You are on page 1of 72
2.2 Deve he mannan ey guts oa veo flowin tg fr, Show ‘taal ie comeiton eesone cin) ment, tn ey ca eo ina supe peer egal fom CHAPTER MATHEMATICAL BEHAVIOR, OF PARTIAL DIFFERENTIAL EQUATIONS: THE IMPACT ON CFD No mong cam be certain fi mot based som mathematic 7 apo some ete neg ‘ich il based pon the mathemati sm Va (125119) Mates sthe quem ofthe scene. 3 INTRODUCTION ‘A rosea rose is rose." as Geri Sten wrote nn patil ferential ‘union is paral diflereial equation i apa ie equation—or i i? inthis chapce we will emphasize the ansverrot real: We wil nd at beyond jist finding solton ta gven pata dierent quan, we must be avare tht ‘ch solons have mathematical behavior which canbe quite diferent fom one Sicumsane to anoter The sme governing ow equations, when solved in one ‘epi of ow fel can exhibit completly diferent solutions n another eon, ‘ve though the equionsthmelves emai identically the same equations is Jat their mathomaaca behavior tha i lfeent. This mysterious aspect of| erential equations was alinded 4 stoma in Sec. 211. Te purpose ofthe present cipro remove (we hope) some ofthe mys: 9s “The goveming equations of Mud dynamics derived in Chap. 2 ae iter intra forms sich a Eq (219) ebained deal om ait conta vole] ot Partial difeenal equations such as Eq. (225) obtined deatly om an Infintesimal id element) Before king up 2 stl) of mumeial methods for ‘hesoltion of thes equations, is sf to examine sone mathematical properties of patil differential equations themselves Any valid aumeral soliton of the ‘Suntons shel exhib the propery of obeying the enor mathemati pope ‘es of the goveming equations amine the governing pari diferemial equations of uit dynamics as Aerivd in Chap. 2- Note that in all cases the hiphertoner denatves ooeut linear, there are no products or exporenals of the highes-order dei: ‘ives—they spear by themselves, multiplied by coeficiens which are functions of the dependent variables themselves. Such stem of equations is elled &guaie linear sem, Fr example, or isd ows, examining the estos in See 242 wwe find that the highesvardr dervatves are fistorder, andl of them spear linear, For viscous ows, examining the equations in Sex, 2] we Aad the highestorderdertatves are second-order, and they vay ocr lina For this ‘reson, inthe next sation let us examin sme mae properties of sytem ‘of quasilinear paral fret equations In the proves, we will ech 8 ‘lasfcaton of ve types of partial dierent equations duce of which are ‘rooinered in fit dare ina the road map fer his chapters given in Fig 3.1. Here, we map out ‘airy strihtorward course. We wil cus tv sepraetechnigues for deter ring the lasicaton of paral erential unions the method using Cxmer’s fle, describod in See 2, and he eigemae method deste in Set 3.3 Bath these methods lad to hese reste We wil se that many parti dierent equations can be clsifed as ether hyprtlic, parabolic, or elit these ‘efntions aswel many other deals wl be given n See 32. Other equations ‘aot "mised" ype, Weil hen contest the mathemati chavo of sltons ‘of tse dierent clases of equations, giving examples for acta fan amie fw ‘“Gronoad esis SS) 32. CLASSIFICATION OF QUASI-LINEAR PARTIAL DIFFERENTIAL EQUATIONS, Fee smpliciy Je considera fui snl system of quasilinear equations. They ‘lnat be the flow egations, ut hey are sila some respects. Therefore, this eto sens at sinpliied cane "Consider the system of qua linear equations given blow uy, OF an ed (1a) D4 Og aap entrant e.15) where wand var the dependent varisls, functns of x andy, ante coefficients USS baba ented fad cm be factions of). ad Fathemeore, ‘hae conaimous facto of andy: e can imagine that wand represen & ‘tninous veloty fed thoughout the my apace AL ay given pot the Sor theresa unique vale of wand wwnigue alu of; moteoves, the dates ‘ivan ide Ow, Onde dv, afte vals ahs give point We could Stacnegomg ino tis fw eld i were st up io he laboratory ad measuring Boa the slots and their desiatves at any given pont However wear now ging to make a sane statement. Consider any point in they pln, sth a point Pia Fig. 32. Lets sdk the lines (or destin) through this point (any exist) along with the dervunves of o and » ae Ihdterminant and actoss which may be dscontnsous, Ths sound almost conta 0 Tworeantisic chances exis tough ach it in hey sla, The sje of equations gen by Ee Oe) and 210) 8 called hyperbolic. ee 720 Soe eye om) an 0 ae pote <0 The dae ln ae imaginary The system of Fay (10 and (3.16) is called elliptic. - ae ‘The clasicaion of quater para diet cqatons as cer tipi, poole, © Iyprblie contain he salsa cquaions Its tis classileation which hs oon the major focus of his section. These de {lass of equations have lly dient behavior. wil be ics sory. The Srgin ofthe words celiptic® “parabolic” or Ayperolc™ usd 1 bel hese xhuatons psmply a det sualogy ith the ease for conc sections. The general fxpaton fora cone secon fom analytic geomety is a thy to? tate sfn0 whore, tac > 0 the conic isa hyperbola B dae =the cone ia parabola B hae <0 the consi an elise ‘We souk! end this section er, as far a5 our purposes inthis Book ae soneemed However, the empltion to extend the thought this ection one mare Stop it too overvbciming, since i pertain to one ofthe classic methods i the Solon of compressible Row poslams the metod of cancers. Rear © fq). Note hat, only [4 were zr, then idx would be infinite. However, the deiniton of cirri nests tht ule be indterminat alg the ‘haters, mo infin. Ths, for ude to be indtemnas. |B] in Ea, (3.6) ost also be ro. Then, Ou i of the form a _|Bl_0 au 1B 6x a [4\70 y ely an indeteinit frm which can havea Fite ale, Hence, fom Ea. (3.5) fon d& li ha & ‘ I= ae ay 0 (0 |~° e4a) 0 ae a Expansion ofthe determinant in Eg, (14) yields a ordinary dirt equation In terme of du and dy where dr andy ae restrict to hold alg a characteristic line eee Prob. 31), (Since |B] ~0 ix a dict consequence of |4|~ 0 fom Fa G3} then whatever elation i derived rom seting [| ~ 0 must be ested 19 tok long» sharers ine | The uation for he dependent vars wan y which comes ren Eq (314 called the compat equation. Is an equation Involving the uknown dependent variables which olds ony along the charter ite lne the advantage ofthis compabalty equation is that iis im one Tes ‘imenson tan the orignal part diferent equations. ince the model equations treed inthis eto (Eqs, (la) and (3.1) se patil diflrental equations in tho dimensions then the compat exuation fem one dimension hence san ‘rdnay deena equation andthe "one dimensions log the charac ‘eton Since onary differential equations ae in general simpler to oe than arial iflrenal equations. then the compatibiity equations provide some vanage. This ads to 4 soliton technique fo he erga system of uations {ag @.10) and 6.10] wherein the chances lines ar constructed in the xy ‘pace, and the simpler compusbiity equations are solved along these chance, ‘tes! This technique is called the method of characterises In gente ‘succesfl implementation of the method of churn regues ot aay ‘tarsetrniediections trough any pia the xy plane, wit eifient compet ity equations applicable o each diferet characters ine, the mae ot harass is fil forthe solution of typerolc paral ifeevtal peo ‘ony This method i ily developed forthe soliton of ivised supers Regs for which the system of goveming Bow equations jp hyperbole, The praca plementation ofthe method of characteris request te of + higronea ‘inal computer and therefore may legitmately be considered 9 part of Cos However, the method of characteristics Is a well-known casa tctogue forte ‘plition of iis superionic flaws, and erefre we wll onside ang ea in this book. For more infomation, see Ref. 21 33_A GENERAL METHOD OF DETERMINING THE. CLASSIFICATION OF PARTIAL DIFFERENTIAL, EQUATIO! THE EIGENVALUE METHOD In Sex. 3.2 e developed a method based on Cramer rule for analyzing system of ‘quas-ingar equations in ode o determine the classifcaton of thos: spe However, there is a more general and slightly ‘more sophisiued meet sesing the clasaton of quas-iear pata diferent equations tose on oe igemalus ofthe system. This approach e developed in the rest socon ete Assumed to be familar fo roost junior or reir engineering and scene set Fora basi review of mats algebra, see, for example, Ret 22. ‘The eigemalue method is based on a diupay of the system of paral, ‘ifeental equations writen in column vector frm For exage It us eee {haya in Eas. 1a) and (3-16 ae 2 for simply, such that he equations become ou OO, OY aT Matai =O (6.150) 20 OE Or OF agree am 0.159) Deng Wa he cotunn vectr v-(} the esem of ution sen by Hs (150) nd 2.158 can be writen [2 S)+[h 4]g-0 9 cancer maa mre sya 108 . wet at an in Ba. 16), Mulipying where (A) and (Mf) are the appropiate 2 2 matics in Eq (2.1 Bg (17) by the ivese of [Kl we bane su 20 1 OF yh B.19) . OF sy " tions writen in the whereby dfiniton [M1 = (KE. With the system of ea inte lomo 19 te egies of] determine econ of een Tr ognic ale he cuton eyeball Complete equations ae elie. This statement made without poo see Rel Boe more deals sample 34. We wil in hi str ing an lin of pins tomas amc Cee nal rer evi ay Ae fa cope gs Ie ow Held ony Spy perro He Hes nism sch the Bo eth body 8 stall rls of ack, ab fe fete Mick eumber cir sone a spenone (a Hanne OF hypenonich he gneming cocina, omen a ee) ego can Be aM 4m om) my me a oa whee and’ rsa peta eles, mesure elie the festa ‘ele. For exam wala Ale B82, Mth fin Mh her: ce reo ‘Reco tuo G21) a nme th he Aoi roa Fo ‘Siam an Gah adm dn of ep ota Sores stp Dor Ree ap are S wn a ore he, we simply we thse equations san expe of system of gus Tee ns (0 a Ome py neers, {Sue re ine don emer need tp ae me Fi tt ea: Ho oe ity. 217 fn a ee et tS Cap Ey 6.0) ont (31) Remprety fat) nd Sty on [a eof oe eso (a) and 18) act tao Resting thee ahs ia terms of nd € a gin Ea 3.10) we ve (ae) ene, E411) yes ty _ ey aga)" ae ee) tora cctv sues phe ta pee Sse pen) Ga ed ay wget Sy PEER AEST, Ra SC aoe iS oe wr sgeepye enone onal tae Siete iba [ues] gee eet : 2 ag ‘where we {i} 1d ("op de of i el ft, ig [> 14a] The em fe oe ao [she] eet 1 “t=. He, ean sen ce eee OT eat ato 108 ele epee ete rae ees wreani-0 ss seo ne ty nat b= i600 aff 20) on (2) ys preci he ee sti iF (3:22 te ‘mans of ean pl the spe of he chat ns, Mere, Foe Srmicsudaboe tits {tun fom 6.4) gees al a and ‘Sete feouons gen by By 20) and (21) hyperbole I Me ‘hen fon Eq (22) he gear sel lag and he yom of gains it ‘it T itor he Cesta method an be aed clasp “ire gens ‘As il not in ths stn, hngs ne ot say 0 laut For some system eens the eiemalus my te miko bt el and complet the tnt ep le sys tater ypetole mr eli The mathemati ivr ofc sone ml flee me Gonsquety, plese tep in ind tt systems of paral lee equations Canola eeemencny paced inj ne of he Client of ype fara o clipe somes eos hve mit tev, metal Boe 34_ GENERAL BEHAVIOR OF THE DIFFERENT CLASSES OF PARTIAL DIFFERENTIAL EQUATIONS: IMPACT ON PHYSICAL. AND COMPUTATIONAL FLUID DYNAMICS. In the previous sestions, we dsussd the clascwion of pial ifeentiat ‘usin, leuding to the difaton of peel, paraboi and lip equations ‘Why doce care abut raking such diction? Wha ference des ake tems of te atl of fl dyaric problem whether he govering euations 106 sen son ema se RAAT NOD are hypetoi, unboli, tipi, af se ied atu? The answers Wo thse {estos are the suet ofthe present section. The snore om th nt at {Sch ope of equation has seem mathematic! bch and this eee Site phsieal bevior ofthe fw elds wl In tats ipl ta Sie computational metinds should bs ao er lng squats suit With he dient clasiestos This sabato hein CFD, ad t's Teson why we are dicing sich mates bale we ads ty pacar america