Scribd
Upload
69 views51 pages

Ot&a Sensitivity Analysis and GP

optimization techniques and or

Uploaded by

AMAR SAI
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
69 views51 pages

Ot&a Sensitivity Analysis and GP

optimization techniques and or

Uploaded by

AMAR SAI
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 51
comm (18) (23) and 24), we have 2)X +1 IY AWOL 1p is Misa ce of % andy Tl ers, Ths An MEMEE ean nt bey gm in which ease clearly, i «BFS. ha of ri 4 one fo one correspond, the set of vertices ofS) 0 solutions and degeneracy. To see thi rem Letr= p 0(iLis non zeraby previous ss 201fa/ <0, Theultiulty comes only iad >, ‘Toensue that all tcrmsin (28) are > 0, we tke, any term in (28) say sould eve in rer that when sew the new soh nee a Ag should leav (25) at once gives he leavin ¥ Scanned with CamScanner Thenew ere is changed 19 ‘wenave, 0-07 4 Ig" = entries inp row of Table ‘Thus, insy—row of Table 1, we enter ‘According 1 above ralue in rp ~ ow. Here x enters, Sincs and by Ry Table? 9,20.Hence for maximum Pe, af 4 eg, = Cf a! column for whi ‘on xy This determines rule FA, with respect to B. «entering variable and then about the leaving v oF As— he pivot element. These fac +ea/tt eal) e9, columns (= m + ich 4-0, because c= ¢,~ ¢," O being the cost of basic variables sy. 5 lake the variable to enter to be the one having most is example only on 4, and Linear Programming Preble 03 olumn of Table 2s changed 10 com by 3, Land -3 and add row, otherwise the linear syste wil change the asi columas are 4,4, and Ay The non-basc variables whose value «zero. If inthe then solve, we will get the BFS of Table 3 namely 4 je matrix of Table 38 B= ‘use only the piv tn Table 3 the basic variables ly angle are c0-ordinate vectors ) For a formal proof see Append f-(0,3.0 3, 0)(82.— 12, -$ay=-32. in yf, entry below x inxpFO™- “ag tow are relative cO3ts =e, in “ros, we sce that he esis below va preteen, orsaring ON 9S Hida roa fomal rotate Append Scanned with CamScanner Introduction to ops in a that the lower part of last colum, From (22) we ives From arproof see Appendix I-A; (equation (9) qn 7 in the last column is the value of the op — tig Oandx,=9, th proces, The most negative valve in x5 ~ OW thai my 112; hence x enters, For leaving variable we consider, 7 (ooly>O entries in, ~ column ae taken). The minimum sf jable by (J), leaving variable by ( to make pivot ted row operations, The 6). Thus npn hata Ofoallcaluns Tas ion from optimal table 4 is, oda hein ™ A Aye : HM and maxzy=11. © 41~4, 0 being non-basic vari first entry which "maxinizaion LP bn of pe 6) he rion. This happens beca unique value of, ons of simplex method converge giving, «and consequently no bass wll ever With a basis) If degeneracy is present Seetion9-1), We have seen in Theorem 'may correspond to many basic feasible ots and Line ‘conwen = Programming Problem happens tha theresa cycling of simplex iterations coresponding s corresponding to some vertex. Due to th fgeems to oceur inact egeneracy without ¢ 4-5-6, Theorem Maxf(W)= CTX, subject (oAX™=b,X2 If fora BFS, Xy" all 420 other words, the optimality condition foreman columns of A (Note Proof: Let where 39 (i= 1 tom) is somes, he cost of sf Let Ybe any other feasible solution. ton. Y= Tito n), we get : “+ OiGqtY2%e sore yan) A =b. Given that ¢<3f forall)~ tor. From this, we have FON= eit eave tn Scanned with CamScanner ‘This proves Tostartthe simplex method we MUST have: ed in terms of on basic variable, — (0) entries, ot ening ons) sins toware > o allen below variables nro mn ssfotons ntslon robles, mn, canbe converted int amaximizaton pen Maximize — 1X, : Me ea nee ~ Man fo. 28 owe the simp en it plex method directly to solve a minim Simplex method format lum) with me " : ensues in gn Pesitve coefficient in xy — row enters ener largest arabe Bes decrease inthe objective function. ars. Semined bythe same rule asin maximization probes ties, tow condition ot esoM tables) are <0, the solutions optimal 5 ~ 480 forall columns of 4. » " ° 2 2 Entering variable: In Leaving variable Min (12/3, 10/3) fe pivot row by 5, then with the help of thi tion, We get Table 6, Second iteration is shown on Tables ° 4 ° o _ ‘Now all entries below variables in xq row are <0; hence the above table gives optimal solution as, 8 y= 157x,587 and Min, xy=~30007 Scanned with CamScanner ee EXERCISES BrERCISES 1 cot aye ttt Se sont ayo $8yt 2 0, gay Byt tS gy3yt FAS 20, tie be Tord ate O and 6, (0, # 0 ;)respectvey. Which ein order