1426 % OPERATIONS RESEARCH CONSTRAINT Consider a general non-linear programming problem having one inequality constraint of the type Maximize 7-1. subject to KD SI N20 A Re tp ote Introducing a slack variable § in the form of S’ so as to ensure that it is always non-negative, the constraint equation can he modified to A(X) +8? =0, where WX) = A(X) = 5 <0. The problem can now be expreseed as manximize Z=f(X). subject to h(X) +87 =0, X20, which is an (n+ 1) variable, single equality constraint problem of constrained optimization and can be solved by the method of Lagrangean multipliers, ‘The Lagrangesn function can be constructed as LOX S, 2) =f(X) - Mh (X) + $4]. The necensary conditions for the stationary point are a a F 7 ~ WOO + $4] = 0, a and =e. = a 2S = 0. aL @ ‘The condition —— = 0 implies that either S = 0 or 4 = 0. If S = 0, then condition * 0 gives W(X) = 0. Thus either 3, or W(X) = 0. ° he. dA) = 0. Since § has been taken to be a non-negative slack variable, A(X) < 0. ‘This implies that when A(X) < 0, 2, = 0; and when 2. > 0, MX) = 0. ‘The necessary conditions for maximization problem can thus be summarized as #. = ox, : an," MX) = 0, WX) $0, ; R20 Pertrpentborere 8 sen called Kuln-Tucker conditions. arg minimization non-linear programming : Minimize Z=f(X), : problem ‘Subject to BX) 2b, X20. @ scanned with OKEN Scannerjparoduction of h (xy » I R(X) pe introduced so that D, rede ean the const in theca * the conetraint ws & cx) > o The aunphoe variable ees SAI =F OME HN) SF =o The appropriate Lagramucat NOMUNEAR programming & 1427 Following an analysig rion £0 1H ization Seeriatizatiin gration, thee Rulher: Paekor ums ation prisstions, tlhe MC NE Prolstony cin fe obtained ae KX) = 0, anditions are also the sufficient ¢ ramming problem, the necemary Kuhn. Tucker ‘onditions for = the maximization problem, when “the minimization problem Stet 0) concave ant HX) is conver both AX) and A(X) are conver AMPLE 16.9-1 Solve the NLPP: —— Ze sey—xf + Dep subject 10 nutmsh Xp. X22 0. solution The problem can be put as F(X) = 4x, = xf + 2x, WX) =x 4-1. The problem is of maximization, and Kuhn-Tucker conditions are G(X) _, Gh(X) =0, ox, ox, DH(X) = 0, WX) <0, 4.20. Applying these conditions, we get a (a) (uit) (iv) Ww) 1 = 0. These results satisfy the conditions (iv) From (i) 2. = 2, therefore from (ite %4(), Solution of (i), (i) and (i) viel 2. 0.8165, a3 bo = 0.1835. ~¥3 ve in X, while A(X) is & convex function. timizes the objective function which comes to and > It can be easily observed f ent (x) fe (0.816: 835 Yi i the solution X* = < 3.0887. @ scanned with OKEN Scanner4428 % OPERATIONS RESEARCH EXAMPLE 16.9-2 a gd Z= 10x, + 4x2 — 2Xh —P Maximize subject f0 dey #985, xp p20. Solution an We have, xy + dey = 2eP — 22s 1 xy 5s ‘The Kuhn-Tucker conditions for the maximization problem are AX) _ , HOY. ay Oy A(X) = 0, A(X) $0, 120. Applying these conditions, we get 10 - 4x, - 24= 0, 42x - AQx ta -5 2x tx - 550, 120. From (iii) either 4 = 0 or 2x, +x. -5=0. Hiren). 