Process Systems Engineering a Retrospective View With Question for the Future SARGENT 2004

7/30/2019 Process Systems Engineering a Retrospective View With Question for the Future SARGENT 2004

1/8

E u r o p e a n S y m p o s i u m o n C o m p u t e r - A i d e d P r o c e s s E n g i n e e r in g - 1 4A . B a r b o s a - P 6 v o a a n d H . M a t o s ( E d i t o r s )9 2 0 0 4 E l s e v i e r B . V . A l l ri g h t s re s e r v e d .

P r o ce ss S y s t em s E n g i n e e r i n g - A R e t ro s p e ct iv e V i e w w i t hQ u e s t i o n s fo r th e Fu tu r e

Ro ger Sargent*Imper ial College, London

A b s t r a c tThe paper reviews the development of Process Systems Engineer ing f rom a personalv iew-point , select ing concepts f rom the past which have g iven r ise to new paradigms,and hence led to improved unders tanding or advances in so lu t ion techniques , but whichei ther st il l pose unreso lved problem s or hold the prom ise of fur ther developments .1 . R e v i e wThe systems engineer ing approach seems to have had i ts or ig ins in the 1940 's in theef for ts to explo i t the possib i l i t ies of the newly invented electronic d ig i ta l computers ,which held out the promise that , i f a problem could be precisely formulated , then inpr incip le an algor i thm could be developed to so lve i t , and hence implemented andsolved on a com puter w ithout the need for fur ther human in tervention or expert ise.I t considered a complex system as a co l lect ion of in teract ing "units" , whose behaviourwas well unders tood, a nd concentrated on unders tanding the in teractions between them.In these ear ly days , chemical engineer ing was of ten descr ibed as "systems engineer ingapplied to the problems of the process industr ies", and it was certainly true that theconcept of "unit operat ions" ident if ied a l imited set of basic operat ions used in a widevar iety of processes , and made possib le their sys tematic s tudy, independent of thesubstances proce ssed . T his in turn led to a better unders tanding of these operations , andhence to the development of s imple models , such as the " theoret ical p late" or the " f i lmtheory, ' for heat and mass transfer , to predict the essence of their behaviour.The recognit ion that mixtures could be separated in to f ract ions contain ing an arb i trar i lylow level of impur i t ies by use of a ref lux s tream in a mult i - s tage countercurrent systemwas also a typical "systems" idea.Equally impor tant was the representat ion of the process by a " f low-diagram" or " f low-sheet", sh owin g how the uni ts were in ter l inked, and the recognit ion that conservat ion ofmater ia ls and energy provided suff ic ient means to determine the uni t in teract ions .However these ideas were developed long before electronic computers , and the f low-sheet calculat ions had to be devised on a case by case basis .With the advent of the computer , the immediate task was to replace the many elegantgraphical so lu t ion procedures then in general use by numer ical a lgor i thms, andengineers were faced with the need to unders tand avai lab le general numer icaltechniques or develop new ones of their own. As the s torage capaci ty of computersexpanded, i t was possib le to s tore more and m ore of the program s and the data , much o f* E-mail : r .w.h .sargent@ ps. ic .ac .uk