chases ‘The maternal ebavor of parle equations i lengthy uk ‘wos eas car be nd in syste mht nok caro Bans Indep tn wl cy dscns itt pol mee eset fears of the behavior of hyperbole, parole, td eligi ya site egnon and wil slate this beh i eps of he ow ond Se anpuct on CID. 34,1 Hyperbolic Equations ‘To begin with, let us considera hyper gun in wo independent variables x andy They plane is sketched in Fig. 33. Consider a given pnt inthis plan Sine we are dealing witha hyperbolic equation, there are to real carctertic curves rough pot P thse ae ibe ale and ight roming characte, respective (The nomenclature “let and igh-runnng” tems fom the flowing idea Imagine tat you pace ig 3.3 onthe Moor and tht you sand! on pox facing in he gene direction. You have o tum your hed othe If 1 eo steric cave running out in font of youths the lefeunting char ‘steric. Simla, you hive fo tum yout head tthe ght 10 see the other athe ey ‘sa semen TR aE a NTL Tne 107 hurts running ot in fon of youths isthe ightunning characteris) A ‘Spiiane of thee characters that information wt point P nuences oly te ‘gun been he wo characerscs. Fr example, fing 3.3 we bed point P (ha pin ier if we stupa sal dirance at point Phen this disturbance is Tita vey point within pion Lin Fig 3.3, bu ont n that eon. I hs Sense, region Tt dhined athe gon of uence of point Now smagine thee ‘Sursteristies though Pestonded baka the ans. That portion of they xis ‘lich intercepted bythe two characters is laeled ab. This has olay ‘Hest on boundary conditons for hyperbolic equations. For example, assume tat ‘Boundary conditions ae specified cn the y axis (= 0). Tat the dependent tavables wand vare Known along they axis Then the solution can be obtained by “raching forward nthe distance sting from the given boundary. However, the soliton for and a point P wll depend ony cn tat part ofthe boundary ten and ba shown in Fig. 33. lafamaton a pote, whichis ouside the Intra ab propagated along characestcs ough cand infeaces oa region inFigure 3.3, Pint P isouade repo I and hence does ot fe! he infomation from pont cPoiot depends on ony th pat of the boundary which sineeeped ty an included between he to oes ease ines hough point Pat ‘intervl ah For this easo, te repion the lof point egon tin Fig 33, ‘eile the domain of dependence of post tha properties at depend only fm what is happening in eon In tems of CFD, the computation of fw elds that are governed by types equstios is sc p a “marching” sltons. The algortm is designed ‘ost wih te gives iia coions, sy the ax Fig 33, and sequemally Salelte the how Held. step by sep. marching. inthe dretion Tn aid dyranics, the following types of flows are ovemod by hyperbolic pial ferential equations and hence exhib the behavior described above STEADY INVISCID SUPERSONIC FLOW. If the flow is twodimensona, the tori ihe tht aleady dicated in Fig 33 Imagine a supersonic ow rer a ‘me-dimensinal clara sro ae sketched in Fig. 3 hes can beat Sng of tack 3, but mst not beso large aso caus the edig-ede shock wave fo become detache, or ee thre wil be pockets of lealy subsonic Now. (n 2 Steady Row field ay. pockets of subsonic low wil be governed by elipic ‘Suntons andthe downteam marching procedure crignaly established fr the “hin ofthe hyerblieequaons wil be mathematelly posed the com ypuer progam wy vblowap" under such conditions) To be move spec, ‘eft nin onthe Euler oqusios given in Se 282, Ley Fgs. (282) 9 286) When wnten for seady flow in ether conservation of nonconservation frm, ‘hese equations ae Fyperbolic when the focal Mach number i sapeson (se Example 31) (A eases point is Eq, (3.20), which is deed from the Euler ‘suse forthe spec cise of iotional ow with sal pesos. We ‘joved in See 3 tht this equation i hyperbolic when Mf > 1, A mae general tals ofthe eigenvalues asocaed wid the general Euler equations fra steady flow demonstrates tat this sytem of equations is hyperbolic a every point wer the Laca! Mach number > 1 Hence, a Figs. 34, whee the ow s assumed o be 108 sure ssa fen aE AEN EAM OP locally supenonic everywhere, the entire flow field is governed by hyperbolic ‘uations. The general Row dictions the x decton. Tero the ow Bd an be compute staring with given inal data at some location inthe Naw aed ‘hen solving the governing equations pumescly, marching stop by se. in the general diction downstear ofthe nial data. Ts locaton of tenia eine 'Sinvluoeed somexbatby whether shock aputing or shock iting is ing used the clelton, Real the cussion of shock capturing and shock ting in See 2.10) Ifshock caring is being used line ab uptea ofthe boy canbe usd ‘he initia data ine, were the ial ata are snl frestream condons long ab Ie shock fing is being used, ines cd and fs downsteam ofthe none and reaching across the ow fel rom the Body surfce to the shock sue, can be ‘ed th inal dats ines. In thi ae, he inal ata ely specified alowed ‘orf ae that associated wth clei slistion of he oblige sock flow once a ‘edge, with wedge anple equal to the body angle atthe nose flv to the fresteam diction. See Ret 21 fr auch casa wedge solutions The esl of hese class solutions yield a Set of constant properties along and another silent st of costa properties along fn ars, the remainder of the ow ikl in Fig. is caleulted by marching dowesream rom tes nil data lines. These raters wil be made clearer when we discuss aa ppiction in Pat Ho this took. To extend the above discussion 19 sreedimensiona, sey, supersonic tvs ows, consider the pict shown in Fig 3.5. ths tre-dcosioal 2 ‘pace te characte are sucess sketched in Fig 3, Conse point Pat ‘Even (2) locaton. Information at infaences th shad volume contained ‘within te adhancing characters surface. In alto, ithe plane isa nial ht suri, he oly tht pion ofthe nats shown she rosshatched ares. cna ton es aa on 108 2 Fan ul he on of hype: ego, Tse ety fm in they plan, intercepted by he reeating characterise suri, has any aft on inf. 3.5. the dependent variables are saved by Sarting wit ata gven nthe ye Plane and “marching” in hex den, For an nici spersonc ow problem, {he gene flow disction would abo be isthe x diction, LUNSTEADN, INVISCID FLOM, Examine again the Euler equations sumarized in ‘Se 282 Ifthe me derivatives in hee equations are ite, a would be the ease ‘ot an unsteady ow hen he govemingequabons ate hyperelc, no mater whether the fois loelly subsonic or supersonic More precisely, we sy tht such flows art Iypetolc with respect a tne (The classi ofthe unsteady Euler equations {byperbolic with eect ome ir denwed in Sec. 11.21.) THs ple hatin Such unsteady flows no tater whether we ave one, to, of thre spatial Sinensions the rmuching diction is always the time dvi. Let us exarine tt ine The ineral abs the oly potion othe ntl dat log the x nis upon ‘which the soliton at P depends. Extending these thoughts for twoshmension Unsteady fon conser point in the yt apace as shown ip Fi. 3.7. The region Influenced by P andthe porn of the Boundary inthe my plané upon which the olution at P depends are sown inthis gure tating with Know inal ata in they lang, th sotion “marches” orward in ne. Indeed the extension othe ‘Giensonal unsteady Bow fs made inthe sme fashion, although we cannot easy ‘daw a akech ofthe case since we are dealing with Tour independent variables. In the cn, he fl tre-dimensional Euler equations summarized in Se. 28.2 ate ‘aed. and the solution fs sll avehad forward in tne skp sd Pundes fr te lino pete euros. On mens tly fo san bp tn ‘einen nea ‘When do we encounter unseady,imvscd fw? The classic ese of one mensional wave motion 2 dic ae example: ee, We ar aly infested in the rani aristons (ae fr example, chap. 7 of Re. 21), The wo-dimentonal ‘unsteady Now osera lapping or plunging ails another example However, by far the most commen use of steady tme-marchingsoltons in CFD 16 0 ‘imately bina steady Now result the fina of lage tines, as long 3s the ‘unary condone ae time-invariant Here, the time marching smply 9 means tothe end~the en being a studs flow fd At fis glance, his would seom inefficient Why calculate a steady fow by going fo th oube of introducing ime itanoter independent arab? The ansner ie that sometimes tis the on Way {o fave wel-posed problem and hence the only way wo aban the sendy ate “ovuton conpuaontly. The soliton of the supersonic blunt body problem (escbed See, 1S tone such case We will se many oer examples of ‘he approach i Pat 342 Parabolic Equations Lt ws conser a parabolic equation into independent vale x ad The xy pine vs sketched in ip. 38. Conse piven pont Pints plane Since we ate ‘eling with» parable equation thee oni one character dection ough point P Furthermore, in Fig. 3, assume tat ital condons ae gives slang tbe Tne ac and that boundary conditions ae known along carves ab and ed. The hracesic direction is gen by a era fine trop P Then information at influence the ene region on one side ofthe vertical charateritc and consned ‘within theo Boundaries: Le we jab P with a eee the efit of hsb set ‘hrouphou the shad region sown in Fig. 38. Parabolic equations, ike hyperbolic Dron bd bmi Gr tn ole pa to en 112 sore ma maT ren ON oo tion a icusedin S41, nd temas marhing solos. Stig ‘ith he inal line a, te auton betwen the bounces of sal a Sbsin ty mange penal ecto. eens cata es lineson tigre sv cnn ig 39. He te paso aes hats npnde rates a= Cnt on Paced ae Aa ht anion he ea oe noah contin pc slong te a stones se aed and fe whch exten nh gee eco ey Fs spree the ntl ds rice Then, fron Fintan be ce oc sino oie rh of conan win fe und arses ae Crome in Fig 39. String mth he mid ne sed en ‘ede ern Aus a eal te toned heres mang Ope slates, ange ae hyperbole ‘equations, “ types fd nai Now ele greedy abo ein ‘Before answering this question, recall that the whole analysis is aed upon the: vem ean deed Chap. thea pal ar of nh oe Sieh tn Thue ein ca a fe on vans Smif (and aly pone) fro te Ree, guns have been we Sepening oe ace Row Bed Se ‘htowg he Nave Sikes Squat heats exhib sae ue ‘chim of he apocina forms daned fom he Noe Ske a pant eqnion There, when nek what Qper of ad dpa fie aegoemed by paleo, we ely ne ake *oooxinatfowld modes se peal Wy pba eens nek wer implicit following similar pilsopy in Se. 3 4.1 dealing with hyperbolic ‘Suns, sn the example ven the to subsections all erie othe Euler ension, whch ae simplified venion of the Navier-Stokes equations teen applied oan nisl low) we delve eo the vanows approximate fms of The NavierStokes equations, then the following types of Rowell models are fsered hy parole equations STEADY BOUNDARVLAVER FLOWS. The concept of dividing 2 general Bow Fld nto to epion, (0) ahi per dice to any solid surface wherein all he ‘Yscous eet ae contained and (2) an nies Now ouside this hin vious ye, ‘reson ofthe most found developments id jmamics. twas presented by [aie Panda the Third Congres of Mattias a Heidelbers. Germany, fe 904. The thin vicos lye agjcet oa surtac called a boundary ayer is ‘sume that you have bee rode othe Sea of Boundary ayers if 4, you te efered othe inteducory dcusoon cap. 17 of Re 8. A schematic f the under ayer on genre serdymanve body is given in Fig, 3.10. ier the oid assumptions tit is boundary fiver sha thatthe Reyols number, Re based on ody length Lis lage (Re ~ pV xLit), he Navier-Stokes futons reduce fost approximate set of equions call the boundary-layer ‘uations Sie to yore thst the boundarylyer equations ae porabalc iid ioe nie ent, ae accurate) fashion the flow in the thin shaded region sh ne varios ort SS eal Me ari ssc engi Sreseomcnetcm sects {ayer equations ae solved by marching downsizeam inthe sdtetin, whet ss the distance along the surface of the body measured fiom the nose, as shown in Fi 31 A feos pen hwnin