thas ager increase (4ete0) (6) Enuytelow i (@) The BFScoresponding to above v2 neva, place ofthe ee, 13, Solve by problem \ Mia 24 By subjectto =42S 2 Byrn s 2m SPA but bounded, Scanned with CamScanner Introd ee i u = for abasic variable A, Show at =4) eee fag TF : reg sda wie 15, Coie the LPP “ . a tet db. z0i nd GT. LL3™ Bean on py eae et, fo) taicbechaned © AC, >. WiLAY Bean orimal BFS of changed pra taibe pie be chnettg += Hon WiILA® be an optimal BFS ray ACfa'-Ae,=A(CFa’—c)) Meo, in LF as soon, leaves its: oral vibes cin bass, selon bet + ARTFCULvaRAaLES TEcHMaUE ave en ar he fan optimal BES. Then show that the LPP Ais prensa H 2epton of Simplex method ieee LP ett ety mai ase weet opes Bain nite te identiysubmatts bene conta 14. Ecampes 7 Pose. For example, consider ibe problem, Min ijt, 7 B42 rite 2 2) PE convex Sets and Linear Programming Problem ‘Transforming to standard forin we get Min apt x,t 2ny @) ince the objective function is 1), But the standard form ‘as a sub-matrix, So simplex method can not be ix artificially by adding artificial variables R,¢0 xyt3q 4" 3, lack (surplus) above, we demand R, > 0501 ions of simplex method. ify matrix is given by Ry.5, and , columns. To: rable, The difficulty is that a solution of ables, we shall denote the a inat the above system satisfies problem by adding to. rnumber. In the case of mes sum of al this problem to be the bjestve function > 0 is abig number. We wi he modified problem of(3) is, ign St 2 MR +My, Min subject to, 44-819 R= 2 ytd th ro) Scanned with CamScanner ystem (4). The above lex method Min. Fa can no 4 rvige the objective function will become sufi si haste given piblem bas solution, then the modified p ‘a solation with ati bles equal to zero, The simplex method soltin, For applying simplexmethod to system (4) we make Table 1. ‘Table! ings ow Ryan ing Mies Ry tw then ute here 4.88. 4483. Remark: An arf basis, dueto obvious ‘Once it leaves the basis its purpose is served. Therefore their columns are not them the temply, This is because then fe penalty M for R, would have given i ‘Two-Phase Method Inthe Big-M method, its not define not be used ina computer be suitable for use in a comput to get a BFS Of| se this as starting BFS of 2) and apply simplex method ‘we obtain a BFS of (3) wth al 3 variables ea , mn ofthe original problem. Starting BFS ofthe given problem, we obtanthe optimal sol Phased ‘We formal Min the following auxiliary problem ables all variables > thepr introducing an appropriate 0b aratng & BFS of given system. Now we ue simplex method to solv systems (3) Ra ° o a 0 alo 11) and Stating Table (ST) are shown in Table 3, To get the Thais the table in which, ~ corresponding to basic "The most positives, ~6 in Table 3 is for). (eR, leaves. Hence enters and by minimum rat Scanned with CamScanner ‘astubleoftable isthe otal able of Phase, giving a BFS of given prose 47 12,470, 5. Remark: Remarks I62and 1163 of Big-Mmethod hold for Phase- also Remark: As in Big-M method i original problem has a sol | bt the Phase problem is always a minimization problem, $64. Remark : fin the opt ictal variables are not | Shia cgaeio (leary en Minry 0), thea the original problem has no soluticn thang Spisempry, ives starting obasic variables 2,3, 2er0 p-Fow. “The crossed entries in TOW are table of Phase ‘Table 4 isthe stating simplex table of Ph le. In this problem the above is also the o-Phase method. 4.64, Example : We solve the following problem by Two-Phase m Max a7 Ax, 39H. subject to, 2x, 442, +644 Ry 64,44, +645, Ry bles > 0 a Phase-I : Phase-1is shown in Table 5 Tables is gives the optimal = ‘esse rere 3, 203.45=0.MaR x9 =— 1 Scanned with CamScanner Untrodvetion to Optin Nethey vee get Ry" 49/8 #0. This plies that thre isnot required, m0 Suppose we Ax Bisposibleonly conti eit a ‘asc variable at some sage Ry- Ganda =O forall san-basc veeton i Be New exep the enty a= 1 Thus all columns of A ate expressible asa lings eobinton ofthe clurnscoresponding to remaining m ~ 1 basic Shows thatthe rak of Aipm =| and not m. Since R, = Othe rank oF augmented mn {UE isalsom = 1, Tis intr implies that the constraint is redundant, Thus the evr? row ante removed from the simplex tale, The column corresponding to atu deleted. one ot more artifical variables remain inthe optimal table of Phase. level, theneiter the ania Phase othe iterations, ifpos ‘arable becomes positive at any fromthe optimal able of Pase-1 we ean us either of the (2) Woptimal table of Phase, shows any redundancy at opie