tubes, manufacturing process; a readymade boiler bank tons. pipes, downcomers, headers etc. Table 1 shows package for transportation. In big recovery boilers the weight of steam drum can be 150-200 tons, its length can be about 20 how much heating surface tubes are there in Furnace lower part is made by using composite m, outer diameter 2.4 m and wall thickness 120- world’s biggest recovery boiler. As can be seen material (AISI 304L or Sanicro 38) which con- Heaviest single component in recovery boiler is 150 mm. total amount of tubes is about 500 km. Distance sists of carbon steel base material and outer layer a steam drum. It is principally heart of a recovery

from Tampere to Helsinki is 170 km and so of stain less steel material. In above mentioned boiler. Purpose of the steam drum is to separate this boiler has three times more tubes than this boiler there is about 18 km this composite mate- water and steam from each others. Figure 10a distance. There are still many other pipes in rial. Material grades of those heating surfaces shows the steam drum of one big boiler on the recovery boilers which are not listed in the table 100 101 a) b) Main fuel is of course black liquor coming Table 3 shows an example which fuels are burnt from the pulping process. Nowadays, very typi- in certain big recovery boilers. cally as fired dry solids content of black liquor is 80-83%. Black liquor systems are therefore As can be noticed one big boiler in Finland has pressurized systems meaning that for example designed for burning really many fuels. Boiler black liquor burning temperature is controlled in has been operated now for five years without any heavy black liquor tank to targeted value which problems due to type of fuels. can be at the level of 140 ºC. Oil and natural gas are typically used during boiler start-ups and Materials in xxl size recovery shutting down procedures in start-up burners and load burners. In some cases load burners are boilers used for generating more steam to steam net and Furnace floor and walls c) d) controlling its pressure. NCG gases are collected XXL size recovery boilers have principally same around the mill from atmospheric tanks/equip- materials as other recovery boilers have. Furnace ment (DNCG) or from non-atmospheric tanks lower part has been made by using composite (CNCG). DNCG gases are mixed with air and material of AISI 304L/SA-210 A1. This mate- burnt together with air at the secondary air level. rial has been used already nearly for 40 years. CNCG gases are burnt in a dedicated burner. As Reason for using stainless steel material is a a support/main fuel in the CNCG burner can corrosion risk in reducing atmosphere in lower be liquid methanol/turpentine mixture, oil or part of furnace. If material would be carbon steel natural gas. In some softwood mills also soap is and we have a high pressure boiler which has burnt by mixing it with black liquor. tube surface temperature 320-330 ºC then tube corrosion is very evident in environment which Secondary bio-sludge from waste water effluent exists in furnace lower part. Figure 10. Steam drum on the way to the erection site and its lifting to the top of boiler support structure. plant can be mixed with black liquor and burnt in the recovery boiler furnace. Floor tubes are typically made by carbon steel Recovery boiler as a multi-fuel boiler system and so mixed with air. CNCG burner is and that is possible because they are protected located at secondary air level. Variety of fuels is causing challenges for boiler during operation by frozen smelt layer. Earlier Modern XXL size recovery boilers are multi-fuel optimum design from emissions, corrosion and it was common to make floor also by composite boilers. Boilers have latest advanced designs Figure 11 shows normal fuels burnt in recovery fouling point of view. Therefore when boilers tube (AISI 304L) but in some boilers stainless from environmental point of view; all NCG and boilers. are designed it is very important to know all steel outer layer cracking has been encountered. vent gases are burnt in recovery boiler. DNCG, incinerated fuels in order to select furnace size When cracks exist there is always risk that they dissolving tank and mixing tank vent gases are and materials properly. Especially minimizing penetrate trough carbon steel base material typically introduced to the high secondary air NOx emissions in XXL size boilers is sometimes and therefore water leak may occur and worst causing challenges because many of those fuels scenario is smelt water explosion. Reasons for are including ammonia which can then generate cracking are not 100% sure but they are initiated more NOx emissions. Table 3. Fuels in different big recovery boilers.

Fuel Boiler A Boiler B Boiler C Boiler D Boiler E Black liquor X X X X X Biosludge X X X Oil X X X X X Diesel X Natural gas X X Methanol X X X Turpentine X CNCG X X X X X DNCG X X X X X Figure 11. Possible fuels in recovery boilers. Vent gas X X X X X

102 103 On Figure 14 there is a lower part of furnace • High dry solids firing and good air systems during erection of one XXL size recovery boiler. with effective air/flue gas mixing have minimized risks to sulfidation corrosion in furnace. • Optimized material selection with increased knowledge of K (potassium) and Cl (Chlorine) chemistry and their effect to superheater corrosion have decreased corrosion findings in superheater areas.

In new recovery boilers there is a clear trend to have higher steam parameters in order to increase electricity production. This has caused Figure 11. Composite floor and cracks in tubes and membranes. special challenges to superheater material selection and how to prevent sudden corrosion. In due to tensile stresses prevailing in stainless 38 outer layer is closer to carbon steel than in Figure 14. Furnace lower part during erection XXL size recovery boilers there can be 150 km steel layer when tube surface temperature has 304L which is beneficial from thermal stresses stage. superheater tube materials and everybody can been high and stainless steel part has yielded, point of view. Figure 12 shows main material image that what could happen if tubes are cor- Figure 11. alternatives in recovery boiler furnace. roded and boiler will have unplanned stop due Superheaters to heavy material losses. Big mills can produce One of the most challenging material selections pulp 1.5 million ADt/a meaning about 4200 ADt in recovery boiler is superheaters. Typically per day. If pulp price is US$ 1000 per ton then in recovery boilers many kind of corrosion losses of income can be US$ 4.2 million per day. phenomena can be detected in heating surfaces Very typically when superheaters are repaired as is shown in Figure 15. Many of those corro- due to unpredicted failures then outage may take sions have been managed to minimize in modern at least 4-7 days from liquor to liquor depending recovery boilers: of course on how severe damages are. Figure 16 Figure 12. Tube material alternatives in recovery boiler furnace. shows secondary and tertiary superheaters hang- • Acidic sulfates are not formed any more as ing from the furnace roof when one XXL size easily as earlier due to high dry solids firing, recovery boiler is in under erection. Length of Due to cracking problems in 304 L composite Figure 13 shows modern material selection in good air systems, low sulfidity level in euca- elements is about 24 meters. tubes new materials have been developed for re- XXL size recovery boiler floor area. As can be lyptus and acacia mills. placing that material on floors and in some other seen the floor heart is made by carbon steel ma- critical areas in recovery boiler lower furnace terial. Floor bends toward front and rear walls, like in primary air port openings. One developed side most tubes adjacent to side walls and pri- material is Sanicro 38 composite tube. In this mary air port bended tubes are made by Sanicro material outer layer has about 40% Ni and 20% 38 material. This kind of design has proven to be Cr whereas in 304L material Ni is 10% and Cr a very reliable solution. 20%. Also heat expansion coefficient of Sanicro

Figure 13. Material selection of floor and lower part of furnace. Figure 15. Occurrence of corrosion in recovery boilers. 104 105 As mentioned earlier the knowledge of corrosion Table 4. Recovery boiler references with NOx emission targets have been managed to theories has increased a lot during the last years stainless steel materials in superheaters. reach with normal air systems and when lower and therefore it is very important to know what values are required then more air staging has are K and Cl in black liquor when new boilers Project SH steam Stainless steel Ash been used by installing Quaternary air level. are designed because those components are most outlet material in handling Figure 18 shows a couple of examples from one critical from corrosion and superheater fouling temperature hottest stages XXL size boiler operation. In Figure 18a there point of view. (C) is black liquor load and steam generation when Boiler A 505 8/T Yes capacity test was carried out. Virgin black liquor How then to avoid corrosion? When designing Boiler B 480 N/A Yes load was about 7200 tds/d and corresponding new recovery boilers it is very typical to deter- Boiler C 480 4/S and 4/T Yes as fired load was 7700 tds/d. Steam generation mine maximum K and Cl values for black liquor during test was targeted to be 300 kg/s. In Figure Boiler D 485 1/T Yes or ESP ash. Those values can be controlled to 18b there are reduction rate values when boiler desired levels by handling ESP ash with Ash Boiler E 487 2/T Yes Figure 16. Superheater elements under erection. had a performance test. Measured values in Leaching or Crystallization processes and also Boiler F 460 1/S and 4/T Yes smelt exceeded 97% which is really good result. with ash dumping. One alternative is also to use Molten phase corrosion may be very severe in Boiler G 487 1T adn 4/Q Yes Reduction rate is one of the most important per- better materials in critical part of superheaters. hottest superheater stages. It occurs when ash/ Boiler H 492 1/T Yes formance data which is followed during boiler This better material means higher alloyed tubes. deposit melts on the tube surface. This corrosion operation. It tells how much sodium sulphate For example Sanicro 28 or corresponding mate- is controlled by the first melting temperature Some operational experiences from (Na2SO4) is reduced to sodium sulphide (Na2S) rials have been used quite often in recovery boil- (FMT) of the deposit. The FMT is the tem- xxl size recovery boilers in char bed. Figure 18c indicates what has been ers which have high steam parameters and/or perature, at which the first melt appears in the oxygen level in flue gases after economizers. black liquor is containing a lot K and Cl. Sanicro XXL size recovery boilers have been operated deposit (also known as T0). Main contributors Low value means good combustion process in 28 material has Cr content about 28%. High Cr extremely well from process and mechanical to FMT are chlorine (Cl) and potassium (K). the furnace and it also indicates low flue gas heat is favorable for preventing corrosion to happen. point of view. Flue gas emissions like SO and High amount of carbonates (CO ) i.e. carryover 2 losses. 3 TRS have been typically at the level of zero. lowers also the FMT of the deposits. Even small Table 4 shows a couple of recovery boiler ref- amount of sulfides (0.1 wt.-%) in the deposits erences with stainless steel material selection lowers the FMT about 50 °C. in superheaters. Stainless steel has typically been whether Sanicro 28, AISI347L or overlay a) The most challenging superheater stages are welded tubes. In the Table 4 there is shown hottest ones from where superheated steam is steam outlet temperature, how many tubes have going to main steam net. Figure 17 shows an stainless steel material, code S means secondary example from the corroded tubes in the hottest superheater, T means tertiary superheater and Q superheater stage. means quaternary superheater. Table also shows if ESP ash is handled or not for controlling K and Cl in black liquor/ESP ash. Black liquor In green liquor In smelt b) Acacia 100% (7232 tds/d) 97,2%, 96,4%, 97,5%, 97,4%, 96,3% 97,1% 97,53% Euca 100% (6994 tds/d) 97,0%, 96,8%, 97,8%, 97,4%, 97,1% 97,5%