2/8

which was common to many different programs, and the need for more systematicclassif ication and storage quickly em erged.On ce past this phase, interest turned towards m ore systematic organization of the f low-sheet calculations. Here the use of graph theory [1] revealed a technique for identifyingthe groups of units linked by recycle streams (the "strong components"), whichtherefore had to be computed together, indicating both a decomposition of the problemand appropriate candidates for iteration within the groups. When "sequential modular"flowsheeting programs, which used subroutines for the unit calculations, were replacedby "equation-based" programs, these same techniques could be used to exploit thestructure of the resulting large sparse system s of equations.Natural developments from the process f low-sheet were Ferencz Friedler 's "P-graphs"[2] for represen ting process sy nthesis routes, the concept of "super-structures" [3],which incorporate a variety of options within a single flow-sheet, thus providing arational (but limited) basis for process synthesis, and Bodo Linnhoff's representation ofheat-exchanger networks, associated with his "pinch technology" [4] approach foranalysing these netw orks and for wider applications.In Computing Science, "object-oriented programming" represented the ultimategeneralization of this basic idea, but in the effort to achieve generality perhaps too muchwas sacrificed. As the success of the equation-based approach illustrates, efficiencydepends crucially on the choice of the "units", and it is not always convenient orefficient to deal w ith very black boxes.Optim ization quickly becam e a central guiding principle [5] , eventually applied todesign, planning, operation and control, and permeating the whole field. However itsma ny ramifications really require a separate review o f a more m athematical kind, whichwould carry me too far from the themes I wish to pursue here, though inevitably it willintrude at various points.Analogous developments occurred in other branches of engineering. For example incontrol en gineering the "units" were described by transfer functions, and block diagramswere used to represent how these were linked; Manfred Morari 's "internal modelcontrol" block diagram structure [6] provided for example a unified representation andtreatment o f the m odel predictive control problem.Use of the concept o f "state", wh ich originally formed the basis for the development ofthermodynamics, caused a paradigm shift in control from an emphasis on thetransformation of inputs into outputs to one on the evolution of the system state underthe influence of the inputs, leading to the development of the Kalman filter [7] forrecursive estimation of the current state and revealing the duality between the conceptsof observability and controllability [8].The state viewpoint also provided a link to differential equation theory, and thepossibility o f tackling control for nonlinear systems, w ith another paradigm shift to theviewp oint o f optimal control [9 ]. For some y ears the consequential compu tingrequirements meant that this could only be used off-line to generate optimal controlpolicies, but eventually Dave Cutler launched his "Dynamic Matrix Control" [10],showing that simple linear models could be used to produce a discrete-time linear-quadratic on-line optimal controller- and proved his point by launching a company tomarket the approach with outstanding success. Since these early days the increasingspeed o f computers and improvem ents in numerical techniques for both state estimation


3/8

and optimal contro l have ma de possib le on-l ine use of mo re sophis ticated nonlinearmodels and more real is t ic object ive funct ions [11 ,12] , making th is the most promisingparadigm for the fu ture .Meanwhile contro l engineers have become more aware of the need for pro tect ionagainst undue sensi t iv i ty of the contro l ler to modell ing and predict ion er rors , and the H-infinity approach to "robust control" [13], based on a worst-case analysis, has donemuch to produce a unif ied v iewpoint .The robu stness issue raises the whole quest ion o f how to deal with uncer tain ty , not onlyin contro l but in any s i tuat ion where we need to predict the consequences in order totake appropriate action. Catering for a worst-case outcome is safe if i t is really possible,but i t is usual ly unnecessa r i ly conservat ive, and i t would be preferable to cater only foroutcom es a bove a cer ta in level of probabili ty , and w ith in th is constrain t take act ion todeal with the most l ikely outcome.There has been much d iscuss ion on whether there real ly is such a th ing as a randomoccurrence or whether we are deal ing with chaotic responses ( responses to such acomplicated and far - reaching set of inf luences that they appear to be unpredictable) .However predict ion using probabil i ty d is tr ibut ions has served us well , and in theabsence of predict ions o f such d is tr ibut ions f rom chaos theory , we mu st be content withdis tr ibut ions based on empir ical observat ion , with some assurance f rom the "centrall imit theorem" that many independent inf luences tend to combine to g ive a Gaussiandistr ibution.For large uncer tain t ies the computat ional requirements may s t i l l be too heavy a pr ice topay for the addit ional assurance obtained , and in some cases th is is bet ter obtained byuse of a weighted object ive funct ion , based on a small number of postu lated scenar ios[14]. (Of course th is can be in terpreted as a crude wa y of generat ing an approximatedistr ibution function).In sp i te of advances through the use of "parametr ic programming" [15] , one cannotpretend that the problem is sat is factor i ly resolved , but we have to be content with thetools current ly a t our d isposal .Process engineer ing has a lways had to cope with relat ively complex , poor ly unders toodprocesses , exhib i t ing essent ia l ly nonlinear behaviour . An ear ly d iff icu l ty , which had tobe deal t with before successfu l dynamic s imulat ion , le t a lone opt imal contro l , waspossib le , was the problem of numer ical in tegrat ion of what are now known as " s t i f f 'sy s tems- - sys tems whose r esponses to inpu t changes a re a combina t ion o f r esponses o fd if ferent par ts of the system operat ing on very d if ferent t ime-scales [ 16] .Happily th is was resolved [17] , but the extreme form of th is problem, the inclusion ofcomponents with ins tantaneous responses , g ives r ise to d if ferent ia l-algebraic systems[18] , and no fu l ly-proved technique for deal ing with th is has yet been published . Onecan take refuge in the fact that no physical processes are real ly ins tantaneous, and m odelthem accordingly , bu t the severe s t i f fness problem pers ists , and the fact remains that theassum ption of ins tantanei ty is the s imples t way o f mod ell ing ma ny real processes .A related quest ion is the paradoxical behaviour of symplect ic in tegrat ion techniques[19,20] , which broadly speaking al low one to maintain the constancy of invar iants ofthe system (such as the conservat ion of energy) dur ing the numer ical in tegrat ion .In tu i t ively one would expect th is to improve the accuracy of the in tegrat ion , which isindeed the case for ex tremely long t ime in tervals , but these m ethods are usual ly infer ior


