SAIS local aM enon isiora! Ou SULESRnST SNe eT Process_ flowsheeting A.W. WESTERBERG | H.P. HUTCHISON R. L. MOTARD P. WINTER CAMBRIDGE UNIVERSITY PRESSPoe te Saf he Cone Ue Pr i pac cma CE cro epi 1079 SE Cie tyr wih epi et pane 9 trary of Conger Caaouing i Pblention Dats TP Ghent poses conta 2. Chem! pas — 11 00/960129 118988), Contents Preface LI Steady-tte owshceting and the desien process 1.2 "The total design projece 2 lowstetng on the computer 22 Developing simulation mode 2.3 Approaches to flowsheting yeems— examples 3 Solving near and noninar algebraic equations 5. Solving one equation a one unknown 5.2 Solution methods for near equations 3.3 Geacral approaches o solving ses of soinar equations 1. Solving sets of pare nonlinear equations 4 Physical property servic nites SA Te data cyte 42 Computerized physi propery systems 43. Physical propery calculations 15 Degrees ffraedom in + Rowsbest 5. Degree of feedom 5.2 Independent stream variables 5.3 Degrees of teedom for 4 unit 54 Dopresoffeedom ina fowsbeet The sequen mode approach fflowsbeting 10s us m 130(62 Other featares 163 Convergence of tar steams {64 Partoning and tating a fwstect 17 Mowshetng hy eqtonsolviag methods based on tering 7H A simple example 72 Am example stem based on equation Ning 173. complex example of electing decison and teat vibes fra flowbect 14 Handling the erated wriables 75 Discussion 4% Simulation by Hear methods 1 Intedacion to linea maton £82. Application to staged operations 443. Applicatin 19 a management problem [84 The SYMBOL system Sor materia balancing in near tems 185 A simple example af the we of SYMBOL 6 A complex example of the use of SYMBOL, 9 Simulation by qusétnear methods 4 Intodtion to quasinese methods 92 Simulition offs in ipe networks 9.3 Application wo dition 94 plication to multiple reaction equim 195 ‘Towans proses simulation by quasilinear methods 10 Porter seding and erature references Index 138 13 a8 16 162 165 180 189 m1 94 198 ws 200 207 a3 28 as 20 22s 2 2s 240 24s Preface “The Computee Aided Design Centre in Cambridge conceived the idea for ths book in 1973 andthe groundwork was completed in 1974-5 when Profesor Are Westerberg and Profesor Rady Motatd spent "eal appointments the Centre, Origaly, the alm wt 12 poblih = short monograph rummariaing the stats of fowshesting {ecology but soon esame evident thatthe subject demanded ‘more considered treatment. Contequenty, moch ation effort brs ‘een devoted tothe weg and Art Weterber in particle apie bi enormous enorgy to vvsing the text ProctrMowahecting there ‘we ope it ss stimulating i reading 2 has bees in wr “The materia ia ths book describe flowshetig in chemi engineer ing. As ued here the cere owsheetng shew of compute als pesorm sedy.stte hen and material alncing, nd rings costing Eatelatins for a chemical proces. Moat previous presentations of Sloat material Nave eoneentated ons singe approach fo owehecting hen infact many iit ad each tae st dane advantage ad Aisadvanage. This hook presents four sppreache, and for esch oe the motivation fri development is analaed and i we estate through a numberof peta example. "Ths book cn be wre introduce the Wes Behind flowsheeting the novie; i ako contsine miter of interest to anyone wh i previously wed a Rowshectng sytem, Lay some materi aimed Sree at those persons who may be thinking about designing 9 new system of have & burning dese to understand How such ystems operate "The book is oreaize into four base pas. The fist wo chapters are Inroductory, chapter | defining the role of lowshecting in proces esgn and ‘chapter 2 presenting an inroskiction 10 the acinus approaches that are wed in lowshecting atoms se ip and perform the design calculations. ‘Chplers 3 to 5 pve useful background information common tall spproace. Chapter 3 is extensive and als with solving the large ‘umber ofimsttaneoos near and nonlinear pare slgeale equations ing inthis tecoloy. Chapor 4 draws atetion fo the calelatonvii Prefice ‘of physical propeny data in process design. The treatment here is Inzoducton since thsmporant and comples ace deserve ae nin it itsown ght A companion hook which wil cover ti top indepth ‘eelealy from the viewpot of computer-aided design planned, {CanperS inoiuces the analysis of depres of feedom ina flowshee', [problem that requires understanding to appreeiae fly the concepts Droblems of designing flowshesting system “The tox four hsplrs,chapers 109, deal in some detail with ach of four approsches used by dilrertfowshecting stems. Chapter 6 ‘eserber the moet common approach the quent modular approw’ Which underes such systems at PACER (Digit Sytem, 1971), (CHESS (Motard and Lee, 1971), CONCEPT (CADC, 19733), snd FLOWTRAN (Seater, Sidr and Pais 1974), wll a many, many ‘thers (ce Flower and Whitshoad, 1973, 6). Chater ? desis Tonsheeing by equatin-oving methods based on tearing. Few sy tems exit which ue thik approach; SPEED-UP (Leigh, Jackson end Sargent, 1974s perbape the mst widely known. Chapter deserts smlation by linear mtd, whereby the enti Bowaboct i daserbed ‘only approxinately wins neat equations. This spproach pct rai sessment of process altematives where detuled analsb ot required. The SYMBOL stem (CADC, 19738) istrtes ois jproach. Chapter 9 desces sling Rowabecting problems ving ‘tasinear methods such ay Newion-Rapisor based methods, and feleence i made othe MULTICOL sytem (CADG, 1977) The ast section ofthe ook, caer 1, ia guide 1 furhe eading material as well ea Int of rferencer used throughout the tok, “A sate! before, this book can e used at several Ives Itc be used to introduce Nowsheetng to senor students 8 design course, ten chapters, 2,4 and 6 wl be of value. With this akground fhe ident wil appreciate the ole of owahosting i proses devin and ‘will eodertand why alferaive approaches ex. ation they wil hve stud the sequent modular approach in some deta The more vanced mater rom ston 6.3, nward may best Be ote ara ‘introductory cour of thi Spe. For those persons with a more advanced background in fowabssting, chapter maybe omitod or st most skimmed through guy. Chapter 2u however, represctsessendal reading sit aotates the organization ofthe bole Depening on the persona interests ofthe reader the rst ofthe hook may be approached citer by reading chapers 310 Sof by proceeding Aety nyo chapers 6,7 oF 8, ech of whic hs ben ween a Seltcontined uni I shou be pose to understand al of chaps 6 Without fst reading chapters 30 5. However, the ender sould omit ‘ection 7.3 ifehupter Sand perhaps chapter Shave nt been tad he Prefce x ‘remainder should ote litle difiaty. Chapter & can sso be read and ‘most of tbe smporane points ‘ondesood witout the tater in hapers 3 5 ‘Chuper 9 is bes read ony after rening the material cntsned in ‘chapter 3 onthe sling of tt of liner and nolicar equations “The nate ofthe ook eo ltroguc the concepts involve lrgly by example. Many ofthe Ideas ate therefore not developed hee 0 the same extent as In Some existing erature; however, the book i intended to cover more than is curently ready aval. hope tht the motivations for the varour approsches to flowahecting in fevistence and for thse being developed willbe made lea othe reser Se he can more fly speecate the strength and weaknesses of he plethora of ternative he wil oe eta fae inthe ner ftare October 1977 P. Winter1 Introduction 11 Steady-state towshecting and the design process Flowshoeting may be defined atthe ase of computer ids to perform steady-state heat and mas balancing sizing and costing caleltion for § hema process We shal now det oar defnkion ofthis term by Investigating for which tps red within the derign proce. gure [Lh ilostrates the patting of he design tp. This gram sin ae tefl for desipn in any Fld Hf the word “owsbect” is rplaced by “Sroctue and ite wefl for sore than vewingsteady-sat desi. However, here we sll ste ourviwpon to stead state chemi process design ater inthis chapter we sal dus amore compre Five view (eg. to consider dynamic havi) bat sal not pts beyond that diseusion igure I- portions th process design effort into thos base tps: syottess, snags and optnization, yates isthe sep where the Aowshect stucture i chasen, Lc tho particular equipment tobe usd fd ts inferconnection are selected. Ao n the syntess sep one ovides il vales forthe variable which one it fee to set The ‘Second sep is ‘analysand itis often broken ino three parts: solving the heat and mtr balances, sng and costing the equipment, and valatng the north (and perhaps sey, operably and 30 forth) “The final steps te teed optinization snd, ie pore, paramote andstrucural optimization, In th cour of analysing pve owes, ‘one aualy discovers some particular peste o temperate level ei Profound infence the resulting equipment sites ad thus the flow Sheet evaltaton. Or, one may decte 0 slter the egepment andor is lnereonnection teeause an improvement is obvious or the eutrent ‘version appears tobe very eos Changig te equipment ype anor fis intrcomecton i stuclural optimization whie simp altering temperature oe peur levee within fixed flowshest ® parameter foptinizatio. The thory to do the later automatealy& in a more ‘vanced tate than the former ‘The final flowabect with ite decison variable values and resting Sos et epreents the fins Ssign.2 Induction =F | = an For this book then, the materia to be covered ie that describing the ‘computer aids needed only Tor the analy step: hea snd material ng szing and costing, and preliminary evalston, The elect of the approaches possible for analysis onthe syathesis adoption ‘ops wll be ideated where appropriate, bu these last to step ae "ot within the intended scope ofthe book. ‘At this point might bo weft fo describe the we of fowshectng programs Within the contest of doing process design, The scenario foowing sony one of many was oe nach ystems, 11 Steny-sae fowshecng and the design process 3 Inthe nti stages of pros development, the engineering tam may ss to pelo some teal Blane calculation on a erode How ‘et "Theflowshet wl ely conan reactors which do mot ope st 100% conversions the flowrbet wit conan vey, purge 0 forth, and in ths asi the apresinate lows corresponding to Aiferent comcrion ves may be desied. Abo one tay Wh 10 Snventgate te regeled purge aes ose what ls f prod may be ‘ling there ae than Using move sphsteted separation “Wiha proper problem formulation, one can often deste the proce sont forth ype of ais ase nea gusto Soving them guity a eps for sight hangs to he sete fndcomseron lee aloe the egiocing tem fo pt el fo the owshect, even if he fel more qualitative than quanéttiv. The ‘ting py ad solving ofthe inca equations by had eos ad tor prone, ant often inpossible becuse the problem ito fe ‘iso fing comin nt oo nyo to ew requirement an he Aowabee ica ere tht one an vse the wets ta ompeer programming system To aid in inputing the ows, the ‘ended atthe rege ows sd 0 forth a then 0 sole ging “Signonieni aur srs. The computer foe dong something ere Ut the eagncer coud ot don theory, Dut certainly reing tbe {sium sited with he ob ‘Athi evel of development be engnsring eam may acover he ce to So moe sta iaborwory evelopment wo improve the ow ‘Set. This aid med to perm gros efecto be caught eat and ‘counted for if possble Management muy tow athe a more deal Howse sting te performed. The design lest wil a tore det tthe Howes, converting each simple scparator Block in subsystem of actual Separators which will opel operate as necded, Now more aciate Dina property ata wl be tceaed sx wel te banding othe eat baances’ Simpl heaters and coolers andsinpte peste changers il ‘unipuntetompeatre and peter eed gig the engines nes ofthe Heat abd power requirments within the proses, "he fa mre sopbitated owsbecting tm egued one wtih hth sy to eluate phys property of mares and Mich conans te progans to soe tbe egeatons sling fom the ‘pret of equipment tthe proposed weet. TR considerate eft wl be needed to sc vp he problem, pat tay itsome ofthe ekemalcomponest physical property data te Pot Svlabie wii the system an fsome ofthe equipment types are ne4 Iavoducion “Te ater wil equi programs tobe write, tested nd then ined in he sytem bythe engaering texm, Ifthe problemi age enough andthe bet physical property estimation salable deste en hs fot will very probaly be les than that for hand caleltions, If fowshet varnins are foe ted, the ell wil amos cetily be ‘The ft owsheoting clelations tobe tied might be those wag the sox sophisticated design programs avalable tthe stem, tether with best procedures fr estnsting phyial properties Also the Aovabeet may’ e heat and pressure iterated, ging ie to proses Interactions which are reflected in computations iterations The oat for such a eacuation may well be tortor onder of magnitude greater han the inal near materia balance elations. By this im, however, one i aly feed onthe lowaeet aad ering that ‘operates at needed, tothe stent tat computer programy an ae ‘ally model the proces behaviour. “The sdranapes for performing Nowsheetng culations onthe coat puted ot always include asavingin manpower anddeigncots Rather ‘he eal edvenages may be tat the reals ave mich moe coset because the physical property estimation proceuree al data are ‘onistet hrouphout the Nowsheet.Parsmetervariaoneare mors key topive meininpu ends because of hiscoasstncy when dong mle ‘ns Abo the chances for simple errr in arthmete ne scaly ‘ediced, ging one a etter opportunity to climinate such rors, Once 8 program has been made as eror fre as posal, wil Bing, ut forte perform the elelatons it represents fom hea on, ‘Comparisons of alerative designs by aflerent design teams are roe likly obese hecause, ain the ame design equations aod ‘Physical property data can be wed for both, “The disadvantages are not tobe overlooked The seem at and may be ile for the problem under consideration and may never seach tm answer for if oF Go 0 only afer very considerable effort A sid ‘engineer may, however, be able to we adequate short design methods by hand te get an anower by knowing inuively Bow to the problem. Or the design team may be temple toe too ato ith this excting new took avaiable and ay never get atleast one ‘erin of the probiem solved by the scbedled tie. Flowehecting ‘ystems are minmade dar auch contain bugs (ero in the Prostam) which the ser may be unaware of, or alltooawate of mt {hese hugs’ may lad o incorrect aswers cr no newer. Having bee “wumt” one, the user may wel be unwligfo Uy ain. As wil large-scale programming toms the syste ikl tbe wel ony i maintained and i the er a torn to expert Bel if problems 12 The oul design projet 5 Following, of aloag with the steady.state owsheting effort, the “dynamic sharaterincs have to be investigated 'o establish controler location and (pes and to establish operability ducing star up and in the foe of expected pets. Conrllers are usually designed OY an lnaiys of smal uctustons i the proces olen wing linear dyeamie "talysls Operabilty studies are well concerned with Inge proses Auctations and these sts ae needed o lente ond ste sare process boldups and the lite whieh cannot be esublshed by Sendystate analysis 1.2 The oa design project Flowsbeting, as we hae just described i is een to occur dering the cay tages ofthe tol desig projet Figure 1-2 san usation sving the. ope which eccrine projec, starting withthe conception fot the project and ending wit the eonsracton of the ceslng plant ‘Thvoughout this ene process many instances oceut when sppeopnas ison the computer do or wllin the future, ext. Looking st figure 1.2, tho project starts whon the process isin conceptual frm, Lethe need and an approach to meeting are stated [At this point the procs desig step nave and may be vewed 25 We have already dscsed in the previous tion, The end rest of his forte the pros foe digrom (PED). A PED contains the process ‘ets and major conta lent in fancionl form (ited fnction- sly a pumgs, columns, acs, cle) and the coectiog steams. AS te have sce thir stage wil have’ the proce beat, materi and Dresure balances solved snd will have all units roughly sped aad ough costed "Ahi tape the fort moves nt the step termed projet engine ing, wre the futon! description forthe procea om tho PD is ied ino Tt of acta equipment to be purchased or constructed, Bndliter enables st of taeprintsorogivlet to be prepared for he ‘cual consructon. During project eapicering the PED is fist com ‘rte inte et of piping snd instrumentation digrame (PIDs), often to termed engineering ine digrams (ELD) ‘The PIDs ate a praphicl summary ofthe actual hardware elements in 1 chemical proces plant and ther interrltonships, Le. bow they are ‘onested form a operable safe and reliable plan, The PIDs, the hardware reality ofthe PPD, incl vessels (otomns an ans) pes valtes, pumps, heal exchangers, reactors, feraces, compressors, Spandets eit and drain valves, tap, ters, conveyors, hopres, purchaood ebeystent, semorn, ialaion requirments, contre (Gow, presure, temperature, lee), sptes and other mapulatured ‘ems, llin logical configuration. The PIDs alo include esses,rn ee —— eee] eater") S| ‘gw 1.2 The lt pt. ‘volumes, pipe diameters, maria of contraction, temperatures, pres ‘res and posi welghiso ements ipinglengts and ends. 6 tot included “nt ‘Wile the impression may be conveyed that the PID ae developed 12 The ttl design project 7 in sequence with other projet engineering activities, n fat, much of the project work proceeds in pale For istance, wih reerene 12 figure 1.2, to obrain pipe diameters an estimate of length is needed 50 thot preaure drop maybe clcelted. Tht imple that primary ‘lant avout hax Ben done whic nfm depends on the routing of one Se mote eical pipes hgh temperature, vapour o opi trailer lines). "These itu have been oblained through «yeannary pipe esgn analysis He also happens fequely hat the PIDs cannot be, oF fre pot, completed uni project engincerag is well advance. ‘On figore 1-2 an addition! iagram ype siniae tothe PID i indicted; Ht ella the wity le dgram (ULD), nd inlades Surdware detaie of the stam and mater piping and contol ten “Thea diagram are even more sense o pact yout than a PID. The ‘lstbaion of ly such se team by 4 common datbution pip, vith each unt requiring the wily drawing its sappy from tht pie When uit hac taken requirement, the tty Sstibtion pipe en then be reed in sie a ft ontnice to the next unit requing the ity Cleary the order of servicing anit alot by Inyo, and ily tin ies ae affected bythe order. Wii serving 26 graphical compendium of lant elements, the PID: relly are the mos important documents wed ving project engine: ing. (loeientally, a substantial chemical plant may actually reuie 40-60 related PIDs) Tey serve asthe basi documents for eommunia. How among » variety of specialise whowe jo i fo analyse fontonal ‘dequacy in elation to contol and operat, corrosion and manten- Soce Each itm serine establish that it wl perform ax deed i the intend service. This document iv ofen ued ana Bai oe onrac negotiations between clint snd engineering Gm or division, ‘The PIDs are abo the starting bass for vessel design an dealing Finally they are used to determi a materia take and purchase cede Ist forall tandard elements, seh as aes ad pnp ‘At an erly stgein their development PIDs and ULDs te ued 10 aable work om the fll stages of plant design to proceed. the feted layout and piping design. Durns ths inate design phase, Pipe routes, postions of alves and pipe fitings, and so forth are fctemined exaciy. In recent years it has been common practice 10 build model of wire and plastic parts to abst inthis exerse and lacy the plant layout sod its plications. The aplication of compat ‘rst this aspect ofthe design proces tl in it infancy Dt number ‘of systems, notably the PDMS Pipework Design Management Sytem from CADC (1977) and IPAC trom IAL (1975) have append recent From the design work undertaken in this has, lformetion cn be sed on fo enable the production of pipework isometries for we inthefabrication shops and on the construction sit, and $9 enable sock control. These aves are commonly computedzed, ISOPEDAC (Cr, 1967) and COMPAID (Davy Computing, 1969) being the mos well nows examples “Another design acti of sina magia to piping, nd conducted. in parallel with I the detaed oetunial desig of the equipment items sucha clumos, reactors, presure vessels and eat exchanger ‘This e normaly underakon by specialst nanuactrers The end result of ll thee imtrelsted activites i the construction, commisioning and sale operation of a chem plant. Pom thi ‘mitedly bet discussion one ean te thatthe lowshoctng phase i ‘nly a small part of the total tort. Nonetheless, it the Key stating point and how well ts done has signin repereassonsthroaghowt the remainder ofthe design und construction pases, and throughoet the operating lifetime ofthe pnt. Tn ther won the rn af fhe flowsheting phse ar rein determining the economy ety and rebabiity ofthe venture a 2 Flowsheeting on the computer ‘he fst teen yeals of history Ia the development of Nowsdecing pronrams can be broken roughly ato thee five yese periods. The Sat fevelopmeats ocurred inthe 1955-9 period before computers were af ay size and speed, and before langues suchas FORTRAN became fencralyavalale. The programs which had any degree of ise Ha Tite scope mally to penton the design caletltons for angle wit ype for example, heat exchanges. To cea ager lem wat really too dfeul to expec uch susces Tn the late 1980s Kelloggs (Kesler end Kes, 1958) tepan to discuss ther Fesble Fowsheet Program. Dusiag the next five years 960-4), several simulators appeared. The sucessful systems where “succes means thy survived the peiod anaes ine — were tone sick were developed ins Hig levelIngaape sich ab FORTRAN, were bull in a modular ation, provided rigorous ysl property Cortelation, and were mado roast so thatthe enpcer cou almost sivas get his simulation to work. “The simulators devcloped wete now put to the test of being wed ena ‘arity of instil problems, Many fated, discouraging te wer and ‘management which funded them Thos the next fiveytr peril (CoGs"5) was a proving snd a weeding out time. Note tha nrgesele computer power al become generally avaiable by this ine. 