pas COMPUTATIONAL PROCEDURE OF SIMPLEX METHOD ‘The computational aspect of the simpiex procedure is first explained by the following simple example. Example 2. Consider the linear programming problem : Maximize z = 3x, + 2x, , subject to the constraints: 41422 S 4,4) — x2 S2, and x,,120. [Kanpur 2000, 96; 1AS (Maths.) 92] Solution. Step 1. First, observe whether all the right side constants of the constraints are non-negative. If nat. ican be changed into positive value on multiplying both sides ofthe consien = s allthe b's (right side constants) are already positive, ints by — 1. In this exampleg Step 2. Next convert the inequality constraints to equation, ci surplus variables. The coefficients of slack or surplus eee jhe non-negative slack: ag function, In this example, all inequality consrints being "ery Tack aiken Zero in the objective Therefore, given problem now becomes: nly slack variables s, and s) are needed. Maximize z= i) +2 +05, +0, Aitmts 24 MX + 5f=2 on ne 114245) 45220 1. Now, present the i i Step. Present the constraint equations in matrix form 1 et 11 ofa, 4 hal 0 4 “Gh Step 4. Construct the starting simplex table yes? Hr should be remembered that the values of poe "Lions alreag 41=42=0 here. Column Xp gives the valyen cease Varial Y explained in See 5.2, Sere. Thesecr & les are and s, = 2here, asi al ‘$2 = here. The complete Starting basic feas; in, Variables ag i means Zero at cach iteration. So 1, $2= 2, x) were Olttion can be pei“ in the first column. So Subject to the constraints : Note. In is step, the variables "can be immed; bash vaio. Ober vars sent *oncing i Netign 201 lately read from Tatiee dee , en 8 coke 8 2270, ‘abl Ase varabiog ms ot Basis matri ‘mich aways nave gent va ts, 80 willbe calledTable 52: Starting Simplex Table ida 2 ° ° aoe cr) 8 a mene xe MIN RATIO VARIABLES G6) Xy/Kifor Xq>0 BASISMATRIX 4 wetted bead lero ‘TO BECOMPUTEDIN NEXT STEP. 8 © 2 Tee Talia fi ren) TP GXy | aia-3 1 Step 5. Now, proceed to test the hasic feasible solution for optimality by the rules given below. This is done by computing the ‘ner evaluation’ A, for each variable x, (column vectorX,) by the formula GX,=6 {from (5-10)] Thus, we get A= Cy) —c4 0,0) (1,1) = Ox1+0x1)-3 (0x1-0x1)-2 =-3 2 RemaricNote that inthe staring simplex abo sare same as (sso. coresponding othe columns of unit matrix {bass mati) are always Zero, So here iano need ocala ther Optimality Test : f (i) Mall 4, Gz, ~c) 20 .the solution under test will be optimal, Alternative optimal solutions will exist ifany non-basie A, is also zero, (i IF at least one A, is negative, the solution under testis not optimal, then proceed to improve the As=GX3—03 +0)(1,0)-0 (0x1+0x0)-0 0 solution in the next step. (ii corresponding to any negative Aj, all elements of the column x, are negative or zero ($ 0) , then the solution under test willbe unbounded. a ‘Applying these rules for testing the optimality of starting basic feasible solution, it is observed that Hoey both are negative. Hence, we have to proceed to improve this solution in Step 6. Step 6. Inorder to improve this basic feasible solution, the vector entering the basis matrix and the vector eee rset eee wats are determined by the following rules. Such vectors are usually named as ‘incoming vector’ and ‘outgoing vector’ respectively. ane ng vn savas ietecorespnng tthe net egaive vabe of (6, Hares ein (S,-As}= min [=3 ,~ 2} =-3 =A) . Therefor, k= 1 andhence column vector x, must én the basis mix. The column x is marked by an upward arrow (7). Sac es aeutgoing vector ,isselected corresponding tothe minimum ratio of elements of Xy ee ing positoc elements of predetermined incoming vector X. This rule scaled the Minimum Ratio Rule. In mathematical form, this rule can be written 35 fens 2, x49] or a ris equation, we ge =2. So the vector Be, Xs marked with down Cappasing oth sides of hs a ae The Starting Table is ow modedio Table goes % ay (Y should be removed fromTobie 53 . eal " ————yiN RATION it S) Xy/X1) biel GY “a 2 —— = CB) 6, SF O%=9 | Gin ay ~Fensngvedor Piva vein 1) nity must occupy in the marked Step7. inorder to bring Bs =[ ingle of incoming vetrX [if Reet position and zero at al other places off Ifthe number inthe marked 1" position is othe ae divide all elements of that row by the ‘key element’. (The element thc intersection of mini (©) and incoming vector arrow (t) is called the key element or pivot element). ‘Then, subtract appropriate multiplies of this new row from the other (remaining) rows, so as to obtain zeros in ‘remaining positions of the columr Xy. Thus the process can be fortified by simple matrix transformation as, follows: The intermediate coefficient matrix is m n x x 4 1 a ica] Rs | 220 = I Apply Ri > Ri —Rz . Ry > Rs + 3Ryto brain 7 ‘a z % 2 ° = ‘ ‘Now, construct the improved simplex table as follows. ia q Tables 4 Pete oWHlina. . Ucolieans ease] | x Rey Re Se VARIABLES By + 4 =) MIN-RATIO A o ama |ino) 50% > 0) * 3 omar ° Peale : =G%e=6 [0-3 —— {negate ratiois ers ot counted) key column Sey improved value of <= s Itis of particular interest to note here that es rr meraved bls fase auton a read a : = 2a ate also co method. However, the correctness of 4's can be verified puted While trans . formula A, = C5X)— 6. °’ computing them eee table by matrix Step 8. Now repeat HPS trot 7 ond hen ntly by using the Table 5:5. pa! un 1 optimum sotut M=mon mite ag a. ion is obtained in& ‘Therefore. k and hence Xy sho et Meult be the catering veetor hoy column). By minimum ratio re [xfs 0)-atin 2 _] (since negative aio is not counted, so the second Sino ist ratios minim, remove thefts ala et cont a Dividing the first row hy key clemer er ae coe from the basis matrix, Hence the key elements 2, nt 2, the intermediate coeicient marx is obained as = maa ren :s Xe ° a ! o 1 © o aay o 0 a ‘Now construct the next improved simplex table as follows Final Simplex Table 55 BASIC care XG) i S [ atisites Se 7 eae =e 2 7 ° 7 2 =A | os 3 \ ogee 12 22) Ne ° ° O. the solution is : x, =0 , xz = 100, x, = 230, max z = 1350.2 18. 1 8 ¥ 5, Max 2=3x + 242 subject to | a bect0 rece Max 255 +7, sutecto Ae Mac 2295 + 282% aes | wegse arms noms mene nese A? sana ore aes Baitaso. uasaniogy ad 20. Mi M20. Ans. 1 = 6, %2= 12. 2 = 60} Upon. =O 04, mar 2m hee eta Max 23x + 2a, subjectto Max 2724 «on sagecto 7 Max 2-33) * 4, ut Putas 10 2ns duces nomst mre ss me deeee sons? eso Aoeett aim20 | Reet [Ans. x)= 24/5 , x2 = 2/5, = 76/1 {Ane Souiton i unbounded) rr gn es. ni26. 2. Max z=2i¢ Sx sues 40 Max z= 15 jc ae X+3y53 e425 18 memes Sates nee race me 56 | = ee, oeest | mimeo | Yana. =2.29=0,2-41 (hns.21=2, 506, 2 =36) (Anson) =8 19-2. 2° 1) | Max 20245), cbjeate 12, Max zai e+ Sq cube to 16. 