Figure 17. Corroded tube in superheater outlet. Figure 18. Some operation parameters from an XXL size boiler in China. 106 107 Summary New green field pulp mills are huge. Pulp production starts to be 1.5-2.0 million ADt/a. This means that also the size of recovery boilers has come very big. Boiler size for those big mills is about 7000-8000 tds/d. Valmet has delivered many boilers which have had a capacity of 4000-5000 tds/d and experiences from those references have been extremely important when developing even bigger boilers like XXL size units. We have also seen certain difficulties in those boilers but all problems have been successfully solved and therefore we have been very confident with our design when bigger and bigger boilers are built.

XXL size recovery boilers have operated really well and availability has been good. From process point of view we have not experienced any setbacks. Mechanically certain big components have caused challenges mainly due to unexpected movement of elements but they have been managed to fix in normal outages meaning that no abnormal availability problems have been encountered.

Bigger and bigger boilers are planned to build and therefore we feel very confident for those coming Magnum size boilers because the references which are working well are a must in order to get customers trust on your technology and on you as a reliable supplier.

108 109 high energy recovery boilers

Marja Heinola and keijo salmenoja ANDRITZ Oy FINLAND

Introduction

The kraft recovery boiler is always referred Since the focus has earlier been on the chemical as the heart of the pulp mill and is the critical recovery and not on the power production, the component in the closed loop producing energy power generation efficiency (power-to-heat- alongside recovering the cooking chemicals. As ratio) has always been rather low in traditional long as the heart sustains well, other systems boilers. In order to improve the power genera- excel. Today the mills, constrained with rising tion from the recovery boilers, several technical energy cost and security of supply strive to solutions have been developed and adopted. It maximize green energy to substitute fossil fuel is not, however, only in the recovery boiler area and attain self-sufficiency. where savings can be done, but it is possible to achieve significant savings also in other process Modern non-integrated pulp mills are nowadays areas. totally self-sufficient in respect to steam and power production, and the idea of producing and selling not only pulp, but also electricity is already a reality at many mills. Even though a mill is self-sufficient in energy, there are normally still many areas where energy could be saved and electricity production could be increased. The market is globally demanding less pollu- tion per produced megawatts (MW) and better utilization of renewable resources.

The change to increase power production from kraft recovery boilers has been quite rapid. Figure 1. Possibilities to increase the energy effi- In the 1990s, when new mill or new recovery ciency at kraft pulp mills (Salmenoja & Heinola boiler was supplied, energy efficiency was 2013). not a topic. Today almost in all new cases,

however, energy efficiency is the topic number Power production in modern mills is presently one. Figure 1 shows the development of power around 1100 kWh/ADt at a softwood mill and production from recovery boilers equipped with around 900 kWh/ADt at a hardwood mill. Power high-energy features and the possibilities to consumption at softwood and hardwood mills is increase energy efficiency at kraft pulp mills. To ca. 600 kWh/ADt and 500 kWh/ADt, respec- be able to increase the energy efficiency of the tively (Figure 2) (Salmenoja & Heinola 2013). pulp mills, efforts to decrease the consumption of energy at the mills has to be on the focus, too.