'B) 1970 mast (but coral ot ll) proces designers were accepting ‘hat Dowsbecting programs Were uf and cost elective. Thay were becoming fst enough and elable enough to be pret! for ven ‘moderately large process, Companies buying hemi plants fom the ‘ontrating companies began to request tha the processes be smalte, tnd, even the contractor dl ot peter toate hi oe a oe to say his estomee "The real sueceses, where considerable enginserng time it sted, ‘longo thos, often special-purpose, programs for hema ans hich are repeatedly contracted and bull, such a8 ammonia Plaats Sometimes wehin a day, the calculations cap be se up and rn, replacing several man weeks of hand eaeustions ‘We wish in this book to deus pensrak purpose fowshetprozrams.10. Flowing onthe computer 1 the last chapter we dncssed the steps taken in the design fort and some of te elevant computer developments, We shall now tempt Seto expan why owabeeting programs ate as they ary casing in fences the vats aprossies wh ah be ae tread te tase systems 21 Motivation for development igure 2.1 gies us basis for discassing the developmneat of general- purpose fowsbeeting programs. The ins! programs developed for Si socom ann | Siemliz™* | te mtn L ! \ Pe 21, Deon itn eae tm, Sowshccing end to deal with specifi examen. The programs were tts for single problem and ven they were succes they sl made I evidont that they were gonealy too esicive. The Rowebeet ‘ould not be ready mode, at al A lack feo or pga property calculations existed within the progam; for example, etal Pies eaculated in ope unit might nat te hoe found in ent. The america techniques tended t0 appear over and over again, fF example, fo find the root of sigle Teton one unknown, The ‘lgnosties tended not to be org ‘Adraniages existed oo. The programs were quit fast, They could be 211 Motivation for development Ut played with until some speia- purpose device made them work, even bo theoreti jstieaton could be provid. But, sine pra requirement of dein st iy alternative the med lost nonste {ent ability to make flowshest changes wn far oo restritie 1B the euly 1060s companies wore balding irae of nit oper on programs capabe of doing many ofthe desig easton. It as ‘vient that thse could then be pat into aster which sould det {he calculations as necded for an ete flowabeet, considerable eiecrng tne might be saved The experenses of writing these unt ‘pertion programs and complete, bu special-purpose Nlowsheet pros re gve the wecesry tacked far the next ep, “To design a programing sytem which woul low performance of the ealealation for an ent Howabest ater tha sage ut meant fone ad to make dessions abot a major sottwate system Where a0 one ‘se had: could one make these stems coaverge forthe genes ese how moch ofa numerical analyst would the user have to be; what fom ‘tinputand outputs needed to make the ystems il poop ee ie Examining the special-purpose programs made it clear tha the ‘robem olde divided rough into thes parts oe pure 2.1). Pat {tthe problem is ver specie the particular design problem at had “This incdes atte very feast the Nowshet, te components Is, and the design requirements that the plant mast mec. Pat of the problem, and hopefully the major pet, common to all problems. The unit ‘operation, type and thew calculations should to lange extent be ‘mmon, he methods to calculate physical properties can hopefully De ‘made common, costing information mayan be common. Undertying these, brary of numeric techniques could prove very ust to op them from being rewrten over and over aga, and abo to cette the approach 10 later improvements might be cvs to moorporate, A omimon approach to diagnostics might aa be Beef “These obvervations, viewing figure 21, provide the bas for the design of t lowsbeutng stem which segues apart to handle the Problerspeifi features, «means o red io the computer a desrp- tion ofthese fears, and st of souls that enables the pera. purpose parts to be aplled inthe content ofthe poble. ‘Alan potion ofthe stom consists of irae that contain routines for the common ealulation. These itludes ange of unit operation, routines forte generation of physical properties of varius ehembea ‘compounds and thelr mixtures cating oatins spd ste of mathemati fal sevice and dignoste outs. In addi wo he routine aris tere ate data braver (or deabanks) thst provide Base physkal propery and costing data for we bythe appropiate routines. All the hove are problem inependent and can be applied a eguted.12 Rowshecting on te computer 2.2 Developing simulation model Figue 2.2 highlights the steps taken to develop #simaation model ‘when using lowsheeting system, Ava fist step heuer has o prepare nd eater the information partir to hi owahoet, He bas it he “tits and their iterconsection He vst supply the list of chemical components in his plant The design speciteatons giving certain temp fratues, mesures, flowrate, compositions andthe like nue Bo mere "The second sep tke the wer information ao rete a problem Aefntion. The requested physical property methods and support dat forthe components must te asembed The cleans deed fr the ricalar uns within the pant mart be collected. Thee include Beat ec ae I I = Pia ey ~ “ +t aie Pe 22: Denlomes of nutans 2.3 Approaches to fowahewing sons examples 13 nd mattial balance equations as well as equipmentszing 1ype of ‘culations. Gos calculation med sod thet sepport data mi 0 te brought in This set of toms can gt very large very quel Ako the ser in is eotbusas, may ty to overpeefy the problem, which is "sualyegivafent to giving an inonsten sto specications, pretud. Ing any solution wi aly sates them, ‘Once the problem has teen defined, one has to develop sliion ‘procedure to save i This, rom moe elementary viewpoint, ean fe has to develop a procedure to solve the andres to thousand of ‘uations defining the problem. The compleaty of this sep rare Siifianty withthe approach one his ake i the problem efniton Sep Ione is very restive in problem defi, thes this problem fae fay simple. In other instances can become ety complet Indeed, Later in dhs chapter me sal we the range of comply at me Ascss some of the allematives which oe can take 10 feat 4 Aowshcting system. “The etal calculation tp of couse elles the approch taken to sevelop the solution procedure. This tp, agin becaune one song large sts of noninear algerie equation, face any geome of = numerical ature. In parla, the system ha to gt the poles to ‘converge oan answer ad, in dong, tas to dei, fat, whe he [rablom has converged. It has tov overeome, or detest and vara Os ‘sen the problem appears co be poody specified. Chaper 8 wil sve Some insight int tone pole, The inal tepsin a conventional owshocting problem are to perform equipment sng nd contig, and proces ov 2.3 Approaches to fowsbeting ystems — exatples We shal introduce ia this section four diferent approaches to the evelopment of fowshoeting stems In ater chapters (ehapters 109) thre ofthese approeces ps one addionl one wil be svestpted in nore deta Am example problem of someulat abstract aa wl ‘our dicusion, sod each approach wil be dissed wth spect to Figure 2.2. Many extension fo cach ofthe approaches ae posible, ad these extension are often included to overcome the partie dfisen- ‘Ges Chapters t09 wll diese some af these extensonk in more det the dacanson here wl be lito othe set interpetaton that ope night give to each approsch, “The lst part of thr chapter wil compare the approssbes, examining qualia the relative cme with which one can develop and wie the owshecting stm, and with wich the engineer can the stm, Some dncason on prorat zt and computation tines wil expone further aitferences,14 Powsheeing om the computer 2.3.1 A imple Rowse igure 2:3 tlustrats a very simple fowshoet that we may wish to male. comprises fur futons unis (wo miner resto and 18 Diner Lt aaa HE lene separator) and contains a resyle scum, The equations we can writ for the Mloweheet st flows (1) Mode equations Klewaeya) <0 Beswrma) 20 ea) Heeserirm a) <0 Hewtarg) =O ing the folowing notation: ay Input) oun By output) tom uit unit parameters fr unit (2) Connection equations sano em 0 e) n=O sacra =O oe I | 2.3 Approaches lomshecng syems examples 1S “The mode! equations are writen fenetonally and, ft, represen the beat and outta balance, tho plyeal propery and otter Goma ‘Sone, and forth tht allow ne to relate the Ui inputs and oat. or each unit numberof equipment parameters wlio exist which ‘nus be speced to complete the wnt desertion. For example, the ‘eactor ylume and is operating temperature ad pressure ae com ‘monly among these parameters. ‘The miter ax cosiered ere hat no ther parameters, and hat is why none ae cosiered in factions specication for our simple problem can te given 1 te computer as «listo interconnecting stems. For each unit, then, {he input and ouput streams ean be named. The following tists would reproreat the Rae for hi problem “This fom of dat is ealy encoded and input to the compute. The sodel desist sed bere oll the computer which equations the tziccr les To we to model each unit, the Nowsbecting system ould have «numberof ferent separator types or reacor tpn, etch witha fret designator and equation wt ins Hbeary of uni types 2.3.2 The sequential modlar approach ‘Seppe ne make the desig decision fr our fowshetng program that ‘uit mode wil he waten so that oe clslts the output stream Talucsof und, given flop steam vals and the unt parameter ‘als Hor «mies, thisrogarement easily see: by mater balance the low out ofeach chemical component readily obiaine, and by fet Balance we san fix the enthalpy ofthe outet steam ~sine Inet16 Plowsheting on she compuer tapes must be given, For the reactor, we can us the ints. he fnetons sa wll wales forte it parameters eu he ‘utposysnilaly fr de eparator. Th sean anooaly hat we rewrite the unit model equations ae fllovs Aieooesy) =O yy muted Began yanas) =O yah = salea an) Hiss IIa as) =O Ye = Rae) Ss ante Mise esa) = 0 a= Released Each set of rearranged equations, y= gui), 48 be represented by n computer bron, Forts cat inp variables Ye anda a i ult vases wl bey Teer telaion would be se Gives a inp ure By material balance: Caculat the exiting flowrate ofeach chemical component Cafclte the oxi! ently of the ei sreaan By heat balance: ‘These ealing Hw and cotay ae the values represented by the steam valet “The second decion we make for dein » sequal modular pe i atl apt reams a) forthe fats mt be Spefied bythe wer The tid eee that al yes lb CSleuste by gung vas for thm ad tang on ere pes For our singe Nowshect in figure 2 we ht Sle the Peat and ‘uteri balance equations a flows 1. Gen apt steam alas tr 2 Gucn eyes nea vals, 3. Use motel egetione yy = eis) © cela yp 44 quate he seam vale fora 3, (sing te pope consestion equations) A se mode ents ya.) 1 eae ya Busts ie steam vide ee ey Use motel eutions yaa) nd Jy = Ry) 0 cleat the outpt steam ves foray andj 8. Etamine te diferente in etm welds for Jy ond ap whi Stout be (went equ we hve wed ur poles. Then (©) Wotescatiy egal, se the vale fr te values for Repeat tom ep 3 (©) Hessen equa the analy forthe recycle has converge. Contin 1 sp | | 2.3 Appromhes © flowshecing sytem —exampler 17 9. quate steam values for sandy. 10. Une model equations yu ~ garage) Ho alae yy 1 Bee We have not inated in tp 80) how to ess new ales given heyy and oy ales. Tas ope willbe covered chaps 3 and 6 ‘The inter wil ene +more detaed dcusion ofthe sequential modat pproseh ve ca ow examine Cs approsch andthe we a 0 develop 8 solution procedure ss we Jcuued wen consdcring figure 2.2, We Shall by dca each ep of the Mure "The wersupplied information tas sted The fowsheet needs © te specified and is equvlet to tinting cach Wit nd I sa routine togetier wth bow ll te nis re inieroomected. THe ‘Speciation ae the values for al steams owing the posse. fbr each un, the equipment parameters , The parca ehenicl Secs preset wit eet Be ted ‘Given this uso spied ip, he computers ble reserve storage spies forall the seams and for vars calelted y the unis ited Tan collect from a thermodynamics databank the needed physic [ropentyenimaton parteer forthe cormponents ted, td tly, {Gens com routine nod, can cole the appropriate cnt data ‘Contesting solution procedure may not be required ite Now sheeting sem defer this kt the er, He may be Tee oft {he order in which clelte the unit The wer in thseae may have to ony all the rey stream el exp indeate when a how {hey at tobe Rerted and converged. Fo his cave the wee must then {np the computational seucnee forthe ul, whch fr our example fowanest in gure 23, might be Unit 1, ue 2, ant 3, erat rom unt 1, unit 4 ‘This sequence implies hat team S2, the ys inks the reyee or ‘om stream, ithe team wick ast be pecmed tnd whone Yalace are ered The sequence Unit 2 wit 3, ant 1 Rest fromm unt 2, wit ‘equally posite, and implies steam 83, the yr lity she “ora stream, Note we do nt calculate unit unt the eye ivan as Ito ate solved comet. ‘As an allornatve, the Towsecting system can contain spots ctesnty to consist a computational sequence and ently the omesponing tor seme. We dco tic ops frter i chapter 6, [As the fina step for contracting the solution procedure, the Yao mathemati, unt operations and physisl properly subroutines are18 Rowsheeting on te computer athcred togetiert@ make up he computer program which wil be [ecuted duit the caculaon sep ‘After identfyng the fom Mca nisl guess fr each nat be sept and ne epithe etter lo, iterates exist within the eomputaion for unin guess fr these oo, ny Have to be spp. Te vale to be gure are ‘oy nthe ut peas mode enter sie. elation ar simply the exceton a the progr pat together for the problem resem pate TH eaupmen sng an costing doe saint pat ofthe calexatons, the routines to do them ae pt of the progam lend onstrated one separately, sinar quence of eons as hat (0 ‘et up the simoatio needed ost up the elcultons 23.3 The simultaneous medalar approach “This dicusion sto present te essential deas behind ths approach Alowseeting. Rosen (1962) int disused lowsbecing tem of ‘ype. Alo stom of thie type has Been developed commercial 5apan (Umeda and Nishio, 1972), The smulaeeows modal approach requires a unit modules to be writen ae they are forthe sequental ‘nodular approach, Each unit od! will be writen so one ecules he ‘output stream vaines of «uit given st input sean tales und ‘atipment parameter values. AC thls point fendamenta difference ‘ccus A further mule for each ui ype must ss he writen which allows one to relate each output valve approsimatey to» linet Combination of al isptvales. The ekulition to be estore te Indes ty the folowing tworstep computation which io tne pce foreach uni 1. (@) Given the ani inputs. ‘Given the equipment puraciers«, ‘Given any inl guste for rted variables (0) Use th wit model to elle the wnt outs, 2 (a) Given the ut imps age ‘Given the init outs 9d (6) Find nee elation nich adequately mode the unis for ‘small perturtatons ofthe inputs, £2 the eoelcints ay sch et where she aumbor af np streams wit. ‘The complenity of finding the eq can Vary wih the acsuracy a 2.3 Approaches to omshecing sytem examples 19 ecesay and dered for hese linear relonships if they ae to behave ‘Sea tne nel. “The separator model can serve to strate possible simple ine ‘model for it Suppose we find after top 1() above for cola ‘pling Denzene, toisne and yee tha the componente cola split as follow — “upton Top ouputor Bosom suet ow “Tie ceenents the output team values forthe exat model sing these given input seam vues We might then model the output Bows inthe simple near form enzee tp flow ost = 098 benzene fed fw in “Toluene top flow ost = 093 toluene feed How ia Xslene top flow out = 0.09 xylene feed flow is ‘eat op flow ovt = 0.5 heat eed fo in [Benzene bottom fw ogt = 0.02 tenrene fed Bow in “Tle botom flow ost = 0.07 toluene feed ow ia “vlene bottom Now out = 0.91 syne feed ow in ‘ent bstom flow ost = 0.5 eat fod ow in For the benzene top flow ost the fall near equation for thi simple models enzene top ow out = 0.98. beazene fed flow +.00 toluene feed How }0.00ss1ne feed ow ia $0 eat fu low in tnd th tlevant ay ooecients are 0.98, 0,0 and 0. We note two things, The 098 came fom obering the res Of sn accurate model ‘alcsation which converted feed steam values info product steam ‘alas. The form of the linear model was our deco becuse with hs {oem me can rely calelte the a vals, most of Which te 200. A true linearization would ot have this simple form. Interactions, ep- ‘esened bere by all the ge values being nonzero in genera soul fac, occur. Clearly the input of otter Teed wl sft the top and botiom poct compositions Te inca form chosen must present a ompromize tetmeen the sft required (0 ibd the ye Vales apd the curacy neoded To gun, ultimate, the solution to the fowshect quavions20 lowsheeting on the compuser ‘We should point out that we selested mass ad Heat flows rather has mole fnetion and temperstare om pepe the former ne extensie ‘antes andthe Iter are intmsve, One fel more comfortable Stating 90% of « compooent enoriaglesves inthe top seam then S3ying 902% af the moe fraction of component does. Presute of 2 Scam can also te ieladd, Ii an Intense vars requiring some ‘are, Often ts vale preserved throughout astm so may a at, ot case problems “The remaining equations or the system ate the connection equations ofthe form hich te lo inearin al variables andy. Afr the ay are clevlated fom step 20) shove we have two sot of linear equations whic approximately describe the fownbeet and are ofthe fons Approximate models yp Sage Connection equations — y= ‘We collec these fogethe forthe entize owshet, that fo ll units and all connections, They ca, in pipe, be solved excep that they ‘omtain more variable than thre ze equations. The exit number of ca variables i realy determined because these equations are 50 Similar tothe orga set, namely Eset model I Bllekan eo , ‘Connection eauatons Sn Ye We sce that the model equation involve the sme andy variables; ‘nly he form has been simplified to give near relationship. Als the ‘varies, ted in vale, are then oasasts nthe exact mode They Inuence the ay sees found Une step to ithe libar model ‘he exit but ae ao explicly considered lathe near model. Thu if we choos alway of the, Yale forthe exact ode snd an devise "aleulational sequence in which to solve the exact model, tht sume sequeeeshoold work for the approximate linear model equations to. The previous rection we deve just sch sequence rele ts to spely the fed stream values Then the number of equations Was suffent to eleuate all valves forthe test ofthe fowshest. So this ‘observation tls us the aumber of varales we sat spe) for out linear equation oo. “The principal dillerenes hor is that, ecase she equations ae linear, ‘we can really make other specications than the feed sean and st 2.3 Agprouches 10 flowshesing sytems examples 2 Ive fail easy time solving, (Solving tinear equations willbe scsi chapter 3) These speifiations most be qualia abe to the mamber of eed seam artes notre, an they mist Be linear in the # andy varshless9 the compete sproximate model ‘remains near. Not every combination of specications posible, ‘hich we salle later x ebepter 5, 50 one has to be careful eve 3 Tegal et “Weare now in poston to dicus this approach and ite wet derive «simulation model ns ontined in our ener iaseson in seton 2.2 fd figure 22. “The wersuppied information agin incades the lowshoet and a is ‘of chemical components, Te design speciation mst nce all he ‘yuipaeal parameters wand est of linca olauoships among the = fy variables oan number tothe number of fod steam varisbls. ‘They mut