4/8

to c lass ical methods over t ime- in tervals of pract ical in teres t- - -and the reasons for th isare not at all apparent!For partial differential equations, the use of f inite elements rather than f inite-differenceapproximations to der ivat ives can be v iewed as a usefu l "systems" approach, againal lowing the imposi t ion of conservat ion laws over these elements , but again thesymplect ic paradox is a basic cause for concern . The quest ion of the val id i ty of theNa vier -Stok es equ at ion as a val id model for real f lu ids is s ti ll an open quest ion , and o urown fai lure to produce a sat is factory general modell ing program for f lu id dynamics[21 ] causes us to quest ion the val id i ty of the basic assump tions widely used in pack agesfor computat ional f lu id dynamics .Other uses of the s ta te concept ha ve also proved to be f ru i tfu l. T he "s tate-task network"[22] was another ex tension of the f low-sheet concept , which opened the way to asystematic t reatment of a wide var ie ty of batch-process ing systems, and th is" funct ional" v iew of processes can be applied to processes in general to suggest newapproach es b oth to the synthesis of new processes [23] and to the mo dell ing of a g ivenprocess [24] . An example is J im Douglas ' s "h ierarchical des ign" technique [25] , whichis based on decid ing on the inclusion of some unit or feature in a process , or the need toexamine th is issue more closely , on the basis of i ts contr ibut ion to the enhancement ofsom e "ut i l i ty-measure" of the value of the whole process .The applicat ion of the idea to the automatic generat ion of a dynamic mathematicalmodel f rom a purely qual i ta t ive descr ip t ion of the process opens new perspect ives anddeserve s a l i t tle mo re d iscuss ion:There is no such th ing as a per fect model , and al l that we can hope to do is predict theevolu t ion of a l imited set of proper t ies of the system of in teres t with reasonableaccuracy . We also hope that the model wil l not be too sensi t ive to the s implif icat ionsand approximations used . . However the vague adject ives in th is s ta tement need moreprecise definition"The set of proper t ies of in teres t wil l be clear enough to the user , and immediatelydetermines a set of physical laws which might be involved in their determinat ion ( inaddit ion to the conservat ion laws) , . We must assume that these " laws" , e i therfundamental or empir ical , but acceptable as adequate for the purpose, can be provided,ei ther specif ical ly or by select ion f rom a previou sly s tored l ibrary .I t is then a well def ined mathematical problem (programmable for the computer ) toreduce these potent ia l re la tions to a minim al set (or detect that they are incomplete) ,and to ident ify a minimal subset of var iables f rom which al l the proper t ies of in teres tcan be explic i t ly com puted at any poin t of t ime or space using only algebraic equat ions .This minimal set of var iables def ines the ins tantaneous s ta te of the system, and theirevolu t ion is completely determined by the system inputs and the remaining equat ions inthe minimal set . These wil l in general be a mixed set of par t ia l d if ferent ia l , o rd inarydifferential and algebraic equations.Thus w e have ou r mathemat ica l model .Of course the computer has no creat ive powers , and we are s t i l l re ly ing on humanknowledge and ins ight to provide the essent ia l ingredients and make the appropr iateselect ions for the purpose in hand, but th is f ramework suggests a systematic way ofs tor ing these ingredients and provid ing the par ts of the analysis which can beau tomated .