0, the solution of () and (if) gives xj = 2.5 and x2 = 2.0, which does not satisfy the equation (iv). Hence 2 = 0 does not yield a feasible solution. i When 2x, +x)—5 = 0 and A + O, the solution of (9, (i) and (it) yields, x1 = 4 = 4, which satis all the necessary conditions. Tt can be verified that the objective function is concave in X, while the com: in X. Thus these necessary conditions are also the sufficient conditions of maximization of f(X). straint is convex: : es The optimal solution is. xy" = > %2" = i which gives Zax = 2 EXAMPLE 16.9-3 : ' Solve the following NLPP using the Kuhn-Tucker conditions: Maximize Z= 2x} - xz + 12x pxy, subject t0 2x; + Sx $ 98, xp x20. ‘Solution The given problem can be put as SK) = 2x} - Ix} + 12x, A(X) = 2x, + 5x, -98, * The Kuhn-Tucker conditions are Fah aay” 4 mh ea ae @ scanned with OKEN Scannerge Sc is 0 + 1409 perenne ven Seren eee NON-LINEAR PROGRAMMING _ A(X) s 0, i i Reo, Applying these conditions, we get 4x) + 12%) ~ 2)» 0, (i) 12x) ~ 14x, ~ 5, =, Ga 4 (2x1 + Sxy — 98) = 9, (iii) 2x1 + Sxy— 98 <9, (iv) A>o. «) From equation (if) either X= 0 of 2x, + $450 = 0 1 Sy 9 When 2 = 0, equations (i) and . and (ii) give xy = which do isfy condition (iv). hs a feasible solution cannot be obtained for? i. ©, which ost or setSty © When 2x, + Sx — 98 = 0, this f jon, x, = 44 pany 2 vith X= 100 and 20, eraion along with (i) and (ii) gives the solution, 46.10 NLPP WITH MORE THAN ONE INEQUALITY CONSTRAINT Let us consider a general non-linear programming problem of the maximization type Maximize Z=s(X), subject to g(X) =O = To This again is an infeasible solution, Solution 3: 2, #0, 25= 0, From (i), (ii) and (iif) we have lax + 6-2, 10x, ~ 224 and x + 2m 10 Solution of these equations The solution of these equations yields 8 jn M6 4 20 33 = Js % . 3 33 This is a feasible solution giving Z = 114,061. Solution 4: 2; #0, 22 #0. From (i), (i), (ii) and (iv), we have 14x, + 6 - y= ay 10x,~ 224 + 32, x, + 2x - 10 and x -3x-9 The solution of these four equations yields #1) ARR This also is a feasible solution giving Z = 702.92. Since the maximum value of Z is obtained for solution 4, where A; # 0 and 22 # 0, the optimal solution is 8 1 a= Saag = 5 ad Zn = 102.92. +a = 0. EXAMPLE 16.10-2 ede; + 38 chef ad tad) Optimize xy) tS subject to 2x; + 3x56, xp 220. Solution A P(X) =2e + 3-7 a a We HO=ntazh A WAR) = 2m + 32-6 an @ scanned with OKEN Scannersea essential to Aetring yy ) NS dition i on De sx et ror ti HEA ee je iz 0 my raion en ‘ vst uit [pF Fg rains 1 0) 30 “9 0] == 10. * n= ms), 2 0 amsin} 0 2 Je (-1)"*! ie., -ve, While fop B should be ii est 8 <0, the solution point should maximize the chen ie ve. ase are see te(- I ditions for this © Kuhn Tucker © 40)- FA 0/ 00" ist nytt = 9 nso 420. Therefore ead 2 -24-h-2ha= 3-2x)- i 3h nea t2- 1 dg 2 + 3-6 aytm-1s 2x + 3x, - 6 50, day Wg 20. it Four solutions corresponding to the following values of 4, and A, can be obtained: @ =o, i= 0, @) y=0, 2a #0, (ii) ty #0, h=0, () 420, a0. Solution 1: 24=0, 2. = 0, a Parent and (ii) give x, = Solution 2:2, = 0, 2g #0, ‘The solution of equations (i, (i) and (v) gives = 1, x, = 3/2, x3 = 0, This solution does not satisfy 12 18 ATS gs = u This solution, re oe PS Ty BAO Mas Ts: a tty equation (vf) and is, therefore, discarded Solution 3: 2, ‘The soh lution of aes ©. 9, Gi) and (iv) yietas @ scanned with OKEN Scannerf ll = __—_—_ 4s NON LIAR proarmumne # 1498 1 sella eee Re meoned ‘his solution satisfies all the conditign § 2 nn ms, and gives ution 4: Ay #0, ay #0, ‘he solution of equations (i, % Bo 3, GD, (vy and (v) gives S40 dye 34, ay 2 13. yan is thus infeasible and is, therefore, discarded. e SCE M=~3, ‘his solution violates condition ( nce only one solution satis si fies all Conditions, the same is the optimal solution. and AMPLE, 16.10-3 Use the Kuhn-Tucker conditions to solve th ie following