be jckouay selected 0 they sonstiite a et which ca ead tors slation, “The probe definition i sine to tat fr the Sequential mode spproath The pyri property dat module equation nt, cont data must be gaered topethe. ‘Construting a solton procedure rater diferent rom the sequen til modular approuc. The order in which to caleulate the unis a ‘more sompleted problem besaute one stn place specitstons on = fly variables which ace pot simply feed tea valle 'W is posible to analyte the equations within the computer 10 temic tin ordeing, a parca, i fhe feed stream valet are Specie, the ordering of computations should he the same at fr the ‘quent modular approach. For the moment we stream values ar inf specie and the orderings Unit 1 une 2 anit 3 ~ as. eee group Unit Tote et “The pamiular sequence af eakulations might be the following: 2 Gueae allege for wie 1, 2 and 3. The sounds Hike » neatly Impowsible tsk at ist However if the linear models are kept reasonably simple, the a Ylues can offen Be guessed without (00 smvch ely 4. Soe the linear form for the model equations together with the connection equations for unis 1, 2 and 3, This sep wil ge Spprousmate values fal the acy varables fortes thre units 44 Using the s-varabe vafues jst founds the wat feeds and the siven equipment parameters solve the exact model equations oF {nit ourpuvrtable values for unis 1,2 aad 3 4, Recaelte he ae coe or ech at22 Flowsheting on the computer 6 Gompme te nea sett fo the ots previo suse (@) I esentialy equa, continue with step 7 (@)IC not esenialy equa, Feguess Weir value step 3 7. Since given eed to wit ad was jut cleulted a he ‘outpat of uni 3, unt con be cleated icy. Use the exaet, model equations to solve, 8 Ee i eepeat from Steps 2and3 requ ter dicusio a they ay seem to be aber Atal For one thing the number fy sett large Bi te loeat model adopted are spe, mon wil te zero. E100 spl, this ih may no sonverge. Telsabo posable ogc steam values, for example th inputs to all, miso dace hs mambo by gsig ony sept cam ves, ‘We can consider the renting sloth for ths st aproach for ou sample problem 1. Given eed steamy, 2 Goes the eye stam values ony for si3 3. Using the sequential mar apron sive unis 1,2 nd’ bn that ore, wing as inputs to ot? the op auc wnt 1 and at input fount 3 the jst ealetaed outputs nit 2 4, ind the my cots for unis fy nd 3 5. Solve the Hear form forthe model equations fgetbr withthe sommetion equations for unis ty? and. This scp wil ive new "aloes for alls andy vale excep thse ofthe given fed, 6 Compare the rele ste values sy jal calcd To tose fess. Then (6) at eseniay equal, pues and repeat rom step 3 (O)itescatay equal goto step? ° 2. Solve unit 4 usng he at mde equations & se Athi poi itis obvious thatthe choice of which ales to ges and to iterate (eto tet) sot unig Ir the feed teams arent spectid, but arathor more general bu lines) specteation provided terms st otor¥ andy values, the segues of esleltiont ferent and ‘must be worked out by the computer or ty the ser. A deat option Simpy to she ll ws cach eration, Ti option may be cosy. a sample inthe abo ese mt eltons fad tobe Jone one uy iter thos for wits 1,2 and 3 were complet, unit 4 wats complet tn (sequence of uit) nt jun simple miter. thi deft 10 colette he tert op frais 2 an 3 a ts ould every cosy an 23 Approaches to Howsheeing ems examples 28 In any ease, once the algosithm to sole i ted, the variables, which ‘a the Heras, a all Wentied and values wil have to be entered ‘Again erste within the unit model routines ar lo Yo Be iced the st of Herts for he flovshest. “The cleltion prion sto execte the above procedure which sin the form ofs computer program put together in te previous steps ‘Seing aed costing caleltions are ab dscused nthe soqental modal ie 23.4 Rquatonsolvng approaches Inthe previous tro seetons we found hat each unit was modelled by ‘witings computer subroutine which could convert input eam, {nd eeipment parameter, vlus ino output steam vals. We ‘ow examine the inlicatins of removing this requirement. “The final lowest i represented by a colton of moana ‘eqations which mist be solved simultancouly. The equations nluded te the model equations and connection equations, fr out sample problem, equations (1) snd (22). The design spetcations aren fact Stother st of equations. For example, to require that a certain flowrate FFin our problem be set to 200 kgmolh i equivalent to requiring the equation F300 = 0 to be stiiod, The temperature Ts in one pat of 8 Nowsheet might Ive toe 30°C preter thas in snot Tys-Te-30= 0 In the las two methods described twas generly required hat all sauipment parameters aceded forthe model equntions ad to be pected, Ofc, snot ually, one wisest spec) many ofthe values for both anit ints and oatpts The ual equipment parameters are then calelated 10 gre these desired transformations of Inpus 10 ‘utp by the wi nother word, the une is dsgned fo meet these ‘quirements. This stiction shat which diferentes ating ale lens to “design calculations, Foreating, the unit specie ands ferformanc ir alsatd. For dep, the performance specie and {he unit designed mect thi perormance, i it can In the ‘uation solving approach this ease The requirements are writen ‘seqeations and added to the et dela the problem. I they represent 2 legal specication set, then they shoul! create no problem, in ince ‘Gace the fl st of equations are together, a softion procedure for ‘hem derived automaticly hy the computer, This efor consider24 Rowsheeting on the computer bly more complex because of tlre sie than having the computer find a solution prosesure forthe smultancous smo appreah, owerer, many ofthe underlying wens ae simi, instend of heat duty on an exchanger being inp it maybe calculated. Inthe next exchange it may doe be np bocase the wee wishes to specity ‘We can now deus, as we hve forthe previous methods, the ws of is spproach for developing and using emulation Model Api We stall lok athe sep sted ln igure 22 “The wser supplied infration the fgwshet andthe Ia of hema somponeats or the proces. The detgn specfations ae inte form of 4st fons if Sead) eats whch yh iy orl “The problem definition steps the collecting topeter ofthe equations implied by the ult in the owsheet. These wil be kept na brary 98 ‘uation sts For the two modular approaches they were Lept ot Subroutines, The flowsheet connection equations will be added are the usersuppied design spelicaons. Tht collection coud capris ‘tousands of equations “The connecting ofa solution procedure means the Mlowshetig program hs todo th equivalent of weit largecompter prog 0 Salve thousands ef equations, The approaches pose wl be cued i ehapter 3, and they lnetue prosedres Heed on hath tesog ee Leigh, Jckson and Sargent 1974 and Westerberg Sha, 1978) snd ‘imulianeous Hinearation (se Hutchison and Shewchuk, 1974; and Katice, Hlavacek and Prochaska, 1976). The calculations set Ist execute, a Before, the soon pro cedure just set op. Sing tnd song ca be hand a «rho pat ‘(the simulation apd can wse equaling or modular iss. = “cision is (0 supply the cost of sont as an input ~ nota totally ‘nrensnable idea ~ then the sing and conting eauations could be included above so that the solution rocedare ean ssenan for esiing ‘ost as an inpot rather than produkt at esl ofthe cleltions 23.5 A comparison of the approaches ‘The approaches given inthe lt sction are only sampling of the variety of approaches one can visualize for a flowshoctng sem. They range from methods Which requ the wer to be very rgd in hi ‘robiem detition ones which allow hit be very Hehe, We: tow discuss some ofthe impiatins ofthese diferentes on the eesti ‘ofa lowshceting stem and its we bythe design engineer. A with all : cal | 2.3 Approuches to flowshesing sytems —exampler 25 ‘hing, ature has extracted its tll The easier it ist build system, the smote ufc i or the degn engineer to wae i fora number of Comton probes he may wi o salve "Th i! topie weal examine ls the ease with which one can wie te flonseeting executive system foreach of the approaches. The ‘owsbecing execute Is the program which recive the wer inp Collects the pablem description logetber, es, righ hroagh to exeeat ing and costing asjtem. "The sequential modular xocuive the east to wrt. The asamp von tat cach unt operation mode wl nelle wait ouput steam ‘lcs piven sit np steam values and euipment parameter valves tolves many probs forthe stom executive system. Tn contrast the ‘Suaton saving executive can besome extremely compan. The pods tems are thse of garatecng thatthe dvignenginer will give Ieptimate problem dfinion to the sem ad thal the solution ‘rocedure usd hy th mse eam be made Uo converge to a0 ane "By nsiing gt form fr wring a uit module asin oth mode ‘approaches, = sibrotine for cach module can renly be writen (rendiy"m a comparative term here). The weer has to make no ‘eenions onthe ape of np data tat he lb ven, The preparation ‘of unit module for an eqution solving approach, fat, act uch tor fel an may he cauer. The writer has together togeor te ‘sustionedetining the si io frm that the system can ecognize Sind han "The execute for the sequential modular approoch aso analyse the ftowshet iis dasigned so that the gre ds not have oo th sep, land work out a computational order forthe anit modes, The iterate Mounds wit methods shich will solve thst problem abd, sine a flowehect compres relatively fow ute and steams, most of the ethos ean ive good anrwers uicly. The combinatorial problem of Tooking st most orderings doesnot have co Be an exesie one even for {hil large sytem ‘The exceutve for the simultaneous moe sytem aso Hat 10 nae the flow, tf har to doce only about which ni ae involved in «computational fop, This oop doesnt follow the Now of Iara though the flowaket since the wer an ie specications on ‘output stream value ar wll at inp seam values. The ant a ‘Somputationl op are slved in parallel fron so thatthe ordering wihins proup i ot require. THe connecting ofthe units sone of Tolving the approximating linear equations, These equations conta 2 ‘onputational Toop, in pine, but, being Tear, they ae realy solved by socalled dest metho. For an equation-oing approach two options occur. The mas of thowtnds af equations cam be solved by suecessive Upearization26 Hlowshetng on the computer actos, of they cin be approached by am analysis Which atlempts to Iinimize the number of tetae variables neded, 4 metho allel “Tearing In the former version, the system mat ciclate or extite ‘onvere, fr example. It epresente combinatorial problem iwe-oe thre orders of magnitude more diet to solve than rowdeing & ‘owsbeet forthe sequential modular approach, Here the exset we obs imade of each and every eqution and Where iis fo osc in the Computational sequence mt he determined. Toe choice of ich other ‘zcision variables the wer might well speci toad the computation fst be fed if the wer has not given fie ‘pedifeatons, ‘We can see then tht the section in figure 2.2 titled ‘comract 3 solution procedure” can vey from quite straightforward 1 very com Poe Since the complex form ofthis problem not yt fly saved in the erature, ite not surprising tha almet al exiting foweheetng Sgsems have avoided st and Have adopted the rquentl oe simultane (modular approaches. “The cost for avoiding itis tha general pitti, which will hen require an equipment parametot to be clelted, ant be handled Snlety and the computation time in the step ead “caleltions i hey to be much trp. The advantoge of avoiding i it that the Alowsheting stem can be designed and writen moh more gel, the sysom is smo to specily and the unt modules can be writen, tested and made compatationally robust by taking special-purpose Fecstons ‘The desig engineer alo has no problem knowing what he has 0 seciy forthe sequential madlr approach to make 4 well-defined ‘problem. Thus data checking i mh ee cic for bi sd for the “To overcome the regultement tht equipment parameter: mist slays he spcitied spect uit modes on be but nt the sqrt "oda approach and we shall dscuss these amore dea in chapter {Alo the computation times can be reduced by other clever devices, ‘ich a ot converging oa oltion within the module calutons ual the reeyele computations are ao clos to converge, We stl dice these ens too in chapter 6, number of Ee \ 3 Solving linear and nonlinear algebraic equations Underyng 2 Mowsbucting system must be wre serice routines which ‘an bo toed 10 solve equstions numerical. The simplest pcblem of practical eed Is to sore a single aoliaea’ equation in one unknowe ‘We shal examine is probem fst and develop a common simulation problem where this aries, Although the numberof strategies co solve {te problem are numero, we sal concentrate o singe metho nd develop i slfciely 80 that ie can hance some poory behived pecblens. “The next general service routine Is one to soWve set of esr uations. Sine mest veshniges to soe noalinear equations require {pitertve sluion of inear equations, tis routine 1 ery genealy ‘fa The srest again the practical rguirements fora flomshesting System, Hore the set of linear equations ae often tery lage (ea) ‘oasand may are) abd very spare. By sparse’ me mean tha each easton wil have ew nonzero coefficieas, For example an equation J Eatieu ss = 20 appears to have 1000 coefiints a, For the problem at band it may ae ample equation Oise ISasties = 9 Which in fact has three nonzero coetients and 997 seroalaod oefcems. We shal cover the racial implication ofthis sprsenes ‘on solving linear equations "The thi sction wl dicus the methods one might we o solve et ‘of ponlinar equations. These too are spate for almost all eal robes {nd two methods have been developed inthe Irate! tearing and cessive linearization. We shall investigate these approuches wih Practical enginering examples 211 Solving one equation in one unkown ‘This problem sounds simple, yt the Hteratre abounds with paperswe 3. Hd a a sop aon. ined ety sing The rte dig eh hat Ie) = cm) =O rane en shown gop n ge 31, Arenson way 0h pt st ee tert Ofcoune eens a tr he cme Ae suet ovate of) ame es spn 1) fect Tape ett cert a rey ner() et) 69 £76) isthe linea function which hs the same value and fist devvative "Fas fe) ands usrted in gue 3.2 We en fad the sation ue 32, A tea won) we printer) aan =, 14 Sb on ection on wtnown 29 9 P6)=Oesiistnerins 20-8 (Ao 0/sshomin ge 32, ean ao tnt Joy Tebeppents tao th omc a ng eet OF 4 fancdon. Appiing this ea to our earoe problem, we gt he following resus wih an ila ues for # of OS foeret) £-r0= 120 0s ae new erations fone digs of sceacy ae found fr the awe Ts bviows from toking feu 3.1 that fi) well aproxinated by Fr), an this uk eoovergence ie a et ‘One commonly eannot or not wing 10 do the algebra to evalate , i that matrix hich whea (pre- or post) male ty A, yields the eat Itt In rat, the inverse abained by carrying oat Ov, of Gnise-Joréin, proces with a special set of RES rectors tepeteting the identity mata: The proces tuily modied to the extent of ‘ving the pivot equations bythe pivot value asthe pivot selected ‘The complete proces for 33 matte showin able 3.6, the sep aie ona oe oss: OH Sis tr HRS 0 8 Thad —nssan “S228 “Oesi0 Oe Las the 2 cept the a step, Belag analogous to those stout ia table 3.5, The fina step consists In eatranging the rows ofthe fourth Dock nl he nonzero elements ofthe fist thee columns le on the dagonal of the Iatrs formed bythe fst te rows and columns, tha unt they form an Kenly matt The Inverse mat then formed by te ost tree rows and columns, as thos rearranged, end ndizted bythe Box intable 3.6. Once an inverse har been bane, may be wed to obtain solutions forthe orginal equations with any numberof RHS vector Tb tlgcbrai for the equation Ax = ha sluions =~" which, pen AM fo maliptiction proces thre ate a number of RHS vector, they may be represented bya rmaisic B (aot necesarly sguare) in which tae the solaions te represented by a mati whe 112 Solution methods for linear equations — 81. xe ‘which i gain « muipieaton process Tei never advantageous to develop the leverse matrix if nei oly intending fo solve eer finear equation wing Caren practice 0 for equation (3:46)), royden's method gute at AN: = APs 64) forall vectors: which are orthogonal toe. Updating wil reuie the few A'to map the latest 4x ato the resling de® AI detions ‘rdhogonal te wil map through the new A Unchanged, whereas, the everatzed secant metho, all previous 4 map trough th new Snchanged. ‘The change appears sigh, peraps but the ccc Signifeant onthe computational alors, As wil be wen the advan tage ere is tht the previous As slope need not be sored. AU the Devious istry thats kept is jst hat i sit isupdated. Again wing ‘rank 1 upd, we get oP = AED = AMPED Equation (3.49) also gives AO WOW = AM forall vectors #0 which are oxthopona o 4x Clea if we select 0 40 then the inner product (P's ~ (42) =, and ony(Gt Solving tinar and nominee algebras equations £40 orthogonal toe ” For this eoice (el~Ada Cana) snd AON ADs mAs) (4eO)7ECal9%440) (3.50) hich is one form ofthe Broyden update formu, Given an estimate fora" the prodisted net step 43°") fr ether ‘he generalized secant method o Broydens method given by APH = (4080) ag = me os ‘Thus me must sove these near equations to fad”. Another form of Brojdon's formals ures Householder’ formula to slate the inverse OfA6"D co the ivere ofa ee two this version Koeps tack f A ‘guaton (81) becomes de) AED 6s ‘hich is a simple matrix matipticaton to get x, eliminating the ‘eed to sole the set of near equations (351), owseholers formu ie 19 = Ae = OHDOT) = A=) WHHETAM AE (FA) 0 HOW oO HOTE+-9)NP) od the ope formala for becomes 9) = HO 469-0) (PHONEHPHDE) 3.53) Fo Boyden method fis Fo tho genera scat method, itis a orthogonalized with expect many sot) previous ecto dat Tye algorian soared (ih enblabcat) 1. Estimate 1 = (0-1 = 1-1 1 teeny mati. Ur op plies x0 2. Braute th err vcore stanton of 2,1 (97 ‘mal enough xt 3. Enimate a0 HQ, Lat = 1 4 Late 50-0 and else 169) small enough, 5, Braue de = 6-1-9 6. Evalante ¥ (a elder 0 for Broyden's method of as the i 3.3 General approaches for ses of nonlinear equations 65 ‘orthogonalized component of de wih respec to prevow veton 440 forthe genealied secant meth) 7. Evatt the denominator ofthe last tem i equaton (3.53, (yt denon to sal yl than 10° tne {ny clemer in A) sip to step 9 &. Update 10 ge using eatin (3.53) 3. Endte nex dst = “HPO 10, Seer =i aod epeat rom step “The advantage of Broyde's method one a ll gencraed scat plates astted ar that no previos vets need be stared ‘Fhse vec, for! = 100, would rere 100x100 ~ 10.00 sorage focatlns, The disadvantage hat Broyn's method has onl pe Iincarconvergcnee properis (Deans and Me, 197) wheres the fered secant sehod & qudratealy convergent. Computational {Exprienc sem oahow that Broyen's metho gre gute favoule vergence mos instance ‘Onc could evisnge aritons onthe shove, sch crthogeaing vitonly be atk secre, ch where ke. Then ony k econ {say 10) need be noes Weystetnaceerton ‘Anacelerationnetod which i very commonly aed ia Sowsbeeting it ‘Wepstin's meta (Wepsets, 195). ‘Weestinscslestion ir tmply #0 teat each variable by the one- ‘inenional scant method by driving H with @ aiguely asoited function interaction ith othr vatabes are ignored. We might ty fon ur previous general secant example. For imps iteration ops two puewes, and 2, are neo to got started I the loops were le we could fave sted nth ony x0” and sed the resuling| ‘ccenive sbetiton values for, Wemextcheos thems hat both ve afe-l o-Ele-(] 2. Toya spy ea meio afr = 1-222)" (9) w .an86 gene (2G) nyo ss65 Solving lncar and nonlinear lgtrc equations “Therefore 14286 13333, eon | 20044286)41.3835-30 ]_[ ooaa2 51.4286) + 504133333860 || - 2.3667 Note that by step 2 we have sea sg redace the ezors in 0 from thse i.e and Note how cows xi fo 20 determined 1. a] ane rm y= pe eye Panera ap betimedraaeer elie raanes omecoee ars =H" («1 a0 cs aus) and we gt much worse estimates than by doing it the coset ways 201.2) 41.7143-30] P3441 =s(1.2)+-s0¢1.7143)°%-e0 | “| 15.2166 ‘The erors are only malay beter th ‘We ow retura to a base theme agsin If cach vaible strongly Inlaences a particular and unique Tanetinn, and if we at fortunate rough to sete tha association, thes this ype of approach wl ule Tey enhance the convergence we are pareulay fortunate, it nay ‘even ease convergence in many fewer than erations, where # x the umber of variables and functions. The we might manage to achieve onverpence before the fst lrations are completed, that before ‘we can apply the gncralled secat method ax nally described or expt iteration loops we Would sles the diferencia the svesed and calculated valve for 2 variable athe asncnted ero 3.3 General approaches for ts of nonlinear equations 67 fuetion for that variable. For impli eration loops the asoition of variable to a patclar fenton ray be evident ffom apy interpretation ofthe proba. = probably best