5/8

The idea i s a l so capab le o f ex t ens ion to make use o f human ins igh t t o s impl i fy t hemo de l t hus fo rmed . Con s ide r fo r exam ple a s imple f la sh vesse l fo r sepa ra ti ng l i qu id andvapour :I f we a re wi l l i ng t o a s sume ins t an t aneous separa t i on , t he re wi l l be two zones i n t hevesse l, one occu pied by l i qu id and the o the r by vapour , each desc r ibed by i ts own subse tof t he s t a t e va r i ab l es , and l i nked on ly by r e l a t i ons a t t he common boundary . I f wespec i fy t h i s a s sumpt ion , o r equ iva l en t ly t he f ac t t ha t t he re a r e two d i s t i nc t zones , t hec o m p u t e r c an m a k e t h e c o r r e sp o n d i n g d e c o m p o s i ti o n o f th e m a t h e m a t i c a l m o d e l .Ins t ead we migh t a s sume tha t each f l u id i s d i s t r i bu t ed over t he whole vo lume of t hevesse l , w i th t he r a t i o o f l i qu id t o vapour va ry ing wi th he igh t and some empi r i ca lre la t ion for the shear force per uni t volume between the f luids a t each point . Set t ing thisfo rce t o ze ro y i e lds our o r ig ina l mode l , bu t i f we wi sh t o de t e rmine t he r a t e o fsepara t ion we m igh t i ns t ead t r ea t the m ix ture a s a si ng l e pseudo- f lu id , w i th po in t va luesof i t s s t a t e va r i ab l es de t e rmined by loca l ave rag ing over t he two f lu ids i n a smal lvo lume a round the po in t . Th i s l oca l " sys t em" can aga in be mode l l ed i n t he same way ,imply ing a h i e ra rch i ca l mode l l i ng p rocess . For example , i f we a re wi l l i ng t o cons ide r as ing l e bubble and ignore t he i n f luence o f ne ighbour ing bubbles , our shea r - fo rceparamete r cou ld be ob t a ined us ing S tokes ' Law. Of cour se t h i s s imple mode l cou ld besuccess ive ly r e f ined , up t o t he l imi t o f a f i n i t e e l ement approx imat ion o f t he two- f lu idmix ture . - - -bu t a t t he cos t o f a cor r esponding inc rease i n t he computa t i ona l e f fo r tr equ i r ed .Thus by p rov id ing qua l i t a t ive i n form at ion on poss ib l e s imp l i fi ca t i ons we can con t ro l thecomplex i ty o f t he genera t ed mode l o r even genera t e seve ra l a l t e rna t ives and choosebe tw een them on the bas i s o f t he i r impac t on some u t i l i ty -m easure fo r the wh ole p rocessmode l , a s i n t he Douglas h i e ra rch i ca l des ign approach . An advan tage i s t ha t t he use r i sfu l l y aware o f t he approx imat ions used in compi l i ng t he mode l .C lea r ly t he i dea i s capab le o f fu r the r e l abora t ion , and i ts implem enta t i on i nvo lves a hos to f subs id i a ry i s sues t o be du ly examined , bu t I have sa id enough to i nd i ca t e t heposs ibi l i t i es .For a number o f yea r s p rocess sys t ems eng inee r ing was focused on the p rocess i t s e l f ,bu t p rocess sys t em s eng inee r s have s t ead ily widen ed the scop e o f t he i r in t e res t s, f i rs t t owider a spec t s o f p rocess management , t hen to mul t i - s i t e opera t i ons , and even tua l ly t ocons ide ra t i on o f t he whole supp ly cha in . They have a l so expanded the i r i n t e res t s t o awider r ange o f p rocesses , such as me ta l l u rg i ca l and b iochemica l p rocesses , de lveddeeper i n to t he phys i co-chemica l and b iochemica l founda t ions o f t he knowledgerequ i r ed fo r improv ing the i r mode l s , and a t t empted to embrace s i s t e r t echnolog ies suchas computa t i ona l f l u id dynamics [26] and even molecu la r dynamics [27] .As t he scope o f t he i r i nves t i ga t i ons has expanded , p rocess sys t ems eng inee r s have hadto ex t end the i r i n t e res t s t o a wide r r ange o f r e l evan t t echn iques . Compute r Sc i ence hadmeanwhi l e p roduced t echn iques fo r encapsu l a t i ng , s t o r ing and us ing qua l i t a t i veknowledge , g rand ly t e rmed " exper t sys t ems" , which p rov ided eng inee r s wi th an en t i r e lyd i f f e ren t way of us ing compute r s , eager ly p ressed in to use fo r such purposes a sprov id ing or check ing the i n t egr i t y o f opera t i ng i ns t ruc t ions o r t echn iques fo r sa fe tycheck ing and ana lys i s . " Q ua l i ta t i ve mod e l l ing" sys t em s were dev e loped in an a t t empt t oex t end the scope o f exper t sys t ems and avo id t he heavy inves tment r equ i r ed fo r t het r ad it i ona l quan t i ta t i ve m ode l l i ng sys t em s .