2 subject xeyse00 nveoe sto 42 Sedge 18 Osx2 400 2m me? 1051+ Pagans 12 osye%00 2-24 13550 andy 2.520 {Ana Twoteratona ands 030 | X= 300, y= 300, max z= 2100) tans. as=0 noa5.0 Max z=81* 190-7, 16. Mar oiesae 3 heomsan ce | sivecto sxiecito subjetio pernerer Shree igs BRR oa Ze wr Ber on eto 2haeson se 2m + 2y +560 pee) Aunonee eats | INTU(BE commen) 21.2} billets to Tans. x}=7/3. 9 =9, 15 <0) (Ans. = 0,42 223) (Ans. x, = 2007 | Max 2= 43+ 510+ 9m 411% 1 Max 2-26 + 45 ag 2 = 2280/19}. etna pats M+ A+ ay, 20. am) + Sep Bate IS 2+ e+ On sauet2 Sibiectto he constraints Tay + Sug Sgr 2ay 5 120 Bis aaron tan ces 3h Ses On See 00 2h me ts, 58245 10 ee! mone ans ees t2 ! (Aras =8077 men0.rys87, I Rinse, Feed meen ar thst ns =2 5,0 nae Max. 2=0%, 13%, Msc fpcnteans ue subject tote contains Hn t (ee NO Me constant, mi | 999 57,44 3m 28.45 ,4420 pe ssa, } Ane. Two taratone %n 3/0 29 = 17/0, max 2=201/2) Max 262i 4% +6, Hsbc a cong HE a Poems tao 3 9012, s-0,mar reine, “TRH KS 4 Mar z= 10n On ies wieconang Mt RS2.2n+m shu One eee wk wime0 {fna.One terason ony. BASS 59,5, eg | cto 2319 8Oheto te conan manage Peon cea oa 8 Slack var *Y (Coeticentot xy the sapsan ann eaten raps ston‘One iteration only, x; = x, = ee" ~ sabjectto the constraints Nyt Be + 2X3 S10 24 +4%+3% S15 4, %< 20 Ans. One iteration cnly. Xj=15/2, = %=0, max. 2= 45/2} BB Max. 2= 7X1 + % + 2%, ssibjectto the constraints : H+ 2% 5 10 4x, + 0 + 5 < 20 Hy, X, 320. [ans. Two iterations. x; = x2=0 , x= 20 max. 2= 40] 3 Max R=2x+4y+3z subjectto the constraints xt y+ 22< 60 2uty + 22< 40 x+3y+2z < 80 x,y,220. [Ans. Two iterations. x= 0 , y= 20/3, /7, max. R= 250/3 ) 30. 32. OX; + 23% + 29x Subject to the constraints 6X1 + 5x2 + 3xq < 52 6x; + 2k +51 < 14. 1% 20. [Ans. Two iterations. x; = = 7% = 161] Max. R= 2x-3y+z subject to the constraints 3x 6y+2<6 axtaytzsa x= ye2zs3 x,y. 220 [Ans. Two iterations. x= 1/3, y=0, z=8/3, max. A= 10/3] Max. 2= x1 — X24 X3 + Xa + X5—X5 subject to the constraints : xt x +6% =9 Bx + 4x5 + 2x =2 x +2x3 +X 42x =6 x20, /=1,2,3,4,5,6. [Ans, One iteration. x; = 2/3 x2 Xq = 25/3 , Xs = 16/3, X= 0, max55. ARTIFICIAL VARIABLE TECHNIQUES: 55-1. Two Phase Method [Garhwal 87; Kanpur (B.Sc.) 80; Rohil. 90] Linear programming problems, in which constraints may also have ‘2’ and ‘=’ signs after ensuring that all b; are 0 , are considered in this section. In such problems, basis matrix is not obtained as an identity matrix in the starting simplex table, therefore we introduce a new type of variable, called, the artificial variable. These variables are fictitious and cannot have any physical meaning. The artificial variable technique is merely a device to get the starting basic feasible solution, so that simplex procedure may be adopted as usuel until the optimal solution is objained. Artificial variables can be eliminated from the simplex table as and when they become zero (non-basic). The process of eliminating artificial variables is performed in Phase [of the solution, and Phase I is used to get an optimal solution, Since the solution of the LP problem is completed in two phases, itis called ‘Two Phase Simplex Method’ due to Dantzig, Orden and Wolfe. Remarks : 1. The objective of Phase | is to search for a B.F.S. to the given problem It ends up either giving @ B.F.S. or indicating that the given L.P.P. has no feasible solution at all. 2. The BFS. obtained at the end of Phase 1 provides a starting B.F.S. for the given L.P.P, Phase ils then just the application of simplex method to move towards optimality. 