uppity he person ing up the problem "a the fllowing dseasion we shal ok i more depth nto what ne might term one-dnensionsFiype acceleration method, ntoing over ‘elixaton, he dominant eigenvalue method and, ally the Wepsein for “secant method” (which we have Just dscssod We shal ty 10 tract from thi dissin the reasons why some of the numeri fapeicace len encountered in lowsheting is whe sping theve methods Overelacaion Large linear systems, of ncarized systems ae often solved by erative method to cconomize on compute memory allocation ice most are ‘Stems are wal spars, Tlertie methods avoid completely then that secompanies matrix inversion or elated procedures. Certain Seducons about Usca systems benavour ako apply 1 the tertve Solution of sequent Nowsbecting models [As defined carbet, ah approsimate linear moet of an iterative Aowsheetng model is (Ge equation (3.37) en arth ‘To dius the implations of overeazation methods, itis convenient so pl the A mate: AaPd eso where is an deatiy matric. The eror veto is now en Peat 635) Jn fact, equation (3.55) provides » hase forsusesve sutton on cexplie erain vaiabes 2000 = ROD 650) provided we interpret the ecor ea the diflrence between stscesive Nalucs of (he enor vector the th tp it naa os Substring equation (3.57) nto equation (3.55) and supercripting the -Fectrs in equation (3.38) with () yields equation (3.56) ) Since he flomsecting model i nonlinear, a mai A may be obtained by iting, a shows carer, or By paral dilereatston ot evuration. However, snot necenary to ave an explily dined ‘aw Testes cat given an estinate ofthe erate variables one(65 Solving incor and nonlinear aiebraic equations an generate fom the program, 0° = fe) oss) "The sequence of vector wil considered a having been generted ‘ya loealyincaized mode ice equation (38), "The sucesve sation proceduse, vz equation (3.8), aalog- ‘us in that for othe Taeobr method for solving incr sytem. Tf ‘ell koown that the Jasob prosedue wil ot comerge woes the maximum eigenvalue of i fess than 1 in magnitude (Youre 1971. ‘Wie asuime the eigenvalues tbe reat and pone. Iti ao ery common experience thit mass tance strats in sequen modular fowshccting converge monotonicaly Hones, the implied linea ert tion procedure of equition (3.38) docs indeed have » maximum elgecvaue les than unity fr systems exiting this behaviour. The Same expeince can genealy be extended to simultaneous energy a mas bales trations, but exceptions do aie in the ater ese where tbeethalpy estimates of stream may have adrstceffet on the pase ‘ondton, introducing dsconunuises inthe fanctionsf of eaustion (G58). We wil igeore tose exceptonal eases (whieh mist be restored to good behaviour iter by moe esteful sequencing of the azuations cor by bounding the iterate variables) and apply the ides all stream fterate variable, “The rate of convergence of the procedure of equation (356) ie {inefon ofthe lnygos eigenvalue of Young (1973) demonstrate tha the numberof iterations required to rede the Eaciian noun of sn ‘ror vector toa Fotion p of Iii value i 11 log pg 6s) here i the ares eigenalve off Alternately, p= lee ae 6.6 ‘where the equality woeld hold for sin age ‘The simultanous overreination method JOE) is sn atempt to accelerate the eration procedure by scaling ech tha the manu tigenvale ofthe mofiedyrocedere seo eration pls santa) lsteduced to a minimal valle. Tho result aeeserated convergence, _Asue tas ew ert vector bie weg ie ih 097 = an D4 (to) 6a) 200" =o) (-0}e0 6.) 3.3 General epproaches or sof nonlinear equations 8 ears 09P = [oF +(L-wy 4 ee sd the eigenvalues ofthe JOE procedure (i) are realy shown to be roti Gs whore are the cgenvaoes of Remembering that we are assuming all elgonvales 10 be real and poste the opium choice of isthe vale which sales or mapa of ‘he into the smallest ineral [MM "Te two most important ‘members ofthe st {A} are the largest, M, and sales, since any linear tanstrmation ite equation (3.64), which edesthe interval oo ‘hse, wl map the ftermedit values Between A? aad M0 wl be ‘mapped onto Mand sil bemapped onto ~M" The obvious saation iStominimiz the sum of Mand" t,t ths ut 210 M'= (M11 = = 0f-1)41 ted AHEM) =0=0(M-1) Item B65) ‘Tre optimum reasation ctor a, is thew 9, =210-Mom) 666) Values of a tond to be larger than 1, hence the JOE method xt the cob ethod As we wll 20 ater examples, values ‘tc only slighty greater than o, lead w uastble eration procedures ‘Soc the resaltan JOE has eigsvavestrger an uy. (We shout note thatthe problem of finding elgenvalus i harder than sling ‘cquations ro we canaot koow Af and m by det methods) "Acceleration factor for JOE are not aboluey bounded ot tnd an upper lini of 2. Optimum Improvemest i convergence rate then "pgnoeches 2, Le. ttl tertions to convergence ae halved is compa Ing OE to sscemsivesobettton. This degree of reduction in Heaton fount the Best that one simensional stoder methods ike JOE can “chicve on muldimemsonalprblems when asslraion i applied = very terton ep Example 1. Consider a lnes sytem with two erate variables whose IMAL hasegeavalues of 09 and 0. Using Young'srevat he number ‘Of iterations to ese the nal errr by far of 1000 fom ‘equation 6.5%), n= lepton Me Flog 000g 08 6670 Solving ince and nonlinear eigebaic equations Usiog JOE with 0, = 22-09-0) = ‘the maximum eigenvalue i reduced to MW sogMt—tyst S132(09-1)+1 = ose Note tbat no accelrtion corresponds to a = 1. The number of iterations to convergence with IOF acceleration i, = og 0.00100g 0.8182 23 ‘Tolstrate the sensitiv of this proce if aseeraton factor oy 0f2.0 were use, the maximum egeavale ofthe JOE procedure would utespond to = 10 and Mr = 2(1.0-10)41.0= 10 “The prosedure would nt converge Example 2). A pathologies examples the following linea stem: [i] feces, cata fff whose lpcoraues are 029999874596 and. 3:75004%10°5, ‘Thus 9, = 2.000080001 snd iM’ = 09999749996 ‘Successive substitution would require $52 600 ttions tore the {nial enor by a factor of 1000. The JOE method would require 276 300iterations.Agsin JOE improves convergence by alactor of 23 te “This example will be ten up again late, au "The dominant eigenvalue method ‘The dominant eigenvalue method (DEM) of Orbach and Crowe (1971) is an interesting member of clas of tne-imeosionl methods that lecues SOF and te seeant (or Wegsel) procedure As the work of ‘Young (1971) has indested the rte of cowvergence of a mocensie sabtttion procedure depends on the maximum eigenvale of Fin ‘vation (3.56) Young's formala may be ed to estimate the mak fmm eigcnvalas since the aio ofswxeive etor nor afer few iterations should stabilize to's relately constant vale let D1 pe| at os 5.3 General epproache for set of nonlinear equations 7 for sulciently lege {A stable ratio would only be achieved for 8 ‘Svcly linear system. At Young has proved the rate of approach Stationary In equation (3.67) s much ower for systems wih mul lage eigenvalues. Ralston (196s) describes a formu which accounts forthe effect of two or mors large an lonely spaced eigenvases which topetber dominate the eonverpence rate, Sationaiy x shoved more ‘gonvaue preva. “In nonlnear systems (equation (3.58) the implied linearization fom 4 suoeesson of vectors wll change due tothe nonlinear interactions mong the erate variates, Stationary of saccesive err nora a) ‘nly osc as ne approchcs the comerged sation “Astuing complete interaction among the Herte warble one ean etne the nono he eror vector Interns ofthe agent cement ote ‘ther thn te Buchan norm, and estimate Mf fom, epee G8) ‘oe montoing parses wre te ne of te monument Climrtee AT wa cma henge yt sane ma ibe teionieed rngcugh Hig se Megs terimtenie batons and ota eee ern Sen wan Soon Zeeneaens en a 6.08) Carying ost the estimated iteration procedure indefinitely, we ind ‘hat hee form geomet progzession, Suberacting fom both Ses of equation (269) gies ata = aD HseHAE .0t summing te progesion, EDDM ME 3) om en) quation (3:71) i the acceleration equation for DEM, #* is the ‘extrapoited vl of the erate vector tuned ote bestest UM. From equation (258) we compute x0 wing a the extinate, and ‘extapolte tox using tho sao Ao all erate variables.‘72 Solving near and nonlinear algebra equations “The sceslerition factor for DEM. oeq = MI=M) om i much large than for JOE. sequent to replacing by | guation (3.66). As poimed out by Orbach and Crowe, one cannot ply suc large af each subsequent craton, others he method ‘eeomos unstable (Fr example I, oggyq~ 10.0, M’ ~ 11.0)-For DEM ‘umber of dsec substitution steps are taken before each new ‘xtrapoation i made ‘One cannot se norms defined over the entire vector efor systems that ee decoupled. In a completely deooupled sytem, where cach iter variable independent one would need to we a characteris M for each variable. In flowshecting problems the iterate variables are either ll strongly coupled nor decoupled, All speles are weakly ‘Coupled at lem, through he phase equiam properties, and stone oopled if they priate in chemieal reactions wah each other ‘ystems with more than one tear suc, there Would Be srong coupling ‘between the Same specs vatable in each ea steam. Seca method ‘The scant (or Wepsein method (SM) iis the same ls 8 JOP and DEM. The secant method & ipted fom the Newion-Raphson ut when appied fo muiimensonal system Inf oe “imenssonal form i, ofc, elation method. As we noted eats, Separate acceleration facors are applied to cach variable. We have ‘Aeagy proposed tht this essential for DEM and JOT if he iterate ‘arable are uncoupled or weakly coupe, "The secant method for cch ae vibe j and aus" 1-3) an ‘The dpe obs fom wo see in ith egeaton 8): 1 )-fe) ‘ e759 So long a equation (8.75) eps, computed fom successive substation ape? = fee P= 16%) 23 Genera approaches for es of nonlinear equations 73 wens sfaapen 6) 1 eae ar) ‘Noting sgutons (67 and(6.72) dang wth he above, nee hat Shtnenetyequvelentto DEM, except forthe ec the nm tte to doe fier than fo die ech pre Shins mulidnentorlendronnen east the ane tbe iy eobioms nthe gh of releaton thor) ‘att oe mould coltzac wih DEM if wee ped evry Heaton ep. Hse Schoen feces moch pater tan a, are eacoumerel tort frotcns It as een he racic to bound Sin he following mana ‘Stone! eny y= ey 6-0 a Om qeon ‘where the lower bound — is lly -2 1 ~5. The upper bowed ong ‘murs that SAI never iterpoates(g = Osequalent ow = 1) Let ‘efato to suceeave aubtaton, ad the lower bound ins etapa However, SM lite DEM can benef from delayed apis othe pnt where the lower Bound at Jens, canbe eliminated Let ws lok nin at example 2). Example 2b). SM with delay was applied to our pathologcl example 2a), Four successive subsitution (SS) steps were followed by one tubounded SM step, den two addtional SS seps were taken, The "sus ae shown in table 3.12. Obsiousy, on ths tinea system one SM sep after four SS steps ges very close to the exact oltion. The ok 5.12 Rent orn 2) sat74 Solving linear and nonincaregebrae equations acceleration ticirs of 80 000 versus JOE of 2.0) suggest that SM ‘would bo inberealy unstable for this problem if he same ge mere be apolied every step. The natre ofthe probiem precludes tional selied SM steps since SS-step resus ze changing only inthe teh Aig or, and ninth dit for Most computers using sgl prion fchmet, woud be unable to handle the problem beyond the fst ‘lyed SM ep, Example 3. A peobim by Cavett (1963) has eon sed repeatedly in the literate to study varios acceleration methods mont resem by Rosen snd Pas (1977), Tho stem i shown i figure 3.1, apd Rosen and a " -— es ~ er 941, Coe out eg Paul results in tale 3.13. Figure 3.12 cleat indicates that the successive substation approach takes log time to sabe. The Cavett problem was slved by Roses and Pils sig two diferest e of streams tobe puesed and erate tor, One east was steams Si and Sth other wae seam Ri, RD and RS. Fr ether tse set he Blot of log ewor versus eration number doesnot become inset 3.3 General approaches for sen of nonlinear equaons 75 “313, tendon fr nel of 3.1 fer Ro nd Pi 187) = = ames fete ea ‘empere 33 ewe a) hae 10-20 erations israting the effet of multiple lage eigenvalues andlor online. Also shown in figure 3.12 the onl of sounded ant procedure 2 aap sa-ae) os) vith 0> o> Sand ackrion every fourth Heron. Tor ou urpones we wore mw tle to rqroder Ross and Pan resus on th compute, priiply eee we sd arnt progr foc esinating ppt propre: However, by shagig the {EAh egitim presures to gauge pues rater than te preowres ve worse ovun eee ocd mee ing eat 2), $y an acho coavetgnc fn 64 Kerns ttle eon antes of 00008, Tico with he shor oaks with CR TRS) whch comerged abut St teraony, Sess compstons {ne tot dena, obvious, tthe modi problem has the same ‘hastens On or (era touphr) ble was und that he take of Soop ercee he mr tance vais dot ves obsrved after 8 iterations, although the maxim ratio between ootssne errors was nly 70% of atime val A delayed SM was fppiod every eighth sep, without lower bounding, and The stom fonverged afer 23 Rermtons with only two applations of theEE EEO ':USCC“C“iéi“éc#” 16 Solving tinar and nontinea algebre equations ‘e312, Connyeeafthe Ca pen, unbounded secant aceleratin. The maximam ty factor (e.g) in ‘ach application was substantially larger ange 9825) than Roscn and Paul maximum of 6 ‘Other experiments on the same problem are summarized in tbe aie : Obseratons (1) JOE, DEM and SM are members ofthe sine cis of one = hhh ome) Merge ffm » rou node ©) Suing fifi) ‘We now atempt o extend the string and doves that the group node sf) has no utp edges. 2) Sting: fh Uf ‘ewe @) We lace, font the tp ofl 1, deleting fan fad alledginto and out of them from oar problem. Again, our atemp to extend the string fas 28, has no utpat edges ema 34 Solving ses of sparse nonlinear equations 8 8) Sein: ff case) Wie pice f onto the top ofl 1, deleting it nd al sedges rom the raph, We again ace Stputs fom f, geting (8) Stig: fy +f f we @) ‘Merging with f, and conning sites (6) Sein: Gf) ae) ‘We plce (4) onto the top of ist tan delet the node Uf) fom our sysom. No-nodes remain nd oar string fs empty 30 we termite, ving ereted the flowing st 1 (5f0.panion 1 pasion 2 Gy pasion ‘We find we have found exacy the paritonng and precedence order covered cari. Figure 3-1 tener the above eps they SEC Ot the dete paph ‘Steward (1962) proved tat parttioiag i unique, by showing that the only posible alemative output varsle esighnents must ase from, and ony from, shuling the assigaments among the atiale rd uations within partion. Precedence ordering onthe coatay, nd ‘not be unique, as 80 partons may exit wich may be solvable in fhther order ct, inded, in parallel, Figure 315 ilurates such an ‘xan. The thie partion does nt regi vals forthe varble nda mbich appear inthe second partion, ‘We’ now return tothe problem of asigning output varbles to uations. The example in gre 31a) isa neesting one i that ‘fal prlen, ut ldefat virally all quik and dint schemes ‘small invented to get an asizment. The usta approach 510 select "ist the column (or ow) wil ewes fecidences. The rw aioned at column i anon those i the clams and isl hak the fewest Incidence oi, Following thi goer sebeme we sal assign clam 1 Ast sige thas only two inidees it We select ow 8 because 00 has only two fncdences ini eater than row 7 whieh has thee. We Sete the nipned row and clam and reest uni anormal, we tele withthe asapamentshowa in gure 31605). Only uation fy ‘nd variable x, romain, and variable doesnot appesen euation "At this plat we must uy to find a s-aled “Steward path Teacing between any inideoe in the 1 fr equations fos) fecaence nthe ‘column for variably. Figure 3:16(8) strate sch 2 path It tat With an anasiged facdence, moves 10 em ssgned one, changes ‘Srecion by 90%, locales an ensign fosdence changes decton|82 Solin tne and nonlinar algeraic equations Cro & Her epee egg ctf a sey as nti by 90%, moves 10 at assigned incidence, ete. Asigned and ‘nasiged incidences alternate along such a pth, with he fit ad ast bing unassigned. The path has one more unsigned acence sone it thn signed. Also cach row an column includ inthe pth es only ‘one assigned and one unasiged incidence init. We can therefore interchange the asignment,onassiging the currently signed inc ences and using the curently unsigned We wil have cated one idioma assignment inthis maneer A detaed algo for finding Steward path given by Gupte (1972). igure 3.16)" has the reasgnments made, and we have 4 fll 4 Solving st of sparse nomnear equations 89 sssgned st of equations which can ow be pardoned and precedence ordered. Fda Steward pals for moderately Inge problems, a) 25 ‘juaton, not oo dilate doby tl and ero by hand Ino sch ath exists whon one is needed, 90 ouput esigment exis and the guatons ae structurally singular, as mentioned asi. qi ar ale ft a) ob Fr 945 tance mar mage wh nomi rc ne “The solving of sparse, square equation sets may of course be done by ec lineartation, he sparseness of the equations wil eetvely be counted for ty solving the resting spare linear equations wing tcciniqos sucha thowe oie in svton 3.23. One ean, of couse, partion and prvedence order the equations ls, a indcted inthe previous scion, Each parton can Be then solved in sequence sing Sizes linearization techniques as dscssed in secon 3.3.1. Three fdvantger to parioning ze apparent Fist of course, a Inrpe problem may bom teres of smal probleme Soon the nat fatter need only be gussed forthe fit parton before ft solved. “Ton staves exist forthe variable which may permit beter cu for values fr the tecond parton andlor aid In developing 2 mercy sapere pot seqacnos forthe second partion. Thal i the problem i partiioned and = pivot sequence Is then found oaly ttn hepato, the filin bealne (which ests nore aime Te operation) wil oly occur wihin the partion. I nat parttooed, ‘ne tay select piven within he partons and wil ry Ie eeate ‘more nonzero elements hs way ‘One can ali olve spare, squire equation sets using tearing amotiods. The ides behind tearing aze perhaps best introduced by ‘rample ad in fay, we hve lady soen examples of i hs book1 Qessesteet © non ‘Jo: Fue 316 Same pen ih int nce en siting tpt ae ‘SIS Fintan) Smad pe come ores Fieneoe ht ‘steer oy fw pie) Renee es oleate 3.4 Solving cet of parse nonlinear equations 85 Is section 3.1.1 we st up and solved the equations for a smplied ‘hrcs-omponentwotberma fash un, The rubber a equation soled tas {4 hese ince thee material balance equations thee equi "um relationships, sx psa property reationshps, pia wo oer. Reviewing the algoith,me hid in loop of elelations where we iterated the vapour faction to drive the exprsslon 33,~3, 10 et, thin thi togp were the istewtons 8. Guem ViF thon V = (VIF) 4 (a) 1 not eel 20, erate (bOmerwise, conte ‘There are eight ogetions in eight unknowns inthis oop, and we are solving them by eating ony one variable, This approach the sence ‘ttearing where the vasiales oe trated, ViF in hs eas, ae called ‘ear warble ad are many fewer in nabs than the total number of voviables The equations within the Heaton Toop are said Yo te 2 "roup i ations since they mast be sled sialtanousy Before continang. lt wal investigate a second exaple of tearing that we Have aleady. preseted. To solve a flowsheet using the Sequential move approach, we trated an example wher a recy Sream was quem, The fowabest unis could then be solved a ‘Sequential order ual the unit producing the reyse was encountered. ‘Atha point, we compared the guowed and eaealatd values for tbe ‘eye stream and erated hey ad nt agre. The repel variables te tear varisbles ere, andthe ealculatins forthe units within the lop fe 4 group of equation tobe solved simultaneous shou be avis thatthe partir tae variables wsed vary with tne approach devised, Thar me can “pla” wilh the equations and tttemprtn devi more appealing approaches for solving. Some interes ing questions ase Can the proces of Gading tear variables be Momated? 