6/8

The so-ca l l ed "ar t i f i c ia l in te l l igence" school bent the i r minds to deve loping a wholer ange o f t e chn i ques i n t ended t o p r ov i de com pu t e r s wi t h t he m eans o f em ul a t i ng hum ant hough t p r oces s e s , u s i ng a r ange o f ana l og i e s to b i ochem i ca l o r o t he r na t u r a l p r oces s e s[ 28 ] . Thes e had i m m ed i a t e i n t u i t i ve appea l , and s om e have been wi de l y adop t ed . Fo rexam pl e t he "gene t i c " a l go r i t hm f o r op t i m i za t i on ha s becom e ve r y popu l a r andem ul a t i on o f neu r a l ne t wor ks [ 29 ] ha s p r oved t o be r em ar kab l y s ucces s fu l i n im p r ov i ngon c l a s s i ca l t e chn i ques f o r pa r am e t e r and s ta t e es t im a t i on . Such s ucces s e s s hou l d no t bei gnor ed , bu t I con f e ss t o a pe r s ona l b i a s f o r u s e o f te chn i ques wh i ch can be ju s t i f ied bya m or e c l a s s i ca l m a t hem a t i ca l ana l ys is .Of cou r s e t he r e con t i nues t o be p r og r e s s i n deve l op i ng s uch t e chn i ques . Wave l e t t heo r y[ 30 ] ha s p r ov i ded a power f u l t oo l f o r t he ana l ys i s o f m u l t i - s ca l e p r oces s e s andm a t hem a t i ca l l og i c and g r aph t heo r y have been app l i ed i n a va r i e t y o f ways : Ea r l y u s eo f f au l t - t r e e s and even t - t r ee s t o t r a ce t he r oo t c aus e o f f au l t y behav i ou r and t hecons equences o f a pa r t i cu l a r f a i l u r e we r e ex t ended t o t he u s e o f d i r ec t ed s i gna l g r aphs( d i g r aphs ) f o r t h i s pu r pos e [ 31 ] , m e t hods we r e deve l oped t o check com pl i ca t edope r a t i ona l p r ocedur e s [ 32] , and the u s e o f l og ic in "d i sj unc t ive p r og r am m i n g"p r ov i des a pow er f u l s upp l em en t t o i n t ege r p r og r am m i n g t echn i ques [ 33] .2 . C o n c l u s i o nI have t r i ed t o r ev i ew deve l opm en t s ove r t he who l e s pan o f m y ca r ee r , bu t wi t h s uch ab r oad canvas I have had t o be s e l ec t i ve , and t he r e f o r e been unab l e t o m en t i on m anyo t he r i n t e r e s t i ng and s em i na l i dea s . I have chos en t hem es wh i ch have s pec i a l l yi n t e r e s t ed m e , and s t i l l g i ve m e caus e f o r conce r n o r s eem t o po i n t t o f u r t he rdeve l opm en t s , r a t he r t han a t t em pt i ng t he i m pos s i b l e t a s k o f a com pr ehens i ve r ev i ew,g i v i ng due c r ed i t t o eve r y i dea i n t he eve r - i nc r ea s i ng ava l anche o f new i deas ove r aneve r - wi den i ng f i e l d .As t he s cope o f p r oces s s ys t em s eng i nee r i ng ha s wi dened , s o it ha s becom e m o r edi f fuse , and i t i s more and more d i f f i cu l t to def ine i t s boundar ies or ident i fy an essent ia lco r e o f expe r t i s e . Many wou l d s ay t ha t i t i s unneces s a r y , o r even unwi s e , t o a t t em pt t odo so , but an a rea which cannot be adequate ly def ined r i sks los ing i t s appea l .Neve r t he l e s s I hope t ha t t h i s neces s a r i l y cu r s o r y r ev i ew has s hown t ha t t he r e a r e s t i l lp lenty of i s sues to address , and tha t there i s an increas ing range of t echniques to do thej ob .References1 . W e s t e r b e r g , A . W . , a n d S a r g e n t , R . W . , H . , " S P E E D U P " ( S i m u l a t i o n P r o g r a m m e f o r