3. In Phase ll, care must be taken to ensure that an artticial variable is never allowed to become positive, were present in the basis. Moreover, whenever some artitical variable happens to leave the basis, its column must be deleted trom the simplex table altogether.at ha, X19 729 4) 9 52 9.839 a) DOcez +018, t Ok, + 0-2, Boor eB eee Oe: 0 _— & oO o Ss Det to ton | Yao Yo 2. ° ! ° =) Z= 0K)Example 12. Use nvo-phase simplex method to solve the problem : Mini 15/2 xy ~3xp, subject to the constraints 3x, Xy23, xy apt ay 22, and.x, xy ,4420. Solution, Convert the objective function into the maximization form ; Maximize.2' =~ 15/2 x, + 3x2 - Introducing the surplus variables x, > 0 and xs 2 0 , and artificial variables a, 20 , a7 2 0 , the constraints ofthe given problem become By —m—m—m ta) = 3 xy —x24%5 sks 1 hae? Xyy X20 yy Xp yy 220 Phase I. Assigning a cost — | to artilicial variables a, and a and cost 0 to all other variables, the new objective function for auxiliary problem becomes : Max. z'* = Ory + Oxy + Oxy + Oxy + Oxs ~ Lay — 1a, subject to the above given constraints. Now apply simplex method in usual manner, (see Table 5:22). Phase !: Table 522 o> ema i BASIC MINRATIO VARIABLES | CX mes el Ree =, Baye as! alse 0S Sera apace 2 eee etn 1 on mesa neat faa 0 = 4 t a IS pace VOa NA 7 Deseo [7S La 1 ma nes ea a [ee ee y ” 5/4 1 -W2 O07 -14 -144 0 ripehins/ |e one Sit72 Danie affizad O57 maeeo jee Toe ano eoit WE Trblen all 420 and no artificial variable appears in the basis: 22 °° as been attained. original fance BBS. Inthis phase, now consider the ata oats associated with OP fon thus becomes : Max. 2’ = — 15/2.x; + 3x + Ory + Ors + Ors Now apply simplex method inthe usual manner. | variables, the objective Phase 2: Table 523 : o a o> =1872 3 g ] MIN RATIO. [vats] ey s SOEs oes wy x 82 oa oe ce a ° v4 ° 2 1 eee ss TAA 4 758 iia a4 Since all A, > 0, an optimum basic feasible solution has been attained. Hence optimum solution is : x; = 5/4, x. = 0,.x3= 3/4, min z= 75/8 EXAMINATION PROBLEMS ‘Solve the folowing LP problems by two-phase method: Max.2=3x,— ao mhlce en 0 3 Maxz= x) + 1.5%2+ 240+ Ske subjectto the constraints subjectto te constants with te constons 2a txe2e ant2q28 B+ Det dag + <6 xi +3x52 mytanee 2+ mtn Su sd xesd M+ne55 2+ 61-8 + 4x =0 and x). %¢20 and. x20 W$9m~ 44+ 3x [UNTU (Mech, & Prod) 2004) Tans. x =2 x2 =0 Max 2=6} [Ans. x1 =0, 1-5, max ag=09 max. 2-198) } 4. Minimize r=x,-2x-Sx,subjectio 5. Max2=3n+2u+% +44 6 — Maxz=5%-2y +30 = 2%) +49 +34q=2 subjectio, 5 subjectto 243% += 4X +59 +453) 242-1922 x20,/=1.2,3, 2y\- 34-44 Su =7 ax- 4 <3 | fans. Here all§)2 0, butat the 2+ 449+ 25x42 6 Buses e ‘same time artificial variable a; nim n2d Riera as0 ‘appears in the basis. Hence the solution) oes (given LPP has no feasibe solution) Hh ples Ban 2002 Max. 2= 2% + 3x2 + 5x3, 8 Max. 500 x; + 1400 x5 + 900, max. subject to the constraints. subjectto, Sas 10x + 539515, Mt 14 100 eS gars 12x; + 3514+ 1545225 33x, — 10% + 9% 533, Bx; + 3x2 + 415 2 6; Fah 8 naan esi ent ee, ea 4 . var inable is not removed in the problem) [Ans. There does rot A ann54-4 Big MMethod (Charne’s Penalty thethoey tional steps of big-M-method arene sup. Expressthe problem inihestandard ion em Add non-negative ar SePP Cesta ofthe ea eBES othe eft side ofeach ofthe equations coresponding to corresponding constrains ‘Tc in on alicia variables are added. it