110 111 Due to the originally poor power generation ef- The optimum way to increase the power gen- ficiency, several improving features have been eration efficiency is to utilize all the methods applied to the recovery boiler. The high-energy mentioned above. Although, every measure to recovery boiler (HERB) concept was developed increase power production has a price tag, the for this purpose. It increases the power genera- payback time of most of the measures is very tion efficiency further than what is possible with short. Figure 6 shows estimated annual profits traditional increase of the steam parameters. The available with different options. HERB concept was the answer to the demand to increase the power-to-heat-ratio and power production of recovery boilers. The first HERB Figure 5. The first kraft recovery boiler outside with a steam temperature of 505 °C was started Japan with steam outlet temperature of 515 °C up in 2004 in Finland. Figure 4 shows the side was started up in 2006 (Salmenoja and Heinola, Figure 2. Development of power consumption view of the 2004 vintage first HERB. and production at pulp mills. Mill A represents a 2013). modern softwood and mill B a modern hardwood compared to an old recovery boiler. There are mill (Salmenoja and Heinola, 2013). several ways to increase the power production from a recovery boiler and some of those may The main factor limiting the increase of power also include operational risks. production from recovery boilers is corrosion. Therefore, the development in the steam outlet This paper recapitulates measures to increase temperature in kraft recovery boilers has been power production from recovery boilers and Figure 6. Estimated annual profits available extremely slow compared to power boilers. summarizes practical experiences from operat- with different measures to increase power Japanese boiler manufacturers have been the ing high-energy recovery boilers. production (Lehtinen 2005). pioneers with high-temperature recovery boilers. The first recovery boiler having a steam outlet High-energy recovery boiler concept High dry solids temperature of 500 °C was started up in Japan Power generation, in combined heat and power Dry solid content of black liquor has a large in 1983. Recovery boiler with steam outlet tem- system (CHP) is described in terms of the power impact on steam generation. It is more effective perature of 515 °C was started also in Japan in generation efficiency (PGE) or the power-to- to evaporate the moisture in a multiple effect 1988 (Arakawa et al. 2005). The highest steam heat-ratio. In order to improve the power gen- evaporator with process heat rather than doing outlet temperature utilized in kraft recovery eration, high-energy recovery boilers employ the same inside the furnace. The increase in net boilers is still 515 °C. several technical solutions. steam generation for 65% solids as against 85% solids is approximately 7-8%. Further, with high Recovery boilers having higher steam outlet There are several ways to increase the power temperature than 515 °C are not in operation. concentration liquor, it is possible to operate production from a kraft recovery boiler and with low excess air, which again minimizes loss Figure 3 shows the development of steam totally, some 20 MW more power can be parameters in kraft recovery boilers from 1930s to the flue gas. A comparison of heat balance at produced by different measures. However, different dry solid content is furnished in Figure (Vakkilainen, 2005). Figure 4. Side view of the first HERB started up some of the measures may include operational 7. in 2004. risks, which should be taken into account. The The most common methods for increasing the following is not a comprehensive list, but shows power generation efficiency have been higher the most commonly used measures to increase black liquor dry solid content and higher steam power production: values. To further increase steam parameters • High dry solids in black liquor from present levels are mainly limited by corro- • High steam temperature and pressure sion factors. Together with these, new methods • Sootblowing steam from the turbine of feedwater preheating, air preheating with • Preheating of feedwater fluegas, combining bio-gas firing to the recovery • Preheating of combustion air boiler, reuse of secondary heat streams in the • Fluegas cooling evaporation area, etc. have been introduced. • Hot condensate return • Heat recovery from vent gases Figure 3. Development of steam parameters of With a modern high-energy recovery boiler, • Interheater Figure 7. Comparison of heat balances for kraft recovery boilers (Vakkilainen, 2005). mills can easily double their power production • Reheater recovery boilers (Saviharju and Lehtinen 2005). 112 113 Steam temperature and pressure steam to extraction steam from turbine, improve- due to the higher flue gas temperature. In order Summary of the measures ments to the power generation efficiency can be to improve the overall efficiency, fluegas cool- Another methodology for improved PGE is to use The means are almost unlimited, but the most achieved. ing after the precipitator could be introduced. If higher steam values (McKeough and Saviharju, optimal way to enhance the efficiency of a pulp the mill has a deficit of low pressure steam or 2007). Figure 8 shows the power generating ef- mill is always a matter of thorough studying of Sootblowing steam is typically taken directly hot water, or is using a condensing turbine, the ficiency as the function of steam outlet tempera- the individual mill’s and department’s balances. from the final superheater steam outlet and is fluegas cooling should be exploited. ture (Rahman and Lehtinen, 2013). However, passed through a poppet valve to reduce the The additional production with the solutions high-temperature corrosion of superheaters is pressure from 17 to 24 bar before entering the When cooling the flue gases with 100 °C water mentioned above for a recovery boiler operat- the main limiting factor of increasing the steam sootblower feed tube. (water heated to approx. 150-160 °C) it is pos- ing at 92 bar(a) and 490 °C with a condensing outlet temperature. In general, potassium (K) sible to set the flue gas temperature in the stack turbine is presented in Table 1. sets practical constraints in strive towards higher Since sootblowers consume 3 to 12% of the to around 140 °C. The heat from the fluegases final steam temperatures. Potassium is the main total superheated steam produced by the boiler, can be used for feedwater pre-heating, air pre- Table 1. Additional power produced with various contributor in molten phase corrosion by dictat- sootblowing with high pressure steam can be a heating, hot water production etc. technical solutions in a recovery boiler operating ing the first melting temperature (FMT) of the costly operation. If sootblowers can operate at at 92 bar(a) and 490 °C (Lehtinen 2005). deposits. A rule of thumb has been to keep the a lower pressure, for instance, 9 to 14 bar(g), Features Net electricity material temperatures below the FMT. If molten there will be a significant economic advantage produced phases are present on superheater tubes, corro- to pulp mills. This is because low pressure steam Feedwater pre-heating and + 2.7 MW sion rates will be quite high and unpredictable. is less valuable than high pressure steam, as it interheating can be taken from the steam turbine exit after With regard to material selection, chloride level the steam has been used to generate electricity. Air pre-heating + 1.0 MW in black liquor dictates the need of expensive Internal sootblowing + 0.6 MW materials. Chlorine-induced corrosion below Feedwater pre-heating Hot primary condensate + 0.6 MW the FMT can be handled with material tech- In a high pressure boiler there is usually a large return nology, but it necessitates solid knowhow of margin between the feedwater temperature to the Higher dry solids in BL + 2.0 MW deposit properties and local material tempera- Figure 9. Fluegas coolers. drum and the drum saturation temperature. In - 80% tures. Maybe due to this fact, no one has had the this case the steam generation can be increased courage to exceed 515 °C as the steam outlet Hot condensate return Fluegas cooling + 4.5 MW by pre-heating the feedwater. First the feedwater temperature in black liquor recovery boilers. Total +11.4 MW can be heated with low pressure steam in the The benefit of extremely high steam parameters In the pulp mill, there are hundreds of different feedwater tank to the maximum temperature. For may diminish due to exorbitant cost of materials. primary and secondary water and steam streams, instance from traditional 120 °C up to 145 °C which are either used or dumped to the cooling Practical operating experiences with low pressure steam. Between the feedwater towers. Often the best solution for one depart- tank and the economizers the feedwater can be ment is not the most economical solution for The first HERB was started up in 2004 in pre-heated with medium pressure steam to 160 the whole pulp mill. Often energy balances are Finland and presently 10 such recovery boilers °C. Between economizer I and II the feedwater reviewed only for one department. One example are in operation. Two boilers are under construc- can additionally be heated by 15-20 °C with high is the flashing of the primary condensate at the tion. The first units were not equipped with all pressure extraction steam. evaporation plant. Instead of flashing it down the power production enhancing features, but to 100 °C, the primary condensate could be due to encouraging experiences from operating Pre-heating of combustion air returned un-flashed to the condensate tank and units, new features to further enhance the power from there to the feedwater tank. In this case, the production have been adapted. Pre-heating of all combustion airs to 190 °C need of steam to the evaporation plant increases, with different level of extraction steam is also The first HERB had a relatively high capacity but the need of steam to the feedwater tank a quite cost effective way to produce more (4450 tds/d) and steam parameters (505 °C, Figure 8. Power generation efficiency as the decreases. function of steam outlet temperature (Rahman steam and hence improve the power generation 102 bar). This boiler is still the largest recovery & Lehtinen 2013). efficiency. Traditionally, primary and secondary Reheater concept boiler in Finland. The first HERB with a steam airs are heated up to 120-150 °C. By adding a outlet temperature of 515 °C was started up in few air pre-heaters the power production can be The reheater concept was studied in the research Sweden in 2006. The second HERB with the Sootblowing steam increased further. project SKYREC by the Finnish Recovery same steam temperature was started six years Boiler Committee (FRBC) to further increase later, also in Sweden. Totally, six of the HERBs In older recovery boilers, fresh superheated Fluegas cooling the power output from kraft recovery boilers have steam outlet temperature higher than 500 steam was used to sootblowing. Depending on (Finnish Recovery Boiler Committee, 2013). °C, as shown in Table 2. the boiler and black liquor properties, sootblow- The above mentioned measures improve the According to the study, the reheater boiler ing steam consumption can vary from 3% up to power generation efficiency of the boiler, but the concept seemed not to be profitable for kraft 10%. By changing the source from superheated total efficiency of the boiler is slightly lowered recovery boiler purposes. 114 115 When the first unit with the 515 °C steam outlet Corrosion management Figure 12 shows a practical example of the effect maximize power production from the recovery temperature was started up, steam temperature of material grades on corrosion rate. It shows a boiler. This has led us to the basic questions, i.e. The main reason for excellent experiences from was gradually increased to the target value, superheater tube sample taken from a recovery what are the practical constraints to maximize HERBs with high steam outlet temperatures as shown in Figure 10. This was mainly done boiler suffering from rapid degradation of the the power output. However, every measure that is the low K and Cl content in as-fired black since previous experiences on high steam outlet hottest tubes. The tube material below the butt increases power output has also a price tag and liquor. Proper control of both K and Cl is vitally temperatures were not available. weld is austenitic stainless steel and the mate- they all increase the investment costs. Therefore, important to avoid high-temperature corrosion rial above is low-alloyed steel. The reason for a feasibility study should always be made to of the hottest superheaters. Since K is the main After the first scheduled annual shutdown, the rapid material degradation of low-alloyed steel optimize the configuration to maximize power contributor to the FMT, the only way to avoid boiler was thoroughly inspected. Since no marks was confirmed to be active oxidation due to high output. molten phase corrosion of superheaters is proper of enhanced corrosion or other material degra- Cl level in as-fired black liquor. As can be seen control of K level in as-fired black liquor. High- dation was not found, steam temperature was from the figure, active oxidation can be effec- The SKYREC project aimed at with possible new temperature corrosion cannot be totally avoided raised permanently to the design value of 515 tively avoided by using highly alloyed materials, measures to further increase the power output although material temperatures are kept below °C. Since then, the boiler has been running at such as austenitic stainless steels. from kraft recovery boilers (Finnish Recovery full load and at designed steam parameters. the FMT, since active oxidation is responsible Boiler Committee 2013). One objective was to for chlorine-induced corrosion below the FMT. evaluate constraints to aim at even higher steam Low-alloyed However, active oxidation can be limited by cor- parameters than the existing 515 °C and 110 bar. steel rect material selection (Salmenoja 2000). The target was set on steam temperature of 540 °C and steam pressure of 160 bar. Table 2. Capacities and steam properties of At high steam outlet temperatures, special mate- HERBs in operation and under construction. rials must be used in superheaters. The amount of Several field probe tests were carried out to HERB Start-up Capacity Steam Steam Note high-alloyed materials is mainly dictated by the study the suitability of commonly used materi- year (td s/d) temperature pressure Cl content in black liquor and applied steam tem- Austenitic als and new potential materials for superheaters (°C) (bar) perature. In some cases the amount of expensive stainless and furnace walls (Figure 13). According to 1 2004 4450 505 102 - materials may become extremely large, without steel the studies, it seems possible to find feasible 2 2006 4850 510 103 - proper optimization of the superheater design. materials also for recovery boiler environments 3 2006 3300 515 106 FG coolers To be able to construct cost-effective HERB at increased temperature and pressure (Finnish 4 2007 1000 505 110 FG coolers solutions, a superheater material temperature Recovery Boiler Committee 2013). 5 2009 2200 485 88 FG coolers calculation tool has been developed. It renders Figure 12. A practical example of the effect of 6 2011 1590 493 103 - possible to optimize superheater design and material grade on superheater corrosion. 7 2012 2500 495 105 FG coolers the amount of corrosion resistant high-alloyed 8 2012 2400 515 110 FG coolers materials (Figure 11). With this tool it is possible The HERB has the same design constraints as 9 2012 5710 495 97 FG coolers to minimize the amount of expensive materials the conventional recovery boilers and high steam 10 2014 1750 495 96 FG coolers in superheaters and to assure a long service time parameters must not jeopardize the pulp mill’s 11 2015 2300 515 110 FG coolers with a high availability. The HERB has the same availability. To be able to design a high-energy 12 2016 11600 515 110 FG coolers design constraints as the conventional recovery recovery boiler with a high availability, there are boilers and high steam parameters must not two major things that must be known; 1) mate- jeopardize pulp mill’s availability. rial temperatures in superheaters and 2) K and Cl levels in as-fired black liquor. By knowing these two factors, superheater materials can be optimized and their lifetime reliably predicted. Figure 13. Measured corrosion rates of sev- The main factor limiting the increase of power eral superheater material candidates (Finnish production from recovery boilers is corrosion. Recovery Boiler Committee, 2013). Therefore, increasing steam outlet temperature and pressure requires a sound knowledge of recovery boiler fluegas chemistry, fly ash be- Increasing the steam outlet temperature also havior, and the control of heat transfer surface increases applied steam pressure. This in turn temperatures. means rising furnace wall material temperatures and presently used wall materials may not be Future development challenges used in the future. In the SKYREC project ex- Figure 10. Steam outlet temperature during the Figure 11. Example of the outcome from the isting and potential new furnace wall materials first operating years in the HERB with steam optimization tool to calculate superheater metal Increased interest in pulp mills to increase the were field tested with a special test rig. Figures parameters of 515 °C and 106 bar (Salmenoja temperatures and to optimize the need for high- energy efficiency sets increasing requirements 14 and 15 show the main results of these tests. and Heinola 2013). alloyed steels. for the boiler suppliers to search for new ways to