t he Econ om i c Eva l ua t i on and Des i gn o f Un s t eady- S t a t e P r oces s e s in chem i ca leng i nee r i ng de s ign , T r ans ac t i ons o f t he I n s t it u t ion o f Chem i ca l Eng i nee r s , 42 , 1 90( 1964) .

2 . F r i ed l e r , F . , Ta r j an , K . , Huang , Y . W. , and Fan , I . T . , Gr aph- t heo r e t i c app r oach t op r oces s s yn t hes i s : Ax i om s and Theor em s , Chem i ca l Eng i nee r i ng Sc i ence , 47 ,p p 1 9 7 3 - 1 9 8 8 , 1 9 9 2 .

3 . Gr os s m ann , I . E . , MI NLP op t i m i za t i on s t r a t eg i e s and a l go r i t hm s f o r p r oces ssynthes i s , in Si i ro la , J . J . , Grosmann, I . E . , and Stephanopoulos , G. , Foundat ions of


7/8

Com puter-Aided Process D esign (FO CA PD'89), C AC HE , Elsevier, Am sterdam,1990.

4. Linnhoff, B ., Entro py in practical process design, in M ah, R. S. H. and Seider, W .D., (Eds), Foundations of computer-aided process design (FOCAPD'80),Engineering Foundation, New York, 1981.

5. Sargent, R. W. H., Integrated design and optimization of processes, ChemicalEng ineering Progress, 3(9), 71, 1967.

6. Garcia, C. E., and Morari, M., Internal mod el control 1. A unifying review an d somenew results. Industrial and Engineering Chemistry and Process Design,Development,21, pp308-323, 1982.

7. Kalman, R. E., A new approach to linear filtering and prediction problems, J. BasicEng., Trans. A SM E, Ser. D, 82(1),35-45,1960.

8. Kailath, T., Linear Systems, Prentice-Hall, Englewood Cliffs N.J.,1980.9. Pon tryagin, L. S., Boltyanskii, V. G., Gam krelidze, R. V., and Misch enko , E. F., The

Mathematical Theory of Optimal Processes, Wiley,New York, 1962.10. Cutler, C. R., and Ramaker, B., L., Dyn amic matrix co n tro l~ a com puter control

a lgori thm, Paper W P5-B,AIChE National M eeting, Houston,USA,1979.11. Morari, M., and Lee,, J. H., Model predictive control: Past, present and future,

Com puting and Chemical Engineering, 23, pp667-682, 1999.12. San-Bias, F. C., A Case Study in Nonlinear On-line Optimal Control, PhD Thesis,

London, 2003.13. Lim ebeer, D. J. M, and Green, M. Lin ear Robust Control, P rentice-Hall, Englew ood

Cliffs, 1995..14. Grossm ann, I . E., and Sargent R.W.H., Optimum Design of Chemical Plants with

Uncertain Parameters , A1ChE Journal (Vol. 24, No. 6), pages 1021-1028,November 1978.

15. Sakizlis, V., Perkins,J. D., and Pistikopoulos, E. N., Simultaneous processsynthesis and control design under uncertainty based on a mixed dynamicoptimization approach, Proc. 3rd Panhellenic Scientific Con ference on Chem icalEng ineering, Athens, 1:313-316, 2001.

16. M ah, R. S. H., M ichaelson S., and Sargent, R.W .H. Dyn amic Behaviour ofMulticomponent Multistage Systems. Numerical Methods for the Solution,Chem ical Engineering Science 17, 619 (1962).

17. Gear, C. W., Numerical Initial-value Problems in Ordinary Differential Equations,Prentice-Hall, Eng lewoo d Cliffs, 1971.

18. Pantelides, C. C., Gritis, D.M ., Morison, K. R., and Sa rg en t, R. W . H., TheMathematical Modelling of Transient Systems using Differential - AlgebraicEquations, Com put. Chem. Engn g. 12(5), 449-454, (1988).

19. Sanz-Serna, J. M.,and Calvo, M. P., Numerical Hamiltonian Problems, Chapmanand Hall, London, 1994.

20. Hairer, E., Lubich, C., and Wanner, G., Geometric Numerical Integration: Structure-Preserving Algorithms for Ordinary Differential Equations, Springer-Verlag, NewYork,2002.

21. Barber, J. A., Th e Autom atic Synthesis of M athematical M odels for ProcessSystems, PhD Thesis, London, 2000.


8/8

22. Ko ndili, E. Pantelides, C. C., and Sargent, R. W . H., A Gen eral Alg orithm for Sho rt-term Scheduling of Batch Operations. 1. MILP Formulation, Computers Chem.Eng g. 17(2), 211-227 (1993).

23. Sargent, R. W. H., A Functional Approach to Process Synthesis and its Applicationto Distillation Systems, Computers Chem. Engng. 22(1-2) pp 31-45, 1998.

24. Sargent, R, W. H., Mo delling---A process systems perspective, Proceedings, NordicProcess Control W orkshop, January,2003

25. Douglas, J. M., Conceptual Design of Chemical Processes, McGraw Hill, NewYork, 1988.

26. Bezzo, F., Macchietto, S., and Pantelides, C. C., A general framework for theintegration of C omputational Fluid Dynamics and process simulation, 7 hSymposium on Process Systems Engineering, PSE 2000, Keystone, Colorado, USA,July,2000.

27. Stephanovic, J., and Pantelides, C. C., Towards tighter integration of moleculardynamics within process and product design computations, in Malone, J. A., andTrain ham , J. A., (Eds) , Foundations of Com puter-Aided Process De sign(FOCA PD'99) , CA CHE , A IChE Symposium Series N o.323,vol.96,2000.

28. Stephanopoulos, G., Artificial Intelligence and symbolic computing in processeng ineerin g design, in Siirola, J. J., Grossm ann, I. E., and Stephanopou los, G.,(Eds), Foundations of Computer-Aided Process Design (FOCAPD'89), CACHE,Elsevier, Am sterdam, 1990.

29. Hoskins, J. C., and Himmelblau, D. M., Artificial neural network models ofknowledge representation in chemical engineering, Computers and ChemicalEngineering, 12, 881, 1988.

30. Daubechies, I.,Ten Lectures On Wavelets, SIAM, Philadelphia, 1992.31. Becraft, W. R., Guo, D. Z., Lee, P. L., and Newell, R. B., Fault diagosis strategies

for ch em ical p lants: A rev iew of competing technologies, Proceedings of 4 a'International S ymp osium on Process Systems Eng ineering (PSE'91),II .12..1,1991.

32. Sanchez, A., and Macchietto, S., Design of procedural controllers for chemicalprocesses, C om puters and C hemical E ngineering, 19,ppS381-386,1995.

33. Grossmann, I . E., Review of nonlinear mixed-integer and disjunctive programmingtechniques, Optimization and Engineering, 3, pp227-252, 2002.

Process Systems Engineering a Retrospective View With Question for the Future SARGENT 2004

Documents