130, what instaons ext? Wl te eslng elelations| converge? We can dover some Meas relating fo these questions by Teinvestigaing the equation in scton 3.1.1 forthe sinple hora ‘ah unit, Ft they have considerable spareness, which can be ijosrated bythe leant incidence mtr, Higure 3.47. We defied an Incidence mati ear, bu, repeat, the rows represent the equation {nd the clumns the variables for our problem, A blank enty i row {dcolumnjdenotes tha equation f des nat expliy contain variable jsnoablank entry denotes that it does. Function f corespands then5 Solving lear and nominee algebraic equations AAHK RBS TLE RR GRAB ET u 3 a, f am ay Rew Tiga pedo nko ech hos nace oma the particular equation piven nF son crt pet inthe 0 on xm eet Weeten aa ‘tf hs ee eect f eqn wih wih sels. We ae ted 1 tans a Saree ‘rm ss sno fhe, We eh ee oe Titra! clsatn the we pate ee andrei ties ay nen ed sa cd fm oar pom Fee poe omg ee at ‘Sith eS qe 1 bree Treo now puien pre ne eucon 1 oan serestege’d me shor ge We octet PEPE Patek Kyandy canbe clcsed ia that onderend ere Ged etn aon tga near ct sro figs te ri aoe as Trina psi o conae that oon pce it hich itotresre nee ni ut inane ot Tee iodo one tip eget We asta hat yh toty Ho, Our So ht iy ae we ‘suas re pn me nknone eosiig Sen val fo docky's tn poet fo ew ei quan ght smo A ty sce ad Sil se tanenps cea teagan shonin sgt tree S21'Stynac wee lag ue oe oboe 54 Solving se of sparse nonlinear equations 87 DANSE RRB REM er 34, eto ti i pte RAMA K RHA wa TL sft 0 a) “pte Gy Bp Uh, Vg a ‘ an ‘ vf, 4 eee: ewe 3.1 as th te pane rc od en, oe ur 220 cee mi eigen co fi favs 18 fed aes Ky Ky Ps Phaed P8 Solving lncar and nonlinear algebraic equations variables. Then we could consider these ‘guesed variables as known, be iterated, asi is anfely that we can fal gues tel value ‘coneedy. These variables ar called tear variables. We proceed, "unetion fas only two vale (4 fy ga f,)- We ean decide to ue io cleat Land quite a ths pol thet we ace to ges ‘alee for V. Thus we Henty V usa teat variate. Now the st 0 Columns ean te deleted alongwith the Fourth ow, leaving che mati ia ‘gure 3.2142). We observe immedistlys rather interesting thing. All a afr a) afin hi no aps afr 4 afoo44 ator o 5 ma hi roo ‘ ma B{rvrira aficrcie gue 221 Pl idence mas eine nr Mii) mat) emma oy are the remaining equations ae linear in al thei unknown variables. That ‘we could simply hand them overt spare matrix oatine for solving Note thatthe sets not square; there are seven equations ink ‘unkown so te spare tee soutine wll need to Hnily od ant equation and simply evilust I value i oro if V hes been sven a correct valu; otherwise ts the ere’ which we dive to 0 by ‘stg V ‘ator than handing the equations in fot oa sparse matic outing, wwe can rearrange the order forthe variables spd equations to ve the pattern inthe inldence met in figure 3.21(0. Here iis oboas we faves numberof linear equation pats to solve fore, 02.75 and forse ye Thea fy simply eluted to see iit roar We cin show that ling these equation pale corresponds preciely with the algebre ‘we aid to wate egeation (315). "Functions fe fy for example, ae fay Vonk anh fern Ke 34 Solving se of sparse nonlinear equations 89 ‘Atti point in the computation Vy Ly 2, and K; have vies, so we fe left with two near equations inthe unknowns andy: Wythe, = yoke, = 0) ‘A simple Gaussian ciminaton to eliminate they tra inthe second ‘auton gies oy “Wks = -aF which equivalent to equation (3:15) Thus me have hoon le to derive the intuit soluion dicovered este. The approach we have taken fan be fomalized and made lo ap algo which may hem be rorrammed on the computer. The imps the roatne wold hea at {feqtions snd the otpat a soliton procedore for them the order in Sich fo solve each equation she variable t ave for in etch, the {uation groups and, within them, the ear variables, and 90 forth "The soutine jus alluded to ii practice enor compen Te ha to te designed to handle several thousand equatios; this metas hat ‘ver feature ofthe incidence matrix, wich we ciety baerved for ‘ur simple example, potently becomes a ge combinatorial problen| to Woete. It solved without de care, i could Ted 0 impose ‘computer time requiemcns. Thus clever means are crucial fo store ‘ato sarc them, endo forth The esearch Mterature abounds With thes seas and comparisons there, jatisably ao ithe problem fo be solved realstial ‘Some of the references which ge algorithms asf for solving square, sparse equation sets by tearing include the folowing! Steward (0965), Lodet and Himmetlay (1970), Westerberg and Ede (197, 1, Markley and Motard (1972), Kevorkian and Sneek (1973) Pho an ‘pads (1973), Sopeme and Serer (1973), Cheang and Kh (1974), ‘Guardabas (1974), tad, Gilford and Seven (1074), Kevorkian (0995), and Sith dad Walford (1978), ‘342 Salvng nomsguae, sparse nts of equations by tearing methods Not al uation sets are “guar” when the wir as Snished supplying ‘hen No being square means the numberof equations, i ae ual {o the numberof raat, It more variables exist than uation, mon deprecs of foedom exist forthe problem, snd men adionl Specisestions must be provided. The routes for detving «solution ‘Procedure can handle this problem by stately choosing m—n Sarisle oe decison variables whowe vale mst the be seppied tele the pocedare is exceuted we examine he Rsh unit eqeation® G.7}0 6.13) of ection 3.1.1 (assuming three component at me dil ia0 Solving tneor and nonliner algebraic equations this section), we find ive ational peitiations are nesded. We were Solving st aoerma sh ith gion feeds the speintions were loeeyeey PT aod oe ited in equations G1). ‘Suppose the fh nit i to be optimized to meet some economic oF ‘operational cbjetve, and, in fat he onl variables we Wish ofa Prior ae2,,2, and P Then the sauton procedure wll have to fad 180 ‘arales for Ws 1 adjast wen optinang these wil be the decon ‘aribles. But which ono? I we donot have any preference, then hy ot find tw which wil ease the problem of ving the equations? For this probiew the soem ash eaculaions prove toe area 10 Selte a5 anyother, the wo variables Which ould be chonen woul ‘ery llkely be F and. Otten, however, unexpected choices revs the computational procedure derived fo the Best choice my contain 80 iteration hops, obvious avery desirable result “The capably of finding desvon varales ie very weful and ‘gost exst nthe iterate to sole this problem Finding deson ‘lables often significantly complicates the problem of finding the Solution procedures one ots extant to nda bet et Of deren ‘ibles: The problem tends to be combinatorial ‘We can ilustate a singe and vlatvely effective algorithm for thi ‘ype of problem. The lierstare contains relied slgoritme for his froblon’ Gee, ei, Chssemen, 1970; Edie and Westerverg, 19715, Ramirez and Vestal, 1972), The exami problem tobe comederd a ‘hetwo-stage, dee component absorber shown ia figure 3.22. Wesball, assume the presire # and the liquid flowrate L are specified. The ‘uations sale to model his alt are (lor! = 1,23 andj = 1,2). Materia balances syetb tga yl 67) sata panne Fg 2.22. Twos tee sonponet abner xe 14 Solving se of parse nonlinear equations 9 Equi: yu Kah a0) Physical property relationships (ne astme no composition depen- ‘ene Ky KEP) os) other: 1 reo n2 os) Bay drt menns aay) Figue 3:23 san nsdense matrix for these 24 equations in 28 variates "The two preslected decison variables, 1. and Pre our ate 24 equations in 26 variables, two more decison variables are needed "We cannot pation awe did before when the st of equation wae square, We mint ereforeprocoed in difereat manner AS We argued aumeunnrsusroraionsorawnnakutatakekt FE 1 yur 323 Tne mii het cumple ein vases92 Solving lncar and nonlinear algebroe equations ‘ator in section 3.8.1 when deriving a solution procedure forthe fash Sit, we need to Gnd 2a equation with only one vanable in itt get Sarid. We cn aso look fora verable appearing non) ve equation For this later case, tha varuble can be asned (if we like the ‘ssignment) fo that equation and maybe dented the ast equation to be rlved, It output variable canot be foguled sh am input ‘hewhere vo that variable can cat be the last variable alesated For our eximpl, no equations have only one valle, and 80 variates appear i only one equation We must pevforce ‘eat an ‘equation wit only one yrable in for we must ‘reas! a variable hich appears In only one equation we are to proceed ‘We would of cour, Ike to erate as few sch taations as possible boca each ons inencm of wo (Wo ofthe sel ceated willbe elected tobe deco arabes) wil ad to out hang fo guess vale for the ‘arse and then drat on thi gues, Le el eeate ter variables “To minimize the sumber of potential tear and dedson variables ‘xeate, we seach fran equation withthe fewest variables. Abt one ‘ofthe varbls wil have fo be Know oper the remaining on o be Sacuated. Looking atthe eustions, we tee thal severly" fe ‘ontin only two vsibls. In fac, me ne that i value for 7 were known either by peesng it or slecting st tobe decison variable, we (ould alcatel the Ky variable, We could teretore pick T to bea "fnown’ variable apd alge the, varale 8 outps fom equations fis to fig However we sal, 1" consiency with the ret of this problem, avoid making quite this commitment. Intend, we shal sect ‘ne of the equations, yf and sip Tata lows, bath variable. Sppetrng init Our it, which we sal he ereting at we proceed, st folows 1) LP a prin decisions @) TK ©) fis Ka variable appeurson or istit means tha value know, thor a8 decon ort tar vale or ealeulaed. With "know We ee ‘ht we ca sig Ky to yg output valor eguons 10 fi We ai these ou ist: Goo Koad ©) Gok) © Geka) Gen Koa) ©) Cink) “The shove tit i being forned withthe following iss In mind. Fist, ‘we shall Ca variable y tel when a need ors value i required © 34 Solving sets of parse nonlinear equations 93 ‘permit further calculations. Inthe above we needed a value fr T 10 rowed. Actaly,m principle, we needed value fo To” Ky ce {he oer should be mesial avalable fom equation fy We know ven now in the anaes tht we prefer to caleate Ky af ater ‘ha calculating Tom dt equation, but we sll no yet make this commitment So welt both, and fllow th sting With fy ‘We shall svaye ita eqton as aoon a al fs varables have appeared on our is, whether we have js crested tht sitsation or has happened by ists the cours ofthe mali The equation, whe listed canbe used a that pon, perapsas an error function into which ve mbites al he current variable value and then check tose the fsuation io tot, sted. We shal sb list an uation ad ite 'sspned ouput variable the assignment i forced because the equae tn et with oa ane varie nt whi soot et sted. ete (2) (@) spear of oue lt for ut his tenon ‘AS cach valable and equation i Hited, we can dlete i rm the incfenee matrix. Figure 326 shows our corrent reed incidnce matrix. We proceed fom this point apa 1 t We note now that iether, or, mete known, the other could be tatealted wing equation. Simla either, oF ya, Were Known, the other could be eluate using equation We all not make ® ‘comment at this point eto which we shall sume is known 390 ‘whi 0 be calulated; rather we aba Spy ad hem oor I94 Solving Linear and nontneralsbrale equations ©) tutus ao) (1) Xyeras (2) fe Deleting these four variables and two equations allows wt discover ‘hat equation f=, may be soled for, and equation fy for, We ald these fom he 3) Grn) 8) Gorm) ‘and delete than from the incidence matin; figure 3.25 shows oor ‘ent reduced incidence nates, ‘re 325, te me teres co We note tht fg bat oats vavable; we delete it and ait fo erst. os ‘We note that either y,3 or 3 knows, wil permit the other called sing fyi 33 #8 #29 fy 48 Be simaty Weaed, 09) yaa m9 2D he (18) yea soa 9) fit We in now tht conn on x, and vile We sien them, delete them, and add them our Is 20) yn 43) QD Upena) be 44 Sohvng se of sparse nonlinear equaons 98 Figure 3.26 shows our Istestredused matrix, We note fig as no sweocited variables apd its deleted and listed 2) Sia Finding f with only to variables i it, we ist the variables kaown, and then BY con [At this polat equations and fan be assigned, 25) Gootand 25) Gxsn) leaving fg with no varies: fy es 3:6 lees mc te in ton Final, ft can be asianed ouput vasibls 7,107, ening fg ih no ited viable 2%) Goris) 2) Gora) ©) Corsa) OD he Figure 3.27 pve ou Sa it. reed to anaiys this Hist to see what it tllig us. For nasiged variable appearing onthe It must be either 4 ‘esion ‘arible of it gust be cleulted by an unassigned equation ‘whish follows [The reason wh the eyuation most Follow ie that we fae lays sted ap unassigned variable before any equations which ‘appears. We alo listed i only when is valve wa needed6 Solving linear and nomics algebrate equations We carcflly handle the placing of unasigned and assigned eau "og on the ti, pacng ther there as soe a they could be wsed, ‘We must now asign an neinigned visible Wo each unsigned uation; we must abe presebe nrle to ach unassigned sari “Thisrole may bea simply asiged varia, a tear rato or dechon tatibl. We ca vec the role to simpy the calculations requ ‘vate all but the two variables whic Become decison varices. oe mA 2 fy Bk Oa a ta he Wend ens eh) 0% 29 hi) 1K 08 rg Nad Ue — OF Oh OUR) Wns end etek) oH md Mawr OA) — BH sd ans 20m) . yr 1 fi unas ced we nding ne ta in wtih hn ‘We il sto of th nasgnd vases tobe dsion aie but which eo? Went the following observations i Ft ofa several ofthe unassigned vraler ae oor own ration ot we could ve signed hem err Forexompl, me od posed item 2) bul immediatly flowing jy a quintile o Cabelas Ky, the ae oF hows, cond observation that 9 tear var mart precede ste fncin fects we mes i gers the tear vai and then eventual vaste mn eo uncon, sn 'sovnted teat faction. Alo the fear nrableslwe mst ac the Vale ofthe fat function, 30 we mist exiian 2 casa fet {eationship bewoon the two. We might to dere tat the tear function expicty coma theta varie forthe he ca cet an croc Keaton lop. 4 Solving eet of sparse nonlinear equations 97 Just how can we arate the eaue-ndeffect elaionship? Ste ‘ral, isnot too del, We can take our ong inedence mattis, tnd or each signed vartle, perform opel Gaussian elimination, toting en tho lotions ish woel, i poner ecome none tre were saving wet of inca equaions. For example, does Pact? ‘We veal tat fe the equation ra-Buahs ‘We of couse know that Taft the K clelation whch in turn fects fy To prove this eutomatcaly, We could perform 2 logic ‘Gaus elininaton onthe equation and variables fom T to fjon our Tht (Se figure 328) Ifwe ign T the vole of er vanble 30d Ky. 1 | ts ma re, Ape Comin ion ney te ees a ™ ‘se fh to be alsa kom fy, then limiting, fom foul cause he indicated fil Thue would appr ply ia, We note that fy swore wied to caleulate 7 given Ky then Ky affects f expt bt hoes not fst fat "We can fr al aged outs, perform lois Gaussian elimina tion gating the finite by eres in Hgute 329. The order in ‘hich to proeed is that ven in our hs shown in figure 3.27. The Sutput alan rows and clus can then be eliminated ing figure 5130. Note that the expt eceurrenes are indicated by Ts and the Ipc occurrences by exes. We now wish to asin the tar and ‘eesion variables fo ous bet advantage The goal ft make a number ‘ot smal subprbiems ater than one gat iteration loop i pose “The probe appear to have Brea Between equations f 2 fur From fy tothe end we ve sv eqestions and si ew variables, 1098 Solving linear and noninearalsbrate equations Koi Kok Boy Aa A Drakes reba Fahy "00 329 a ine fom gal Gala cnn aie wi [sea |B ue 338, ioe psig ep ‘Therefore we shall atompt to partion the prablem st it pit. The Jas six equations willbe wed to calculate thefts variable, The fs four equations ivolre si new variables; we shall elect two ofthe ‘varlbls from dis nt to he our two decson vais Proceeding. we asian Ky, t0 fy and perform a logkal Gaussian timation. We asin fy gun ptiormig logical Gavin . (ie 24 Solving se of sparse noninear equations 99 clininaon. Each sang ni shown in we 3.30 wih a We Sin 10 yar ae le to ae an plage fo, om song th ies 7, yy and yay ANY of tee wl fice a ompettionl loop, so me howe Yay since the loop wil Sat jut {Site eleaing te arnt we ean make. Then ue ‘ote swan ue 330, wie Computation loop horn he Inner nee brackets The vrubles 7 andy, wll be the dion ava. The lst equations ean be aig mow Uae 7.) G90) a 22) GeV) Gy) 808 U3). Fe 31 ss {hs apoment th eich fil ning tt te ptonmance 04 tops Caan chnnaton nie y 8" ioe: * +o On» ig 331 Ani opt ies cd parton pln Figue 3.2 indicates the final computational order. To const figure 3.32 the equation onder of or sin fiute 3.27 manned ‘The deciion variables are listed fist and then the asiged output variables are Tied in the over of the equations. We teat two eration loop: ff with, being the erated ter vail, ang Wo fog witha at, being thre tear variables. The second loop could be sited sf wo thea flows 1. Guess, 2 Guest 3. Use fot td fot obtain 73549 amd, espectvely ‘$a fav at err functie it ene oro TENE aa and erste om stp 3 5. Solve ft fi by tearing ad erating yp Sate (1 0" Yu Yay AD )93, 1. Balu fc I not esti 2c, epess and Hemte from sep? 1 would probaby be better, however, to solve it by guessing all three ‘arable tad converging the three functions fy fy Ad fy smblane- sy sing the methods of section 3.32, Note th wo futons, fg df ate impli thie ea variables ft expo100 Solving near and nonlinear algebra equations HBR Faha Sas Ya Ba Ra SaY Sabha Yoh Foe 3:8, gin nw rn Co “The example just demonstrated iltrster that finding an effective rotation procure i «lame effort, i avo appears tat we could !ntomate ts that computer could self develop aol posed ‘We shoatd note that all decisons made whem working tough the ‘example were the apparent bert we sould male that point the Problem, but we have no guiramte tht tte soltow procedures ‘ould net be found. Chrbtesea (1970) published »brincand-hound “lori whieh wl grantee the fees ea variables. Whether, fora Inge problem, one can sfford the search which sel tl an open questo, The algorithm described in thi sacon implemented tnd Seems vory flee for small problems with 3010-100 equations (Westerberg and shan, 1978). ‘Wecould us the above approach oaeect ony the decison varibes. ‘Once selected, she problem agua, and we might be tpt then fo reanalyse the problem aba square cnc, parting ist Again the "fort may not be worthwhile, in general but this fest as ot been proved. I i ear tat ome may dicover a prtoning not dacovered bre 34 Solving se of sparse nonlinear equations 01 Forsolvng set of linear equations involving m (>) varies one could use the sparse mati stages fo find a pivot sequence. The ‘rales not pote wl become the decison variates. gai, nce Ihe decsion taable are scored, the problem con be tested. ‘sauare; partitioning can then be attempted “The parpe of this secon bas been to demonstrate the approach one night tate for aonsquare problems where the number of variables caceeds the number of equations. The ides involved under much of the bterature i this sea ‘34. Solving redundant equation sets “The other mannerin which equation setae not square site umber ‘of equations m exceeds the numberof varbles m. Tis casei one of ‘venpecieton snc, n general natal he equations can be satis ‘the same tine. Two approaches are posible: (1) fad a compromise Colton, a done in ear sgoaes formulation, or (2) sam of he votows only and simpy evaluate tbe remaining nm This tate ‘Shyroach was the ope tat we requted eater whem we chose a tear Sarible inthe nothecmal ish example secon 3.1 and left a Donsguare sytem of equations My fact, the equations ae simply edundas, that they areal valid but one or more can be derived “icecty from the srs, the Jest guar error wil be ro eicting ‘hat al the equation are satisfied othe n~m ‘vale only equations wlaave ato vei the econdapprosch sakes A danger irks fo {he second approach: one cannot normal octet veda equa ton cai toler the equations ae lear. The redundant eqaations tay be inte in the Stam equations we propose to solve, ming thet unsokabe4 Physical property service facilities [A flowshecting program lost always contuins a physical propery SSermice’ sine the quality of process desig s timaely dependent 00 the woy in which the laws of physi ad chemistry are apie t the robin. According, the qualty of ths fervice Is an important Eomideraton to the wer ofa sheeting stem. The provision of thot fees sy however, a tchnology i Hr owe Fgh and in thi took we shall nly ook at ovealquatativ characterises of ysl property stage ie the Tlowsherting context. “The physical property services to prior a umber of tasks, but the mot tefl of hes are! (2) suppl estimates repetitively fora aumber of deren physi properties while the simulation isin exeeton. {@). to provide the ser with values of properties of intrest during she ‘Steulation anion smlaion completion, for subsequent wen other ‘Slelations (@)_ tallow the wert input his own special data for new components Sn ttenform i into the ove required by the system ding a ‘mulation, (4). to supply the wer with ameansto estimate properties where itl freept puihaps chem stature, sown abot a patel chemi ompoun. Again he system mst put Use esas Into the frm Seeded by the simulation diag excetion. 