e310h Heal coo resect terete 0 nae SSratie ede ese ah sn = megane atari aie aioe necae Cane cand ennt e mais sciparmntsaeted Cr ai ere mary apc mney 23 FEA. te Maes rr ete spent by Zs Co eae amo a ner tre oe ga ge ao SOR Sarason Saami cn [eS ABOLIZED” VISCOUS FLOWS. Wht happens wen he bu is otha when he ete fw elie ly ves? Aneel sown Fig 3, whe appeal es td sy ods ues low enough, vic fs wl each wl te fl irmay fom te src. Inde te flow fel ewe tt wi ad ‘the body surface might be tty viscous For thi case a Boundary -lye saution is {tappoiate: the boundary layer equations ae not valid for this ow. On the ther Tan ifthe ow Hel des wo exit any repions of ead reversed separated fw inthe stare decom, sll anther spl version ofthe Navier Sthes equations nay slice For example, al he viscous tems in Es. (2.580) {0.2.38 and Eq (21) that involve deviates inthe seamwise detion (sch terms as (Os) GW = V+ 2p Oud), (009) Ovo, and (Ds) OT) are ‘Sued fo be salad can be neglected, andi the flow is sumed tobe steady, ther the tevuing cqutions ae called the parabolzed Navier-Stokes (PNS) [Sains This is because the reeling sinpliled vein ofthe NaverSokes ‘Rintions exhibit parabolic mathematical beavie. The PNS equations ae derived ie Ret 19 and ate displayed and Gacused io Re. 2, to cite jst two of many utes. The avarges of the PNS equations are (1) hey ate simpli. contain {Gs tems, than te fill Naver Sokes equations and (2) they canbe sled by thea of downstream marching procedure, On the ote: hand, erase dhe vneous {Gms involving devaves i the Row direction have been neglected, and hese ‘ervainesrepesent the sical mechanism by which informations fed upstream ‘cto viscous scion inthe ow hen he PNS equations arent apropi othe Catcalaton of viscous flows that involve pins of few separstion in the Strammise Greton. The isa severe limitation for some applications tn spite ‘ofthis drawback, the dowasueam marching aspect ofthe PNS equation is sch a ‘Sompallng advantage that this methodology si widespread use. The type of wel: Fehaved fly vneos ow kethed in ig. 3.11 perfec fora PNS slaton and the aturacy of sich solutions is ustlly quite acepable LUNSTEADY THERMAL CONDUCTION. Considers stationary Bui (quit or gs) ‘nbee heat tanefered by themal conduction, Moreover, assume that the Cirpertre patents inte fd are changing ar fenton of time; this ean be Iimpined as due fo a time-varying wall temperature, for example. Athough this ‘rample isnot «fw per se, the govesing conduction eat tansfix equation is ‘hy obained fom Eq (2.73) applied othe ease where V ~ 0 In this case, Eg ae Emi 2 6) 306) -208) 00 urea, if there sno volumetric Ret addon (j= 0), and assuming the tate reition =e. Eq, (0.25) becomes SAI ($0200) 0m gation (3.26 isthe governing eatin forthe timewise and patil variation of T throughout the uid parabolic wath respect to time, which allows 8 ime 16sec sence ma en ee OE a Sl ene epee ee ea constant, then Eq. (226) 8 Weiter as Ben ws aes sees serum tet soi ania cian ame Crom = ees th and wi hel ad a = TT wil ba uy hangin See ee ci ot te ne aves nl pees Be ea hie ae Poed Ey 27 me sce sa Soonioa on_ or aoa Sc into dbo of T ema ase n i212, ing Sse tae ial emery Bad ten, popes eo ce ee ere in sextet derby SOE anid 2 ha te (028) 343. Elliptic Equations ct us consider an elite equation in two independent variables x andy. The sy Pane is shetched in Fig 3.13 Real fom Sex 32 ta the characteristic carves for Uhl eqstion ae imaginary forte most part, the methodology associate Ati he metbod of characte is Beefre ales for the Slutin of elipic apations For elie equation, thee ate no mie regions of infuence or ‘limams of dependency ater information is propagated everywhere in ll Ciretions For example, consider poiat Pin the 3 plane sketched in Fi, Sib: Assume thatthe domin ofthe problem is define as the rectangle abcd Shown in Figs. and ht Pleated Someer inside this hosed domain. This isatealy in Contato the rather open domains considered in Figs. 33 and 38 for Iypstolc and parabolic equations respectively: Now assume tat we ab point Pin Fig 313 witha nels te, we introduce a disturbance at point P. The major huthematcl characters of elite uations i that this disturbance is fet ‘Cepyere thoughout the domain. Furthermore, because pint P inlucces pti the domaine in um the solution point Pi inuenced by the entre osed boundary ahed, Therefore te solaion at pont P must be cared oat Simaitaneoly wih he ston a all other pits i he Join. This en stark anasto te "machng”sltons gemane lo parabolic ang hyperbole equations. For ths reason, problems volving elie uations ate fequeny called jay fpoblems,beonse the soon within the domaia depends onthe total boundary ma dns fre on ft eis Si iw pp fe ie speng Boo ho ante ite ry a in, oun conditions mus be spi oe th etre ound abd Tse ‘boundary conditions ean take the following forms: = 1. Aspeieation ofthe dependent arial and along the boundary. This ‘of foundry condition elle te Dischi conditon, nn MOF A specication of derivatives ofthe dependent variables, suchas Quy, lon the boundary, This pe of Boundary condo called he Nemonn cones, 3A mix of both Disc and Newman conditions What pes of ow ae gover yi ston? We wl ans ‘such flows, a8 follows, il tae STEADY, SUBSOME,IUISCID FLOW. The ky epee wad ee be sinc" tn soon Hodson ohh woth es oe fc) an ysl wr hry peta ra a ty wee, 4 ine dtc nvidia ao seen dae es, thems costo, oman finn wt pops my nal ae For ecu, many fu afr ite Sanit pcs ae flows overs sho soa he Pap n ig Ss Nate Be stains no he irl ie deed ua an thos eid ie te {deed ownnard Te dons nied by te Pen of tesa inthe sae ee oughon ect fow el teats fr parca Ewe 8 al peacoat ihe ment eee a ct eqs ivi ow gems ty he ake eqns pe 22 280) ntun ets fee 7 ane at en Ele ans flip nie the os Mach safaris ow thn i te anahe y ieee, the pesece ofa abn, isd ow sol Be cessor hot te fo nd Fi 3a an cumple of eh ee INCOMPRESSIBLE INVISCID FLOW: In allo, an incompresile fw i 3 limiting ese of sso Row wherein te Mach number goss (The Mach umber i dfned se =m whew se pel ef od ne ee tree compen Row, he compres hoe, koe oe found jira b inayat oe oe pe ee ot Therefore, m0 suis that an incompressbe inviscid flow i govermed by ‘pic equations: adc, sac fame ae the “queen” of lic Deaviog All the ‘Sor stead, sbsonc incl fe desl above cas overt the se {rincompressble inviscid Bow, and with a song effet a ht 344 Some Comments: The Supersonic Blunt Body Problem Revisited ‘One ofthe most important problems in moder high-speed aerodynamics the ‘Slaton afte invseid flow over supersonic or hypersonic blunt ody. Some hokground on this problem was provided in Sec 1S, where the diicalty ‘Ssovnted wih caning 9 soiuion for the mised susonicsupersoncslea’y hw ws underscored, would he very pertinent to reread he ast hal of See. 15 efor propessing fier. Also, tm back to Fig. 130. There, along wit the fasted tex, you wll finda discon ofthe mixed nature ofthe supersonic Hank Bray ow fk, where the locally subsonic Now ented aban elite region tthe cally supersonic flow eed asa hyperbolic eon. The problem in Sbtvng this steady sis How fell is dae ene othe extreme ical in ‘hung solton techngue that vai pba eon. Now, without vanage pnt afer our dieasion ofthe mathematical behavior of patil intial uations nthe present shape, we can fly understand and apreit the source ‘hihsicaly Bocuse af the totaly diferent maleate! behavior of elite “Sh hyperbolic equtons the sudden change tre ofthe Euler equation oss ‘heson ine virally presley praca steady Now solution of the bunt body ‘robe involving uniform ete of oth he subsonic and supersonic exons. Flomever, in Se, 5a breakthrough inthis problem as mentioned, which tok place inthe mide 1960s, We are aw in poston undead theatre ofthis Freaktrough, Recall fom Fig 3.7 that unsteady inviscid. flow is governed by Iypetolie equations no mater wheter the fw i ll subsonic of sopesonic {Tis provides the following opportunity. Staring with rater arbitrary ‘nial onions for the Bow field they plane i Fig 130, solve the unsteady ‘wo-dnensiona inviscid ow equations, marching forward in tine as sketched in Fig 37. At large times, te sltion approaches asad ase, where the time ‘Setvates of the flow vanables approach 2. This steady sae isthe desired Fenland whit yu have when ow aprech his steady ste fx solution forthe “ue flow Hil inlading both the subsonic ad spesoic regions. Moreover, his Solution i obtained withthe same ifr method throughout he ete flow. The ove dscasvon gives the elementary ploophy of te time-dependent echnique forthe solution of ow problems x praca! numerical implementation by Mort fu Abbot in 1966 (ace Ref. 12) cone the major cee Breakthrough foe the solution ofthe supersonic bint body problem as discussed in Sec... At Fist ‘lnc te use ofan aon independent varuble, namely tne, may seem he Ext baggage bu ning soul be fre om the rh Witbout rogucing tine ‘San nlependent varie the problem cannot be solved. BY intoucing tine an independent varible- the goveming Esler equatons become hyperbolic with Feapet to tine, tas allowing srsghtforward marching solution in time, with Ther proper steady flow results appearing inthe limit of large times. For the Dh 120 ua some an NO ED body problem, his steady state obained a lage times i the dese wet the time-marchng procedure is simply a means toh end. Hoe lsc example the importance of wndestaning the mathomcalbehaver of varios pes of arial diferntal equations. Inthe ban boy pcblem, an cleat sppeation ofthis undestanding nally ested na racial ston, mheees none existed oor ‘The tine marching approach decribed hee, where the final steady sate at luge times isthe primary gals widely used inser CD fora whole Bot of | dierent appcaions—it i no way unigue tothe blunt ty problem. For ‘example, the mathemstial bebwor of the full Novier Stokes equation’ fr steady lw isnot easly plead in a single calepoy. Rath, the Naver Stokes ‘equations have both parabolic and lpi behavior The parabolic behvioe 1s through the time dervatives of velocity nd ncn ent. much inthe sae ‘manner asthe feat conduction equation, Eq (327), is paabic vi the time erative of 7. The paral eliptic behave stems from he ica terms Which Provide a mechani for feeding infrmaton upssem in the ow. However, i ‘pie ofthe mixed nature ofthe Navier-Stokes equations, te-marching slain is well-posed: most ofthe exstng aurea solton tthe fal, compressible Navier Stokes equations use the tne-arching meiology 35 WELL-POSED PROBLEMS ‘We end this chapter wih a efnion but a definition that we asin 9 positon to appreciate Inthe solution of pai diferent equations it sometnes easy 0 attempt a solution wing incre! or ines! boundary and ital como, Wheter the soliton is Being attempted anaytally or nomena sch an “il posed problem wl usualy lado spouses a het and no soon t work. ‘The supersonic blunt body problem discussed show ws lassie example, Whe onsidenng the mixed subsone-sueranic ow fm a stead flow pot of ew, any attempt obiain a unformy valid olution procedure for both eons Was posed Therefor, we define a welposed problem as follows: I the soon to rial diferent equation existe and (unique, and tthe solution dopends entinwously upon te inal and boundary condition, then the problem swell ‘posed In CFD, Hi imporan tht you eabish that your prob swell posed before you tempo carry ou a numerical solsion. When the Wut body problem ssl up using the unsteady Euler equations, and ine marching precede Was employe to go the steady sate a age times starting with exentally biey ‘sumed inl conditions a time r= 0 the problem suddenly became Wellposed 36 SUMMARY amine spun the road map sketched in Fig. 3.1, and tink bout the rather stnighforeard course