[S| . Revised Simplex Method 6-1. INTRODUCTION fcatee eedenitinion of Ge The usual simplex method described so far is a straight forward algebraic proced oo disadvantages : sequence of calculations in the usual simplex method, however, leads to the follow Re a0td edtigeien. (@ It is very time-consuming even when considered on the time scale of el Hence it is not an efficient computational procedure. ‘i ther never needed at (ii) In the usual ‘simplex method, many numbers are computed and stored which are ei the current iteration or are needed only in an indirect way. At “4 fify the ordinary ii) Tt does not give the inverse and simplex multipliers. Although it is possible to a. sobetions Simplex method to give the inverse and simplex multipliers, but this would in general increase effort. Keeping this in view, a revised simplex method has been developed to eat cok te peed, consequent, yy calculations by lucii ired amount of a Bpproecn orth aise simpler meal eee hada tos odin snler waetaod i eta Proceeding from one iteration to the other in the simplex method, it was eecemny i eonioo s ; 4%, Xs. 3G and z at cach iteration. If fact, all new quantities (B"?, Xp, Cu B™?, 2) can be computed ceecey from their definitions, provided B* is known ; that is if, only the basis inverse is transformed and only such X; is determined at each iteration for which the vector is entered in basis. Thus only the parts of information relevent at each iteration are : @ coefficients of non-basic variables in the objective function z= CX ; Gi) coefficient of the entering basic variable in the system of constraint equations AX =; and (iii) right side of the equation ax = b, that is, the vector b. 62. STANDARD FORMS FOR REVISED SIMPLEX METHOD There are two standard forms for the revised simplex method : Standard Form I. In this form, itis assumed that an identity (basis) matrix is obtained after introducing lack variables only. Standard Form I. If artificial variables are needed for an initial identity (basis) matrix, then nwo-phase ‘thod of ordinary simplex method is used in a slightly different way to handle artificial variables. The revised simplex method is now discussed in above two standard forms separately. Revised Simplex Method in Standard Form-1 El T YARD FORM-1 [Meerut 70, 69 (Summer)] +m Subject to +61) FORMULATI169 Unit 2 : Revised Simplex Method AX Hayy + + Oink Hota ori (6-2) GX + AyQX7 +... + ay ky thet aoe mi 21+ ayaa ons + Gay tine = Om A635) and MID ee Tne me Oy where the starting basis matrix m is an m x m identity matrix. straint in In the revised simplex form, the objective function (6-1) is also considered.as if it were another cons which z is as large as possible and unrestricted in sign. ‘Thus, (6-1) and (6-2) may be written in a compact form as: EHO Om ,, = Cunt Oty 4 1 +OKp sg teem +O A FED tink yy at any, + ants + 6-4 ee a : :, Ary + Ayaka FoF Anne +n = Pm sagan which can’ be considered as a system of m+1 simultaneous equations in (n +m +1) ie otal 7a (©. 1, Xa. Zp-+m) - Here our aim is to find the solution of the system (6-4) such that z is as large unrestricted in sign, Now, the system (6-4) may be re-written as follows : a Latg + apity + aqnt2 +... + amt + 0,n ein ett +40, nmin sm = Oxy +a, Hayat, eo bane tLe tet One 6 Ong dg gaa os Fang + Ota tiem yj (= 1,2) 109m). ‘Again, writing the system (6-5) in matrix form, 1 An A9,n61--280,nem] F xy) [0 e ( Ayam OT 1 a9) fx) _[ 0 ipealieriie=s se where a9 = (doi » 402, +++» om a aa Sk hats ba ssa nennee iancxeugieee 6-7) can be expressed in the original notation ok de “i « 2) sla , az Using the partitioning of a matrix,Operations Researcy, hall represent the (m+ 1) componen, 170 veet a) -sponding to m-component vector b in Ax =b, we ® (HD) Similarly, correspondi 4 : A = tor by b™ given by ent unit vector which is usually cubscript 1 will show that itis of 0, by BPH 0, bys by oes Bal =U i ‘comp (ID The column vector corresponding to £ (or ag) is the o be , ches denoted by e and will always be in the first column of the basis mm Pac he correspondig #/ ae linearly order (m+ 1) x (m+ 1)) whose remaining m columns are any aj” SUC independent and denoted by 8" ca.cm (in some order). (6-10) a 1 boc Bi Alyn ADU BEs BPs If the basis matrix B for AX = b be represented by Bu Bio... Bim Bu Bao.» Bom |, Bor Bua ~- Ba then, from equation (6-10), bf eueae ie mee Deca cpa eee Coe (6-11 BS D where — cg; (i= 1, 2, ...,m) are the coefficients of xp; (i ET EPA C92 = 0 = Ogg and Cy =[Cp1 , a2» ---+ CBml - Thus, the basis matrix B [in equation (6-11)] can be represented in the partitioned form as 1- =| ol (6-12) Now the right side of (6-12) can be frequently used to obtain the basis matrix i, in revised simplex method for standard form-1. V) To compute B;". Since it is very essential to find By’, smu hy apg te lowing le mailer (6-13 (6-14 (6-1 in tion of column vectorsAWD RVI Nino Motto wn AR) TheneaKG, WN) BN ny 1 agg BLD ya” from (610) WAR yaks m1 Boj) PS BE oo BEAR Bost , Ae. XM a”, (ore CVD A very freresting result o WAN HH HO), ine rewu can bo ba ined by using the formula (6-15) and (6:16). Substituting a! front 0 SP Poe itetteh el i x ys Ts Prteresting & note from result WANE deonte the optimality, (6-17) that the first component of xf" is (ej —«) o (Ay), whol is Always Note. Tho oratest aavantage Sethe desis can de. ‘of treating the objective function as one of the constraints is that, 2)~ «oF (A) for any Am W9E ‘easily Computed by taking the product of first row of Bi !, with aj? not in the basis, that (6) Ay= 2)~ @)= (fist row of 8) x aj not in the basis, wn ‘TE (+ D-component solution vector x," is given by Xs? = Bib” CORY ; eel, SrlsHeeee | =| iss [because Xy= I! by Cy Xp = x-|S]-[ 3] ote) (619) Qn (S19) & has been observed that X{° is a basic solution (not necessarily feasible, because x may be ssepative also) for the matrix equation (6-7)' corresponding to the basis matrix B,. Also, the first component of XS" immediately gives the value of the objective function while the second component Xp gives exactly the bile ‘Geasibile solution 29 original constraint system AX=b corresponding to its basis matrix B. ‘Thus the result (6-19) as of grea importance. : ‘Now the results of this section are applied for compitational procedure of revised simplex method, 3s. [TO OBTAIN INVERSE OF INITIAL BASIS MATRIX AND INITIAL BASIC FEASIBLE SOLUTION =.51. When No Artificial Variables are Neeced. Ss discussed un section 64, the i é exse of initiaOperations Research Gat B 1m 1 ot o OL ffetaet Igo to start with the snl “Thus, itcan be concluded thatthe inverse ofthe initial basis matrix 8 will Be revised simplex procedure. ‘Then, the initial basic solution becomes =. = By 0 sal X= By bY = Aas w=[8] and an initial basic feasible solu implex table. tion to start with which is feasible. ‘After obtaining the initial basis matrix inverse By'=Imo1 the revised simplex procedure, we need to construct the starting revised si 6.52. To Construct the Starting Table in Standard Form —I. =z) should always be in the basis, the first column Bg” By"=I,.1 will not be removed at any subsequent iteration. The remaining 8.8, ....BO. ‘The last column in the revised simplex table will be x=) where k is predetermined by the .) of initial basis matrix inverse column vectors of B;' will be Since xp formula A,=min A, (for those j for which a, is not in By). Note. If there is a tie, we can use smallest index j which is an arbitrary rule but tic ‘strong. Finely, it is conclnded that only the column vectors ey, B¢?, BX” .... BC) of Bj?, Xs? and X{° will be needed to construct the revised simplex table. Now the starting table for revised simplex method can be constructed as follows. Also, for convenience, form an additional table for those aj” which are not in the basis and will be useful to determine the required 4,'s. Starting Table in Standard Form-I Table 64 Table 641) Bit PP -nit 2: mereen oumpees onetnod 66. Application of Computational Procedure : Si Ocedure : Standard Form-1 Yolems of linear Now apply the computational . procedu a ramming. All necessary steps j re of revised si ‘ n PS involved in thi simplex method to solve numerical problen’. of problem. All the necessary steps are pit Procedure can be easily understood by solving ee Mn feach of them t0 ees aied | : following illustrative example so that each step endo pete oe by. spolying oie wed more easily without any : Example 1. Solve the following simpl Mustrative Example Maxz = 2%; +, subject to 3x, + ax : a programming problem by revised simplex method. . +x S53 ud [Kanpur 96; Det 1+ ay S3, and xx 20. : hi (3. Se, Math.) 99, 79, 78; Madurai (BSe. Math. 84; Kerala (MSc. Math.) 0, Solation. Step 1. Express the given problenitn Stendard Formel Math) 63; Meerut (MSc. Math.) 80] suring th: m in Standard Form-1. i ees brain ae and transforming the objective function of original problem rl ii be remembered that the Hees sleek, ee to convert the restrictive inequalities 4 . is also treate if i i . “Thus, the given problem is transformed to the following ss oe ee the first constraint equation. Zn2—x a6 Bxjt4mtx =6 A) 6x; +22 +44 =3 Step 2. Construct the starting table in revised simplex form. Now poet, to obtain the initial basis matrix B, as an identity matrix and complete all the columns 0 tarting revised simplex table except the last column x,” (which can be filled up in Step 5 only). Applying this step, the system (i) of constraint equations can be expressed in the following matrix form. for maximization to equations. It a EB a” af? a BP) al EBD) z 1 2 -1 0 0 x 0 0 3 4 1 0 \=|6 0 6 1 0 1 | 13 xs sere the columns Bf”, B and Bs” will constitute the basis matrix B, (whose inverse is also B,, because By = rere). Now starting revised simplex table can be constructed as follows : Table 62 Additional Table 62 ro P 2 Ep 3 4 6 % (-2)+0x3+0x6]=-2 1.operations fase he 174 in sivifie tip 0 Remark. Instead of computing each required A) soparatoly, wo can also compute then sinsiert follows : f-2@ o-4 (44, Ao) «erat row of By ' fa, af”) = 11 soa { i IxCR4+0x940x6) [2]. 4 a LCD Nanay Oe tfe[a gee eet Ri dew tas obtained earlier, aie Me solution (xj =x) = 0, x5 = 6 x4 = 3) S08 OP ating basic feasible solution is nt optisnal. o which gives the values Ai Step 4. Apply test of optimality, ‘Now apply usual simplex rule to test the startin Since Ay , Ay obtained in step 3 are both nog Hence we must proceed to determine the entering vector af” ft) Pa Ge ee af”, spat a dikes oe wo 'o determine the vector a” entering the basis matrix at the subsequent iteration, which the criterion : Ay = min, (Aj) for those j for which a" are not in the basis is satisfied So, in this example, we have Ay = min, (A, , As] = min (-2,-1)=-2=4, A= A) => k= 1, : that the corresponding variable x will enter the solution. Now, in orderto find the leaving vector in Step 7, first compute x(” fork = 1 in the next step. Step 6. Compute column vector x!" (for k = 1), Since x, =m," a =1,, af therefore, x{ maj” = (2 3.6). by waiting {= af =(—2,3, 6) in that column. Now complete the last column X{” of starting Table 6.2 So the starting Table 6-2 grows to the following form, Table 63 rative, so the st| nit 2: Revised'Simplex Methog 1715 ‘Variables in the basis “s z xf? ‘Min. ratio rule : ane min. (5 m= x4 ) a Le Moe : T Note. It is interesting to note that the et Leaving vector Bf" Kay coturnn more column in Table 6.3. vero oht® process of 1 we observe that the rarbary a8 of Stop 7 can bo mare convenianty partrmod by aching ore must brin, the col (as we have seen in ordinary simplex mathod). in tablo 6.4, ) 1g unity at its place mn X4") comes ; : table from which the new (i the ‘key olement or pivot element. 60 We Zero at all oth = tte [= Improved) norton eer aaa ti cokan Xt” n onder to determine he transformed femark. W(X > 0] is atta degenerate, So, in order ned for mor than one valuo of i, the resting basic feasible solution will be degeneracy. nv to ensure that cycling will never occur, we shall use our usual techniques to resolve the Step erminatic i ae ion of the improved solution by transforming Table 6-4. In order to bring uniformity with the ordinary si : : it ary simplex method, adopt the simple matrix transformation rules which are easier for hand computations, Here the intermediate coefficient matrix can be written as = et wu xi! xf? Ry o 0 is ag Ue if ° 6 3 aS 9 a 3 © [Itshould be remembered that the column ey will never change. So there is no need to write the column e in the above intermediate coefficient matrix, Also, because the vector xi” is going to be replaced by the outgoing vector BS, the column x;" is placed outside the rectangular boundary]. ; ‘Now, divide the row Ry by key element 6. Then add twice of third row to first, and subtract 3 times of third row from second. In this way, obtain the next matrix. Now the vector Bf = Q {has been thrown out of the basis In this way, the process of entering a(” and removing BS” (i.e. a tmatrix and it has entered in place of X” - a X ‘write the column af in the additional table given below. from the basis is now complete. ‘Accordingly, ° 0 1 Additional Table a a 0 =1 0 4Operations Research 176 ‘The improved solution is read from this table as : ' z=1,%=9/2, . in this solution is table will be complete only when the further improvement 1 30 ‘Srey 8 (if tion, Repeat the entire procedure starting from” ® jon with a finite or infinite value of objective fu Second Iteration Step 9 Computation of A for ag” and a$" i.e. (Ay Ay). 3 1x0+0x0+5%1 1/3 -2/3)" O-«! {44 do} = (first row of By) (af” , af?) = (1 os] : + Hlice neoeeepe 1/2, =%4=0 ‘The last column of thi possible. This completes the first hecessary) to obtain an optimum sol Thus, we get y=, A2=~2. Since Ap is still negative, the solution under test can be further improved. Step 10. Determination of the entering vector aj. To find the value of &, we have Ay= min (44, As] =min [:,-2]=A). Hence k=2. So a” should enter the solution, means that the variable x) will enter the basic solution. Step 11. Determination of the leaving vector, given the entering vector af” . Compute the vector x{” so that the column x" for k=2 in Table 6-5 may be complete at this stage. wo _p-1.@_{! 9 13][-1] [-1+0+1/3] [-23 Xe =By'az”=/0 1-1/2] al=[0+4+-1/2/=| 7/2]. 00 16]) 1] | o+0+1%] | 176 Now. instead of preparing a fresh table for Performing necessary steps in second iteration, increase on ‘more column for ‘minimum ratio rule’ in Table 6-5 shich is the Mk it ion). The ‘minimum ratio rule’ shows that 7/2 is eens on) So remove the vector B{” from the basis, to bring it in place of X, by matrix transformatio: Step 12. Determination of new table for improved solution. z = For this, the intermediate coefficient matrix is : Be a xg x R 0 V2 71 ~2/3 Pe 1 -12 ae 1 fz i x 6 12 V6 t >RoG(FR) oe getOperations Research 178 First Iteration Step 1. Compute A, for n{” and a$”, ie, (dy, A;). “i poh [ gfertna {41 Aa) = (first row of BS) x (a , af?) =(1,0,0, °| a . = =6,2=0 i: Hence Ay=—1, Ap=—2. Since Aj and A, both are negative, the solution x3 =3, x4=5, x5=6,2=0 is not optimal. Therefore, we Proceed to obtain the next improved solution. Step 2. Determination of entering vector af”, To find the entering vector a”, apply the rule : A= min [Ay , Ag] = min [-1,-2)=-2=A, Hence k= 2. So the vector af" must enter the basis, ‘This shows that x, will enter the basic feasible solution. Step 3. Determination of the leaving vector BY” , given the entering vector af” . Compute the column x,” corresponding to vector ae) 10 0 Oy-2) 2. Meprtam[0 1 0 off af_] 1 % Et i ar Diels Ooo wala Apply the minimum ratio rule by increasing one more column in Table 6-7 . This tule shows that [2] is the ‘key element’ corresponding to which BJ? must leave the basis matrix. Hence x, will be the outgoing variable. Step 4. Determination of the improved solution, From Table 6-7, the intermediate coefficient matrix i = ; BO pi? pe x” xg? 0 oO oO 0 —2 1 0 0 a, 1 ° 1 0 a 2 co 0 1 6 Ey q iUnit 2 : Revised Simplex Method ae nto} (4y,A9=(1,0,1,0)} | f]=(0.1) 30 Hence A; = 0, Ay=1., Since A; and A, both are 2 0, the solution under test is optimal Furthermore, A; =0 shows that the problem has alternative optimum solutions. Thus, the quired optimal solution is x, =0, x)= 5/2, maxz=5 Alternative solution can also be obtained ai x) = 1, x) =2, max. 2=5 Example 3. Solve by revised simplex method : ‘Max. 2 = 6x, — 2x; 43x subject to xy t+ ey $2.4) +415 $4 and x) 12.5 20- [Kanpur BSc. 95; Madurai (MSc. Appl. Sc.) 89 Solution. The given problem in the revised simplex form may be expressed by introducing the slac! variables x, and x5 as £- 6x) +2m)- 35 =0 2xj- yt2x +m =2 x 4x +5 =4. “The system of constraint equations may be represented in the following matrix form : aS seer ture? as Bm BP z x 1 -6 2 -3 ° 0} s,| 7° [ 0° 2 =1 2 1 0| 3 =| 2 2 4 0 1 oO 4 O =) a5 > — at | {-6,2,- De. A!So 6shi2 in fave dga astern =612 3) Fe the basic solution. ill enter () This also means that the variable x will ¥ ©) | given the entering yector a} 1g to the entering ‘vector at Hence k=1. So the entering vector is found to be ar Step 3. Determination of the leaving vector 8 ( correspondin; First we need to compute the column X1 1 0 07-6) [-§ aT Ene ooh lt in Table 6-9. This rule indicates that [2] is the trix. Hence x4 will be the outgoing variable. x= Br Now apply the min. ratio rule by increasing one more column -key element’ corresponding to which i” must leave the basis mat ‘Step 4. Determination of the first improved solution. BP Bi? xi? xi? ° 0 0 1 ° a o 1 4 To transform the Table 6:9, transform the above intermediate coefficient matrix. Apply usual rules of matrix transformation to obtain 3 0 6 oO Ww 0 L 1 =12 1 = oO Now construct the transformed Table 6-10 for second iteration. Table 610 Additional Table a aff? 0 2 -3 4. =i 2 oO 4 3 2 |={3,-1,3 ; ) ‘leaving? {est is not optimal. Hene! 5 auins vectors in the next step,Unit 2: Revised Simplex Method Step 7. Determination of the leaving vector p First compute the column xj" corresponding (0 vector ae 1 3 0 72 0-172 =x,” of Table 6:10. umn of Table 6 rye xin" a Now complete the column x4" ‘The ‘min ratio rule’ in the additional col corresponding to which the vector BE? must leave the basis. iE Hence x, will be the outgoin o piven the entering vector 82 - “| 172| 10 indicates. that | {1/2 is the key element 2 variable. Step 8. Determination of the next improved solution. 2 Tranaform the Table 6 10 into Table 6-11 from which the next improved sotution can be easily read. Table 6-11 Additional Table: ine Bes x PP? pep | x | xi? 0 0 =3 A 2+2) 1 : ipees ° Hence Ay=2,4s=2,43=9- ae aton unde teats optimal because Ay As, Ay areal postive, Thus the required optimal sokaom 5: xy =4,%2=6,%)=0,max.2=12. Ams. Example 4. Solve the following L.P.P. by revised simplex method. to the constraints : Max 2=3x; +x + 215+ 7x4, subject xy +3xp— 5+ 4tq S40, — 20) + 2x0 St3— 24S order to make the lower Solution. Step 1. In 3,x4=y4-+4 in the given =n t2.m= yt 1 Ht eed subjectto 291 +3y2— xy22,%21,%23, 0424 bounds LPP and obtain the following modified problem = 2 =3yy +y2+ 2ya+ Ty4 where Z = 2-41 35, x, + xp — xy + 3x4 S 100, and of the variables zero, we substitutewee — 182 Clearly, the problem is of standard form-I. In matrix form, the system of constraint equations ean be written a8 : v » gt gem n° Me ay Piln| po % ah a oa ta ye a ee er he lvadapae 6 Bh alee ee eee a 2 Beem ndichink ty faint eaPipe Lee eal g oot Yo te B Seba e: > Step 3. To find initial basic solution and the basis matrix B; . Here X,"=(0,20,26,91) is the initial BFS and basis B,=[ Bo”. BA”, BS”, BA] =1y (unit matrix). So By "= Ly. Step 4, To construct the starting simplex table. matrix B, is given by Table 612 ‘Variables in ] Solution |x=xf thedass |p ap] | mt a | Min Ratio a af? ad) of? eee I tae ae some 5 (min) a 2a “ oO NSF Site teanelee incoming vector Step 5. Test for optimality. Compute A, for all aj” ,j=1,2,3 , 4 not in the basis. (Ay, Ap, As, Ay) = (firstrow of By") fay”, a” yas” , ag?) -3 -1 =2 =I : =(1,0,0,0/ 3 3-5 _t|=@3,-1,-2- L 1-2 3 Since all Aj’s are $0, aaa ponding to a” is given by 71,3) {= af? is the outgo by 4 and then add 613, pia183 Unit 2 : Revised Simplex Method Vartobes inthebass |g ae gure Cla Cpa toed 5 tals WA aire beter H ba 1 6 Otgotty vector Step 8 Test of optimatity for the revised solution Table 6:13 « ‘We compute (A; Ap , Ay, As) = (first row of By!) (af, af”, a4! f=3.=1 -2 9 ‘a 2 3-1 tL 0,, the solution under test is optimal. So the optithal solution of modified LPP is, 1 = 63/4, y=0, y3= 23/2, yq= 0 and max z/ =281/4 . ‘Transforming this solution for the original LPP, we get the desired solution as, 4) =y) +22 71/4, =y2+1=1,x3=93+3=29/2, x= 4 +424 and max z= mak (¢ +41) =445/4. Ans. }8 SUMMARY OF REVISED SIMPLEX METHOD IN STANDARD FORN-1 (COMPUTATIONAL PROCEDURE) [Meerut 90; Raj 81] ‘he computational procedure of revised simplex method in standard. I (when no artificial variabl eeded) may be more conveniently out-lined as follows: enc Step 1. If the problem is of minimization; convert it into the maximization problem. Step 2. Express the given problem in Standard Form-I. After ensuring that all b, 0 express the given problem in revised simplex formas explained in section Step 3, Find the initial basic f ’ = a iG x method as ole € Oban the keUnit 2: Revised Simplex Method ae (@ Te find in-coming vector. The incoming vector will be taken as af? if Be=min (4) f0F : which aj are not in the basis B, . GD To find out-going vector. For this, first we ‘compute x," by the formula = Tee ee BE fe rte ano kad ‘The vector 8,” to be removed from the basis ratio rule. That is, it is selected corresponding to such value of r i tiene amas eee 8 min Banoo] Ark Note. Hewat pence chro ecm er copra a. Gi To find the key element. When a" is the in-coming vector and ;” is the out-going vector, the Key-element x, is situated at the intersection of th row and kth column of the matrix. ww To transform the revised simplex table In onder to bring ax” in place to B,"), we proceed simplex method ——— the new (revised) ao table. it ra eo aa In this manner, we obtain the improved BFS. E Step 7. Now again test the optimality of above improved BFS as in Step 5 If this solution is not optimal, then repeat step 6 until an optimal solution is finally obtained. @. Give. brief outine for the standard form | ofthe revised simplex method. [Dethi BSc. (Maths) $3, $1, S01 EXAMINATION PROBLEMS Preeti eee ee - Max z= m+ Max. z= x1 +20 3. Max. z-Se+3e ‘subject to the constraints : ‘subject to ‘subject to Su+Se Ox =0 yt ey =3 xy tim tm 5 3x, + 32 +45=6. ‘Then express the system of constraint equations in the following matrix form : of af ey a af Pius o a? om mL uae eee ee 0 One eal (osc | Ale 0 1 2 oO 1 oO * 5 ek Se ee ee