41 The dat yee 1s perhaps sel to zeview the flow of physcl property data from Sours o ue in lowshecting Pare. iustrats thie ow, The new ‘Sate eiginate from mearorement i the laboratory and are often Published ia he open iterator To slat «physi property caleulation ‘tem, both sures are ued to buld up what one might term brary ‘Of aw data, The bulk of dats fr for pure componens; some $6 TOF fnutres of stall no moe than Evo or res componeats over Lined 4 The date ele 108, ue 4 Te gent mangement nd we af hcl pope neni ‘ ranges of temperate and pressure. These data must be evsusted ‘ton wth some being seected. Vall checking» combination of ‘Shnssteney cocking and subjective evalaton of he data sure. Some liborstories and some people are sowed ax teing experimentally siperior to others at producing da, 30 comics ip yalues must be ‘considered agro this assessment. Most abortory dats are preseatod ia tables and must be condensed tor form more etdly tae, Alo confiting dts, where none canbe feiected out of hand, rege that 4 compromine set of values be proguced. The resting emfunted data are reduced to equation form104 Physical propery serie fetes sng some means of static iting, such as leastsquares, to establish ecto parameters for we nthe quan. The form of ach equations Soften determined om theoretical consideration, and, of courte, ms eit ae to be extpated with ny confides heyond the ange ‘here experiaeatal values are avaible. "Tne above comiderations apply to data onthe pure component, bat tng bulk ot pips propet rege for designate for mines of fomponents appeating Together in one, two and sometimes more hace, Few experimental data exit for mints relaelyspeating, nd tually all design need are for mixture popertes. Therefore those properties mit te eximated, and cousierble theoreti effort idevatd to ths one gost im order lo enable resomaly scurate prediction ofthe Behaviour ofthe mixtures. The eificlties ae evident hough For example, how does one estimate tbe beat capacity of & iiguid misure? One might be tempted to moke-racion average pure Component Igoe! het cipaciey bot by nol welghttration oF Sofumesacton average? And what fone of the components he Tiqid at temperature well above its ritcal temperate? No lq het capacy exit for To deve frtce nto these ining {heston is byond the scope ofthis chap, 30 mesh Tn vow ots enormovsxorta efoto pemit accurate physi propery predtion or mixtures the modes used for data fitingand for Siiuiaion of the mitue properge ae in cather a sate of x Nonetheless many models bo exit soar used tay secs parcuarly on wel-bebved fames of eomponent suchas sonpolar Epdroarboa, The models equ et of parameters Wo chataterin {he pure compence thx act oxtally cosa of Ete parameter of the ype ducted above, ps baie thermodyamie constants nd may th clade sional data that characterize the interactions between binry mints of components, Examples of base constants re the ntl properties To Py and Z, (ental teeter, pressure ad ompremibity facto. renee), Abo many dita are cotelted Spain the acetic factor @ which an lndator of the nonspbercl nature of a molecule’ fore Bld. ted parameters ince sch item as heat capacity curves reduced to polyno form aga temperature, Other data ae fo he tan Pom properties such s viscosity, thermal conductivity and dies Thos ae noxmaly in popnomia fom. Thus the dating step to-comert rw dit oa compact fom stable for we the property teneaton soins that embody the theoretical model Thsome instances there sel or nora” dita on which to bse te creation of these parameters, Estimation Of mass dat s normaly Toudimppingpeoble. Many methods sualy eat for etrating pried pscl propery abd they often use diferent information 42 Compucrized physical property systems 10S shout the compound a a starting pont forthe calelton. To find an nimate of physi property mey rege knowledge of estimates of propertce & and'¢ by ove eeimation procedure and knowledge of Dropertes cand by another These i ar, if estimated wl que ‘ter physical properties, The etinaton eer propagate through sich sors of calulations, and one ean keep tack ofthe approve {magatodes. Tis eror estimate ean give ote x means, though No & ‘ebive one, to choos among the methods wsed. One may prefer nssteey withthe estimation of other properties, and choose 3 path ‘onrlered lee accurate heeawe the other propertisn use much ofthe Sime pth and are temelres ore acorns te way, Tho las mount fof data required for some systems the chemical suucture of the fompound. From thi starting pont ecital properties are estimated, ‘hen the vapour prssire cures and sofort, These type of estimations ‘tay permit ne to dos design and then atts wheter ny ational seuray is needed. “The fiat spect of the data ec of curso, ts ws in the desi step This we oculs pica inthe analys phase where Nowshoet- ing occurs, I at thi stge where the qualy of the data rally ‘Ssese since is only uel it enables the engine tofu the Asian needs. This is abo the stage that acts ab a springboard 10 imate the demand for new data and beter method, So compu the dna jet 42 Computer physi property systems lowsheting calculations tend to havea worsiou ppetite fr posal Property estimation To mode! dvilliton eam one may request ‘nimats for chemi potential (or fps and frente 10 000 (Gat wally many more) nes. Dependiag onthe effency ofthe Programming, thew caleaatons could represent 803% or more ofthe computer time taken to. do 2 simulation. With thi reqiement (0 handle an enoemous number of request, the design of the pial oper estimation system must be done wih exteme cre All Sonesivahlepreprcesing sich mipht be done hae to be done Ii such too coal daring 4 mulation to have 3 request for the heat ‘capacity fs mistue Wigge off eaeution reaching al she way back to'the ravadan Mary. The steps lang to the physical property poramterUbary must if ata possibe be done once and forall fore {he simulation, and the point generation outines mus beable find all the data they ped in this ary. For ade efficiency the ata fr the emia! species actully in te proces should be extracted ilo 3 Smaller data ary fr wie during the simaltion. "igure 4.2 summaries th gnetal features of what one may expect 0106 Physical prope serie fies Figs £2, Goer eats fang pe oper te be supped inthe type of physial property Sta seve provided fr te wh lomsheeting mystery. Ft consider he central column of he ‘ingram, which seprsents the bus set of facies. "The simulation meses ofthe Nowabesting system need 10 make frequent eguasts for properties st spel tempetatures, presses and ‘onposiions These eal for data are wally made ina rigorous) ‘defined manner winch sindependeot ofboth the point dats generation ‘models and the particular components. These point generation outines reve the propery vals uning the selected methods tha base tit | | | 42 Compueried physical propery sytem 107 calculations on a et of parameters for each component as discussed ia {he previous section. 'At runtime its desirable to store only the parameters for those parca components inneined i the nation ands het of se ‘Seta normaly copied over ttm the permanent system paraters ‘etaban nto run-time dala locations. Ia peal Lownbeting podem {his involves collection of purtcula parameters for 5-20 components {tom databank with extensive sets of dts for 250-750 compound ‘Use of sch data system is very eas, provided tat the permanent systom databank has ches forthe component of intrest. This is ‘often of importaoe tothe wr tat the databuek extensive an nt restricted 10a small elas of compounds ort x particular sto data {eneration metho. Unfortunatly, the rumber of chemical species i ‘Enormous and expanding at apd rte. Accordingly, for wide apples ti itis necmury thatthe physical property ata serie OF the Mowsieetng ssiem embodies fair for the tiation of we supplied data rom the systems point of view the casest way to accommodate thie requirement infor th ser tobe able to create ha on equvaent of te petmarcot syeem databank by explicitly entering data in the sane format. This option isealy made availabe and allows te user to Dull up hs own private databank for we independently or in conjuction sch the stem databsa Daa supply inthis way normaly requires thatthe ase ha (has sccess to) expert in physical property data and possibly ako of| omputer wre, In these eeumrances he may also Miah Yo provide Gata im tabular or polynomial form for se by a propre tof {ntrpoative point generation rostines. This fact is shown a the op fight offre 42. Tn adn these facls for supply of data an expt orm for slvetwsge by the system, thee ae bo those option designed forthe ‘culation of the parameters used by the sites point generation routines. Two obvious entegorie ofthis type can be ented and ae Ince at the top let of figure 4.2. The st of these applies to the ‘coreation af a dats and is most commonly app ote entination| ‘of binary iteration parameters. ‘The second category difers from those discused above since it ‘eles im the mal, to those stations for which 0 ata oF only harsctrsng data ext In sich nse thie eval st of chartering “ata, of i absence, structure data, are ullsed to eximate ast of parameters ofthe spe required by the pont generation rotins, One otable spite example of this type of fay I the ereston of data ‘et fo petroleum boing actions rom information om average boing point and density.108 Physical propery serie foes Al the above facilites form pat ofthe spectrum of options that, sudivon to the permanent system databank, enable the engines to gt {he mot out of 2 Howshovtng stem. If we consider thse as comp ontsafa whole system, figure 4.2 lusts u way by which the vations ‘options together could be used fo provide weightings for creation and ‘pdatng ofthe parameters databank. a noe = = | ‘al pe oar om 4 Sin efor tl os Ss man, od A general physical property service ofthis type cou allow put from any of al of the soures for piles property data. The service ‘fneeded, merge all of hse pat to gts best sto data for the pial property parameter library. posbly by predicting the propet= ‘er vesulting from each tpt and patting ther ito a table of pews raw data. Ten these generated data cn be ited a they were det ‘er input. The data sours can be weighted so tht the wacr con fmphasize those soares tht he fels re more elale 443 Physical property eaeaions 109 Hiavng discussed te overview ofa psi property service the nest section wil seus rather braly the ype of propery cleans Fequired by Nowshocting astm 443 Mysia propery eaeuations “his setion wil sca briefly typist of phys properties which ght be provided with 3 fowsbeting package Table #1 Tit anys Bot most, ofthese. “Lite 41, Moeonmon pial pope se indigo syle me lap a ror pha aes ap el ive mot ha i ower os wg oy ‘Tho asters indicate thowe properties quite commonly availble to the engineer who may be wing hi own mode fo model spec it for his plant simulation. The semining we les commonly provides ‘A substantial amount of the calculations for fownecting package involves two-phase sytem of mixtures. Figure 4.4 ahi Stig the someaeliture for able 42, which gies & numberof the vapour gd ‘ulm services commonly provided. The composition ofthe it {ure characterized overall by mole factons, and foreach a ite 2 pluses by compositions y, ay Avot tta amount of F, with V bring the numberof meen he vapour phe, andthe numberof roles in che gud pase. We can stew as am eqiiram fash unit Plc ina seam spliting the Steam into wo Hows and then remiing110 Physica property serie fates We oto that the vale off, Land V arent relevant only their relative talus, cratered by the vapour ition p= Vi ate needed ‘Returning to table 41, we have partitioned the piyscal properties into thermodynamic popertis and tanipot properties, whith charac tere equilibrium situations and rats stantoas, Te thermodynamic ‘mopetes are grouped roughly to reflect their commen exigl oF Fe 4 rwopeie ma fr 4 mis F VL: amo (a) name ‘Sons miyyg f= Soe fn Ym prs an 6 = pa et (2 ape 8 a Te iyo mare “Te 2. vey epltv pope erly ne me tm coum = iw ary Se be ; ome, bea £ Bicces es Bee met te Set et, Beare, ne, Retail ae rat Seton nese ee meee scene ae purpose. The st group are PVT relationship, and the second are for Felatng the enthalpy and entopy af amature to standard tat forthe pte components. The properties ofthe third grouping allow one 10 Feit the sander sates of pure components to one another and ate ‘seeded only if chemical eacon cakes plac. Vapour-guid equi 43 Physical propery eaeations 111 sre commonly charactested by using the properties ofthe fourth st Surtace tension is «property We seem forced Yo It by iss ota bk propery ut rather an interac propery andi fen agued a sign eotelations. ‘We might aote that many addon and vse caleustions axe va ifthe pysleal properties sted are avaiable on request. Single pas mitre i charicterzed by i emperture 7, posure P, pate ndistor p (= 0 or 1) andi compositions = 1,2... We Fay witht supply the entlpy HP and 3 and find the soot Temperature (or example wen solving the eqhations for = eat ‘achonge) The piyscal property package wil alow w osove = Flt, Pos, (a0,0) lect. By wetiog gC) = Ho gl, P51, @) = 0 wwe have an impit faction in T which can be soled wing a Tbotfinding technitue, euch the bounded secant method described in fection 3-1 I this povided a a ostine, i becomes» well part of the paysical property package. Many ober general routines ate alo tet (1) tnd P given entropy 8, > std, all); (2) fad T given Sensis, Ps 9 and (ll). “The tworphe talcultions are nots szlghtforward at so many “diictiecan are. For example, what ifthe conditions given a inet Festina singles rather than a two-phase macure? Tas situation ts tobe tapped ane fequenly ose. The other ah es) trap a8 one ean caluate the dew pot temperstare soe ifthe ‘speed temperature is above and the mature ten ll apour If ‘ot one cau compare tothe bobble pon temperate and below that, {he mintre all gu. Otherwise st must be to phase, “The aisatc fash calculation is smal checked. Here one i djing both the temperature andthe vapor fraction 90st he beat balance and the equilrom relations Ifthe beat balance Insts tat the mite most be hte dani at de pol, then ‘imate all pour Simi the het balace atthe bubble pot Indistes tnt it mst be colder sl the mixtre mst be all iui tere, ap wae ath sorption which might Be fcladed in the physi property servic lsat for'a tre pase fash =o ig nd one Vapour pas, This serve mens thatthe vapouriald eqllbiam | (VLE) mores used to pred the chemi potenti a each species in| ch phase mut also predet the exsteace of more than one tid ‘hase The Gis fre enexy is at minimum at equilibrium 50 the Iodel mt predic lower Gibbs fee energy for thee phases Than for12 Physica property serve faces ‘wo and this comparison must be made oalow one to determine the umber of phases, "Asan teri the reader might devise sluion to the felling problems. Be sure to conidr the pose dfclice which ight are Do not worry about tice phase ocuring. (2) Find the toperatre ofa mistre given HP ad, (ll), The pase ofthe mixture snot know, (@) Three suoams fora countercurrent heat exchange ate known (Le their, Pep, and als value re known), Fn the conditions ‘of the rernining seam. Wil the hea exchanger work? (@) Fora compressor, the inlet steam conditions are known andthe desced outlet prewue Is speced, Th comprestr is hentopi Discuss how ta ealeslte the ole! stam condos. ‘We stal aow a some quton read othe problem ueionreaing to pati 1 DM! you ws te ata fash soi dss pec theo, ueston niag pron A ai ane ote the streams rove the ety Ofte Tenting sam, pre dpe are Kore, the outet pres fave an we apa eq the Servet of te alate fh eesti. It peste roe are Sport th eto sould gt mesg We Sl pth poten ere“ OF nce, you ate a he tempest mht hve Sr one mang the Seg moni? er seam coset thor den or hobble pin whine eashnge scene ay ‘cur, nd hs maya te ade apa by extn te et ond {i tcmpertces How dt you ec or te (You ne ough a " ‘Gurion tng poem 3. The revo proton shoul seve =r warming, Dio you thatthe eed frvce tue W ns al pout Gonprenot itdaret he let tenn eo pase. The Spe ‘eproach hore, given sn allapour eed to note ta the enya Ge inet eq the enopy ofthe outee. Thus we mee 1 fl tpty Oy Uadiag oct BA(0) =S-FAT, Pa 5 1D. 9= 1) 0 here F(T Py; al), able cation nthe physical ropes pack Ts beyond te intended scope of his bok ode farther into the catalan of pysealpropeics The atenp ee hu bee io ge an Aedeanding of their inporince and Row thy te Wo proses owoosting, Much weft formato one aot sented the bok by Res: Pmt aod Shore 107) 5 Degrees of freedom in a flowsheet ‘We have continually mention the degrees of freedom fra lowset. Tic wey df problem in flowshecing i specif the problew otha the lowsheetng propram solves the dered problem. Tied up ‘wth this problem roe of comely idling the number of depres ‘of freedom and nding a lepmate ments to satisy them. Tn this ‘hapter we shal investigate this problem in some deta, mos by Spec exampe, We sll it give the most isc ew af he umber ‘ldegresoffeedom, then exblsh Bow many mut exist fora proses ‘iream. Next we shall conser the mer which may oecut fora unit ‘ode, Unatetng with several simple un eyes nit, a ash un fn extent of conversion reactor, 2 simple How spliter and pressure ‘hanging devices, Wah these mod we wl Bald 8 small Howe ‘Smmrting them sid examine ts degres of feedom. 51 Degrees of freedom The simplest definition has already been given for the number of sdegres of freedom of 2 problem involving the solution of = et of ‘sustons I our protest ene win inependeet equations iam (ea) variables, then the uber of deprees of Heedomd sien 8 A on “This formula ie based on the iden that welbbshaved independent uations can be ued to sole for exzellyn variables. Thus mmo the ‘aribles must recsive ther wales from other means Before me en tive the protien: Whatever the otber means they snt be equivalent {0 writing sn ational tof mm independent equations. ‘We can corider the asic re for degrees of feedom the phase rule. The phase rule sys thatthe numberof degrees of freedom for 8 miiphaseyatem, wbere all phase in equi, i daenpt? 62) ‘where cs the number of componoas and p the numberof pass.114 Degrees of feedom in a owas “Tis fle applies only forthe tensive variables ofthe system and ‘sys aothing stout the extensive variables. Intensive variables re toe thick are quanty independent, such at 7, P, mole faction hd homicl potential Fatenie variables ae quantity dependent, doubt ing in vale ifthe sl ofthe system doubles, ad icude tal volume, ‘ofl mass and total enthalpy. It spot oo ieult to rove tas rule ‘ing our definition forthe degrees of fReedom. The equations model Hing miphase sytem are Noster ‘ton ‘Teena equim T2227 80 >» 6a Presure eqivu Po Pan ee oP EP. > 64 ‘Chemie potenti: By hyo @ 63 Relnng conical potato mesure quite Iy= AAUP d= 120) o 6 tat, a Lda oe fection vm sation Sythe othe REND oP 6m ‘Total n = 3p+2ep equations For each pose ne have the vaibles TP, Hy ty = 1,26 “Tis gives 2042 vilables per phase plus +2 ena overall variables Pend = 1 Zoos) for aot oF m= (42) pte42 Anping equation (5.1) then iver the phase re: ron =e-pt2, Wile the principle behind equation (6.1) is srughtformard, the appliation os a hm snp. 1s nt am cay task o wre these ‘fequations which motel uit operation even! the theory undetyng the equations is nt a problem. One often wales too many eqstons, not eling that one oF more of them canbe derived algebra rom the others. An example for he materia lane equrtone of un FFguee 51 ilsirates a eit ith to apt and one output (4 ig ‘mt for example) "We can write the material balance equations 5.2 Independent seam varabies 115 SabitmaFe oF JA Qowe (8) ther 69) ‘us the mole fraction sum equations Biot (510) Bart oan Bigot 12) ewe 5, Ainge mi we do thi, we nave weten one eqeaton 100 many. Equation (512) fan be derived fom the others by the following Sum equations ($8) overall CDR = Eds 613) “Then substituting equations (5:10) and (5.11 to equation S13) ves Ft: = Gy (14) sod his combined wits equation (5:9) gies ws equation (5.12). "Fara lowet the problem canbe ery felt and ons inion anor loge may easy fai. Ifyou ares heme enincer, aly ‘ocesary to rel your stoichiometry cours to rocopnze the uth of this oberon {52 Indapandent stream vartaler What ic the minimum numer of variables require to spel fall 2 ‘treat? We define «stream ashe flow of material between Wo wit {fleet The arables we soemalysssosate wih a steam are is {emperatre, presse, toa ow, eral moe schon, pas actions tnd phase mole factions, total ethapy, phase ontalpes ad forth How! many of tase ate aceded to fk completely the stream if we ‘ssn phase and coma egitrum exis?116 Degres of edo i a flower Intuon ghos wb the eoreet response ere. We aow, without counting equations and varabis, hat, It we speci the temperature, fresue snd inva! eomponeat flows, the steam must be fall Spected. We Koo this must be tue Because create sich a steam Inthe laboratory and fxs tempers, ts pressure and te somponent flows, the steam wil break into phases (if ii muliphased at qui) anno forth (We oul et into dfcly wih tomperatre {or pure components If for pare component we fx the presere and the Individual eomponent ows, and hen we chose tbe lempoature to be the folig point temperature, we ate at alos to state the Vapour freton fort In this ese one ea replace temperature wth ental, and then one as no such problem} The question now abe ws 10 ‘whether we an prove our inte response. Wecan proceed writing the equations ateded 19 mle our ston, Duhen’'s theorem (Piggine and Defy, 