ay ane. then that room the al (6) Toremove abasic aifcal variable from the optim rghit optimal tble of Phase-, wecan se ee conves 1 ets and Lineat Programming Probiamy 3. Min subject, 4, Solve Exercise 3 with Ma, 5. Min 6. Using simples method 1 find following system Use Ph 27, Solve Example 1.246 by both 8, Solve Example 126 withthe fl Outcomes Inc abjctvefenetion by using 110 AX=b 20.820, jurplas variable o each constraint. Let, Scanned with CamScanner lehasto be used forthe new Tressow nt ix~Fandstyne cn aol SOME UE FE Rew yyy ee ucedsimomne nda procedure disused above _Prceane S 1:7, EXCEPTIONAL CASES IN LPP We discuss here some important exc sinopex method. The eases discussed are (0. Non-exitng feasible solution. Degeneray. he ptional cases which are often encountered inthe beforehand. Ifthe standard form ofthe problem doesnot involve any able, this case will not occur. Ifuse of artificial variables is necessary, the problem has no easbl solution (that is 5, isempry) if, (0) Anartfcial variable is 200 zero inthe optimal table when Big-Mmethod is used (@) Minimum value of the sum ofthe artical variables obtained after the fist phase ‘of the Two-phase method is noe 20 The fellowing example wil explain (o) and (b) respectively (Solve by both Big- and Two Phase Method), 174. Example: Min ag" 3,-p subject t, athe h 244322 6 mz ‘See Additional Examples 10 nd 1 178. Degeneracy ‘one of them to leave the basis but the new basic soli mers ‘ew basic solution thus obtained will have the ing such basic variables at zero level and s this new solution wll be = nama will be degenerate. () We may obtain a degenerate optimal solution. (©) We may obeain a degeneracy) noo-degenerate optimal solution (This is called temporary .71¢. Unbounded Sol (o After ap appear again and ‘There are metho section because The following two examples when solved will exp 1.72. Example: ‘Max, subject to, Unbounded so von ay ay not Be unbounded nty refered above inthe y-row ina minim “The ease is explained by following examples: 4:74, Example: 45,42 <2, ‘452 4 %20 (unbounded unbounded sol 4-75. Example: Max. subject 0, ago 48, (usbounded solution space but bounded ‘optimal solution) See Additional Examples 8 and 9. Scanned with CamScanner edu Open 1. Finda he ternative cera enpresion | Max subject, the 12g Hey by Hay FS 10, athe § nSh 520 Max, = 6-4, . | subject, athe h | Bons | mats h 520. a) 4) +45 3, 20,4520), Then Shas no venex,bectuse (@) Sisempty eco () Sisnot bounded ruse x) af 4 22 <1). Then Stas : | ee 0 tenner ot vere, 2 (S~ Wensd f+ 3 St). Then eve intro pit of Sisa cleat 7 ‘exactly two vertices of S, ‘ nize (€)_ more than three vertices of S. oe ® 4 hete= (sya) ix, 42,2 [Ser adingts2 (©) Sis ot bounded fom © Sisnot boundes, ‘it vertices re (1,0) and (0,1, 2 0,4,20),Then ever (ry.3,)€S, isnot ace ofits vertices, ©) Spisnotconver, o (5, isunbounded in, drection, o cost sc 2era fora nn-basc variable, then there exists optimal simplex 49: Lethe optimal ofa LPP occur atmo vertices (BFS?-X, and X, Then we know that it aso tccursatevery X=(U= A), + 2M ORT ) Xie abasic solution, ()Xisa Bes, abasic solution. o 10 ) Number of BFS Number of vertices }) Number of BS = Number of vertices, )) Number af BFS > Number of vertices. o 5 crresponds to sme Vee u bein simplex table, then (a) negative, (6) positive, « simplex ethos ines the objective Faction ofa given problem, isunresrited in sign o 13, Consier the system tnd S22 2 “Te slain ofthe above system is basi tion, (6) base feasible solution, (©) notataie solution « 4 lea LPP in standard fm there ar six variables and fur constraints Then he numberof BrSare wos. @ <1, (@ 215 0 Scanned with CamScanner tngoductonto Optimization 14. emerson of we comen SEES IS () wseaseanen st, mt Scanned with CamScanner sex oc may ave Bo Sopeiag wp other coOUFCAS wo @ ie Soe Foci Tees ER yr gen Panes as simi me me aS varie reals a pORvG level i te @ ein = Remmmone Pare e Sy momanat 2 2 he pening ime wt tc ® e © » E as ciher only coe point (at is siagletoa) oe has infinite somber ef ints. Sand Ye S.thenall eco (abich rife in pomber) belong to SABE (0) sxeconnt 2 leeimig eitlonc ene ter ion ofte sixplea aedat e 54, Le i simpen tbo the variable, leave the has in some teraton. Thea, () 3cxa exer the sis se next ierion, (@) 3 cannot ener the bassin the next iteration, (2) Bosh (2) nd (6) ae posible depending epos the LPP. of, willbe neguve ifLPP if Min. and postive if LPP is Max o lat: The news sets and Linear Progranming Prien, my pesmme agin sereice 4 eras the Bass The. {septal soliton snd Xi fea soliton, We hve om the defies of Ox OK 2) [Next 2 is optinal solution and. ea feasibe solution (wes com vectors C.). llz,> 0 and integers ESOS Example 1: Soe Exercise 1 of Section 13 To prove tat Aye Sisavenex of SUTA=5) 5) iseonen w A= (-94,44¥, 05151 “ Ath thes X=, 1.