116 117 Recovery boiler capacities have been increasing Summary References extremely rapidly during the last few years and There are several ways to increase the power it seems that this is the trend also in future. This 1. Arakawa, Y., Taguchi, Y., Maeda, T., Baba, Y., production from a kraft recovery boiler. Totally, is due to increased pulp mill capacities. New 2005, Experience with High Pressure and High slightly over 20 MW more power can be pro- Temperature Recovery Boilers for two decades, mills are in the 1.5 million ADt/a capacity range, duced by different measures. With the HERB Pulp&Paper Canada, Vol. 106, No. 12, pp. 89-92. which means that the recovery boiler capacity concept, mills can easily double their power is around 7000-8000 tds/d. The world’s largest 2. Finnish Recovery Boiler Committee, 2013, production compared to an old recovery boiler. Internal FRBC report (SKYREC). kraft recovery boiler is now under construction (Figure 16). Its capacity is 11600 tds/d and it 3. Lehtinen, M., 2005, Boosting power from the Proper superheater design and optimal heat Recovery Plant. Paper Making and Distribution, will be started up in 2016. It will also be the transfer rate are the key to predicted performance March/April. most energy efficient kraft recovery boiler ever and high steam production. Increasing steam Figure 14. Estimated maximum corrosion rates built, with steam parameters of 515 °C and 110 4. McKeough, P., Saviharju, K., 2007, Enhancing outlet temperature and pressure requires a sound production of energy, power and fuels in conjunction for 3R12, HR11N, Sanicro 38, Super 625, and bar. It has all the features to maximize the power knowledge of recovery boiler fluegas chemistry, with chemical recovery and simultaneously reducing Sanicro 67 materials according to the SKYREC output. fly ash behavior, and the control of heat transfer green-house gas emissions. Proc. 2007 International field probe test (Finnish Recovery Boiler surface temperatures. Operational experiences Chemical Recovery Conf., Part 2, PAPTAC, Montreal, Committee 2013). have confirmed that the HERB concept is pos- pp. 385-390. sible to realize without endangering the mills 5. Rahman, F., Lehtinen, M., 2013, A New HERB availability. (High-Energy Recovery Boiler) in Asia. IPPTA Vol 25, No. 1, pp. 139-144. New ways to produce more electricity have been 6. Salmenoja, K., Heinola, M., 2013, Operational introduced during the years. However, all the Experiences On High-Energy Recovery Boilers. the easy ways to enhance power production have 8th International Colloquium on Black Liquor and been adopted and to be able to further increase Biomass to Bioenergy and Biofuel, Belo Horizonte, the power-to-heat-ratio of recovery boilers may Brazil, March 19-23, 2013. include high-risk measures. 7. Salmenoja, K., 2000, Field and Laboratory Studies on Chlorine-Induced Superheater Corrosion High-temperature corrosion is mainly limit- in Boilers Fired with Biofuels, Academic Dissertation, ing the increase of steam outlet temperatures. Åbo Akademi University, Turku, Finland. Figure 15. Estimated maximum corrosion rates Therefore, efforts have been put to overcome 8. Saviharju, K., Lehtinen, M., More power from the for 3RE28, 3XRE28, HR11N, and Sanicro 38 material issues in kraft recovery boilers. The Recovery Island, Paper and Timber, Vol. 87 No. 4, pp. materials according to the SKYREC field probe Finnish Recovery Boiler Committee has also 226-231. Figure 16. Side view of the largest and most test (Finnish Recovery Boiler Committee 2013). been active in promoting material studies and 9. Vakkilainen, E., 2005, Kraft recovery boilers – energy efficient kraft recovery boiler in the material research in different areas. The coopera- Principles and practice, Finnish Recovery Boiler world. Committee, Valopaino Oy, Helsinki, Finland, 246 p. Further increase in the steam outlet temperature tion between boiler suppliers, mill operators, as requires larger amounts of more expensive well as with research institutes and universities, superheater materials. In addition, largely used Emission limits will be tightened in the future. has led to practical solutions that have already lower furnace material 3R12 (304L) may not be New European legislation sets new and more been exploited in practical applications. any more feasible at higher operating pressures. stringent emissions limits for the recovery boil- Therefore, new materials need to be searched ers. Other countries are following the European and tested. The question still remains: could the route towards lower emissions from kraft steam outlet temperature of 540 °C be possible recovery boilers. This trend is not a problem in the future? with other emissions, except with nitrogen oxide (NOx). Increasing power output brings more power boiler features also to kraft recovery boilers. Aiming at high power output with low NOx emis- One practical example is cooling of fluegases sions is somewhat contradictory. Maximizing after the precipitator. This increases the en- power output requires high furnace tempera- ergy efficiency and power output significantly. tures, but minimizing NOx emissions requires Today, fluegas temperatures after the coolers are as low temperatures as possible. Therefore, the around 140 °C. But how low can we go with the outcome is always a compromise in this respect. fluegas temperature without losing the boiler availability? 118 119 CFD Modeling of Kraft Recovery Boilers - A Retrospective