1958) provides os with a needed esl, and ts proved in just this way. The thorem ates ‘Whatever the number of phases, of components or of chemical reactions the equilrium sate of dosed stem, fr which we Know the total inal mance of cach component, completly detemied by two independent varbles"Can we casi steam asa closed system? ‘Wo cand soy taking fied amount oft (yi ow fore second) tn tapping tin contin. Beexane we brow the compostions ta total ow ofthe steam, we know the total nfl macs we have trapped The theorem says we need onl fx wo addtional ndopendent ‘arable, such a temperature or enthalpy 1o be safe) and pressure 0 Ts the equim rate ofthe system, “The proof fellows. We shal oor a particular system to redace the hance of geting lori the notation, Phase equim imple tht the ‘equations which we wrote inthe previous section to prove the phase ale bod tro. These equations ae equation (33) t0 (27), Added to hese te materi brlance equations and eeactonequilibriom equate. “The example stem we shal consider wil lie tbe = 5 species 1 Hi ry Br and HBr related by ther = 3 independent reactions. =H HB, 2 2nBe Iris convenient o rewrite those resctions inthe oem Reoetion SM snhotey re stoichiometric coetcens, are species names and is Oar (S19) 52 Independent steam varables 117 the number of independent reactions. We wil hive vy pone for Products ofa reatan (apes on the right and side ofthe chemical Fenction a writen) andy egatve for ceactant. For out ee reetons we get Ha =0 ran =0 2HB-H,-By = 0 For reaction 1, equaton (5.15) tly expanded is 18-2848, +08r+ Br = 0 2 4s = 0, ad ford Wie next introduce the conept of ‘extent of reaction’ denote for reaction f. We sy eath reaction | occuls to teat, and foot ‘ramp this means 2 moles of H dsnppear to form moles of Hy Feacton 1, 22; moles of Be disappear to forme, moles of Bes by reetion| 2, and ily mole exch of Hy and Tr disappear ofan 2, moles ‘Hr byrealon 3. An oveal bate alan on each specks ives whore np is the origins mambo of mole of species j present These (Sratods ae ofthe frm peal ‘Reaction eguilbriun occurs when the Gibbs fre energy G i a saisinus, Any varition in @'around the equlitre pot moet ‘heretore be 22, be Bone 19) cence CT and Pare constant this reduces to Saviny 6) However he dy re constrained to aie by the reaction slcomty118 Degres of fretom in a fowsheet sand matoil balances given by equations (5.16). We take variations of eqution (3.16) fr out example aad get Inger ten tie a of te on bya B oy Poh oe ean Sshting se int gen (5.1 sve fr or name (A) Hy -28 rg 8B) tip — 202) + Haan (25 3) Wy Aes Gln, 2 ae nares" Sic th etn re nen cach an vied nde Stn. To bet uli he eeticenteoch mas tee unto =9 an san Hiy Hon, = 0 on poner Bryn better 19) ‘These Took exactly lke our rection equations (5.18) with each ‘hemicl potential replacing its corresponding speces name. We should ote cuacly what he did ere. Equations (5.17) exprem the reton| ‘quilbium felaiostips we need; however, the variations ip the umber of mole of each af thee = Sspecee reno independent. They Imus Susy the teaton stoichiometry ofthe #3 independent rear tions Thus equations (5.17) donot gies te = $ reaction equllb- ‘um relations but rather r= an we have jt demonstrated “The nt equations we can write are the material balance relationships sameong the phases, ad these relate phase mole ations, tthe a {otal tumnbaraf mole ofeach compost each phase hat, molec ine can wt foreach component (520) _ | ‘52 Independent sveem variables 119 Equations (516) and ($20) ave the required. mateia balance ‘equations, and equations (519) the ceacton equilibrium equations (Callecting them ll ogetter we have Number of equations “Therma equilib: ete -te7 ’ Pressure equilibrium: Perea > Chemis potenti By By moog P= ve ° Relating chemical potenti! o messrable quantiles: bg = ls Potgi = 12-050) * Beh 2 cP feb ne Moe action um euations: futtgt eo thy et Bebo ’ Material alan: name, . Sgvmmn, ite ene . Reaction eqitriv: Boyd bt, oor , Preece , ren j=L2. ne ¢ ‘Toul = Ge+3)pte+7 equations Number of ‘ries Poreach pais = 1,25. 0p! Prep y J ve Get For each wie, = 1,2... map *120 Degrees of feeder in fowsheet amber of ‘arlbles For cach reaetion, = 1, 2,047 6 . ove: ne 2 n= Qetsip Herr v2 variables ‘Te number of dogues of freedom is then d=m-n “The total degrees of redom asocited witha team therefore 2in sdditon tothe mols flowrates for each component, wheter rection ‘ecu andor whether the steam api fto more than a sing pase ‘The varales normally specified fora steam are FT, total moar fw fnd-1 overall mle factions, equivalently Psd the ind ‘veal eomponcat mola Nowrats. ‘ore fowshcetiy nytem,reston i ot permite fra sceam. If into osu, the unreacted sreses becomes 9 fed to a eacor ‘nit, where eacion fe allowed, and the reacted fed (pethaps at ‘tuilbrium) becomes the reactor out steam. 5.3 Degrees of redom fora unit We shall conser several rather simple unitsype models ere to Stobsh how one susersine the degrees of feedom associated witha it A related development appeats in Rodd and Waton (1968), “Our fst unit wl be te snple mixer anit showa in figure 5.2. The ‘equations we sal wi fri ae the ftlowing new |e tine asestin nes rant oe 52 A single mera ‘Material balance for each componet: sgt naky (= bt oe rr 5.3 Degres of fedom for annie 121 Heat balence: ARSE, = Hh rssre balance: Py= min (P,P) Mole faction sums: cy Bg Bg Paya propery Hy = ll Paste oA) 113 il Ptyno vhs) Hy bill Pisin hy Wo hive writen +8 cqutins a the 3-12 wnknows 595g 2th Fy Fyn Py Py Py Py Ty Ty By eH He umber of degrees 0 eden ttre d= 2412-(8) “This nut coapns io into whch says the fed came reilly speed, th outfall spec Echt ren ake? degrees of teedom as we proved in the previous section, It would be somenien here fo wie ou uni model digrammatilly win gure 3 Fg 53. eat dara sea a ge 52 “Thee#2, es numbers onthe reams, indzate the numb of egres cot ecm foreach ream The nmberc+ isd the block indtes (c+2 relationships eat among thee variables whin thie wit We, however, jst wrote c+8 relationships in 3e+12 arabes whic snot consistont wih ths olan. The et decor of cedar ate he sae H+ 12-48) = 2044 e42)-(e42) = 2e122 Deqrer of feom i a fowsheet . “The itferene is that we defined our model sting more than the rinimame numberof arable foreach seas. Ip parte, we wed he i variables Sy Sy Sey Ply Hand Hy fn ees of the eest needed. We then ssh tree mole ration sum equations and the thre phys property definition equations. These variables and ea tions cancel eachother in number, ving the base block diagram of figure 53, which indents the "net umber of relatoneipe among tbe fenest numberof sre variables needed ‘One sonal observation that We are not at liberty to choose any cobiaton of 24 seam variables to be the delsion variables, that isa the oncs we spc by ater means Is ebvously permis 0 ‘choose the 2 variables fo the inpat seams a8 we can Fadl devise ‘means to use the model equtios to caeulate the e+ variables for the outps seam. We cou ao choose to specify s the deckions, 252 maiables of oe input ream and c+ variable, erlang te presse of te ott stream. The peste balance equation ax writen hove caant be reverse, in general, since Py is specie, and if Py= P, we ean say nothing about Peace th Is greater than, of ‘equalto,P, Also we enpot spect the toll flows ofa thee seats “These ows are related by an overall material balance (which canbe rium, the sor specifies (a) ter undependemt reactions to oeatwihin the rsctor by giving the stlehomtty fr them and (0) the extent of| rencton foreach reaction. From our dicssion of Duhen theorem insecton 52, we se tht we ate simply fading the regula ‘elaoasip hat wll ot bold Erm 0 tt.a.cor for a se of ser-supplied rection extents Gyghen 112 phase equiva esl assumed alte same egwatons esl with the exception of the above unde ofr equitions for rater equations, ‘The ett of varlables ae the sme, Many reactors fave Beating oF ening possible, x0 agin we sould allow heat to te put Id of be emoved, Two eations not listed when proving Dahen'stheotem can Bl be writen Hat balance: HE, HEA RAMESH where (-l the Ret of ection for reaeton Presrre tle: Pyaar “These equations intoduce two new variable, Q and 4P, and thus do tot aer the depres of freedom. Our asc block diagram model tart in igure 87 for this reactor mol isnot uneoemmon or the est of reaction o be unavailable hough te physi property package. Lo theory they cou ben pact they fay not be Therefore the see of this module may have to supp) Cctnintes for thems his would 27 adtona parameter othe ait parameter ino labelled with +2 in Pigue 57 Fe8ibudoonebo04P) Fig 57, Bsc harm ml ext cme etre 56 ; 5.5 Degrees of feo for a snis 125 “The fourth module we shall conser i simple steam spiter. The only task this nit cap hands o spit the Mow of team toe pats, ‘Weesn daw the base Bock diagram for diet figure 5.8. The nik parameter laos labelled 4-1, the number of fadependent spit ——=a re 5, Bas gran fr ge et pti. tton ane can seo a iter wt Te ow pt acting a the equation Bena 23-004 ut the gate related by the equation Soynt so only s~1 of thom ae independent. The (¢+2) equations comprise {hes oral low balance equations just gen, pus the oiowing (e+ 1) suatons: JERR gH REL 2 ged “The fifth unit-ype we shal consider sa counter-curont het exchanger ‘nodule Figure 5.9 ilosrates this unit type alongwith ts base block poe 52, Aileen xan ote i ected a Heck dra oe126 Degrees of freedom in« fowshet ‘agra. For this uit we permit the heat doy 0 be a variable which “se ache Fe 510. Spl emp nd ic re a, ‘eye opener parameter alowed, and this device will model a vale, and approx Iately model pump. A second pressrechanging deve compres- tor, an ilstated in gue 510. Thi simple sompresor hn he nit pfametes of 4P and ffiency permite. The effeseny i the rao 1 = ADIs where (41, the enthalpy change st constant entry and (gust the ethapy change actaly encountered “The att modsle we shall consider ism special one. Suppose he engineer analysing # owet wisest ad» speciation (involving two or more variables) ons seam. An example might be ry hich says hat the rato ofthe noe facton ofthe fit component ‘ream to the mole faction ofthe second component mst bs 3. This ‘requirement might ocsar when one wites the feed to neato 0 Bake Some components na stozhometrcraio.A more genera specfaton Set may comprise equations nvohng variable from ome te8 fy ‘tom another and so fort. Figure 811 gives base block diagram rtructur which wil handle thi requlreme 54 Degrees of feedom in fst 127 owe 1, Af a dna nto yn mi en ‘We stall consider this problem by anaying the simple owshet in figure 5.12, Of interest, beyond the degrees of freedom, wil be 10 Investigate how we might derive sation procedure forth owt, find thi topic wil be covered in chapter 7 owe S12, Aspe eet “The lowsheet hat a hgh presare feed stream of aeous component ‘A contaminated wit sal amount of BH mine ist wih #feqee ttream conssting mostly of A, and pases into a rector where =n xathermic reaction form Com A takes place. The sueumn i coled {o condense component Cand passed through a valve ite a fash unt torescted A and the contaminant Bash of aving = fay pare Co be witha au heli stream. Part of te eye i bled of o keep the conceatration of om buldig up inthe ster,128 Degres of fedom in «flwshet ‘The rs is pressurized ina compressor and mined, as sate eer, with the fed team, ‘We can dra ata block diagram for this owsbect, gut 5.13, and sirectysucertain the total aumber of degree of feedom fr it We Inbild eseh steam with +2 in the base Block diagrams in the previous selon, where crests the mumber of components i the — ae ASS : Fy oy AD Fer 51%. Baek gr fr he Bwnet i fg 5.12 “rile. Doyo pom ani or fone fe $1 E i ‘team i ‘he wes d= 4-48 = 1 eg eda 54 Degres of fedom in a fowtheet 129 sszeam. The fed steam has two components the cling wate stream has one, and the remaining streams bave three. The numbers of ‘cations fr the mer snd the eat exchanger refi this method of labeling. The numberof uations namie eau othe numberof ‘omponenisin the utet stream pls two. Fr he eat exchanger 29 the mumbo of equations sequal othe su of the variables i the two streams leaving, ths ease, te plus thre "The rats forthe gree of ireedom say ar ted in tab. ‘The lowsheet has 16 degree of feedom and determining his mumbo se obowly very stuipiormard6 The sequential modular approach to flowsheeting Section 2.3.2 ntodoed the sequential moda apprch oHowshect- ing. The base philosophy ofthe approach i that each uli mods ‘wten forthe spc wil cleulste all outpat steam values piven ales forall input steams ad forall equipment parameters. Also fundamental othe approach that recycle ream rarables wil be he literate varias at he owseet evel of calelatin. The unit modes wil bs slving the set of nonlinear equtions which model the unit and Wil in general alo contain erated variables internal “Toreview ome ofthe eas in setons 232 and 2.35, we can again consider the obvious advantage and isadvantages for thi approwch, ‘The major advantages rest om the very claret statement af how 2 unit modal wile written, The person or team of persons writing the fovsbecting sytem have very stalghtforward problem definition, "This stem Is mh les cat to rte than one which mas ante ‘very comple formation flow. The wer of the syste alo knows tht ‘he mast supply values forall tho ip streams anal the equipment fertmeters; no other optoes ere peried. Tas lnitaion bes the Pecular advantage of Being well doin. The unt modules canbe tlen with considerable effort put into them wo make sre they are Fobust~ thats they ca lost lvays sured in clealating the ott ‘ream vlus. This best ality may roger conser a tines ed hoc test ap avergenceand'50 forth. Also the routines can contin 1 cas) The unit has no farther outputs. Pat onto st of group and delete it completely from the problem. Continue mith forming stan Shing 1: (CDERA)=(¥0) se) Te sou FG ay futher apt. Pa ont op of fp ‘Delete from problem completely. = a oe Stig ts (CDEBA) os) ‘he poup (COPA) ha ose onits:pt attp ois of gous to dete om problem ample Sng vy ee, Ua e Jetio be comaired so wes ong ah Sting 2: ome @) Unit Has no feds to any unit wich til exist the Rowshect, Delete i and put at the top ofthe hist of groupe. The fomset i completely Aeleted so we stop Our Tit of groupe i List of groups H, (CDEBA), (FO), 1 Which sgrees with our eater observation. Tas algorithm is rea Programmed so partioning and precedence ondeing vey sah Forward and very a 6.4.2 Tearing a group of units Next we have fo determine just ho to solve the groups wth nore han ‘single ni in them by chosing the set of tear steams for each, For ample, the group (EO) can be voted by gucsing scam $10, ‘lelating ni G theo Fad geting an updated se ot vats for 10.1 inalso be soived by guesing SI clelting uit F then G and geting ee 64 Parioning and tearing @ fosheet 181 an updated sot of values for $10. If we intend to wie seeesve Substation the two sequences ae similar Sequence 1 StO-+G8Y PAE G YF Soquenee2 SIF 2G FG “The sowly catelated value forthe gised stream become he opto the next eration. Clearly the ony ilfernce which steam essed (o str the calculation, Ime intend to acelerate the convergence, then the sequences sre quite deren Sequence I: $1059 F acter GS F Sequence 2: SHE acer? POG. ‘We might choose between them by picking the sequence which puss ad accelerates the fest varies Tn figure 6-11 steam SLO has ve Sarlaler and stream S11 hat four varales 30 we would shoowe Fequence 2 ‘The troup (CDEBA) offer smc larger challenge. We know the Impl tear S13 mst become an actual teu. How many addional tears to chose is ota all obvious. Forex teat 83 and S7, using the computational order i gue 6.12: ter 813, ASSL tea 83, ear 7, C513, 55,86, D, 54,58, EST, B, S3 where he ‘aleulsed steam ‘ales for $13, 87 and S} are underlines. We noe bere thatthe sequence (ar S13, A) ensld have Been moved to jot before uni a which ese the ciclted value for 1.3 appears before ai152 The sequential modular epproch o flowshectng the tear 13 lathe sequoace, Hut caine this tar is implicit, wales halted for $13 in © are only inky podesing values forthe np to A. They ar obtained only tter the etective looks a the S13 values fiom C tose if they ae aero, and then i regueses he inputs Ao dive these tps from Cto ver they are ot wero, The SE tly torn ienpectve ofthe aequence red ete TS Ui the best ede o abe? By what enters can we fe the odes posible s0 that we can choow? Can an elective algorithm be Implemented to elect the comet order? These qucaons hae heen approached inthe Ierture, and algorithms exist for the flowing 1 Choose the order giving the fest umber of tear reams, (The ‘bea algsthm i probably tht of Barkley and Motard (1972), 2. Asign' weight to eachstteam, Thiswelght ea elec the expovted icy associated with eating the sre and may for example be equal fo the number of signiennt torn variable vals in each stream, Find the order which minimizes the sum of the weight ‘sociated withthe torn scam. Two algo to dle This Droblen are avuable (Christensen and Rel, 1969; and Pho and Lapidus, 1973), 5. Tear to give the best convergence characteristics for suscesive stron. The algorithm is that of Upadhye and Grens 1973), ‘and we sal consider iin more deta ere, Upadnye and Gras define decompositions of 2 Aowsbeet, that isthe hoosing ofthe ea seams s Belonging to wo class edad a Sonredandant. Their work angicsqualatvely and demonstrates that ‘noneedandant decompositions have beter convergence ropes than ‘edundat ones for the method of sucomive station, Fortis i 5 relatively easy to locate 4 nonredundant Secomperiion fr small roblem abog a rplcement rule they define. For arer probe fould become teas, but they deseibe a mote sppropriste algorithm 12 their paper for such problems. We shall apply the rule, since st ‘eatvly cy 0 the group (CDEBA), The algoithn follows: 1. ing any tear set forthe group 2 Apply the vplacment rl which tates: all he opt steam of ‘mia represented in th ea sta ea sein the sme family of {ear sce wil be produced i these output steams ste replaced by {he input stream for that nit {3 Ifasteam appears more han once nthe tear tt at any sep, delete albu one oecurence for The fama wth the repested team i ‘redundant one. The ew st isin a ew fly lt over with as the tea et rr 6:4 Partitioning and waring 4 fowsheet 133 4 Repeat steps 2 and 3 uni familys reached where no member bat repeated scam in sea ot gute 613 itraes an spptiation of the alps for ou example oboe, We sued with our obvered terse in gee 612: $3, 3? End $1, Steam $3, being the sole outpt of unc B canbe replaced by ‘he input steams toB, Stand S4, Abo scam 87 cn be replaced by $6 and 38. The nex replacement along the et branch of our ee” 5% fubuttate SI by SI3 bat th pver fo oecurencer to S13 inthe eat Set We sik one ou, leasing ws withthe tear setS#, $7 and S13. The feat of figure 6.13 fo enerate all the tear sets thin the family onainiog 8, 57 and S19. The complete tear ses which ate shown Wrth an seri have appeared slewhere and requis no further Shalps. Nose pines ie t's epeted govurence os team, ad hit fam is non-tedundant. It sicconive subsiuon were uted, say ‘ember ofthis amily would have the same coovergece carters. ssaay ther memberso we can now apply a seconds terion to slet mong the memes "We want S13 to be tom so we ean fok aman thou et containing S13. These ate ($4, 57, S13), (4,86, 8,813) and ($5, $6, 13). We ‘ould choos among these thre alteratves by choosing the one whch Iinimies the total ofthe ssocated weight ive i gue 6.1 Sis7818 34342 = 8 58.86, 88,S13 3444502 85)86,815 54442 " ‘This terion gives $4, $7 and S13 as our best tent set. The onder to calalte the ants found by deleting these steams fom the roup and Sppling the pattining algorithm to the rest. Since these are eechely th streams with 4 backward direction figure 6.13, we ean ‘Se the competational order is (ABCDE), “Motard and Wetetere (1978) present an interesting perspective on redundancy in tearing. We can define, fora iteduaible grou of is, “he computational fps which are inde within it Fatt we define & "fu sing’ which comprites a soquence of wits and steams (UG) SU), LEDs +9 OG SM. Ube $0) sch hasta) an output steam ro eit UG.) and a inpat Seamta wit Uji). Ancusap os al ae ae Figure 6.12 6 (A,S1,B, 88,6, $5) (lstsing example) ‘We define a unit op fr afl string as one where the fs niin the fll sng is abou tobe repeated. An example for Higire 6.11 is (A,S1,B,83,€,$13) (anit loop for fal string)64 Partoning and arin a flowsheet 155 ‘The next unit which would appese i uit A. We can define a stream Slog a ata sing withthe ont names deleted and one ca fay Se tht seam stings are onetorone with fll stag An example erved from our fst fl sting example (1,59, 85) (ream suing) “Thay ar more convenient to wet down, A wit op for stream string (ve are being ite careful here with someneltre) stream tig