=1 ten X= X, These eonrdict (2) and (2). 1F0-<2 <1 dea 1y. Formulate an LPP give (4) conrad tat i vere (definition of vere is covtradited). that * thatthe numberof complet Teas vehwe stom at Xe Sando, Hace e Tha neon retin Neat we assume thigh hand ie of (1) and prove te left hand side ee ne ‘Assume that Ais convex: Then prove ta Xi vertex ofS. Assume on oor) tha isnotayenex ofS Then threes 2 onde Ssh hat Dui per production ran he From this, we have A #hX,0CLED Hie SandX, 2%, hence £4 = He Sando hence Ned | inc isconve, he above tw show ht ence rove “ ‘Te remitof this exercise canbe generalised that. smote eeneralized that is conv Scanned with CamScanner be number of rodeo un at vrehumer ofA pars manufactured = 35, +342 + red = 45, ‘The purer of C parts man 226, $4n #555 =%5 “Tre number of complete ems made from above are [4 42, } esses ence tee akoe Cr Tesin stmt ing nr tte fe} sa, nye drssneane 1 2 omdinege “To conve the above into an LPP, use the method of Example 1.2-6, observe reuteneat That the Example 5: Consider the following system Bt yan 42q = (o) Cans and, be basi vibe reason for your ans ‘Equatng a zr al variables except and the system (1) reduces to Be ed Introduction to Optiny — Pinion 1 and Linear Programming Problem (oy Answer (a ight hand side of 2nd equation in Then we get at ye? o ayo 222 aft ‘sin 3 Hence det(8)=0. Thus, and, can note basic variables. In tis ease the solution of (2) does note Can x, ands, be base variables? vaables excep. and 40260 the system (1) reduces to ea dat td Now [i] =-1. Thus Bis nor-singula. “ Observe that basi variables be, and degenerate basic solution beense one ofthe 0. Thisisa though al values ae same. Thi fated wi aay tayt 3st St sobjectto, ey 6Qy~S1y 45> 20, dey - 26, ty Paes 1O, fny-3ept Bx, +2545 20, ails, 0 jn abe is Table 2. ‘one ofthe simplex Scanned with CamScanner column in A) + 6, ~ cal [- ~ eM 02,727. Simi he cal (a b6) of in Table 2, we have Pee ing LPP by us 24 804 unvesited in sign Metn1"3, 47134), 54 The above LPP becomes Ma Sy-ty, bjet 0, tH s 0 PW Sy> 6, Yotyy2 0. Now solving by Two phase method bettnae sp tions are Mateos in Tableris 2 Hence column of, in Table 2is (0, 3,2) 2 & - F Cresent A oo +o fo : et 6 s # [ 7 oO 3 x oo 4 S a i oo 7 a | as 3 7 = 6 = - ° o ° 3 s = 1. rr ‘ ; 3 Tesi rat | | -ele|-elef> Scanned with CamScanner + 06,404- 2 4,6 08,4 OG Ls Wn ata Hy 5 *2-TRen above LPP econ sie, Wt yt dy Ty4, "5, Wo Ran tyra. allvaables2 0 5. The solution is unbounded Tables vex se and nea! Poaranming Potem os Stee 1 ° a4 v 3s o4 4 ag0 Be, #24, 3M, Scanned with CamScanner DXA ; jee ‘Thusthere exis nfiteumberofallerate optimal sl 1Obverve tha X and ate sie feasibe solution but isnot r staked = + Then, 2 nota vertex ofS, bu and X; are vertices and. segment are alternate optinal entering x, we find “This apni optial ond Scanned with CamScanner gets ana nen” PRA rein r wren, FLOW DIAGRAM OF siMPLex METHOD Gon) Anorrranny coon 7 none irr mee a " Convert above into the startin lel sy on eperein. Then sro becomes ENED FOR ALL BASIC VARIABLES. 5 - Scanned with CamScanner t ts and Linear Programming Pobin Introduction 6 Opting conven S00 808 ara | qm ‘SOUTION corral GD OUTION s point of Sif every end of Ky ong wo $1, acd not be x poi led am interior pont ofa 4et Sc Bi there exits an end of ih Lssounoeo CD fro (2) FSAACE PeASIC vasuasie or ABLE BY PROTING ON ELEMENT of Increment in component of X ves of /(X) with respect to alls, (( = 1 40 exis then the gradient pad fAD= BSED (OPP nne Df Scanned with CamScanner at Duality in Linear Programming and Applications of Linear Programming DUALITY IN LINEAR PROGRAMMING With every LPP we shall be asso’ ing another LPP. The earlier one i called the primal problem, while the ‘second is called the dual problem: We shall see that the ‘olution of primal and dual are closely related. See Additional Example 1. To be specific, it is sufficient to solve either of them. ‘We solve the one which has smaller basis matrix (that is less number of constraints). Besides this, there °° many other advantages of duality, which will be discussed. To define the dual problem of a LPP, we first define the canonical form of a LPP- ‘A linear