andrew jones international paper united states

Introduction

Progress in the field of kraft recovery boiler Clearly we need careful reflection about, and modeling is assessed by retrospectively look- search for, enduring patterns and critical turning ing at the proceedings of the 1991 Timberline points in the past, for these are the historical Colloquium on Recovery Boiler Modeling facts that everyone needs to know ... (1991). The focus being on: the players involved then and now; the extent to which gaps have ... ignorance of history--that is, absent or defec- been closed since this colloquium and progress tive collective memory--does deprive us of the (or lack thereof) that has occurred since the time best available guide for public action... of this meeting. This gives me my cue on how to go about this The 1991 Timberline Conference was the first presentation: how has our understanding of large gathering of those interested in developing CFD for kraft recovery boilers changed in the kraft recovery boiler CFD capability, using CFD intervening time, what does this tell us about models and validating the results. It was a com- the problem of modeling, verifying and using ing of age of this technology that had its infancy CFD models; can patterns be extracted from the in the late 1980s. application of this technology that we can learn from and guide us moving forward; and how can Organized by Weyerhaeuser and B&W, two this presentation be best used to retaining this early leaders in this development activity, the knowledge so we can be guided going forward, conference had 51 attendees. I will delve a bit so that this valuable tool can continue to become more into the attendees and their affiliations more useful. later in the paper. I would be remiss in not mentioning the location, Before diving into this topic we need to ask: just outside of Portland, Oregon; the Timberline What is the intent of this retrospective? I believe Lodge is located above the tree line on Mt. a proper study of history can provide valuable Hood, a dormant volcano in the cascades range. insight. The following are a few quotes that I bor- The lodge was completed in 1938 by the Works rowed from the American Historic Association Progress Association (WPA). Many of you may (1985). have unwittingly seen footage of the lodge in the movie “The Shining” (red rum, red rum…). One . . .the changing perspectives of historical un- of the snowiest places on earth it averages 14 derstanding are the very best introduction we meters per year. I encourage anyone visiting the can have to the practical problems of real life. Portland area to make the trek up to the lodge and stay at least one night. An old photo taken in 1943 is shown in Figure 1. 120 121 The other theme that is also worth considering is this meeting. Within this group are two aca- on in the development of kraft recovery boiler that of physical flow modeling as a basis for the demic organizations (The University of Toronto CFD models. validation of CFD and also as a valuable trouble- and Åbo Akademi University) that have become shooting tool in its own right. Rolf described a critical contributors to model development, This session ended with a panel discussion on technique where acid-alkali reactions are used to taking over from IPST. I believe some of the char bed combustion. What was most interesting simulate mixing as a way to predict NOx emis- catalyst for taking on these roles was provided in this session was the discussion for a need for sions on a utility boiler. This theme will also be by this colloquium. a dynamic char bed model that could predict the touched on further. influence of operating parameters on char bed The topics shape. It took another 17 (!) years to develop this The other early pioneers on the attendee list in- type of model as described in a recent paper by cluded: Dr. Tom Grace, who led the development There were 5 topic areas: Combustion Engblom et al. (2008). of the first comprehensive CFD model at the Fundamentals, Physical Modeling of Recovery Institute of Paper Chemistry in the 1985-1989 Boilers, The Role of Models in Boiler Design, The next session on the physical modeling of time frame; Two of his team members were in Computer Modeling of Recovery Boilers and Figure 1 - Old image of Timberline Lodge Circ 1943 recovery boilers is a research area that most attendance, myself then representing ABB, Model Validation. - USFS photo #424587 by George Henderson. young engineers are probably not very familiar and Allan Walsh, representing the J.H. Jansen with. Before the rise of CFD, physical models Company. Allan continues to work in the field I will not address all of the individual presen- were the only tool for predicting gas flows and So let’s look back 23 years, before some of the to this date; Woody Fiveland was also present, tations, but rather just touch on some of the mixing in combustion equipment. Early in the audience was born and when most of us had far he was working for B&W, another company highlights. development of recovery boiler CFD models, less gray hair. that has successfully developed and maintained validation of CFD results versus physical mod- their own CFD models to this date. Rick Wessel The combustion fundamentals work highlighted Attendees and affiliations eling results was an active research area. It was has long since taken over Woody’s efforts at research on the burning properties of black this work that allowed us to gain confidence in As mentioned previously Weyerhaeuser and B&W; Tampere University represented by Reijo liquor a topic that has been a key research topic the CFD results, and in some case grave doubts. B&W were the organizers of the Colloquium. Karvinenen and Tampella represented by Pekka to this day. Many PhDs have been received in There are still some cases, such as very complex The introduction to the colloquium was given Siiskonen were already active in sub-model an effort to fully examining this topic. A full geometries, where only recently has CFD mod- by Denny Hunter and Joe Barsin. Denny has development using Fluent. understanding of these fundamentals is essential eling advanced sufficiently to replace the utility recently retired from Weyerhaeuser. Most if accurate predictive CFD models are to be of physical models. recently he was CTO at Catchlight Energy, the There were no representatives of Fluent at the developed. A vital partnership exists between Weyerhaeuser-Chevron Joint-Venture for biofu- meeting, but there were a number of end users these developers of fundamental models suit- In a subsequent panel discussion on the role of els development. Joe Barsin, a long-time B&W of the commercial code, Gotaverken, Ahlstom, able for CFD application and those using CFD CFD on recovery boiler design the remarkable employee retired a number of years ago but still Tampella, Jansen, ABB and perhaps some oth- to predict recovery boiler operation. The need fact that B&W has been doing computational remains active as a consultant. ers. I will touch on Fluent’s role in kraft CFD for this partnership was recognized very early flow modeling since 1971 was mentioned. Some development a bit more later on. early kraft recovery boiler modeling results The two keynote presentations are very in- are shown in Figure 2. ABB, Ahlstrom (some formative and were given by two very well I’ve attempted to describe the attendees based on early results shown in Figure 3), Tampella and known researchers: Dr. Rolf Collin of the Royal four different categories at the time of the meeting: Institute of Technology (KTH) and Dr. Martha CFD model user, CFD model developer, Salcudean at the University of British Columbia. Validators – interested in proving value of CFD Both of these institutions have continued ac- modeling, and other interested parties. tive research into CFD development but with CFD Model Developers – IPST, ABB, Jansen, distinctly different directions. UBC developed B&W, KTH, UBC, Tampere University of their own code, using multigrid segmentation Technology, BYU (for Utility boilers mostly). and advanced solvers, eventually spinning off CFD Model Users – Gotaverken, Ahlstom, Oy a company, Process Simulations Limited that Polyrec, Chemrec, Tampella (all using Fluent) continues to provide modeling service on kraft Validators – Paprican, Quest Integrated, recovery boilers and a variety of other pulp and Sandwell, Weyerhaeuser (also a user of the UBC paper process equipment. KTH based on subse- code), Advanced Fuel Research. quently published research has made extensive Interested parties - Longview Fiber, University use of Fluent a commercial code as a basis for of Toronto, U.S. DOE, International Paper, Åbo the development of useful sub-models. These Akademi University, James River, Kimberly two different pathways will be touched on again Clark, Swedish Pulp and Paper Research Figure 2 - Ahlström Recovery Boiler “Slice” Figure 3 - Early B&W Recovery Boiler CFD later in the talk. Institute. None of these attendees presented at Model. results. 122 123 Gotaverken also presented examples of CFD be- mercial code of vendors had its genesis at that can be then adapted by the do-it- ing used for recovery boiler design, some of the about this time. IPC and Tampere University yourself developers; earliest references to the work of recovery boiler of Technology had already begun these 2. Academic organization have played a key vendors in this field. These modeling efforts efforts and in the next few years activities role in sub-model development (usually have continued to this day the players have new at both Åbo Akademi University and the freely available), the commercial code us- names though, Metso (Valmet now!), Andritz University of Toronto began. The DOE rep- ers can more easily benefit from these and Alstom. It has been a very interesting history resented by Stan Sobszynski at this meeting “open-sourced” models. This synergist of recovery boiler manufacturers, but that is a went on to play a key role in funding the relationship has many benefits and must be topic for another day. early code development activities in this supported going forward; field at both of these institutions. Contact at 3. Academic organizations need sources of The next session on computer modeling of this meeting certainly played a role in these funding in order to play this role. In the recovery boilers had a paper from Bob Horton efforts; U.S. the DOE played this role, in Finland of IPST, Bob led the modeling effort at IPST in 3. A number of topics were not even discussed the support of Tekes has been vital. the early 1990’s. Work was completed on a DOE in this meeting including simulation of soot- 4. What we did not even see in 1991 are now project focused on improving droplet modeling blower jets, modeling of fume deposition. active areas of code development, not really and char bed modeling in kraft recovery boilers Modeling of fume formation and modeling a surprise but reassuring. Additional novel and integrating this work into the Fluent code. char bed growth were on the wish list but no modeling applications must be pursued. Martha Salcudean also presented a paper in efforts had been started at this time; One such idea is the ability to model growth this session that described the innovative work 4. We have made significant progress since and removal of deposits by sootblowers; that was going on at the University of British meeting in a number of areas: validation, 5. Validation is still elusive, some progress has Columbia at this time. They were using a code using CFD as a design tool, using CFD as been made but more work needs to be done. called Teach, that used high order differencing a troubleshooting tool, modeling gaseous Getting validation data from an operating schemes and multi-grid techniques, improve- emission levels – NOx, CO, O2, and carry- recovery boiler is a challenge and we need ments that would be adopted by commercial over predictions; to continue to pursue means to do this. codes. Woody Fiveland of B&W then presented 5. There is a surprising continuity of play- cold flow modeling results using FORCE a very ers in this field. Many of the names from On final point is that I have a copy of the pro- early version of the (COMO-PR™) code that this conference are still active in the field. ceedings from the 1991 Timberline Colloquium, they have extensively developed since this time A number of new players have entered I would be happy to make additional copies for for kraft recovery boiler modeling applications. including the University of Toronto, Aalto anyone interested. Please contact me via e-mail The last session on validation presented a few University, Åbo Akademi University, if you would like a copy. ideas on how to take in-situ measurements in Tampere University of Technology, and a recovery boiler. It is unclear if either of the Umea University. techniques: acoustic Doppler, or FT-IR have References ever been successfully applied to recovery boiler Conclusions 1. Engblom, M., Mueller, C., Brink, A., Hupa, M., in-situ measurement. Jones, A., 2008, Toward predicting the char bed shape 1. Two valid paths of code development have in kraft recovery boilers. TAPPI Journal, Vol. 7, No. Observations been proven: build your own code (e.g. 10, pp. 12-16. PSL, B&W and using a commercial code 2. http://www.historians.org/about-aha-and- In reviewing the Timberline Colloquium a few (e.g. Andritz, Valmet, Jansen). The second membership/aha-history-and-archives/archives/why- general observations can be made: method has the advantage of being better study-history-(1985) American Historic Association. 1. Fluent played a key role in kraft recovery able to support the inclusion of user defined 3. Timberline Colloquium on Recovery Boiler boiler CFD development, initially by pro- sub-routines as the work product from oth- Modeling – April 10-13, 1991. Available upon request. viding access to source code (IPC developers and in being able to rely on the software ment), and then later on by the use of user vendors to implement improvements to defined subroutines that provided a window the general code. The “build your own” into their code. While obviously a commer- approach allows for possible competitive cial company interested in selling modeling advantages in the quality of results and code it seems to me they went above and the control of the overall code. History has beyond this role in the level of support they thus told us that there is a place for both of provided to code developers; these approaches. Thus any user subroutine 2. The concept of using academic organization should be available as both user defined to develop sub-models for use in the com- subroutines and as descriptive algorithms