causes volation ofthe that atficial variables wii fo CU is removed by introducing a condition which ensures exists). On the other hand, if ah a final solution (provided the solution of the problem variables will appeariniie tothe Problem does not have a solution, at least one ofthe artic Targe price (por unit eal olution with postive value Tiss achlovedoysesigeng a vey bedestenated by areal) these varies inte objective funeton, Suc rg ce wl sp8, nthe lat eet okmstizaon poems (at formniesionpobien) ees routine until he optimal stu Sarg Solution and proceed with the usual simplex inthe use of Big: M-methodin solving LPP, What are ils charactodatce [Kanpur (8.80, 92,01) ple 13. Solve by using bie-M method the following linear programming problem: fax £=— 2a) — x, subjectto 3x, +23 =3,4n, +3 26,14 207 4, and x, 220. [UNTU (8 Teen) 2000) equations become : Solution. Step 1. Introducing slack, surplus and aificial variables, the system of constait 4utim-n xy +2 + 4 whichcan be written in the matrix form as te fsa 00 ees Tye Som Step2, Assigning the large negative price ~M tothe anifical variables ay and as the objective function Yecomes: Max. 2 = ~ 2x; — x2 + 0x5 + 014 May ~ May Step3. Constcuct starting simplex table (Table 5-24) ‘BASIC Ce % | ™ % x TA Aa “MIN RATIO Dates aa 4 o 4 2 se) ~ io o 4A | Toapply optimality test, compute Sp opmaT COM 4.1) -2)=24(-aM—au 40 aX — C1 OM iy 3.2)= E214 M340 CM a0) 0,-1,0)+0=m ein 2, 14M, Mh = 8, Theo Xl be entered ag ee ae) nd the Key clement 3 which indicates tht Ay shuld be removed. Now the table (Table 5.25) is obtained in’ usual manner.Mw | MINRaTIO >) sa | 7 a | Meneses © % s Ba meme ae |__| rg 5 I ——————— La 0 asm t —_+-._—___] 1—5m)/3, and similarly, r Again compute, a; = yXa— c N= Ms = (24 Jan73 x ‘| Biceausi., Sear ‘and minimum ratio rule decide the key element 5/3 so enter Xz and remove A,, fore, the second improved table (Table $.26)is formed. Table 526 Qa 2 peOURL Ocala BASIC o Cn! A, | MnN.nAT VARIABLES, a | Scie ROM Morava 175 Bs | Core RIES MENG INC ca, a ° 1 ChE gee L Ewe O Ce re To test the solution for optimality, compute 43 = GyXy~ c= (2 4s=Gahi~cs 46= GyAi~ cg Since m is as large a8 postible, A, ¥1=3/5,%:=6/5,maxz=- 12/5 Example 14. Solve he following problem by Big, Xj +2t)+3x5= 15,26, 4344 Se 1.0)(1/5,-3/5,1)-0=1/5 1, G75 lethod : Max. Solution. Since the constrains of te get) 4,20, a;20.The problem thus becomes, °°? Max. 2= 4) + 2x4 35 —‘ the usual sit applying the usual implex method, the solution i obtained as given in the Table 22° Table 527 (Example 14) ee ee = = = etd at " % Xi a | MINRATIO 8) . x == somes 4 20/5 — E 10/1 Mec | ss ° fearon 0) | 04-25 0M, ° © 7s t 7 eA] 5 G as = 5 3 un] 1 ° ene | so] a 1 Sac oie camo fa o i er eas ea | i u | = wey T 7 v6 Te * Re ss|. 2 0 1 v6 _ fabeiisse|e ° ° us eae a ° ° re Se ED | ° _e atambasteneininelsss aa basi Be ise penalty (Big) meted fe mat? ‘ay, +4922. +3%2 yor S4,and. 2220 rie izing ne sos le 82 20. at variable 20, and slack variables 520 ,the problem becomes MEX: Et 32 Ty + Ory + Ors May subject othe constrains: ee a cl a eee 2 ee maxz= 15 2=3m—%2 ali; 2 0, an optimum “subject tothe constraints © ts od 485912090/ OPERATIONS RESEARCH. 15) ee able 5.2 exam 8) MIN RATIO Sees | hws) | BASIC 5 x ar ane |vaRtABLes | C™ = ret Ww “a 7 7 peepee u 0 3 1 3 U 0 1 ay 0 a zo 1 0 ts 1 7 T 0 Q = ° ‘ pier 06 a tw cir 9 a n 7 7 a” sy ° 4 ° 5 1 2 os 0 0 0 D ow