programming problem is said to be in canonical form if (2) it #8 2 minimization problem, () all rroastraints are of > type (c) all decision variables are 30. Note that there is no restriction rea the vector B Iti clear, that every LPP can be transformed into canonical form Consider the LPP in canonical form : Min. x= SO= cx, AN) subject 10, 2) 2% 3) where C= (Cyrnn6als X= FA = (A,)mewe 6 = Ore .b,) are defined 2s earlier. pets. Definition : The dual problem of the above problem is defined to be the LPP. Max. yor 00> oy, 4) subject to, ATYS C, “6) 6) where, Scanned with CamScanner -- fed the primal problem. The variabj 2). te bes ihe o t. The constru dual ae be flowing eXaMDIS: | sare: Forse i al ofthe allowing LP ws 207 ST , 9H % abet i nat 38 6 aihns 2 422 sph efit converte Fst problem into canonical form. The eanoni Min, ge — Styt Tx, ical form ijeett, = Z—% ” dej-342 6, i axj-82 - 12, a 422, y sym O We atsocite dul variable y ¥9, 0) Min subject, >, _ ramming & Applications of Linear Programming us dual, by definition, is Max. p= BY, a subje AYE C20. ‘To formulate the dual of (i), we convert it into canonical frm Min, -9()= -6'% subject ~ATY2 ~ C20, ‘The dual of the dual that is dual of (i), by definition of dual is Max v= CO'Z, subject 10, CAN Zs CY, 220 or Min, -v@= C2 AZz 6220, subject The above problem isthe same asthe starting problem This where Z proves the theorem "as a consequence ofthe above theorem, the primal may be inthe either of the problem with > type constants and non - negative decision re canonical forms. and (6) bot 24-4, Example : The dual of Max. (00> S546 subject 0, nyt 9p 60, det Bes 45, S30, syne Min, @(2)= 609% 45¥3 #2005 * 304 subject, BATES responding 10 es problem as primal Scanned with CamScanner enn se peer Beate 215. Eramle: Max, stats . o. inequalities and bring the LPP i eal the xoog We singe Be 2) Then since IP i eastle secon of Let Vibe any fess (eye base foe Theres tatis This proves BH Sesto ncerents = - ee EET SS etl me le pc Te san oC Oe Sate Se aoweg eee NE Frbless a closely related We sal take pcan! . 4 een aE, i as 1 focal feasible sotatioa FoF CM Ma 1-1 : eae . ctx, yes that 1 i the opeinal station fC Bieta At2 oo eave: rite print prosen (7) ss an optimal san ee 7 ex") des} prablen (00) oso has 2 optimal slacion 2d 6) ee "fay = Mex 00 yaaa Scanned with CamScanner = rpine ry actually contig 1 optima si, prove above es on eign of pina. Iroding PIS Varley proot ordeal for 2 0% oF were pay v rysmplex method. Her Bis not necessarily 2, that Farripostveb, we shall have fo add ar vex method format and then solve. Firstly, we take columns of , hence = Cfa’~¢ £0 forallj= 1107, ° CEM, < 6 forall that is cfetas ch Vector or, Having Tat psi cfe'gz0 < r=¥0, This prove that Y= Cf 5 that "= CF 5 is feasible solution of dual, unt mune Proper & ecko of Ut Now it rematas to show that iis a2 optimal solution of doa!, Consider om= ur = b= C]e1 b= Cf Hy“ SO- “Thus the feasible solution Xan Y of primal and dual ae such that (80) "9 n:beneeby Theor 217, Ys optima slton of el This roves teeoren last one Example 9). dual has unbounded optimal 50 Primol soe dt subject, 41-4 5-2 nate Max between pri vz an optimal soltion then dual has also an oPt 18). an unbounded solution then dual has no feasible ¥ empty). See Exercise 10. tation, then dual as imal has no feasible 1a See example 2-19. This follows from unbounded s and Theorem 2-13. In applications, in general, We primal as fa the application of above Theorem 2:1'8 shows [Naturaly, the one with less sumber of © ihe advantage of LP. The problem we solve is concerned. the primal problem. is preferred for ing duality theory. i The algorithm for solving a ality theorems. There are many other is one of the powerful concept of LPP. rentary slackness theorem brings the 1g theorem known as complem more closer. ‘appears as a basic variable in Scanned with CamScanner sponding sal variable ¥ €077¢5p6 Sin he optimal ‘Then nthe optima able of pri lum inthe bass. Since by hypoth +0 because cost of fis zero. ‘This proves pan (0) ofthe theorem, ‘Nex Hehe varisble x bea basic variable inthe optimal solution ofthe prima! Wetne 44> Cat 6,= CPB, 6,=0, eis FA" Ft Pa tI a= Teed? cel consaiot is 14, < c, Inthe above = CEE isthe optical at solaion and eqaliy olds. This proves the result. Tos shove theorem canbe verified fom the optimal tables of primal and at, ofthe sem soed example To serify some intresting primal-dual relationships, we solve below both 22 rials wll as al by simplex method. 