You might also like

• Get Chemical Engineering Process Simulation 2nd Edition Dominic C.Y. Foo - Ebook PDF Free All Chapters

  

  80% (5)

  Get Chemical Engineering Process Simulation 2nd Edition Dominic C.Y. Foo - Ebook PDF Free All Chapters

  41 pages
• Systematic Methods of Chemical Process Design

  

  71% (7)

  Systematic Methods of Chemical Process Design

  808 pages
• Douglas - Conceptual Design of Chemical Processes

  

  100% (2)

  Douglas - Conceptual Design of Chemical Processes

  608 pages
• Conceptual Design of Chemical Processes - J. Douglas

  

  67% (3)

  Conceptual Design of Chemical Processes - J. Douglas

  619 pages
• Design of Thermal Systems - Stoecker 3rd Edition

  

  92% (13)

  Design of Thermal Systems - Stoecker 3rd Edition

  568 pages
• (James Douglas) Conceptual Design of Chemical Proc (BookFi) PDF

  

  50% (2)

  (James Douglas) Conceptual Design of Chemical Proc (BookFi) PDF

  309 pages
• Biobased Converters

  

  No ratings yet

  Biobased Converters

  23 pages
• Principles of Esr Spectra

  

  No ratings yet

  Principles of Esr Spectra

  16 pages
• Pretreatments To Enhance The Digestibility of Lignocellulosic Biomass

  

  100% (1)

  Pretreatments To Enhance The Digestibility of Lignocellulosic Biomass

  9 pages
• Process Flowsheeting by A. W. Westerberg, H. P. Hutchison, R. L. Motard, P. Winter PDF

  

  No ratings yet

  Process Flowsheeting by A. W. Westerberg, H. P. Hutchison, R. L. Motard, P. Winter PDF

  131 pages
• Conceptual Design of Chemical Processes (By James M. Douglas)

  

  50% (2)

  Conceptual Design of Chemical Processes (By James M. Douglas)

  309 pages
• Conceptual Design of Chemical Processes PDF

  

  No ratings yet

  Conceptual Design of Chemical Processes PDF

  309 pages
• [Chemical Engineering Research and Design 1999-mar vol. 77 iss. 2] Adrian Finn - W.D. Seider, J.D. Seader, D.R. Lewin, ,Process Design Principles—Synthesis, Analysis and Evaluation (1999) John Wiley & Sons 0 (1999) [10.1016_s - libgen.li

  

  No ratings yet

  [Chemical Engineering Research and Design 1999-mar vol. 77 iss. 2] Adrian Finn - W.D. Seider, J.D. Seader, D.R. Lewin, ,Process Design Principles—Synthesis, Analysis and Evaluation (1999) John Wiley & Sons 0 (1999) [10.1016_s - libgen.li

  2 pages
• Process Systems Engineering Chapter 2 Process Modelling and Simulation

  

  No ratings yet

  Process Systems Engineering Chapter 2 Process Modelling and Simulation

  38 pages
• Module 4 Structure of Process Simulator Week 11

  

  No ratings yet

  Module 4 Structure of Process Simulator Week 11

  70 pages
• Chemical Process Design and Optimization: Jan - Apr 2017

  

  No ratings yet

  Chemical Process Design and Optimization: Jan - Apr 2017

  34 pages
• Douglas - Conceptual Design of Chemical Processes

  

  No ratings yet

  Douglas - Conceptual Design of Chemical Processes

  309 pages
• Conceptual Design of Chemical Processes

  

  No ratings yet

  Conceptual Design of Chemical Processes

  5 pages
• Chemical Eng Design

  

  No ratings yet

  Chemical Eng Design

  28 pages
• Dominic Integrate Process Simulation

  

  No ratings yet

  Dominic Integrate Process Simulation

  5 pages
• Simulation Chem Eng Gil Ivan Copia-A+

  

  100% (6)

  Simulation Chem Eng Gil Ivan Copia-A+

  537 pages
• Chemical Engineering Design

  

  0% (3)

  Chemical Engineering Design

  9 pages
• Nlewis65 72 74 Summer School Workshop King 7 No 2 Spring 1973 Cee

  

  No ratings yet

  Nlewis65 72 74 Summer School Workshop King 7 No 2 Spring 1973 Cee

  3 pages
• Steady State Chemical Process Simulation A State-Of-The-Art Review - Rosen

  

  No ratings yet

  Steady State Chemical Process Simulation A State-Of-The-Art Review - Rosen

  34 pages
• ROSEN, E. M. (1980) - Steady State Chemical Process Simulation - A State-of-the-Art Review. Computer Applications To Chemical Engineering, 3-36.

  

  No ratings yet

  ROSEN, E. M. (1980) - Steady State Chemical Process Simulation - A State-of-the-Art Review. Computer Applications To Chemical Engineering, 3-36.

  34 pages
• BK 1980 0124

  

  No ratings yet

  BK 1980 0124

  34 pages
• Design of Thermal Systems Stoecker 3rd Edition PDF

  

  0% (1)

  Design of Thermal Systems Stoecker 3rd Edition PDF

  568 pages
• Chemical Engineering Design

  

  No ratings yet

  Chemical Engineering Design

  27 pages
• An Applied Guide To Process and Plant de

  

  100% (1)

  An Applied Guide To Process and Plant de

  7 pages
• Handout CHE F314

  

  No ratings yet

  Handout CHE F314

  3 pages
• SP Atm-Cap-2-3

  

  No ratings yet

  SP Atm-Cap-2-3

  79 pages
• Conceptual Design

  

  No ratings yet

  Conceptual Design

  95 pages
• Revision: Module Coordinator/ Supervisor: Dr. Oh Kai Siang Email: Office: B1.16 Extension No: 5098

  

  No ratings yet

  Revision: Module Coordinator/ Supervisor: Dr. Oh Kai Siang Email: Office: B1.16 Extension No: 5098

  49 pages
• Computer Aided Process Design and Simulation (Cheg

  

  No ratings yet

  Computer Aided Process Design and Simulation (Cheg

  45 pages
• E. J.Henly and T. D. Seader, Equilibrium Stage Separation Operations in Chemical Engineering, 198

  

  No ratings yet

  E. J.Henly and T. D. Seader, Equilibrium Stage Separation Operations in Chemical Engineering, 198

  384 pages
• Birla Institute of Technology and Science, Pilani: Pilani Campus AUGS/ AGSR Division

  

  No ratings yet

  Birla Institute of Technology and Science, Pilani: Pilani Campus AUGS/ AGSR Division

  4 pages
• Unit 6 PMS

  

  No ratings yet

  Unit 6 PMS

  16 pages
• CHEN 4460 - Process Synthesis, Simulation and Optimization

  

  No ratings yet

  CHEN 4460 - Process Synthesis, Simulation and Optimization

  40 pages
• Chemical Engineering Iii (Mastery) : Dr. F. Stepanek Che.301

  

  No ratings yet

  Chemical Engineering Iii (Mastery) : Dr. F. Stepanek Che.301

  23 pages
• FE Sample Questions Book 2015-14-24

  

  No ratings yet

  FE Sample Questions Book 2015-14-24

  11 pages
• Osf353 Cps Unit 3

  

  No ratings yet

  Osf353 Cps Unit 3

  48 pages
• Solid Works Flow Simulation 2009 Tutorial

  

  100% (1)

  Solid Works Flow Simulation 2009 Tutorial

  240 pages
• Syllabus Simulacion Revised

  

  No ratings yet

  Syllabus Simulacion Revised

  6 pages
• Chap 1 Introduction to CFD

  

  No ratings yet

  Chap 1 Introduction to CFD

  30 pages
• Introduction to Design_1c

  

  No ratings yet

  Introduction to Design_1c

  22 pages
• Conceptual Process Design MANUAL

  

  No ratings yet

  Conceptual Process Design MANUAL

  69 pages
• 07 - Chemical Process Simulation

  

  67% (3)

  07 - Chemical Process Simulation

  16 pages
• Paloschi-1982-PhD-Thesis The Numerical Solution of Nonlinear Equations Representing Chemical Processes

  

  No ratings yet

  Paloschi-1982-PhD-Thesis The Numerical Solution of Nonlinear Equations Representing Chemical Processes

  189 pages
• 4-uom4dvvr

  

  No ratings yet

  4-uom4dvvr

  6 pages
• Plant Design & Economics (2020-CH-243)

  

  100% (1)

  Plant Design & Economics (2020-CH-243)

  28 pages
• Plant Design and Economics

  

  100% (2)

  Plant Design and Economics

  311 pages
• Cl3061-14 Tut1,2 Ass1

  

  No ratings yet

  Cl3061-14 Tut1,2 Ass1

  366 pages
• Database Documentation: Natural Gas Information 2021 EDITION

  

  No ratings yet

  Database Documentation: Natural Gas Information 2021 EDITION

  54 pages
• Short Version Bio Based Building Blocks and Polymers 2018 2023

  

  100% (1)

  Short Version Bio Based Building Blocks and Polymers 2018 2023

  16 pages
• Acrylic Acid Novozyme

  

  No ratings yet

  Acrylic Acid Novozyme

  29 pages
• PP in Automotive Applications PDF

  

  No ratings yet

  PP in Automotive Applications PDF

  16 pages
• HFS Technology: We Make Biotechnology Work

  

  100% (1)

  HFS Technology: We Make Biotechnology Work

  6 pages
• Ceramic Membrane Market

  

  100% (1)

  Ceramic Membrane Market

  28 pages
• ZAVKOM - IPA From Acetone

  

  No ratings yet

  ZAVKOM - IPA From Acetone

  32 pages
• Flavours and Fragrances Basics

  

  0% (1)

  Flavours and Fragrances Basics

  11 pages
• Epa Mtbe Us Economics 2001

  

  No ratings yet

  Epa Mtbe Us Economics 2001

  25 pages
• 2nd Biofuel Geneneration

  

  No ratings yet

  2nd Biofuel Geneneration

  124 pages
• Wacker Chemie Big Picture

  

  No ratings yet

  Wacker Chemie Big Picture

  83 pages
• Patent Analysis Till 2007 - IB

  

  No ratings yet

  Patent Analysis Till 2007 - IB

  39 pages
• Trading Barrels For Bushels: Renewable Feedstocks

  

  No ratings yet

  Trading Barrels For Bushels: Renewable Feedstocks

  3 pages
• Trading Barrels For Bushels: Renewable Feedstocks

  

  No ratings yet

  Trading Barrels For Bushels: Renewable Feedstocks

  3 pages