124 125 Entrained Flow Black Liquor Gasification – Review of Pilot Scale Research 2004 – 2014

Rikard Gebart Luleå University of Technology sweden

Introduction

Black liquor gasification was proposed as early In this chapter the research on entrained flow as in the 1960s as an alternative to the conven- gasification of black liquor during the last ten tional Tomlinson Recovery Boiler for recovery years will be summarized. Most of the research of energy and chemicals from black liquor. The has been carried out within the framework of black liquor gasification based recovery process two Swedish research programs called BLG-1 has the following potential comparative advan- (2004-2006) and BLG-2 (2007-2010) and within tages (Whitty 2009): the EU 7th framework project BioDME. The gasifier design around which this research has been • Elimination of the smelt-water explosion risk performed is the so-called Chemrec technology, in a recovery boiler which is an evolution of the original invention • Elimination of malodorous compounds by Kignell (1989). The current Chemrec process • Improved power efficiency by using -inte is a slagging, oxygen-blown entrained flow gasi- grated gasification and combined cycle power fier with a direct quench. generation • Increased pulp production and increased pulp The review is divided into sections about pilot quality by utilization of split sulfidity pulping scale experiments, durability of containment that is made possible by the natural separa- materials and mill integration. tion of sulfur and sodium in a black liquor gasifier • Improved control of the sodium-sulfur The DP-1 pilot plant balance The construction of the DP-1 pilot plant in Piteå, • Possibility for efficient production of second Sweden (Figure 1) started in 2004 and the first generation motor fuels and high added value hot experiments were made in September 2005. chemicals from sustainable forest biomass The gasifier is an oxygen-blown entrained flow • The historical development of black liquor gasifier that operates at an internal pressure up gasification has been described by Whitty to about 30 bar. The nominal black liquor flow (2009) and the interested reader is referred rate is 20 tons dry solids per 24 hours and this to that paper and to the paper by Whitty and corresponds to a thermal power of about 3 MW Verrill (2004) for more details. The conclusion with the standard black liquor (app. 73% wt dry in 2009 was that the future of black liquor solids) from the neighbor pulp mill. The typical gasification looked promising and that the high process temperature (as measured by the process temperature, air/oxygen-blown entrained-flow thermocouples) needed for full carbon conver- gasification had emerged as the technology of sion is about 1050 °C and the corresponding choice. oxygen equivalence ratio is in the range 0.3-0.4. The operating temperature is well above the 126 127 melting point for the spent pulping chemicals After quench cooling and coarse removal of The pilot plant has several hundred sensors/actu- Pilot scale experiments and the resulting viscosity is very low so that smelt droplets the syngas is fed to a syngas ators and an industry grade computerised control The main goal with the pilot plant experiments any material that ends up on the hot reactor wall cooler operating in counter current mode. The system that records all parameters, including the in the DP-1 plant was to run the plant in the will form a liquid film that will flow out of the syngas, which at this point is saturated with syngas composition measured with an on-line same way as an industrial gasifier to prove that reactor lower exit. Residues that are suspended steam, will gradually be cooled in the syngas GC. The pilot plant design is identical to a full- the technology was reliable and mature enough in the gas are also molten at this temperature. cooler. During this process the moisture will scale industrial gasifier with the exception that to be implemented in a full-scale commercial collect on particles in the syngas and on the heat there is a lack of redundant systems that would plant. Since the plant, due to budgetary reasons, The reactor pressure vessel is lined with refrac- exchanger surfaces. In the pilot plant the gas make it possible to do routine maintenance with- was built without redundant systems the pos- tory material consisting of two layers, a hot face cooler is a tube-and-shell heat exchanger with out shutdown of the plant. The limiting factor sibility for continuous operation was somewhat with fuse-cast material with high resistance to two tube packages separated by a plenum cham- at present is the green liquor return system that limited. The goal for operation was therefore chemical attack and one back-up layer with ber approximately at mid-height. In a full-scale has to be cleaned from fouling with inorganic set to achieve at least three weeks of continu- higher thermal insulation. The gas and molten plant the syngas cooler would instead be a heat deposits with three-week intervals. Two opera- ous operation without interruptions. This was salts that flow out of the hot reactor are quench- recovery steam generator (HRSG) that would tors can run the plant safely. achieved for the first time in the fall of 2007. cooled with several spray nozzles that can be recover a significant fraction of both latent and Another goal was to gradually increase the load fed with water and condensate from the gas sensible heat from the syngas and thereby im- During the break-in period during the first year and to operate the plant at its design capacity cooler. All or part of the spray will be evaporated prove the overall efficiency of the plant. of operation a number of issues typical for until the gas becomes saturated with steam. The start-up of a new process had to be resolved, and pressure and this was achieved the same resulting gas temperature and steam partial pres- Chemrec AB, who holds several patents for the e.g. replacement of pumps and flow meters that year. After these achievements the goal was to sure depends on the contact between the sprays process, made the design of the DP-1 gasifier. did not perform according to the requirements. maximize the operational time during the rest of and ratios of the flow rates for syngas and water. The black liquor to the pilot plant comes from Other than these rather trivial problems the the program to make it possible to expose any The lower part of the quench vessel forms a pool the neighboring pulp mill in Piteå, owned by major effort was spent on optimisation of the unexpected long term problems due to build up that collects the un-evaporated liquids from the Smurfit Kappa Kraftliner, through a pipeline that burner nozzle to obtain a stable flame and good of trace elements and to quantify the material spray nozzle and the molten salts from the hot is connected to the mill’s black liquor evapora- carbon conversion. At present, in April 2014, the degradation/need for maintenance. In June 2013 reactor. The liquid mixture in the pool is called tion plant. The green liquor from the gasifier plant has an accumulated run time in excess of the accumulated operation time had surpassed green liquor and has an elemental composition is pumped back to the mill through a parallel 25000 hours and it is routinely operated 24/7 in 20000 hours (Figure 2). In addition to the indus- and color that are similar to those in green liquor pipeline. The pulp mill also provides medium three-week periods (Landälv et al. 2014). trially motivated activities mentioned above, the from a recovery boiler. pressure steam for preheating of black liquor plant was also used for scientific research that is and other purposes. described in the following.

Closure of the material and energy balance for the DP-1 A key measure of the goodness of pilot scale experiments is how well the mass and energy balances can be closed, based on direct measurements of the relevant parameters. In the present case, in order to analyze the data, the following assumptions had to be made:

• The amount of dilution water entering the system is unknown. The flow rate was therefore adjusted to close the water balance around Figure 2. Operating hours of the DP-1 BL gasi- the system. The resulting value was typically fier September 2005 to June 2013 (Landälv et around 700 kg/h, which can be compared to al. 2014). the black liquor flow rate that was around 900 kg/h in the experimental campaign for which the balances were made. • The water formed by the reactions in the reactor was also unknown as a consequence of the bullet point above. The formed water was assumed, based on a theoretical estimate, to be 100 kg/h. Figure 1. Schematic of the DP-1 pilot scale black liquor gasifier (Courtesy of Chemrec AB).

128 129 All other flow rates were taken from direct optimized thermodynamic data for black liquor The parameter study was done by changing the value of about 0.6 below which concentrations

measurements. The system boundary for the (Backman 2010). Hence, it can be concluded system pressure, oxygen to black liquor equiva- of CO2, CO, and H2, are significantly changed system is defined in Figure 3. that the gasification process is a non-equilibrium lence ratio λ, black liquor flow rate to pressure in the quench. process for the inorganics. ratio and black liquor pre-heat temperature. • For high primary spray flow rate/load ratios, Black liquor Oxygen It was found that all of these parameters had a the temperature is reduced adequately fast significant effect on the product gas composi- to prevent any significant influence from the Raw syngas Syngas composition Nitrogen tion and other important process characteristics water–gas shift reaction and therefore the gas Gasifier The syngas composition from the DP-1 gasifier, Conden- (Carlsson et al. 2010). The main conclusions concentration after the quench is nearly the ser after the gas cooler, was continuously measured from the study are: same as in the hot reactor. Nitrogen Cond. Flash gas with an on-line process GC. The data was auto- Quench • For low primary spray flow rate/load ratios, Flash matically stored in the process-monitoring data- • The bulk composition of the product gas is the chemistry in the quench is highly influ- base for off-line analysis. In order to investigate mainly determined by the water gas CO-shift enced by the water–gas shift reaction and the MX Cooler the influence from quenching and gas cooling on reaction and thus, by the operating tempera- gas composition after the quench is shifted to Bypass the syngas a set of experiments were performed ture which to a large extent is controlled by more H2 and CO2 and a smaller amount of CO

Weak wash Green liquor MB water Condensate in which the syngas was instead sampled from λ. than in the hot reactor. the hot reactor. The sampling was done with a • Increased system pressure promotes H S in Figure 3. System boundary for the DP-1 2 • For high primary flow rates, CO2 absorption specially designed water-cooled probe that could the gas. material balance. can become significant. be mounted in different positions (Wiinikka et • Increased λ suppresses the amount of CH4 in al. 2010). The cooling rate for the syngas that is the gas. A common problem area in biomass gasification A detailed mass and energy balance was done for sampled with this probe is of the order 10000 K/s • The amount of analyzed carbon containing is tar in the product gas. The relatively problem the system defined in Figure 3 in an experimental and the gas composition at the sampling point is gases increase with decreased black liquor free operation of the DP-1 gasifier (very little campaign in March 2009 (Granberg et al. 2009). therefore efficiently “frozen”. The samples were flow rate to pressure ratio. fouling on cold surfaces) shows that the tar con- Four independent measurements were done, stored in a sampling pressure vessel that could • At constant pressure, the heating value of the tent is low but in order to quantify the amount of each separated by around 24 hours from the be disconnected for offline analysis with a cali- gas increase with increased black liquor flow tars Johansson et al. (2013) have made a direct previous one. Mass balances were checked for brated GC. Using this probe a set of experiments rate to pressure ratio. measurement. The measurements were done the total flows and for each major element. The with variations of process parameters were done • Increased black liquor pre-heat temperature both according to the CEN Tar Measurement

mass balance closures were in most cases within (Carlsson et al. 2010). A typical gas composition suppresses CH4 and H2S content in the gas. Standard and with a solid phase adsorption 10% difference between inflow and outflow. from these experiments can be seen in Table (SPA) method. The tar content was found to However, for nitrogen the error was 20-28%. 1. Notice, that the measured change in most of The characteristics of the gasifier were further be about three orders of magnitude lower than The most likely explanation of this discrepancy the components before and after the quench are investigated in another experimental campaign from low temperature fluidized bed gasification is that nitrogen was bubbled through the green within the experimental uncertainty but that the in which the influence from the cooling rate on of forest biomass, 4-9 mg/Nm3 with the CEN liquor to enhance mixing and that some of the COS concentration is significantly reduced after the product gas composition was investigated. standard method and slightly higher with the nitrogen was absorbed by the green liquor and the quench. The cooling rate was varied by changing the SPA method. Approximately 50% of the total tar was therefore unaccounted for. This explanation flow rate of water to the quench (Wiinikka et content was found to be naphthalene. is further strengthened by direct measurements Table 1: Syngas composition from 10 independ- al. 2012). It was found that the hot reactor gas ent measurements in the DP-1 gasifier (pressure composition prior to the quench could either of the dry gas from the green liquor flash tank Other trace components were also investigated 27 bar, temperature 1050 °C, λ = 0.3). The num- be preserved (high cooling rate) yielding the that showed that it consisted mainly of nitrogen by Öhrman et al. (2012). The particle concen- bers after the +/- symbol indicates the standard same final gas composition after the quench as and hydrogen. tration in cooled syngas was measured with a deviation of the data. in the hot reactor or shifted (low cooling rate) particle impactor and the particle concentration towards a higher concentration of H and CO . One important result from the mass balances is Compound Inside Cooled 2 2 was found to be very low (Figure 4). As a refer- The conclusions from this study are in line with the sulfur-sodium split that results in the transfer reactor (%) syngas (%) ence, good wood combustion yields a flue gas the previous results but with added insight into of a significant fraction of the sulfur from the CO 33.9±0.3 33.6±0.2 that has particle loading that is about two orders 2 the influence from the quench: black liquor to the syngas while the majority of of magnitude larger (in the vicinity of 10 mg/ H2S 1.65±0.04 1.71±0.02 the sodium ends up in the green liquor. This is an 3 CH 1.36±0.07 1.44±0.07 • The system pressure, O /BL flow ratio, and the Nm ). important difference compared to the recovery 4 2 primary spray flow affect the final gas com- boiler where most of the sulfur in the black CO 28.7±0.2 28.5±0.2 Eleven different elements and compounds were position of CO , CO, and H after the quench. liquor ends up in the green liquor. In the present H 34.3±0.2 34.8±0.1 2 2 also analyzed using impinger bottles through 2 The system pressure and O /BL flow ratio also case, according to the mass balances, 36-43% of 2 which the syngas bubbled. Out of the 11 ele- COS 468±22 ppm 122±5 ppm affect the concentrations of CO , CO and H the sulfur in the black liquor was found as H S in 2 2 ments and compounds that were analyzed only 2 H /CO 1.19±0.01 1.22±0.01 inside the hot reactor. the syngas. This can be compared to the value of 2 two, Cl at 3 ppm and NH -N at 100 ppb, were • The primary spray flow rate/load (mass flow 4 about 60% at a gasification temperature of 1050 H2/CO2 1.01±0.01 1.04±0.01 found in concentrations above the detection lim- °C obtained from an equilibrium analysis using of black liquor and oxygen) ratio has a critical CO/CO2 0.85±0.01 0.85±0.01 it. Chlorine is a potentially problematic element

130 131 other sulfur compounds. In all, this means that borate. The sodium borate would then follow there is a possibility that the formation of depos- the liquor cycle until it enters the gasifier with its on solid surfaces in the green liquor system the black liquor for a new gasification cycle. will be different in the two cases. Bialik et al. However, no auto-causticizing was detected and

(2010) have therefore done both theoretical and the conclusion is that the presence of CO2 in the experimental work to assess the scaling risk. The gasifier inhibits the auto-causticizing reactions theoretical work is based on a solubility model under gasification conditions. for sodium carbonate, calcium carbonate and pirssonite (Ulmgren et al. 1999) that was used An alternative to borates for auto-causticizing is to evaluate the scaling risk in green liquor from to use titanates instead. Sinquefield et al. (2010) the gasification process. The model was first have presented an experimental study similar verified against samples from the DP-1 gasifier to that by Nohlgren and Sinquefield (2007) and that were obtained under different process con- obtained slightly better results than for borate ditions. In this study, it was found that the green auto-causticizing. However, also in this case the

liquor composition was relatively insensitive to inhibition of the reactions by CO2 makes it un- process variations when the process parameters feasible to use this method in a gasifier similar to were varied within their normal limits. The the DP-1 where the operating pressure is around overall conclusion in the study by Bialik et al. 30 bar. (2010) is that the risk for pirssonite precipitation in gasification green liquor is relatively low. Mill integration After the study of Bialik et al. was done the Recovery of energy and chemicals from the DP-1 gasifier has been operated for a total of black liquor is a necessity for a modern pulp more than 20000 hours using optimized settings mill, both from an economic and an environ- of the gasifier process variables (Landälv et al. mental point of view. The industry standard is 2014). The green liquor from the gasifier is con- to use a so-called Tomlinson recovery boiler (Tomlinson & Wilcox 1939) for efficient recov- Figure 4. Particle mass size distribution in the product gas after the gas cooler. tinuously flashed and pumped back to the host pulp mill for mixing with the green liquor from ery of energy and chemicals from spent pulping liquors. The main reason why this is so is that for catalyst beds since it may cause sintering of the concentration of sodium hydroxide, sodium the recovery boiler. The operational experience the recovery boiler can be tightly integrated with packed beds of catalysts and the gas cleaning sulfide and sodium carbonate. The conclusions from this extended operating period is that there the rest of the pulp mill and provide the energy system must therefore be designed to reduce Cl from the study are that the separation properties will be a slow formation of scaling on the com- and the pulping chemicals that are needed for the to sub ppm level. for the green liquor and the total concentration ponents that are in contact with the green liquor pulping process thereby minimizing the need for of the produced white liquor are almost identical but the deposits are easily removed by a standard acid wash approximately every three weeks. additional fuel and make-up chemicals. Recovery of spent pulping chemicals between the two cases. Moreover, it was found that the causticizing efficiency was almost the Larson et al. (2000) made a flow sheet analysis of Richards and Theliander (2010) made an experi- As mentioned above, black liquor gasification same for gasifier produced green liquor and several proposed black liquor gasifiers that were mental study in which the white liquor production leads to an increased demand for reburning of the conventional green liquor in spite of the used for integrated gasification and combined from conventional green liquor was compared lime mud compared to chemicals recovery with significant differences in green liquor composi- cycle power production (IGCC) and compared to the case with green liquor from the DP-1 a Tomlinson boiler. Several investigations have tion between the two cases and the fact that more the economics to that for a conventional recovery gasifier. The black liquor was the same for both therefore focused on the possibility to reduce reburned lime mud must be used to causticize boiler. The conclusion is that the financial cost is cases and the green liquor samples came from or completely remove the need for causticiz- the green liquor from the DP-1 gasifier. about the same as for the recovery boiler but that the DP-1 gasifier and from the recovery boiler ing by adding a causticizing agent to the black the power production can be doubled to tripled. of the host pulp mill (Smurfit Kappa Kraftliner One significant difference between a conven- liquor before it enters the gasifier. Nohlgren Hence, the power production cost will be about in Piteå). Laboratory experiments were made in tional recovery boiler that operates at atmos- and Sinquefield (2007) presented results from half that with Tomlinson-based technology. which the two green liquor samples were first pheric pressure and a pressurized gasifier that experiments in a pressurized drop tube furnace filtrated to remove green liquor sludge and then operates at a pressure around 30 bar is that the in which black liquor with an additive of sodium borate (NaBO ) was exposed to simulated gasi- After the turn of the century much of the inter- cooled. The cooled clean samples were then green liquor temperature will be up to about 100 2 fier conditions. Theoretically, the sodium borate est shifted to the combination of black liquor used for causticization experiments in a small K higher in the gasifier case before it is depres- gasification with syngas conversion into motor stirred reactor using the same reburned lime surized and cooled to below the boiling point. should react with sodium carbonate in the smelt and produce Na BO and CO . The Na BO fuels and high added value chemicals. Consonni mud from the host pulp mill for both samples. In addition, a significant fraction of the sulfur 3 3 2 3 3 et al. (2009) made detailed mass-energy balance Finally, the causticized samples were filtered in the black liquor will in the gasifier case be would then react with water in the smelt dis- solver and form sodium hydroxide and sodium simulations and combined that with a financial before ABC-titration was used to determine transformed to gas phase as H2S and possibly analysis for seven pulp mill biorefinery process 132 133 configurations, all of them based on an oxygen- a lifetime of about two years for a gasifier that is A., McGowan, K.A., 2007, Recent experience with 19. Whitty, K., Verrill, C.L., 2004, A historical look blown, high temperature black liquor gasifier. continuously operated. structural materials in commercial scale black at the development of alternative black liquor recov- Six of the alternatives also included a low tem- liquor gasifiers. In International Chemical Recovery ery technologies and the evolution of black liquor Conference, 29 May - 1 June, Quebec. gasifier designs. In International Chemical Recovery perature fluidized bed biomass gasifier for IGCC Conclusions power production that was fully integrated with 8. Kignell, J-E., 1989, Process for chemicals and Conference, 6-10 June, Charleston. the gasification and syngas cooling section of Pressurized, oxygen blown entrained flow energy recovery from waste liquors. U.S. Patent No. 20. Whitty, K., 2009, The changing scope of black liq- the black liquor gasifier. Three different biofuels gasification of black liquor has undergone a 4,808,264. uor gasification. Proceedings of the 45th Anniversary were considered, dimethyl ether (DME), FT significant development since the turn of the 9. Landälv, I., Gebart, R., Marke, B., Granberg, F., International Recovery Boiler Conference, June 3-5, 2009, Lahti: Finnish Recovery Boiler Committee, pp. liquids and mixed alcohols. The ratio of useful century. The accumulated operating time of the Furusjö, E., Löwnertz, P., Öhrman, O.G.V., Sørensen, 133–144. energy outputs (steam, power, fuels) to total oxygen blown DP-1 pilot gasifier exceeds 20000 E.L., Salomonsson, P., 2014, Two years experience energy inputs (black liquor, wood residuals, fuel hours (in May 2014 it is over 24000 hours) and of the BioDME Project - A Complete Wood to Wheel 21. Öhrman, O., Häggström, C., Wiinikka, H., Concept. Environmental Progress & Renewable Hedlund, J., Gebart, R., 2012, Analysis of trace com- oil) were in the range 66-74% compared to about the process is reliable and predictable. The syn- Energy, in print. ponents in synthesis gas generated by black liquor 57% for a Tomlinson boiler based system. The gas has very low tar content and is suitable both 10. Larson, E. D., Consonni, S., Kreutz, T.G., 2000, gasification. Fuel, Vol. 102, pp. 173–179. estimated specific capital investment cost was as a gas turbine fuel and as syngas for catalytic production of motor fuels and chemicals. Preliminary Economics of Black Liquor Gasifier/ found to be US$60000-150000 / barrel diesel Gas Turbine Cogeneration at Pulp and Paper Mills. equivalent per day which is about the same as The process can also be integrated with existing Journal of Engineering for Gas Turbines and Power, for a much larger coal-to-liquids plant. Vol. 122, No. 2, p. 255. pulp mills in a very efficient fashion that has the 11. Nohlgren, I., Sinquefield, S., 2007, High tem- Pettersson & Harvey (2012) made a similar potential to improve the pulp mill profitability and reduce its carbon footprint by production of perature, high pressure black liquor gasification with study in which black liquor gasification based borate autocausticizing. In International Chemical fuels that will displace fossil fuels at the same biorefinery alternatives were compared to recov- Recovery Conference, 29 May - 1 June, Quebec. ery boiler-based pulping biorefinery concepts time. 12. Pettersson, K., Harvey, S., 2012, Comparison both from an economic and climatic point of of black liquor gasification with other pulping view. It was found that black liquor gasification References biorefinery concepts – Systems analysis of economic with DME production had the best economic performance and CO2 emissions. Energy, Vol. 37, No. performance for all the energy market scenarios 1. Backman, R., 2010, Equilibrium based modeling 1, pp. 136–153. of black liquor gasification with the Four-Box Model. that were considered. It was also concluded that 13. Richards, T., Theliander, H., 2010, White liquor Chapter in the BLG II Program Final Report. the black liquor gasifier should, from an eco- production - A comparison of the causticizing stages nomic point of view, be supplemented by a solid 2. Bialik, M., Wadsborn, R., Karlholm, I., Berglin, using gasifier and recovery boiler produced green liq- biomass gasifier for additional syngas that can N., 2010, Assessment of scaling risk in a pressurized uor. In International Chemical Recovery Conference, be used both for synthesis of additional DME black liquor gasification pilot plant. In International 29 March - 1 April, Williamsburg. Chemical Recovery Conference, 29 March - 1 April, and for power production. 14. Sinquefield, S., Zeng, X., Ball, A.W., 2010, Direct Williamsburg. causticizing for pressurized black liquor gasification. 3. Carlsson, P., Wiinikka, H., Marklund, M., Durability of containment materials In International Chemical Recovery Conference, 29 Grönberg, C., Pettersson, E., Lidman, M., Gebart, March - 1 April, Williamsburg. R., 2010, Experimental investigation of an industrial Black liquor is a very corrosive liquid and, 15. Tomlinson, G.H., Wilcoxson, L.S., 1939, Method scale black liquor gasifier. 1. The effect of reactor regardless of the recovery technology, the mate- of and apparatus for burning waste liquor. US Patent operation parameters on product gas composition. No. US 2161110 A. rial selection for components that are in contact Fuel, Vol. 89, No. 12, pp. 4025–4034. with the medium must be done cautiously. For 16. Ulmgren, P., Rådeström, R., Edblad, M., 4. Consonni, S., Katofsky, R.E., Larson, E.D., 2009, the recovery boiler, more than seven decades of Wennerström, M., 1999, On the process chemistry of A gasification-based biorefinery for the pulp and non-process elements in systems with a pressurized development and industrial experience has led paper industry. Chemical Engineering Research and black liquor gasifier. J. Pulp Paper Sci., Vol. 25. to material combinations that have an accept- Design, Vol 87, No. 9, pp. 1293–1317. able lifetime and safety level. For black liquor 17. Wiinikka, H., Carlsson, P., Granberg, F., 5. Granberg, F., Berglin, N., 2009, Closure of mate- Löfström, J., Marklund, M., Tegman, R., Gebart, R., gasification the experience is much shorter but rial and energy balance over DP-1. Technical report, 2010, Design and methodology of a high temperature the experience from the Weyerhaeuser New Chemrec. Bern booster gasifier has yielded a good starting gas sampling system for pressurized black liquor 6. Johansson, A-C., Öhrman, O.G.V., Pettersson, gasification. Fuel, Vol. 89, No. 9, pp. 2583–2591. point for evolutionary development. Initially, E., Sweeney, D., 2013, Tar and Trace Element 18. Wiinikka, H., Carlsson, P., Marklund, M., the lifetime of the refractory lining of the reactor Measurements in Synthesis Gas from a Pressurized Grönberg, C., Pettersson, E., Lidman, M., Gebart, vessel in New Bern was disappointing (Keiser Black Liquor Gasifier. In 21st European Biomass R., 2012, Experimental investigation of an industrial et al. 2007). After systematic development of Conference and Exhibition, Copenhagen, pp. 3–7. the material solution and extended tests it was scale black liquor gasifier. Part 2: Influence of quench 7. Keiser, J.R., Meisner, R.A., Hemrick, J.G., Gorog, operation on product gas composition. Fuel, Vol. 93, concluded that the optimized materials will have J.P., Leary, W.R., Brown, C.A., Landälv, I., Gupta, No. 1, pp. 117–129.

134 135