2441. Example: Primal Duat ST Srey 3ry Min. 5 307144005 bit FS 41-55-65 < 403,30, Scanned with CamScanner =09,-0, Ming = 150 Min ye =150, habove lations we verify the theorem proved ee eis ame for primal and dual problers the matrix under the starting basis (ot hein erse bass : issinplen ble for matix "ofthe bas bl, for example the bass matrix B corresponding, a aysy=($ 9) the method offi of finding inverse of Mi a eB then by eleme fthe mati Tne an ati 18" Tab PT, we open ean be veried by ine liectly| roger the coefficient in, ~Fow correspond ing coefficient under and si” fo 15 can be used to obt re shal show tis follows. ion. wi ‘This is done 2s stop 1. ‘simplex Mul ier is defined to be = (ry sna) Cf Tis the cost vector of basic vé sles. In PT2, the basic var! ables are where, Cf Jind hie cost fom objeeive fonction ae 2and 0. fe ny and 15 0 For P12, n= @.0)(""§ §) =25.00= (4%) sc cetiientcoresponding to present bass (REE a £2 where Cy denotes the Co step 2. Coefficient in xp~ row in PT2 below Ry vay (Coeticient off in original xo eauation? 2 2 = 2ean-5em = row in PT2 below 5 “= ny (Coefficient of s, in ori = 0-()=0. ste above ae the entries Below R, and in %9—r0™ i8 PTE rete Remark: To find the optimal station of dual for 6° imal solution Ma eee a follows, Caeulated Cy forthe oni! Petre le. BY + Prvas gives the solution ofthe dual probem. Im the above ‘example from the pina solton of a3 one cf. (1 6% from ODT the optimal solution ofits dual (that i primal 8 20 Coefficient in xy inal zy equation) 5-1 Note that by above the simplex multiplier, corresponding 1 optimal able ECP the ston of ee (ua) bien rer torne ot (50 (4 §} = 6.0% Here simplex mul implex multiplier 11. Take TT fe solution of the dual or not 24-48. Remark : Every primal table has @ ‘corresponding to PT2. Let us see whether this TT is feast Scanned with CamScanner 45, This consraied Cxtesprnds pengascosreit" Zs Ts np sens tis beat Ligebs 4 5 serch wnat 1 S13 spews gpk -se0= 222 ‘Test tal cosas is By hy fis oe 2 2 Tis onset eoteryey 923. This ome 2 genet wh one g Te dine gin Be ATID n wh este wi te beri sey ti 2,85 tum ig YD Oleh site tip comvesyaicn so 5, -G-4- unr th he Nema te sy wegen able ot 1 eve te fone age U6. Fer ate ttt aie in yy ho of A ca conten of ie furatinn ace atin A eresponds to WTA ’ i“ ay below 6 Joe B= NYE 9, 18 Ey be VID P Ae te nes cree ge czry ble ts messes #7 mn yee can VE eis pete O) te, Seas oath, seo eames AE = se tne ny wong ae Jey b= om 89 Scanned with CamScanner sets Remar: In any simplex table fom (14) of Section 1, y¢ . Hh, lowing. Constraint coeicients unt = B" (column in A of x). = BTA, ‘Take for example PT2. -xy-eolum (not including x,- row) in PT2 90-05 | cater (2 9 (3-09 | Similarly verify (18) for other columns of PT2. in use of (16), (17) and (18) will be in sens er any variable that is the coordinate yey, analysis, Ene HCH mG esr EXERCISES Formulate the dol ofthe fllowing problem and. vetify the theorem that d 2 ‘heorem that du of du rim oem S16, i msi 2 6 Beys3e5 45, sum2 0 Formate dual of Execs Min Foe B34 44 bject to, 3 ee, Betty eine 4 Jette 5, yt a, ications of Linear Pr Min aye 24M subjectto, 442 3 nat?’ 2 ay ty2- | v2, Oozpunrestrieted in ang poem by sling te da problem. Veni complemsta?) Find the solution ofthe f sfackness theorem. Max subject, 1: See Adsitional Example 3 Find the sation of Example 123 are matrix, Obiai Su by solving its dul at gonditions on C. a,b under which the Assume that i faltowing LPP is Min, CPN, subject to AX> 6.20 hee variables whichy self Sua: <0, then tou Construct a primal LPP {)_ in Theorem 218 fin pe al solution of fe" CU0 the sign of y, changed. simplex metho t0 Find id Sy denote the set of feasible solutions of ws cafata Xe S,yauch that othe > Twit fia oon Sl eons orm made a aston of sal aye TyAseOB,Ag = 110m 06) In view of ) a (7, we ge fom (18) Faye et om l# AD flan o- 0 6,if020. “as mat lion a Fa = 14, 0BA, rake ight and sie af 19) 10 Pisa solution of dua swe have fom (19) and (10) 7), (18) and (19) show that im + Ito then for 00, isa solution of dual constraints forall 0 > 0. Thus O can be made veiber Even then P will bea solution of dual constraints. In rary eg) follows tat the solution ofthe val wil be unbounded, Thus Hence #7 Scanned with CamScanner Introduction to Opting ive. Tomabe itty those of Testis a Pian Aa 056,040 itis necessary that Tomake sight han wf solution the that to increase where z,-6,20. The procedure tobe followed in dual simplex method is stated below. 