non-linear programming problem: Maximize Z= 2x) xf +x, subject to the constraints ; 2t + 3xp< 6, dey ty <4, Xn x20. [Nagpur U. B.E. (Comp. Se.) 2004] slution : We have FO = 2 ~xP +x, WO) and The Kuhn-Tucker conditions for a maximization problem are GO- 2M hi) = Ah) Hi) < 20 Applying these conditions to the giver functions, we get 2 = 2x - 2h, —2ha= 1-3h-22=0, dy Qx1 + 3x - 6)=0, dy Oxy +x -4)=0, 2x + 3x2 — 6 $0, 2 +m- 40, hy M20. The following situations can now arise: Case 1: ay=0, a= 0s ten 7 dot ® Case 3: a Case 4: hee @ scanned with OKEN ScannerSolution 4: 2, # 0,2 #0 The solution of equations (1), (i), (iii), (iv) and (0) gives MPa 40, eM AH © condition (viii) and is thus infeasible and is, therefore, discarded one sohution satisfies all the conditions, the same is the optimal sotution This solution violates Since only and EXAMPLE 16.10-3 Use the Kuhn-Tucker conditions to solve the following non-linear programming problem: Maximi Z= 2x) -xf tx, subject to the constraints 2x, + 3x25 6, De; +x) $4, xp 20. (Nagpur U. B.E. (Comp. Sc.) 2004] Solution We have SM) = 2x - xP +4, HX) = 2x, + 3x2 6, and PO) =2y +m - 4. ‘The Kuhn-Tucker conditions for a maximization problem are GX) - ah Hf) =0, 24 H(X) = 0, HO) <0, 14,20. Applying these conditions to the given functions, we get 2 — 2x - My — 22 =0, 1-32, -22=0, dy Oxy + 3x2 - 6) = dq 2x1 +2 -4)= 0 2x, + 3x - 6S 2x +m - 450 Day Ia 2 0. ‘The following situations can now arise: y= 0 Case Case 3: Case @ scanned with OKEN Scanner(ii) is vi th mn of equations: (i, (A), Gi) and (iv) lerefor da 2 4and hy = 1. ives it is discarde ations (i), (ii), (Hii), and (ivy viet Is Case 1: When Case 2: When 4 x 200 ‘on does not satisfY equation (¥)s solution of et As this soluti Case 3: When dy # 0:22 = x Ses teraiee, 3 1 the constraint equal ions and is thus feasit 8 feasible sof tion This solution satisfies al ty Z= 229. : ae Case 4: £0, and Az # 0, the solution of equations (0), (1, (i) and (iy 3 el Mes a bin any wel bsg and a= 7 rs sition vi), which {8 1, hr 2 O and ene is ina sible This solution violates © Thus the optimal solution is on 4. n +o giving Zu = g? Siving Zeer = “9 * EXAMPLE 16.1 0-4 Use the Kuhn-Tucker conditions (0 solve the following non-linear pro gram Maximize Z=Txp ~ 6 + 5x7, "ming Probley subject to the constraints x; + 22 <10, xy - 3% $9 Xp Xp 2 0. Solution: “The Kubn-Tucker conditions for a maximization problem are | HO) — YIjOO =0, | a | | nO = Hs az | Applying these conditions, we get 1x - 6-24 Ap Hs | 10x, — hy + 3Qg @ duGe + 2x - 10) = 0, li) Jole1 - 3x2 — 9) = 0, “ta | x + 2x, - 1050, ) 4 = 3-950, | My W20. + i) alii Case 1: 2, = In = Equations () and (i) gi Pp ape nt ge 3/7 and x, = 0. This satisfi . Case 2: 2, # 0, Pea with Z = 9/7, isfies the constraints (),(v) snd Equations (i), (i) and (it) give _ ne Fe BM anda, = SN teecHeeea3 This solution sats fies the oth Case 3: 2; = 0 and iy 2 ot Conditions and gives Z = 95.78 Equations (9, (i) and (i) give @ scanned with OKEN Scannerye N “ ON, x, = 288 LINEAR PROGRAMMING “ 1435 =o : . an his is a0 infeasible solution 17 nd x, = =. case 43 M1 # 0 and A, x 9, Equations (iii) and (i) give x = 48 1 1,274 5 and x)= — : 1,936 and A, = ~——. . : 5 and then equations (i) and (ii) give - a 25 This Solution Satisfies al] optimal sol 48 the conditions and gives Z = 587.72. 1 imal soluti i - Te. Hence pti tion is xX, = : 1 i 5 ith Zmax = 587.72. wae GRADPHICAI jierrnine @ scanned with OKEN Scanner