we mapped in ore to dsc the moathematcal behavior of ‘various types of partial diflrental equations, Tere are two standard metodo ‘deerminng the mathematical bebavior of given equation: the Cramer le cis approach and te sigemalue mabod. dsb in See 32 and 33, respect May ston can be Gly iid a siher Foe para oF Slip: hes sch ahs ustey Navi token qution, have ied Pea ‘Tie major matematlbenvr of hyperbole and prs equates hat ey {End hemseler gute wall 0 marching sltons, bean fo kaon ta tna ple olin In orton lip eanion di not For eli equnons he he varies tien pst be solved melanesy wath the ow rable a al tet poe OF couse tbe dierent mathematica Behavior of Sipe compare to prt Sud pea eqns adie! efletion of he {itt prysica! tebe of the Rows dsenbed hy these equations ia, we ne ate presen cape ting ane at 1 is took We hve examined and dered sme of he ane hgh nd gations ‘Stn tan ucoundaganaplcton f CFD, Weare now dy to ma on ton enplass on the maneal spots of CED_thesabje of Pat I PROBLEMS, Ss. By exung the eet i Eq) ein he compaiy 4, thn the eoe ll progressively ecm lags and il evenly case the numerical arching Solon o blow ‘pon the computer “The above analysis an cxarple ofa general method calle he von Newman taht method. whch used frequently to ty the stably properties oF Kneat dierence equations “The act fr ofthe stably rteion depends onthe form ofthe dierence unton. Forerampl, lets briefly examine the suit characteris of archer ‘mpl equation ths time Hyperbole equation. Consider te Ristorder wave ‘sunt (ce Prob. 33): (478) Let ws cele the sail derive with x cnt dilference Ou _ wy =u Bo Tie Ie replce the ime derive with simple forward erence, then the resaing Aiference equation rpreseaiag Eg. (478) would be (429) 7 (430) ‘This isaboutas simple adiference equation aca be btained fom Ba, (478: itis Sometimes called the Euler exc form. Hower the aplication of the vo ‘Neumann stability arts to E94.) shows ht (48 eds an ata selon 0 mater what the talue of Ar it—Eq. (480) is therefore called wcondiioaliy ansible. inset us replace ihe time deatine vith @ Aistoner deren, whereas epesented by an average vale Between rid pointe #1 Tandy te un =H +) Then au att otsy a (an Susiuing Fs (4.79) ad 8) int (4.8), we hve apr Mantas Mth (4a) z ‘The iferencing usd inthe above equation, whe Eq (4.81) i sed to represent ‘etme erative sealed the Lax method. afr the mahematican Peet Lax nb fit proposed it I we now assume an err of the form te) = eC a dos Previously and subst thi form in Eq. 482), the ampifcaton for becomes 7 = 05 (hi A) ~iCsin A) (493) where C= © AAs. The stability requirement is |e <1, which when applied to Bq. (483) yields cmeMer 4s) Ina. (4.84), isc the Courant number: Tis equation say that P< Ae for the meni solution of Ea. (4.8) be sable: Moteover, Ey (88) calle the {Cont Pridrichs Lewy conan, general writen ths CFL sen, is an, ‘portant stably cienon fr hype equtos. The CF contion dts bck {0928 he original work canbe found a Re 25. “The CFL condo, ithe Courant mune must best than rat most eval oui isa th sti condom which lds forthe secnd- we ‘euation (ae Prob 3.4), Pago wo 3R Tre is cnmeston heveen theca acid wth 9. (4.85) and the CFL. conden conection which hls suche phys spice ft CFL conion Lets pure ths comet, These castes ns Seu. 32) for ta (285) ae given by (2, een (4.850) He (ef-unning) (at) (4485) anon nares 6 and te shetcod in Fig 4136 an nth pars of Fig 6.13 let pit bbe he incase icuming cnc ph pt" and Taning cts thgh pH pon +1 owes poe, tered yo Ison of chr es oh lpi” Tani a a Sig acest sifemce one soci wih te CFL sity conn Sitch atest tr ti Cor pumber C= To se tt mor seh le ‘Recs dee teva of pon by Eq (288) when C1 Then, bom ror (a7) (Cito ashe Te nunca doi cae alte ami dain aon (Poem The met nd ur hal ae ae 164 ma ener sncenznn {Win Fig 4.130 and b we move a distance Ai ety show ri pit, we find ‘usc dicey top a pont 6 This because from te characte Tne piven By Ege (4.860) and (86h) (ss) ‘Obvious, he nerement in Eg (4.8), having to do wth he CFL conn, eb the increment A” in Eg (4.88). having to do with te intersection of charcteiie Ties, tthe same alts, since the ight ides of Ls (247) and (888) a the same. Therefore, icy i eatly the distance hetwcen po band gid pin Fi. 4130 and, Now assume tat C= 1, which pene to the Sec in Fig 44 13a, Then fom Eg (48 As < Ate ae sown in ip 413 Let pind ‘tegen tthe rd pint ety above pot existing a time» Aver Since fropertics at pont dae calculated numerslly fom th diference cut sing {nfoemation at gl points ¢—~ Land +1, the nimerial doman or pind the tangle ade shown in Fig 413g. The anata! domain for poids the shaded tangle n Fig 4132, defined hy the circtersis thwgh point These hartterstis are pal to those trough point. Now hatin Tig. 4.134, the merical domain fp din the analytical domain. In contas consider the ease shown in Figs 413 Hee, L Then, from Eq. (4), ‘eos > ic, a shown i Fig. 13H. Let pot din Fig 4130 cotespond ‘othe grid poi dey above point eising at tenet Ate Sine propeen int far xed numerically fom he difereace equation sing infant Bd pois 1~ Land 1 the mumerial domain for pin ds te tangle ade Shown in Fig 4138 The enact domaun fr pond the shaded angle a Fig 4113 defined by the charactenstics hough pt Note tht Fig 4.130 the numerical domain does ot inl al the analytical domain. Moone, ip 4136 isforC > | which ends to unstable behave. Therefore, wean give lowing iyscal inrpretion of the CFL condition For sbi, he numeral doin mst side al sh ana min “The above considerations dealt with stably. The question accu, which is sometimes gute dierent an alo be exami rm ie point of view of Fig 4.13 Considerable case a8 shown in Fig 4.13 Note ht he asta doin ‘of dependence fr point ds the shaded tangle ia Fig 4.130, From our diction ‘in Chap. 3. the properties a point theoretical depend only on thse pnt within ‘he shaded tangle. However, note ta he numeral pd points Tan + ‘coud the domain of dependence for pot dnd hence heretical shold ak inluence the properties at point On te other hand the mumeria allt of Properties at post d takes information fom grid points 1 tad + 1 This ‘haan is exacerbated when Av chosen to be very sal, Aes Ate In this case, eventhough the calculations are stable, the rests may be gute “ngccurate a tothe age misatch between the demain of dependence of pont nd the loaton of the acta marcia data aed 9 calle propetiey a nlp oft above dco, olde that he Courant mabe ms be ca toes thn ay fr sab, = butt ose me is dese fe ws cee to uy a pone frac ASA, Stability Analysis: A Broader Perspective “The pocetingdcusion focused onthe tba feo asa meant analy the subi chores fa given difrence equation pric, the bebor ie tuna er eat defined Eq (05), as tas This might eave the Sfeouctinpressn tif we ad por computer tho ean eo, he thre would be oo fbi Sok notte ase The Eee comet of mera stabi in reli, bed om the nese Bebe of the sion [itu nti pedo evo anal ero at ‘ramp insted of coneing Eq (460) we could stupa more general vm ‘tm ami whore the salon writen as Fre ees lows U0) = Sac as) where ty isthe amplitde of the mth bartonc of the soliton. In tum, the Unplication factor ie writen 3s G- (490) For subibty |0] <1 = “Tis, a the considerations of stably, are efor your ure studies Our purpose hee ts ben simply to oduce you tothe base tought tha siabity ‘onseitons ate mporant to CED and to give you some ies, no mater How incomplete, how these consisetions can be approached 46 SUMMARY ‘Dscreization” as ben the hey word in he present chap, We ave sen bow Aiscctice pra fle equation, ieluding the governing flow equations. Soch dscretation tte foundation of fstesierence metho. In ation vin robe 47, ou wil show to diereze he goveming fw eqns ning foxm. Such dst he foundation of ftesvalume methods. Both finite {tfeence and fintevlume methods abound ia CFD. However, keepin mind tht fhe dieretztions discussed inthis chapter ae simply wool: they do no by Themselves const any speic technique for the sluton of a gen ow pwoblem, A CFD tecfnigne defied by what ols we choose fra solution, how din wht sequence we use hse ono pure alton, and Row we han the tuna codons. Some ecngs that ave been popular in CF are disused {Chap 6, ant applisns of thes ehmigs some cassie hid flow problems {fe illupied in detain Pat I ofthis book. On Be ter hand nthe present oper we haw futhed a ew important aspects of CFD techniques witout ang the techniques themselves, For example, we have wot hit aby CFD "ecngu falls within one othe ce of wo general xtegois, exc approaches ‘ipl appoaches. We have dscassed some enem jal what We can oy elie and implicit approaches. Futhermore, we hive touched onthe shi ‘pests of hese approaches and have examined the ven Neuman stability analye sbich ives us some insight fo Te stably rstitons and stably cro ee éxplct methods. So with this caper we have taken» int slp i he rd of cD, Before progressing fare, rum to Fig 4.2, the road map fortis chapter, ‘au make cern tat you fel eomforable with he tril whic conan cach ‘ook i this oad map: Note tht we ave emphasized the tedilerence method !nd hase chosen not adress the finite volume ote element methods, tlement methods ace sill ot sed o ary great extent in CFD ine irene and fe-volume methods account for abut 98 percent of alpaca CFD station nme this situation my change, We note th he station is reversed instal ‘mechanics. whan the numerical method of choke is almost alvays the fn tlement method. The rad map in Fig. 42 also emphases thatthe mates of !npht versus explicit methods aswel as ater of ably ae omanon obo the fnte-difeence and fie volume methods Final, we note that we are pot quit eay to go dtly tow dscusson of CCFD technigues. There temains ne item of unfinished Burns, namely the aspects of grid generation ad the necesaytasfomtions which tetas: This the subject of the next chapter We hv wacked he age whe oa he enh tof actly ep a ‘etiam crt psf ows Te cep ms oe ‘at gon it tak tad "ing Jor han ty" a el coat Pre Forth of you wo wast ll mere ep yor bck CPD Bo "ting ampere spy con edo wil eee ae of ‘enc a anheaton he ecu) igen ma fe spn cused Chip 8 Howe fi se fou whoa realy te of raga is ge ‘snd at owt on acne poet teaming ep: tee hr up ‘ett ee toug mts pict alse of apna ncnprene anon hos Fron Cnet Ron To cy ough pion, you ned cena aoe ‘mac Wan ou abate Te, sa tenses Tenet, ‘met sohtn of Cate: ‘he flow of inkracn mci wth the hove sons diagramed Fi 132 ti ously eran a i ane Nae ely toatl es a ‘gen hme wih i xeon, sali het acne cea ‘Spot Sse ice pnp Te mate gy pn fe yo he fo be, te fate dimnsond pst fo Tha era me seman of forte sm ges you oparanty lea cna project wth eae mrs Goto S63 ‘Then g9 1 See 7.11074, NicCoemed's compro of tanto ‘tne sotson tn sypenons eos he ow of infason sci whe abreast granola Fig 1325, which ocshuld rye atte ota eect pert ong ene howe of snemachig teen sats, some of whch move Smeg ke coral ue ined ee othe pit we est om a contin i roel angie CFD Well movant he ref gd getertn nC es PROBLEMS: “LL. Using Talon ees, dee std fwanifience and rar iene reson fr Du 42, Using Tv Sens, dive be sent ne cml dence fo 43, Conder th Kon @ (cy) # Caer th poi 3) = (eh (a) Coble the eat ls of ie andy tpi. (0) Use fine ae fred tence, with Ax = 8) ~ teat appoints ‘oles a 9 andy pot, 1) Caan be peng ie wh ema wit the eat veer ft Pt) (c) Use liste eran irene, wah c= y=, 1c prone ‘alee dean 30 po, 2) Cause he peng diene when Cera nh the eat ler fo prt) (a) Use sco ere conferees, with t= Y= 01, 19 eke approx lirate values for O90 and Dy poe (1,1). Cae the peerage | a by! _| ip av a Be | 5 TR ae of ei fr pe oe omens wns ue ine om A ewan, 227 Fora subsonic fo, Eq (624) i lips, 9d MacCormack’s technique doesnot lye mec amy pace-masching techmiqe wil ot apply. Hower, 2s mentioned 3e hap fora Row that oly specie everywhere, E (6.24) hyperbole Tn the case, space marching i approrate, and) MocConmack’ technique 1 pliable Wi ths n mind oie tat Eg (6.24) is waten wih the x denative ‘Mire on te Ita side and he sre era he dative on he right hand ide Ret to Fig. 63. Assume that the Nowe variables are known along the veri nthe pla hs ne the ina dane. Also ase tal the ow i oaly supercnicevennere. Then olson ca be obtained, stating with the inl dae line and marching dwrsteam io thes dcton, We wil state ths process fora singe spa slop using MacCaracks technique The eas are {he tame dacs in Se, 6 3, e8cep hat ete he pai abablex proms the ‘Sc ole asthe tine vara Fin Soe 3 Forexample, an Fg 68 assume the Bow ‘Stunes are known longa veri a given hcaton (The calevltion was redwing hen dat along the vera ine ~ x9) Ls his ert ise an Through te gd pois yf 1h Gf) and J ~ Lyin Fig. 63. Tat the Bow “able at eset ged pints ae considered known, MacCorack’ techie ‘tows te clelaton ofthe ow varbls a rid pont (1) fom the known Salusat(j-° 1, G,j)abd (4) ~ 1,35 flows. The ale ofthe solton vector Fin Eq (24) a id point (1+ 1,3) cane found from Noe hat, in keeping wih ovr previous otton, the index for the machine ables inthis cave i used as 2 supescep. In B. (625), (OF Ox is 8 ‘epreseintve avenge vlc ofthe derivate of F evaluat between 3b POA eis found fom Eq (6.28) hy means ofa predciorcomesor apposch, follows Preitr step. In Ea (6:28, replce te »deiative witha forward irene ae Sea (4-5 (626) tn, (626 a es on it ile ae ow mabe as he oi Tone Sop ea etapa pe e er Fs poe hytom Tr ee tors ( f a+(Sjm on were, a in Sec, 63, the bared anti represents a pret quantity, Keep 0 Irn tat he shorthand vestor notation shown it qs. (6.26) rnd (627) represents these operons onthe india consi, momen, and carey equations, tne te lets of and ae given by Eas. 2106) and (2.107), respective Thay, Bo" represents the picid ales oF indeidul elements, given for ‘he preset 1wo-dinensional case by ay" ine" Bt} amy"! (6.28) P(-EE) | ee poesia ue a en erie (22 mat waerathnmndce kere atts er et feet: St0 moc oih be Gio eB io tae pe Pale ane ne nue e ha So es Corrector step. Caste prodited value of (4724 "at lean x + An, ented by ("sy icing the pried ales or/and G nto (624, using rearward differences. That is, = (), e620, ea fl we coe a te pied ‘imine variables which been decoder inthe peeicor sep, The {herage vale, (OPOr)q, is now formed a6 an aihnetic mean Oe) = 1 GG A (629) Inu, he fal, ert we FF, Hm Fy (625) peed 1 aap (%), a ws) ‘Clea, hin spi, owas marching slton using MasCormack’s technique iv dice analog of the tine-marchingSalaion ncaa in See 6.3, with the ‘marching varable playing the fle ofthe exer marching arable There afe two nteworhy diflreces asincaled sth the downstean smashing approach compared othe tne marching spec, The fit assay been meson: ts associated wth the ede decod th rimiive vara om the lax vanabls. Tv decoding is simple when a me-marching solton of the gnseration form ofthe equations is empiyed ay rested im Eqs. (2.100) to (2:06, bari move elaboate when spl marching soliton of the con servation form eguiions is wad a8 weed in Eqs @.1112) © 2.1114), OF Some. fora timemorching son using the noncoservation form. of the Sqosuons, no Gerding nese a al the dependent variables are He primitive ‘Shables themselves, ts we have een in Secs 6.2 abd 63. The second derence een th two marching procedures at Iet fr explicit solutions, sth the {dwrsteam machin pose demands the se ofthe conservation form ofthe {fmering equations so thatthe x derivative cam he Sold a a sng Tem 35 [Eipayed in Bq (623). Ths ea abe done wit he noacoaseration form of the ‘Summ as guick examination of Eas (6.1) 40 (64) wl show. Hee, withthe ‘ine deratives et to rar, tse ou of the Fou estos have to tems ech Involving desnanes. and threfoe single x derivative ean aot be olated os he Teh side without anther «deine sil appearing oa the rican side ‘Tis of ourse deste te expt ature of the dowsteas marching approach a dscused her 65 THE RELAXATION TECHNIQUE AND Tis USE WITH LOW-SPEED INVISCID FLOW ‘The relation technigue is «fnieifeence method particulary suited for the Solin of clip paral dient! equations, Low apeed subsonic inviscid ow {s'govemed by elipic paral diferent equations, a discussed in See. 3.43, Therefore te rclanaon seni = equeny applied to the slation of lw pel subsonic Bow. Relation teciques can beer expt o mph. See Ret 13 foran indepth discussion of varios relaxation techniques as applied 0 CFD problems. Inthe pest section, we wll deseribe an explicit eoxaton Tecigue, sometimes ealed pometeratie mths. For purposes of llastatio, lets onside an ise incompressible, t90 imersonal rotational ow Forsch a low he goversing Bow equations rede tov angle paral dierent squton, names, aplaee egutin, in terms ofthe Seale velo potential where @ is dened such that V= W@, We vl oot frovide the deals ere bt rater make the assumption tht you have some Feminty with sich mater foto if you simply eed a eve othe dena, efor example Sev 37 of Rel-& We wil simply wate bore dat he govenin ciao i #008 ae We wish w ae Ha. (631) mmc on the gi sown Fg 64 Replace portal dwar In (31) with secon, eal Son Cite: gen Wy Egy 12) and (213) Ban My HO Hset PEO g (aor (ay Examining the gi in Fig. 64, ta gr ois ough 20 const he 0 (630) (632) Sms sont : boundary ofthe domain. As discussed in Sec. 34.3, boundiry conditions must be splat over the ee boundary eneorng the donner fer the slain of anelipie equation wo be wellposed. In ome of the grid shown in Fig 64 this means that through Oy are own vals, equal 10 the given boundary onion at points though 20, The wakes of @ at al ther eid pois ‘he itral gr pins—arewakaown. gation (32), centered sound id pint J, contin five of thee unknowns namely, ir Dey, Drone Orysie 4, In principle, (632) ean be writen around ach of he eral hd points (ure ae TS such points Fig. 6) leading to a sytem of 13 hear tlgebrac equations wit foal of 15 unknowns, There are several dice meds for solving these simultaneous equations, One isthe sadard. Cramer ae, however, the numberof clelatons required fr the implementation of Cramer tls very age, duet the need oer deteminans of the sie 18.» 18 or the present example For any vel caleultion, hundreds or events of gh ot maybe employed. Cleary the use of Cramer sles ot ofthe question ot uch aplicaions Another, and mach moe resonable, direst solution susan liminton (ce, for example, Re. 13) However, the sinplest approach isto we a Felton technique, as dscied bom, “The relaxation tecigue an erie method, wherein ales of four of the ‘quanitis in Eg (632) are assumed te the Inn values a terton ste and ‘only one of he quasi teaed ae an unknown a ieration sep m1 In Fy (632) let us choose, a5 tat unknown, Solving Eq (69) for, we hae a (PT orp = NP [Mat tay Sa et aan Hayy? 2107 | (aaF (or Inq (633) the superscript mand n +1 aes the orion set has nothing te doit our previous se ofthe sipecrtto designates ine or pacemarching Stop. Indood as we kno, sich marching isnot appopeate fr the solution of lic equation. Rather, im Eq. (633), Qf represents the unknown 10 be falls st the next tertion step, n> Ii ts ofthe knows quates Oy, ‘Or Oty ysaDd Oy, ro the previous ep, (This approach sealed te lca method) To gt the whole process rtd, we ist assume values for ata tei points except oe, at which fs Wate as the unkown. Equation (6.33) i hed to calle that unknown. ARer repeated aplication oF Eq, (633) al te {id pins, we have nish the irs eration, a= I,and we go oo the ext sep, fr 2 This whole process i repeated for 36 many iteraions ase necessary 0 omerge tw asain, Tobe mere specif, consider Eq. (633 applied at grid point Via Fig 64 Assume that we have aleay cared out mations. The, fete nF ieation, Ea. (6.33) yields os GO [04% O40) gs Fayh ya [PF In bg (638), 047" is he anon Py nd te known from the vis tony and ad yar known ffm th silted ndy ond iris ogee at ep ac f © be ed as soon a posse on he sind ef (6.38). For example, afer we ne cele, fom Eq {Shay te moe oo gdp 22 whe an pletion of Ey 1633) els yo tao? fo ees er) 5g at ezer [ae ar] In Ba, (635), O41 i the unknown; @, and @, are own ffom the previous iteration, Roow fom the sulted boundary contin, and" I known from Eq (6:38), which was th immediatly proceding calculation, In his aston, the unknown e a ertion m+ Tre peogesvely calelated along a given horizontal ine. sweeping from ttt right (Ths approach ell the Gauss ‘Seidel mth Theres nothing magi about his sweeping Suction. During the rogtsriv soliton of (633), me could oss well stp sequences that sweep From right to lef for top to btm, of om batons 1 tp. ‘The above procure s repented for number of Keaton comergence i achiev when 87," ~ @",besomes lew than sme pescbed vale stall 5d pois The dere fo which you wish conergence to be shieved i up 0 yous the tore trations You take, the greater wil be te accuracy. Freer the convergence to 1 sluon vmetmes can be enhanced by a technigue cao suevesive nerrtaation This an extrapolation procedure based fn the folowing idea. We interpret. (33) a yielding an itermedie alu of 8, denoted by 0757, where ar. (navi? o3'- ee 0 Fay 20807 232 ene some OME mam Note hat we ve chosen to wre he valu of 7 n B66 a ertion evel ‘+ 1 withthe asumion that we are sweeping om Ie tonight av diese ati, and hence the val of 7°) known at di snge, Silay, yi own ath tage because we ar sing our sweeping procedure athe bon of the wri and sequently stepping to the next higher row of grid pots, Then me ‘se the vale of, obsined atte ed of the pesos terton and, ohana from Eq (636), 1 expla a value for Qf a follow: =, +0(0j"—a7,) (67) In Fg, (6.7), sa relaxation factor whose vei sly found by taneor experimenationfor'a gen problem. Ife > I, the above proces i called Suecese overmavaion I <1. the proces i clled woderlaaton td ‘usually wed whea the convergence beavor sos back ad oth between Some vale For overeanaton gency the value of is ued by T= 2 (Gee Ret. 13). tmany event the use of Fg (6.37) with an appropriate vale ft en Tec the numberof ietatonsnecesary fo achieve comnetgence and therefore Felice te computtinal ie in some probes bya factor of 90 scoring 19 Ree. 66 ASPECTS OF NUMERICAL DISSIPATION AND DISPERSION; ARTIFICIAL VISCOSITY Many aspects of life ste never quite what they appear to Beat ist inpression — CFD is no diffrent. For example, inthe prvent chapter we ive dcused several techniques forthe numerical solution of the governing flow equations We have approached these discussions, aswell af those in previous chaps, fom the potnt of view tat numerical solutions of the Eult o Navi Stokes quations ie teng obtained widun an sceuracy dete by the neation ad roundoff eos The focus has bees an the fot hat wea saving ame pee partial difereal equations but tat the nae solutons ae aays somes ‘Theresa diferent perspective tht we can take on tis matic, oe with « shade of difference compated wo our previous discisions, For smi, let onside model equation, namely, the one-dimensional wae equation given OY (638) witha > 0. We consider (6.38) be the spec parti fren equation tat we want to solve numerically. Let w choose to diese this equation by sing frstorer forward difeence n time an fis er rare aference a space Then bq (638) i reprsonted by the flloving difference eqution (639) ersar seca ein Oar Ty 239 ns 9 numeric ‘rom our previous perpectie, a solution of (6.39) repress olution of (638) han atin ecracy a determined by he ranean aed ound after From our dacssonsin Chap, we know that te acaracy of Eq (60) i gen by Ot4r, A Lets nove ake a highly eet point of vie. To fp establish thi ew, we repceo,"and an Eg (6.39) with Tylor series ‘spunsions a follows worn (Ba (GSE GASP van (@) ace (26) OE (Pa) Be aan Subsiuting Eas. (640) and (641) ito (639), we ave (CG )F-@e+] J-0 (8) ee Pause for & monnent and examine Eg, (64). Te le-and side exactly thee hand side of the rginal paral difretal equation sven by Fg, (6.38); te night hand sie of Eq. (643) i the traeaion enor associated with the rence ution given by (639), Clty this taseaton ers O(8, 2) Let us 00 ‘plc the me derivatives on the pt and side of Eq. (648) wit x derives follows Tit diferente Eq, (643) with respect (We wl drop the subscript fd supeserip since we know tha all denatvs are Being evaluated a point fed a ime) Ou, Pa Oude atu (ane (os) ao HOP 2 ORG : Ou ade ou (oa) wea TW 6 234 sae anc ramets A momen ‘Als, diferente Ea. (643) with respect ox and miply by a oO, Pu Ou ads ou ale? Wo" oS ORR 2 WO 6 Puede duei(de) ae Subtracting Ea (645) fm (64), we ave Pu Pu Puss Tuas? Dy ade Be 2 oF 6 aR DP Ow altel, Ou aa, Oty ata? wane to 2 toa e cn cx i (646) na mre cmp ony Spigot tone Ou _pPu, | Ou, Oe Sine Se AS 0 Pe +o] al, Ow +8 [ee (on) Equation (6.47) provides the expression for Fu? whichis tobe subsite for the ist term op the Fightund side of Eq. (643). Befine carrying ot this subsiton, however, et us wet the second erm onthe right hand side of Eq (6.43), namely the third time dentine. We do ths by dfentating Eq (64) ‘wih respect fo time, yielding ou + o4ar, a8) (648) Difreting Ba. (645) with respect x and muipying by a, we have Ou, ou 2B te MF otar, an) (645) ‘Adding Eqs (648) and (649), we ave Ou 0 4 ota, 8) (650) Equation (6.50) provides an expression forthe third Hime deve to be inserted into both Eas. (647) and (643). Reming to Eq (647), we sce two mined I. That replace in Eg (067) By tf and solve for 7, fal values of fom It Mf sing Thomas ago. Tit rossi oped W times: Le toe are W sweeps in they dtecion eulge ‘Thomas’ algorithm being used WV umes This sweeping in the ston is show schematically in Fig. 12. Atte end ofthe stp, the ales of Tate © Aare known at al grid points (i) hati, Ts known at all), At the end of this twostep proces, the dependent varible Tas boon ‘marched valve Ain the direction of Alough he re wo indepen! seta arabes x andy ination tothe marching varable ths marching heme involves only riagona forms andthe solution as been shire ty th opened application of Thomas’ algoritn. Because the scheme involves tw sleps oe hich the ference equation i implicit «and the othr in which the ere ‘uation i implicit in. the source of the name of the scheme "alereoag “dracon inpis—is eons, ‘The ADI schemes second-order accuse i, andy that is the action enwris of OFAN (87. (ANF) See Rel 13 1917 fr detail sen com ome ruc a ema oe 247 sept Ea athe fo ‘This sheme bus foun appistion a many fi ow robles deh abn pati ih te lion ore te she partial fret eqlatons Also, he scheme descnbed above is & {pec faa gee las of hes ohing ping of 90 oF Pe estonia a pl sion of he geting How ego a cn {ape fos Hens, ADI san ese gee deseo of bole cls ‘heme en whch sen dod th sn, Anthor pvesion Sion AD scene ald apoinae acterzaton: ssa mote aban tie i dead in Ses 113.2. 68 THE PRESSURE CORRECTION TECHNIQUE: APPLICATION TO INCOMPRESSIBLE VISCOUS FLOW numeral tecnique fr the seton of inviscid. incompresle fw was Gasol n See ney einai tees va meee tw meme alii pr tna yn a ean ‘Sng whch eal re pos cl mel {ersoNng ep rts cone, vac, incamgeessBe Bw governed 2 en sone cr mee: nen by the incompressible Naver Stokes equations, which exhibit « mixed pti pri eh and hence the andr Yelnation tehngue as dserbed it Se, 15 i not prea help. The parpese ofthe preset Secon debe an iterative proces called the pressure cretion tchngue bi has ound wide Spread appistion inthe numeral ston ofthe inconpresile Nie Stokes futons The pressure conetion technique Mas been veloped fo praca gineering solutions by Patankar and Spalding (Ret 67) and cussed tenth in Ref 6k The tchngie is embodied in an algortan called SIMPLE (com impli method for pesurtaked equations, pionered by Patankar and Spal ing. which has fund widespread application ener the past 20 years or both compressible and incompressible ows. However, nthe present section we wil focus ewe ofthe pesure crction metho save meompressbl, isco ow ‘Before deseribng the pesure comcton meth there reo considerations sociated with an scompressible Now solton that need io Be aes. They are the subjet ofthe neat to subsections. 681 Some Comments on the Incompressble NavierStokes Equations ‘The compressible Naver Siok extatins ar derived in Chap, 2 and summarized ln Sec-251. The incomprenible Naver Stokes qutins can be obtain fom dhe ompresible form simpy by seting density equa to 2 constant. That ith p= constant, Eg. (229) becomes vv (668) With the fhe ansemption tht i constant throughout the flow Eg. (204) 0 (2.50) combine with Exp. 27a) to (2.57) become Pe ay nk (a) ou (Me) sof (608) OB an (RM) aut (BB) oop, (670) (O28) eats ern Dra a Pt 8. (M Oe) | 2 Man ae * ae) * ay Not that ning Eg. (669) t (61) the ts in Eas 2.5) 25) ‘ety iting VV hve ten sto zr da to Ey (8). Te fs a N= Ofermomprese ow ales ur daton ep (68) (07 om , vive Mg (62) Rearing Ea. (672), we have ‘me esc eee AMEN HE HD tu __ a Ow q--5-F (6.720) Difereining Ba. (6.720) with respect fo, we obtain Ou oy Pe aD Bix dng 202 wth sie of Ea (67) and mpg by 1 bin 4 4 Oe de ne ae Oey Bea Satin a, (674 send eon he gh si FE. 1669) ad Sing a ee ieee 3) oe be bu ve, a De ~~ as ine "Oxo “Bed * eo Pa Bu, Oe Se nS and wiht (6%) oh (638) Ccacsing ters in Eq (675), we obi 2 comveient form ofthe nome equation fora viscous, incompressible flow 36 BeBe Be) ot . . * Wve (676) whee 7 ithe alain ofthe component of velo, Bquatos (6.70) and {6 71) canbe teat ina sna fashion The esling stem of eutions ithe Incomprssibe Naver Sos equaons, summarized below. a wm momen: 9 = 8a yous of (678) Bae ymontn: 9Pea Benders | (6m tome: 92a Benvieveh| (69) Note that Es. 67) 0 (680) ae second hey ae Tour equations fo ‘he fo dependent variables wand p. Through he assamptions ofp = constant nd j= costa the energy equation hs been completly decoupled om the mags The mplcton hee et the ontinty and momen equations ral Tha are eves to solve forthe voit an pres fel im an ncompressibe flow and that i pvenpecblen involves hex rans, and ence temperature rains exis in the flow the temperate field canbe obtained diel fom the ‘teay xa fer he velocity ad pressure Bel ate etn, In hs section, Se wl ot ea wih a tempertre el rar, we wl assume that ~ constant, Sich is companble wth our ater sssumpion thay constant Teen Dp eAD), Hence, Ege (67) to (80) are sulin for eur discussion ere ‘Cleary, fom the above desson we sce tut he incompressible Navi Stokes eaustions are dered inn ssighforerd sion fom te compresibie NavierStokes equations. Tn turn, this mig lead us 10 thnk hat a numerical “olution of the incompressible equations mig! be obtain «sealord fashion fiom + numeral fechngue fishioned. forthe compressible oquations Unfortu ths not the case, For exam f ne write a cmputer cde save the compres Nie Stokes equation ving a tme-marehing McCommack’s tcchngus a descrfed in See. 63. the explictme sep A” este by stably Conve. An apgonimate sy condition for an expt Navier-Stokes Solon iy given i Ret 13 as 1 as — wife |o(/ae-+ay/iytae? + Vay? Fora compres ow, the ses of sound is nie, and Ea (6.81) wl ye a fine value ofA for the naman soln. However Torn ncompresie Bow, the spot of wand theoretically infinite, and hene Eq, (681) woul yield Ar = 0 fr such & cane. Clea fr the numerical solaion of an incompressible fox Something cle must be done This phenomenon ts further reinforced by the ‘sation that compresbe-ow CFD soon fchniguc, when applied 0 2 ‘how field where the Mach nimber is progresiely reduced toward zero, kes rogsessivly re tine sop to comer, isthe auhors expesence that a npr coe run fora flow ich x everywhere ta ea! Mach number ‘faut 0.2 orles kes a prtibive amour of time to converge, an inde has a ‘endency to be unsble at such 2 low Mach number. or sch esos, CED, slution techies forthe compressible Navi ‘Stokes equations are sully difleeat fom tose used forthe sohton of the Compeesible Naver Stokes eauatons. The pressure corestion method 1 be Cfecnbedsholy tanscends ths dificil bas been used seth reasonable ‘sees for compress ow but wath ven mre success fr incompressible Naw. Ie an ascepled and widely sed technique foe inompresibe iscous, CFD sppistions. Therefore, we focus this method inthe presen! section a) 682. Some Comments on Central Differencing of the Incompressible Navier Stokes Equations: The Need for a Staggered Grid ‘Tne incompressible contimity equation i given by Eq. (6.77, which in wo dimensions is (6m) ne mesen conten mame en nem vem BSL A central siference equation represensing Ea. (682) is s tao (633) Tae Ty “This diference eqtion numerically slows the checkerboard velocity distribution tiven in Fig 613, lsatd here i a zgrig typeof tistbtion of bth the fongonent andy compenet of th velocity wand, espetvely. Inthe xeon, vate a8 20,40, 20, 40, ete, a successive gi points, and in he» dretion, + ‘rosa 52,5, 2, ee a ences rd pots these numbers ae situ Into Eg. (643), both terms are 20 at every ghd point Le the dsr velocity ‘dstbuion shown in Fig 613 sss the cet diflewce form of he continuity ‘ution: On the oer hand the cheskerbord velocity distabubon ia Fig 6.1315 scaly nonsense i terms of ay Fal payscal Now fed “The problem described above does et oor for compressible Now, where the incision ofthe density variation i he continuity equation would generally wipe fut the checkerboard pare ilustated in Fg. 613 afer the ist time stp. "A eld problem is encountend in regard to cena difrences in the momentum equations, Eas. (6.98) to (6.80) magne a hwo dimensional scree, ‘hockerbourd pressure pater slated ia Fig 6.14. In parcl, consider the ental difeence fomulston fr the peste grade: Op _psts—Bits Tar opt Ay eo Disc cad wt toa ch pie br he per 282 ne wnt cro noma Ame For the checkeboatd pressure ditibuson istated ia Fig. 6.14, Fas. (6846) and (6.80) sive aero resrue gradients in the x andy deco, respectively. Chea, ‘the pesure fel dsretd in Fig. 6.14 would ot be fel bythe Nave Sokes ‘equations rater, the numerical soliton would effectively se only a uniform ‘pesure iat and In shor, when central dileences ae used forthe incompressible Naver Stokes equations, the resulting dillerence equations are of 4 frm tat, when fprscted withthe nosensical velocity and pretsure dstbtons shown Pipe {615 and 6-14, wl tend to perpen these dsbutons, Amite some eat onal illerence algorithms for incomprenble veo Row jgnored this posi, Sh suetesafl soluteny were still bined, presumably because of seca ‘reament ofthe boundary conditions or by some otberfotutous aspect ofthe funeral procidire However, pen the weakness of the cena diference Formulation described shine, we shoul jtifaly fel wneomforable, and we should lok for some "fix" before embarking on the sfuion ofa given problem, Two such ines are suggested upwind diferences are used insead of cenal suteresces, the problem immediatly goes sway" A dacanson of upwind ifr: fences igen in Sec. 11-4 However, apohe fs oan cena econ but suger the gril, as described below: ‘staggered pris usted in ig. 618. Here, the prestues are clelted atthe sold pid points beled (Tlie Tedd = Malle) — Distr seu ae ace opin edd - Lh (oj "Dates ~ siete Specialy wtecalelsed st points A. jhel4 Ph ‘ee min concen oe mucosa 288 a Stabe et te and vis aula at iferent points (4, + 2.6, — Yee The key fare hei that presses a velocities ane elated at iferin grd poits. Fig 615, the open gid pins ae shown equidistant between the slid gid post, ba thins noc unecesny: An advantage ofthis staggered pid i for enampl, tat when tiayny islet a ental dierece for x Yells (poy pe VA tat 1, the presse gradi based on adicon esse point which iit the posiblity ofa checkerboard pressure pate as sketched in Fig. 614 No. ental difeence expression forthe continuity equation, Eq. (682), centered sound point () becomes Mivnas— Meus Meson s =— a ° (5) Because Ea. (68) is based on aiiacent velocity points, the possiliy of eheckerboard velocity patem as sted in Fig. 13 i ciated 683. The Philosophy of the Pressure Correction Method The pressure coreton technique is hasilly an itetive approach where sme innovative physi reasoning ised © constut the next iteration frm the esas ‘ofthe previo iteration The thought proces Ha flow 1. Start the ieratve process by guessing the pressure fed, Denote the puss pests by 2 Use the vals oF p* wo solve for wand w fom the moment equations Since thon elocies ae tow ascisted withthe ales ofp, denote them by wet and wm 4 Since they were obtained from guessed values ofp the values 4, and w", en substtued ito the coin equation. ll nt necessary say tha quaton Hence sng the cont equton construc a pressure correction py” Wich hon add wil rina the veloc Hed tore nto agreement th ‘he comin equation, That, the "eometed™ pes pis parte (6.6) Correspuing velit comeeions u,v and w’ canbe obtained from such maw (6870) yey (67) wan ew (68%) 4. ng. (6.86), designate the new value ponte let-hnd side a the new vale of | ‘p*- Ret to step 2 and repeat the process url a velocity field 8 fod that, ‘oes sais the entity equation, When ts s achieved the eoret ow Fld ie hand 684. The Pressure Correction Form “The presse cometionp was inode in Eg, (686). The calculation ofthe ale pis the sbjet of th subnecton. For simply, we wil onside 8 40- ‘mensional Now; ie additonal terms asocaed with the tied dimension are Treated in alike manne. Also, we wil neglet body force, “The san pment equations Tor a incompressible viscous Bow are sven by Fay (678) and (679), respectively. Those equations fe ia nonconserte tho fmm in Eonsesaton form. they ae face Sex 2.8) atm Ba e(B%2) omy w= on ey om) {As discussed in Chap. 2, the conseraton fom follows ict fen the mode of ‘infinitely small volume fed in space Because of ts model ate dierence eum of Fas. (688) and (689) wil be somewhat akin othe discretized equations ‘bud fom a fine-olune approach. The original formulation of the presse ‘conection method by Paarar and Spalding (Re. 67 and 68) nol fie ohne apps nthe presen sen, ne wil contin wih ieee {Dp by ung th Cmsnaton foe of he goering ia aie TITSn is tniciRowns aac cues fe sume dred Statist told ebuned ina neem mcd We pce! 0 Serlop seaeteated quien whch ae te be wok oe pore soneton We choi wwe sow fee nn eel ioc oe {fatal tntnce Notte esr someton met rely ea ip cev acer apprsch span nsec 68 andthe of rer dion eto tn ie apy genera ttt. Ta Ihe scheme developed felon hr eo apc hu eae Sore oo seca ‘Comin a staggers ps asin Fig 616, Rel a te presale a he sid pd pte ie een ai Fei cw floes Fp (08 coed snd he pot) Fa Ie free pups sven Soe sch ld de Sin esol alin Fp io) We il sed nrge ar oat he pet Ind on bop andra, espe oe hae eh These dein iearimepetan tence i wo act pot de (aseue t terssial (904) (ujae terial (6.908) A point Bene Ar poneb » | Lit pao keagbntsn || } wt 256 owe ro momen sn Centred around point (+ 1, & difeence representation of Eq (688) is (oof ils voy (ay 1 (svayca] Pay y ay ae se yftlnasZ Moray tM ny, Mouaser May tmnt 4 (an (ae | (on) (otha = oma tA St Moet) | (692) whee, fom Eg (691), flow its [ee< vast Mevay , Meanser~ Moray + Meonas-t ee uation (692) isa ifrence eguation representing the momen equation, [Noe that and in Eqs (691) apd (692) ae those values defined by ay (6.900 nd), 1 ad ¥ Use diferent gid points tan those Toe Tn like manne diference equation forthe jeomentin equation is baie Here, we will dren Eq. (689) centered around point (+ ‘a shown in Fig 6.17. We define average values of w atthe points c and do te le Gnd right sce of the shaded cll Fig 617 a fll Ae paint (6ras+4- ras) A pott d Houyay tmecyayea) Using forvacd diference in time and cena difeences in space, Fa, (6.89) becomes peer ra OMe Opn tew—Lot Ay | 9H : L it feta fous 7 oe (eva) (pelt a yeue OPN jesi2 — OM) 12] | sn in Preyer = Mere tM Witte | Nyse — Mysyet Myon fleshy [Note that and iin Bg, (693) are those vals defined by te average values at poms c and die, « aad a use different grid pots thin tase To "As ulin in Se. 683, a the begining of each new eatin, p= p* For this sation, Fax (692) and (698) Become, respective. Co Say = Ow Many tA Atay Py) (694) andy WN jaa BA Ris 8s) (695) » Subrstng Eq. (698) fm Fa, (6.92), we have rt era Nay = OOF Loe (696) hore Say oe May (00 ay = Oma OH Manny es Preis Pleas Ay Subrsetng Ea. (698) frm Eq (693), we obtain en ys = OM sn tH Ae Be =p) (697) where WN La= OME WN a Nera = Osa Ossie 8-8 se Pet Ply By =Pas Ply as, (6,96 and (6.97) ae the and momenta equations expressed in terms of ths pressure and velocity cometons p's ad” defined by Eqs. (686), (6872), and (687), respecte. ‘We ar now in postion obain fom fr the pressure caretion by insisting tht the veloc el st sat the ony equation, However, we are ‘mind tht the presi coretion metho san trative approach and therfore ther is incre reason hy te foam designed to predict” fom one tration {ote next be pial comet aher we are coerced with al wo aspect (1) the fn for” mst yield he vals hat aimee the proper, conversed ‘elation, and (2) inthe limit ofthe converged slton, the formula for p must {ede tothe physically correct continaty equation, That, we ate allowed to nsf for’ which is smply a mumercalartfce designed wo expeie the comergence ofthe velo Feld fo 0 soluion that satisfies the conti ‘Sato Whan thie convergence i achieved, p> O, and te formula for ‘dues tothe plysicaly cet contin egiation, ‘Wilh the above aspects in mind fet us proceed to obtain the presure conection formula Following Patankar (Ref 6) et us abiranly et, and (7 equal 9 2200 ia Eqs (6.96) and (697), obtaining oh as=— Whey AP (698) ee : ‘ od (Neue = — yor HA (6%) Considering that we are simply conntng a merical aifce which will provide Some guvdance inthe rave procure, the above should not make you tally Uncomfortable. Returing to he Sfintion of (pw fy, given pst blow Eq (696), namely, ha = Cou = WN ny swe can write Eq (698) as (6.100) ont siny Retuing the dion of (pf! nen jut Blow Eq (697, mame (Nis Mia ML we can write a, (6.9) as Wj = OM User w (6100) Returning 1 the cotinty equation pa) at ay and writing the comesponding cena direne equation centred avund point Gp, we have ic 1g Wy yay , Myer = My u2 a ee Substuting gs, (6.10) ae (6101) it (6.102) and dropping the superseips, (owt ny = Atle ny Ah) (yn + BH AN ~ Bas) a (asaya — AAAI =A) Ohya * AUN =) een at Pt o (6.103) Rearanging Ea, (6.105), we obain [ermine tay toh rata e=0) 109) af + fant" a oy An (oy \ 4 = Flo eras Mad gelOM sana OF whore hy-val Equation (6.108 i he pressure conection formula. I hasan elliptic beta, orient withthe fat that 2 pressure disturbance wil propagate everywhere "troughou an incompressible flow. Thus. Eq (18) can be sled orp’ by meas ‘a numeral elation technique, sich a described in See. 6 "Nove dn Eq, (6108 is the cena dierence formulation ofthe e-had side of the comtiniy equation expressed in terms of wand. Daring the couse ff the itemtve process and 4° define «velocity field tht does nor say the continity equation; hence i Eg. (6.108) d #0 fr al but the It terton, In this sense, oi 2 mass sowee tem. By definition, inthe lst eration, the Velocity el has converged wo 2 ld that satisfies the continaity eto, and hence, thercticaly, 0 for ths last eration. In this seme although & ‘mathematical arfce was used to abun Eq (6104) jn the let erative sp ec coma. (6100) at Beg oper py stent of te com is interesting to ote thatthe pressure comection formals Eq, (6.109), i ental difrence formulation ofthe Ponson equation in terms of the presire oy, Feo (6.08) 1 the second paral derivatives in Ea (6.108) are eplcad by central diferences and if Q = aA Athen Eq (6108) obaled (Ths shot ration et a Prob, 61) Poisson’ equation is one ofthe known equations frm lassie ‘inser and muthmatse, andi wortuhle 4 bserve tht the prerare rection formula i nothing more than a difference equation reresenaton St the Poisson equation for p"" We vo ote tht the Poon equations 3m ‘elie equaton, which materatialy verifies he elticbehavir of he presare ‘erection formal. 64S _ The Numerical Procedure: The SIMPLE ‘Algorithm To bring al the above discussion into perspective, we po summarize the numerical step forthe pressure coreton tho. The flowing desiton she ence of the SIMPLE slr a et forth ip Patankar (Ret 68). The acronym SIMPLE ‘ems rom semmpit method for pressurelinked equations. The semiimplict terminology refer to ou abieary Seng of 4 (pa and (peg 220 ing (696) and (697), ths allowing the peur caren formula, Fg, (6108), te hve p appearing a only our er post his rie had ot Bee used the ‘esting premiecoreston forms woud have ince veloc t eighboon {Endpoints Theevsloctis ae i tr inlusnced by pressure comets in hit cjhtorhood and the resuling press eorecton formula would ve reached ‘ch iter nthe How Fel. evenly coupling the etre pressure comecion ‘eld in one equation. This would hive represented “fll pli” equation Instead, beau of the above aif, Fa, (6 108) contains presare cometions at ‘ly four grid points, and Rene seme as only somimplo by Patankar (Re. oo The step-by-step proalure forthe SIMPLE algrtim is as follows Keeping in mind the staggered grid as sketched in Fig. 613, guess values of, (GAP at all he “pesiure™ ghd eins (he filed pois in Fi 618) Aso, ira set vals of (pu an (gna the peer "velocity" grid poss (he open pots in Fig. 6.18) Here, we ar consider te gi points nal © ‘he ow Hed tbe wesbnet of pos onthe Boundares wl be seus ater 2. Solve for (put for Fa. (698) and out" fom Eq, (695) at all, srprorate item i poss Subse these values of (pty! and gv" ito, (6.108), and solve for (Pata inrior gid points. (This alton can e cai! ot by relation ‘procedure sich a described in Sex 65) 4. Calculate 97" at all steal ged pins from Eq (686), ie wate ‘5 The alts ofp obtained instep 4 are used to solve the momentum equations gan Fortis, we designate f°" obtained abo asthe new vale of "tbe inserted nto Es (694) apd (6.9). With this ierpectaton eur to ep 2 and ‘peat ps2 to Sup convergence is achieved. A resol criterion 1 se fora meaure of conwerence s when te mas source fem d approaches 2 18 seen ‘When comergence is achieved the osloctydatbution has been cbianed ‘which saifier the continuity eqution, The whole fancion of the pesare ection frmala, Eq (610, so aim the sean process in such a decom ‘ha, when the velo dition i caleltad fom the momentum equations, ‘wil cventaly converge tothe cores dsebuton which sais the coaauty couton ‘Something nee 0 be si in regard othe saperserps wand n + 1 used in the above equations. Equations (688) and (689) are the wsead) mame ‘ustons an hence the corespoding dieence equations Eqs (692) and (693), tas the standard superenge notation,» fora given ns level and > Yor the ‘nextel. On te oe hand the tr tat were nepeted inthe dexiation of the pressure eometion form, Ey (6108), rel ins epi eration process {the proces esrbed by steps 2-8 above) which inno ways timewe-acarte However. this sno problem, because the presse cometion method is designed © Salve fr sea No, apd we oan this steady flow via a iterative proces, From ths point of vie i best wo interpret the superserps abd nT in the above ‘equations as Simply designating sequoia rain steps, wth sgnfiance “ny veal tnsent anton. Also in ths sense the valve of tht appar nthe hove equations cn be viewed simply a parameter which has some eet onthe Speed at which convergence is achieved On a relted mater, Ba. (6.108) may exhibit a divergent (athe than ‘comergent) Behavior for some applications. Paunkar suggests using some under ‘elation a such ees, stead of using Ea (6.86) in tep 4, use the equation =e bap! (6.106) hese ap ab underlaation fico value of about 0. ie suggested, K may aso ‘bliin some cats to underelac the values of and v obtained fom Eqs. (694) and (695) 686 Boundary Conditions for the Pressure Correction Method How ate boundary conditions specified consset wih the plosophy of the eewure comecton method? This qustion i adiresod bere. For goometic Simply, conse the contantrea duct sketched in Fig. 618; tagged pid Js dtbatl imide the duet Foran incompressible viscous ow, the psa Problem is unguely specie i At the flo boundary, 9 and v ae specified and allowed to oat. 6p Specie the pi Zo the inflow boundary. Hens im Fig. 6.18, Kawa wap =0 sn Fe specied nd beld Bed. {At ihe outflow boundary pis specified and w and v are allowed Yo oat, Hence ‘ern comuenoe ome AUN To MoM CN Hm 26 Shs fir te Sicnino Aonty contin fre peste tn 2 At he walls, the viscous, nop conto held athe wall Hence, the velocity 1 the wall er, 6 = ton = 0 Forthe numerical soliton, wens ne more boundary conn a the wall. Since 9 (6108 as lige behavior adi sled y a elaxationehniq, boundary ondion asocned with p” must be specified over the complete boundary Sotaning the compotion domin From stems I and 2 above, we hive 0 atthe iow and outow boundaries. A condion associated with pat the wall canbe dered a ols. Evaluate the momentum equation at the wal, twee = = 0. With hese velo values ised i Eq (679), we ave athe ae C)fGeS) wn Since i ~ 0, ten in a, (6.107), (0728), = O. Also ia the wea vicinity ofthe veal, srl; Bece in Eg. (107) we ca reasonabiy assume tat 0/2), i Tina. Thus, fom Eq. (6107) we can comfortably ste the approximare (but reasonble)presure Boundary condition at He wall to be given by (2)-o (6.108) Discretzing Ea. (6.108), we have (ering to Fig. 6.18) DI=P pape =P ee 264 se mene ro none acme ‘With his, pressure oundaryeondton is aun spect verte complete eunday enclosing th computational dona. vee ‘ovale spl cas im pio Se 94 mney telnet ‘wy rem ete sin t-eesnl cge NeeSees sons Thewtie you esos 0 Se ap ee ew he Pe ‘Srecon med ou can Hoe sou os it s,m carn ered an othe pesto in "he Baan soe ow gelding whe met sn ps es 69 SOME COMPUTER GRAPHIC TECHNIQUES USED IN CFD. We end tis chapter on spe CFD tenia witha dscssion of some cmpater Bape “ches” that que expo ithe presen of CPD dat Ts scone fom te presen secon int we ae a oi pes a specie nme shige fr he aon af fow probly, Ite prs section weds how amputees suo an eal ol 15 the conpaston Aid dani opp te laa «CPD caleaiaion Therese vrs grat ehnigues st fore prsenton of dat and hen :Seams appro fo cle nhs cope om ecu sh oer och tapi equ tht seen funy encountered in CFD. "We can cll the ways tht CFD dat ave sally reseed under si cargos, be discussed blow The compustonal fa pmamiit usally implemers hese annus modes of grap spesnation vi the we of existing compute graphic sltare her ha developing the dtl of row compe ‘apie pogres bins or hel gently oot he puriow af CFD to be volved wh te dcvelpmet det of cru graphic sofas bu ae Simply us thor a a fol We wl fathead in he presen son ‘Thee are may existing soar pakapes wed by somputationa id danas toy Inte one of hs ators stants, TECTLOT soar page pve ty Aree Engicenng, se. For ths fans many of the ars computer raph pues pesca ths soon were gnc wi TECPLOT, ths mot {be consed man endorcnt of spe pda bu ae simpy 20 trample of standart raph astvare approuch New tlic snd soar fr comput gophis ae evolving apy shoe for CRD tc vo when Jour tie comes, Jou wil watt make our own chief an appopente Bates sshoue page rajnty of ays that CED rel we presented graphically an be classed unde signs ago sate ofthese enegereconae he ‘enuind of th sco 691 9 Plots ‘ou are paps mos amie with y plats; you have been desing with them at least since your fit course nage. On a two-dimensional graph, they represcn the variation of one dependent variable verte another independ varabe Retin foramomentto Fig. sbwof These ae good example of pls. Inthinease they tne pte of presse coefficient vers nondimensonal chordie distance eh ‘ieee plo rom Fg 1 6b of coresponds oa difleretspansise sation Suck Spots ae the simplest and mos! stagforwaedeaepoy of computer sophia ‘eprselaton of CD ress, ARough sich graphs te ot patil sopis- "ste hy stil renin the mort precieeqantttive way to present urea ata fon raph ht aner person an rey ea quantitative dla Hom cues on tay lt without making ny meal or antmete serpin 692 Contour Plots A sadaniage of plots a dseribed dove i that hey usualy donot laste tho elu tre ofa tet of CFD rel al tone view. On the ther han, contour Plot do provide such a global view [A contour ine is ine along which some poperty is constant. We have ley seen some contour plot. For example, returh Fig 1.82, This a contour ‘ot fr presse coefficient onthe trace ofan T-20 Bake silane. Each line responds to a constant ale of presue coeffcient Genel. cools are pote such thatthe diference Beween he quantitative vale of the dependent “able frm one Costou ine on adjacent contour ine eld constant In this Techn, in epons whee the dependent variable rapidly changing space, the sjcen contour ines are closly spaced ogee, neonates whee the ependentvarable is shwly-changing in space, the adjacent contow nes are ‘wiley spaced. In Fig. 1s, the regions whore the contour ines are bunched together ideas repions of age pressive gradiets on the sufics—i this eave ‘ioointing repons where shock waves ae preset on the surface ofthe apne ‘Another example of contour plot piven a Fig 68 od Here, pes contours ‘teshown forthe two-dimensional, viscous supersonic Mow over earwar sing ‘Step The repons of age graders i he owe exparsion wave fom the tp ‘leo th sep andthe recompreson shock wave Farther dowasteam-—ure leat) Seen in these contour plate Tt scler rom examining these contour pots that the plobal ature of he Bow is seni oe single view: oebathe same global feng fo the esl om plots sy fo asutin the lations ofthe shock and expansion waves, me ould fave to examine umber of plots. Contour plots are clearly a sipenr graphical representation fo this post of vie: On teeter had requires or eft Feud precise guanine data fom contour pt as compared to carve an Plot. Ahough each contour maybe labeled st he constant numeral vale of the propeny it represents he obtaining of naraial values heme comour lines requires some mental andor numerical interpolation is space an impretse process to'say the leas The ploting of a contour digg by hand + lng, Isboous proces although th plots were made ey fue by mented souls bette fhe ofthe computer This sin cont 02 pls wich have heen made by tad with aplomb since the days of Rene Descartes i the sevetenth cent Therefore, he proferation of contour plos withthe advent of the compte understandable in CFD, contour paw ae coe of the ros commen found raphe rpresenttions of it. [Letus examina few more xan of coor pots sme modem CFD sppiations, pointing out various nuances and subalegones. For example. consiet Fig 6190 and D. Those ae contour plo of the tansvere scoot (Be > ‘component of velit») in the flow Rel behind detonation wave propsesting ‘trough t combustible mist of My, Ox and argon. The dctonaion wave ‘opdgatng fom left righ; the rol ofthe wave ie san at the ast Perpendicular cuter of contours tthe gh of the Rus, The dstontion wave SS popasting sto uniform ps, which the region othe Hat ofthe ont by dion there reno contours in thi uniform, constn-popery repon gprs tntally clear rpm mth exe right the gues. Combustion ft th hydrogen and oxygen occrs bind the detostion fost Basis ofthe pla! presence of slight disturbances nthe flow Bhi the fot, te low ed acomes vo-dimensonal wth transverse waves alg with ios ship nes as a he seen in the contour plot The prpse of including Fig 6 19a and inthis tiscasson sto pont out etfta the mumber f conto ies coven fora given braph, Figure 6 19e contains 1S difrent contour lvls each contour fled ‘eth a number or ltr and he valu ofthe taser velocity in cetmeters pet Second given nthe ible of cotour ales a thet ofthe raph Now examine Fie 6196 this th sme acto data but plied with 35 contour levels. Clea, FE 6190 gives a sree clare picture Of the fw ld than Fig. 18a. TB ‘ompariton clase he vale of inclading suc lange number oF ‘omtour lines in your plot Figure 194 and b 8 an example of line contour plots. Antbsr ‘ype of contour plats foded contour asta n Fig 6.20, Tis gure shows te sae ‘da forthe ansere veloc resented in Fig. 6.19, bute of sin ines, " eosiat property eno yn sonst itn of coor shading. Ith ase, fry is the olor, and Fig 6.20 allo a gsc color map, Ther instead Stalag the Now th dre numberof contouring, he recs between ‘hes lines ae simply ile wth colo tens tat dene the alo he Now Seid propery the regions bstcen the lines ae “ooded™ with color ites The eolorcode velo scale i shown atthe ight of ig 6.20. “The autor wishes to thank James Weber, ane of his graduate sans the Univesity of Maryland fo ping these Ruts obtained spa of hie doctoral research These cleans wore made using 4 fsitesolume scheme ale the Jccorreced tansport (FCT) metho 4s describe im Ret 6 {et examin comou plots fr anther peo owl sation, in his ‘ase the shock-shock ntracton problem sketch n Fig 6.21. Here, the sight ‘oblique shock from a wedge ina Mac flow mpinges onthe bow shock from 4 ‘vine placed above the wedge. The interaction of the shock waves frm the rn en) ‘open Coors (ar) oe sad omnes Mtns of Savon) wedge and the eylndr crests a comple ow field in the segion of interaction. Notice n Fig 621 that he cured shock wave fom the clider comes kinked in the region where the wedge sock pings. The boudary-fited cordate system shown in Fig 6.2 crete wove the Naver Stokes equations a thinteaction region, using 4 fintediference method. Density comout for this fw Rel (desigted ope 17 shock inracton because ofthe special gsmete fares ofthe angles made bythe two intersecting shock) are shown in Fi. 623. The Incident wee shock enters the computational grid fom he ler let a shown, The bow shack from the cinder Is Menficd by the veey sharp elasttng of onoor ines bordering tele of he inaction region, The fw dowaseam of the bow shock sa complex region of reacted shocks and ip ines, However te ‘ei ofthis type of fw canbe mae cleser by constructing sont pot nt of ‘ensity but ofthe density gradient Such a plot i shown it Fig. 624 This a ode contour pot ona ayscale color map of deny grad. Interesting to ote tht, in physical laboratory simato, actual photographs of shockwaves fan be made by means of a special opal system eal aslren stom In ‘eiiren poteurap shock ahd exputsion waves ue mde visible bythe refacton of light waves tough te Dow, which eeates pater of varus sk and light Imcnstes proportional othe magnitude ofthe local gre of density nthe Hom, Hence, the flooded contour pot in Fig. 6.248 rally» CFD-genraod sheen plore ofthe low-field, anslgost in very sense fo a seren photograph {hat would be obtained in the laboratory. Te llsates acter subestegoy of ‘contour plas and shows he remendous verti ofthe whole concep of Sntour ploting “The author wishes to thnk Charles Lind grate sett the Un vesity of Marland or providing these figures oboe part of his doctoral research, “The conto plas shown in Figs, 68 and 6.19 0 6.24 are made fm two