1 Express the problem inthe following format (Min orMax fun Cx. mijeai, aK AX0, tere on or orb sareneqatv sn conti entity mar 5 sbmai (0) Express aterm foots vars. Teds ingle meio is aplioptinalty conditions sted in above fora ek arog da simpler ale which same a artng stole is er tone or are erties salaton clea ncadng Toa) aceai, 2 Tesbone ges abuso (nt ase estos gies rot ele, ‘variable x, is given by ae (Frm Tg). The leaving > Ma ts,5,<0) a 22-4, Example: As 20 formulation of these che ‘corresponding to he mi ering variable. The details ae show? isthe en “This iaplies that ether #5 OF in Table 2 Scanned with CamScanner EXERCISES tl, SENSITIVITY OR POST OPTIMAL ANALYSIS Dy sensi ‘changes accured i Sa Scanned with CamScanner 8,0) (5, 7.3)" = VTS. 0) B= (5, 5,0) Letus nove change 61 “This solution i not feasible, IF we co "Jone for ealier 4, we wil get Table 2 instead of Table | 5 Then he ne soliton column i is also fasty rations withney changes in Costs thata change in cost coefficients will aect objective func ve must use new GF new entry Below 5 if "= (15, 15,0)(0,- £15, —115.0)(1, 0,0)" f= 8/5 = (815,- 15,0000. ied C range. T mabe changed ‘cams find out the ne So = (815, 15,0) om (16) of Section 21, ¥€ oye U5 =A vero, because they are Basie Hence the new table corresponding, cree in, x, leaves and s, enters, giving Ti is disturbed, it ean be restored by du o a 6 the BES remal jon of BFS, =" 3/5. %2 mplex method to restore optim a5 0 | ys o | 6s a 1 ° same, The value of Gis is new objective fun in Table 4s, enters and 5 a LS Tous | es onus | ws o ous | os “1 1 ° Scanned with CamScanner pram amvarables = The opel ble of this robles is Table & thesrate £ cfzinthe cbjectne = Sepee Ge oles enapeaing pactsk varhbl 5 (= THES Tee seplex cuter fee TaNe cemngad Fm (1, 3.1 2G. 1497 Then ts find the new optimal solstion we precea! xy ~egeatioa 2 TaSte 1 WaT SS silos: op SME? De chante i he cots of 5, the eatry below 5, fn z,—ow # Bip Oar eae 2 row wil not change. So we calcula is chines. sree sense sont “ Scanned with CamScanner Introduction to Optin bed and x, enters the basis, We muy. in Table 1, which is ow “abled isnot optimal ofthe new problem. (Optimal | 296, Addition of a New Constraint int can affect only iy condition isnot Consider Example 2:32. Let us add the cons al problem x, =x; = 0, x, = 230 remains optimal forthe new = 1pm" 0,25" 15, min xy tea fate included, Table 9 isn vara eee oe bare proper con Byron pein 3B eee vari Sato ton a Mpa ot ( wat 23F. ty In Linear Programming & Applications of Lnew Programming Table 10 H 2... maea 3 | o[-@ 2 0 [+ -i 0 of 20 x | of 32 0 1 fo wo 0 | 20 5 fof 1 4 o fo o 4 0 | 420 a | o oo fo -m 0 1 {-” ‘Now use dual simplex method to get the new optimal solution (&)= 437 0, x5 = 200, Max zy = 800) Deletion of a Constraint in a LPP, besides non-negativity ‘occur at vertex. Suppose Let there be three constraints imal solution wil We boundary, hence if Cor C, is deleted from the problem the of change. Note that dat the boundary th ‘or surplus variable coresponding tot that ifa slack or surplus variable in the optimal lack or surplus variable has value 2270, then te cor (tat is satisfied on boundary), corresponding to ths sack or sul the boundary). o riable n, Heace the opt is deleted Ifs5~ 0 in the optimal table, then, fn. The deletion of third constrai Scanned with CamScanner _— ate nest Prograenng § ockcars Unex Popes = Heoce Teble 6 is replaced by Table 11. Now caias ands, aves The oor tile hes te optimal sess of Sr problem in which second coastain Bas tees deltad 25.5 =, 0,75" 3020 357 1720. ,. Deletion of a Variable To makes, leave the bass. The me ie ‘Now use dual smplex method be st Thc em ae the optimal solution of the problem in which , has bes tnd Max xy ~ 460. (See Exerc hat is @) and $2,031 ° Ely ow sin y-1ow = er0{"]-onns 0 Cot 2 vano(2 3 Scanned with CamScanner Introduction 10 p where first (0 CErmS Bive org e hey st p is th ‘te seve ~ columa in Table 6. ‘ esol Hd EN Newstin Table (2,42) a 2 = ptlen-2| <2] a,,| +3" |2 cons oben finion at wi coum oe below starting bas ¢ . 5 | 1. Consider the problem seca coer Min, 8 4,25, ty pl Table 6; Bnet one en ctua in THE 6 \ {e) Lettie tied constrain be deeed fom he above problem. Find o ett caged ble incase whee be of. co original problem We discussed above the case where a columa is changes acia ae sir lum is changed. The above techn Lhe vals be deleted fromthe problem. Find problem, I the cost coefficient cof, is changed from +1 to —1 ditional Examples 9 to 14 carefully. fenot given. Get the original problem, vz, inpace of, ands, 1 bischged rom y + bat ey by +65) then find the lion rains eaten md sh fiom Example 2-1. Find he solution ofthe new problem. +31, 2:4 to Example 231, Solve the now problem. hangin jr tesiln a Fin ind 3, and 6,, such that when a € (6. Scanned with CamScanner tape ns, it cos 7m column (3.5). The opine Now all he ig sama ts 7 ne aad sl cen Ta tn ot incre iota ee ee eee ee ‘Tecaring he sigecine fancier & SETS SS WIE Scanned with CamScanner arly independent. Thus p columns of Be ety flo an conespondece canbe easly ay? sis rit fe ote (2) 88 Be a basis matrix of (4). From (7) "™4 si 1 *) wel rt (8) wnbere a” = (0f. 0 From (8) “hen sry) ‘This gives an importa interpreation of «/* aeate-(! o FQ; 0 aT Uy, Theva nt oiion which may not be feasible because z may be ec ve the basi are determined by the same rule 3 Crm ot suethod x= is always in the basis as the first ‘nonsingular leaving eles art. Form, (ied pi; of the simplex method ensure following form where obj where and 3 computational procedure, ‘The LPP in which identity mtr subject to, 34-1 428,+5,= 7, =2x, 44; +5,= 12, nan, #33, +8, +55> 10, ay t)-3qt25= 0, at Byyo4y 2 st det 4e,= 12 {Asin simplex method 20) A eA b= bye ig isthe column of variabl (19) tem in the 20) Scanned with CamScanner ae sna Testing, ie po mereress Erin iad acer a sb : va ar de coe whch ex te basis ad et fe the coca wig ene tetas Be stove probes ible 1 comesponding to leaving variable i ti ete teneSnfcomen of Eis replaced by & Then from (18). eck en Wee! ra : sain, tt Pefabove rena, we can find the inverse of basis matrix forexep Fetars fo Mempetanenan ata ina sip eats inaboveepesenttiog = sehen acces | oo) ce aspen SPOS) Se ow avarahe Fie rib miner) oo ET punuty n Lineat Progranenny & Asolcatons of ines Payrenmng Now we perform first iteration in Table 1. The muss £ bas already been cakulted From 21) a In above the operations ae performed inthe flowing marae? (04) etercbange ecering and Jenving varabes ata poses, Chr neeshanet fp ‘ccelue eoeplete slice colons (that is ieteting ay = bes sop. Hace, hes determine cern ‘ecuinicg colon of tering veabie optimality condition & ssid Scanned with CamScanner 182190 ut mojaq won ave suoneyn=i29 au. “pauoyod ‘ye suoyyesay x9jduus pastaos fensn 24 Uau s2}geIEAo4seG moIDg MOS — xu SZ ‘uuaq am “3,gen Surwers 198 og “11-880 Jo 91g Ho[dus Se Dues 8 aIaeL om | ee ou] °° ‘ a w- | 0 L o | ° 2 6 € 1 ¥ Laiaen sprays stu Atzanoedsas _A10) pus ,(0"1) 01 p2BuEy> 39s syaesen seg moa (62) Pe ABLE £410 nv uy S386 01 dso 10398 sup902 24) 64 35d 10 188, ee dol eg yo ued sors smu am omy mos Kod —PBueds XQ EN ce rcrego ase ae e024 T3564 29 povnba ze Kun amend SH sae ponenaye> one (Ez) Suis 29h S884) J 2186 aso HALL EL ees enone meer cee) % sn m-_m- | i ont It - 9a1aeL 192198. 8131981 PON, a ie) en- en 0) = 1 ror gn 9 oy (i are O ef olfo em of = zo rt sui one? va. °S AGB 5 age 4H 8 sonia ony ne Pompeu gy fash 1 | oa ec) ole ve MOK on Ssuy-smeyg jo ajge Suman ag 0 wong dn as es rT or soyqeizea jetoynze ppe ox Arssooousti1y zea puepURS, SERGE te a nun 2m poe oR EE 1.2L Fy novenbo ensayo au se pan ngage y00 rat estuoinay soso an pone eyduonsaaiems et {Rye Fone a aor ge eds pono} my Hote AL HOP 154 mofog sodaz ate ata pou a|duNS aN HAAN VL ‘anno eum son € 168 carat hm cage Spica IH 3910 aw sauna soo syne wut pur age vas AH we HN 1zir to o oie s/t 0} = 0 ot sito mm omy "$0 91 onanpenuy Scanned with CamScanner 27/0 Then g |r oO” The next table is Table 9. Table 9 2m Leia om The optimal solution is x, =3/2, x) = 3, x; = 0 and Max. x= - 15, 2-4-2. Remark : In the revised simplex method, we find B"' by premulti ing earlier B"' with E matrix. This brings round-off errors in a computer. In some computers, there is a built-in provision for finding the inverse of a matrix. If this is the cas, then instead of finding the inverse with the help of E matrix, it can be directly computed from the basis matrix B. This is advantageous only if the number of iterations recur to solve the problem is large (this requires judgement). If, otherwise, B"s of higher order, then to invert it is quite time consuming and is never done. EXERCISES, Solve the following LPP by revised simplex method 1. = +x, +3, subjeetto, x, -x4-24,= 5, %y 42x, 3r,+x,= 3, y+ 2t,~ 524+ 64,= 5, all x,> 0, 2. Max. %p™ x, + 2x +3, subjectto, x, +2) 43x,= 15, 2x +x, + 5x, = 20, x, +28, +3, +34= 10, allx,> 0, Scanned with CamScanner

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy