Professional Documents
Culture Documents
Operations Research JK Sharma PDF
Operations Research JK Sharma PDF
(i) 2x, +3yS8 Gi) Sxj+2x,+25<10 |
Gi) 55, 14515,
and AytyH2 0
Dual
Min Z~ 8), + 10y3 + 15y3
subject to) 2y,+5y,23
Gi) 3y,# 29,4 4y,24
and YI W20.
11. Primal
Gi) +5) 25 |
1x, and.x,= units of fresh food I and Ito be bought, respectively
Min 2=3x, + 2x,
subject () 724,230) Sx, + 4x,220
Gi) 2x, + 8,216
and xpme 0.
Dual
_yyeJp and ys ~worth perunitof vitamins A, B and C, espe
10 the bady.
Max Z=30y, +20), + 16y,
subject to) Ty, +42 t2y5 83
and Yin ar¥9 20,
(b) The optimal solution to the dual problem is: y, = 20/52,
GH 2y F494 8 <2
12, Product | = O unit: Product 2= 50 units: Produet 3
and Max profit = Rs. 700,
13, Primal
Xp ty and ry
sO uits |
quantities of coffee seeds, respectively which
te company buys from exch plantation.
Min 35+ 2m +1204 154
subject io () Ody, 024, + O64, 08,2 03 |
(a) 06x, + 0.8, + 04x, + 0.22 0.7 |
sod remnine |
Duality in Linear Programming Tag
Dual
1p; and y,= werth of coffee seods A and B, respectively
Max Z,~ 0.33, * 07),
subject to) Oy, +0643 ii) 0.29, + 08,2
ii) 0.6y, +04y,< 12 (iv) 08y, +029) 51.5
and 9220
14. Primal
xj.% and.y.= number of valves of the types A, B and C,
respectively to be produced.
Max Z,= 13%; + 13.5%, * 10x,
subject to) $x, 3x +2xy€ 1,200
(Ix, + 2xj 44x, 900
(ii) 4x, + 10x, + 5x5 1,300
and 0.
Ams. x, =22,400/217, x9 = 5800/62, x5 = 0 and
Max Z,= Rs. 58,125/31
Dual
Yd} And y,= Work of Chambers |, 2 and 3 production capacity,
respectively
Min Z, = 1,200y, +900), + 1,300,
subject to () $9, +%2-+ 4yy215
() 3y, #29) #1052 135
Gi) 2444+ Syy210
and Wn ZO.
Ans y,=0,9)=— 4135 =~ 645/62.
17. Primal
2X). = number of units of product A and B respectively to be
produced
Max Z,=5r, + 3x,
subjectto ()) 4 +25) 10 ) 24258
and 2320
Ans. x, and Max 2, = 14
Dual
Yj. ¥5 = wort (o shadow price) per unit of resource operators
in three shits respectively
Min Z, = 109, + 8,
subjectto (4, +2,25 i) 2429523
and FoIrz 0.
Ans.y, = 1,y2 1/2 and Min Z, = 14
18. Primal
2,44 = numberof units of product P,Q and R wo be produced,
respectively.
Max Z,=31) * 549+ 4
subject to () 24 +3ry 58 (i) Sx, 42x +2510
(ii) 5x, +4, 515
and apa EO
Ans. x, =4/41, x5 = 59141, x, = 80/41
Dual
Yip Np Jy = Worth per unit of material m,, my and ms
respectively,Operations Research: Theory and Applications
Min Z, = 8), + 10) + (5y, Max Z,=2;+%
anecte ) B+m23 MD Anrsy2s — | sabato 21042120 5Oqje 80420
y+, 44,24 (i). 1, + 60552 240
and, Vedra 20. and p20
19, Prinat | Ans, =2,.)= Band Min Z, = 10
xj.fy~ aumber of unis of food A and B to be consumed, Dual
respectively yi3bs93 = worth per unit of biscuits of type A. B and C.
Mas Z,= 8.505, +4, | respectively
subject (i) 2x, +28 Gi) 6, +212 MinZ, = 120y, + 740, + 240
Gi) 54329 subject to) 2, +509) 107551
and Gi) 10 +809, + 609551
Ans 5" 1.8 = and ds20
Duel 22, Primal
Yin Ys Worth per unit of proteins, carbohydrates and fats Max Z,~ 200s, + 300%
respectively. subject to (i) x, + 2x, $32 Gi) Oxytx $8
Min Z, = 8, + 1273+ 95 and 20
subject © 6) 2%, +69, 495 $ 880 Gi) 9, 493 +3955 4 Dua
and Yarnrs20 MinZ,~32y, + 8
Ans. y,=31/8,9,= 18,5 =O and Max Z, = 657%: subject 10 ( y5+04220 Gi) 2 +9, 2 300
20. Primal and Jn 3220
Yt. %)= mumber of units of artes X, ¥ and Z respectively ‘The optimum slution tothe dual LP problemis:y,= 200,)=05)
Max Z_~ Airy + 3 + 6ry = Ox, = 100, 2 = Rs. 600
subject 9) x +4522 @ y+n25 ‘Note that, y= marginal increasein profit for an addition I hour
and Hoty 20, of epacity. Since y, = Rs, 200, profit becomes Rs. 6400 + 200
Ans, ~ 0,2," 3.4 =2 with Min Z=21 = Rs. 6600. Also y3~ marginal increase in profit given an addition
mat hou ofcepacity in D, Since y,=0, there would be no change in
ss the profit
Ya Ja = worth per unit of resources to be used. respectively {In primal LP problem solution, s, = 8 indicates that there is an
Min Z, = 2y, + 535 excess capacity of 8 hous in D. Thus, a dual solution we have y=.
subject io () yi <4 (i) p83 ual constraint (i) is writen as: y, + Oy, 2 200 or
Dy tHS 6 total value of hours in department D and D;
and yu3gS 0 required to produce | nit of Product |
Ans, =3,¥)= 3 with Max 2" =21 unt profit received from manufacturing
2 Pa (ming enttPeaet |
xj.32~ ntmbor of assortments ofsize [and I, respectivel
CHAPTER SUMMARY
In this chapter we discussed the relationship between an LP problem and its dual. The method of formulating a dual LP problem
of the given LP problem was explained with several examples. The main focus, while solving a dual LP problem, was to find for
each resouree ils best marginal value, also known as dual or shadow price. This vaiue reflects the maximam additional price to
be paid in order to cbtain one additional unit of any resource in order to maximize profit (or minimize cost) under resource
constraint,
CHAPTER CONCEPTS QUIZ
True or False
1. Ifthe objective function of the primal LP problem is maximized | 3. The dual of dual is a primal probiem.
then the objective functon of the dual is to be minimized. |, The value in the ¢ ~ z) fow under columns of slack/sumplus
2. A dual variabe is not defined for each constraint in the prima variables ignoring negative signs do net give the direct optimal
LP problem, values of tne duavormal basic variables.image
not
availableBQ.) Operations Research: Theory and Applications
4
1"
16,
19.
a
92,
TO o8F OT AF 6&F 6T 72T 8T OT WT
Primal, dual 42, Dual 13. Simplex 14. Untounced; infeasible 18. Equality; unrestricted
Complementary slackness; equal 417. Shadow price 18. Slack, basic
Decision variables; constrainis; objective function 20. Dual primal; minimization
(©). (0) 23. (a) 24. (a) 25. (&) 28. (c) 27. fa) 2B.) 2B. (co). (@)_— 31 CC)
@) 3B @) 34) 35. (a)
APPENDIX: THEOREMS ON DUALITY
Theorem 5.1: The dual of the dual is the primal
Proof: Consider the primal problem in canonical form:
Minimize Z, = ex
subject to Ax2b; 120 ®
where A is an m xn matrix, b’ © E” and ¢, x” € E".
Applying the transformation rules, the dual of this problem is:
fa
Maximize Z, = by
subject to Aly sels y20 ®
Tais dual problem can also be written as:
ie Zt= (-W)"y 8%
subject to, CATy 2(- €); y20, where 2; =~
If we consider LP problem (3) as primal, then its dual can be constructed by considering x as the
dual variable. Thus, we have
Maximize Z* = (- "7x = (ox
subject to CAN x <(-b7) or A) s (-b)
and x20 ®
But LP problem (4) is identical to the given primal LP problem (1). This completes the proof of the
theorem
Theorem 5.2: Let x* be any feasible solution to the primal LP problem
Maximize Z, = ex
subject to Ax Sb; x20
and y* be any feasible solution to the dual LP problem
Maximize Z, = b"y
subject to ATy 2e7 y>0
of the above primal problem. Then prove that ex* ; y unrestricted in sign.
Theorem 5.5 (Unboundedness Theorem): If either the primal or the dual LP problem has an
unbounded objective function value, then the other problem has no feasible solution.
Proof: Let the given primal LP problem have an unbounded solution, Then for any value of the
objective function, say + e, there exists a feasible solution say x yielding this solution, i.e. ex — =.
From Theorem 5.2, for each fessible solution y* of the dual, there exists a feasible solution x* to
the primal such that ex* < b”y*, That is, b’y* > + =. As b is a constant number and y* has to satisfy
the constraint A7y* < e7. Hence the dual objective function Z, = by* must be finite. This contradicts
the result b’y* — ©, Hence the dual LP problem has no feasible solution,
A similar argument can be used to show that when the dual LP problem has an unbounded solution,
the primal LP problem has no solution.
‘Theorem 5.6 (Duality Theorem): Ifeither the primal or the dual problem has a finite optimal solution,
then the other one also possess the same, and the optimal values of the objective functions of the two
problems are equal, Max Z, = Min Z,,
Proof: Let the following LP problem represent a primal problem:
Maximize Z—ex
subject to Ax Sb or Ax+Is=b: x20
‘where T is the identify matrix of order m and s is the slack variable.
Let x, = B"'b be an optimal besic feasible solution to this primal problem and ¢, the cost vector
‘of the basic variables. Then according to the optimality condition, we have, ¢,~ z, < 0, for any vector
a, of A but not in B, Therefore,
~e,(B'a) <0, forall)EEE Operations Research: Thoory and Applications:
This is equivalent to € < ¢,B™'A, when written in metrix notation.
Let y* = ¢,B', Then ¢ < ¢gB™'A becomes ¢ < y*A orc! < y* A!. Hence y* is a feasible solution
to the dual of the given primal because it satisfies the dual constraints. The corresponding dual objective
function is given by
Z,= y*b" = ¢,B"'b? = cy3, =Z,
Hence, it has been shown that x, and ¢,B"? are the feasible solution to the primal and dual LP problems,
respectively for which Max Z, = Min Z,
Theorem 5.7 (Complementary Slackness Theorem): Ifx* and y* be feasible solutions to the primal
and dual LP problems, respectively, then a necessary and sufficient condition for x* and y* to be optimal
solutions to their respective problems is,
By %yei2O TL Zee
and mej FQ
where, ., is the ih slack variable in the primal LP problem and y,,., the jth surplus variable for the
dual LP problem,
Proof: Let x* be the feasible solution to the primal LP problem,
Maximize Z,
3 gs,
pale
sunjetto B ayay tage “85 4/20 o
and y* be the feasible solution to the dual of the above primal.
Minimize Z,= £ 4,y, ®
a
subject © Bays nay “G5 EO
z
‘Multiplying constraints in (7) by yj, i= 1, 2, .... m, and then adding we get
5 Lagat ani $3 y,5,5 f= 12am ©
‘Subtracting (9) from the objective function of (7), we get
Substituting the value of Z, and y,,..; from (8) to (10) we get,
~ Barnes Erte Z-% ay
If x* and y* are optimal solutions to the primal and dual problems, then we have ex* = b’y* (or Z,
=Z,) Thus, from (11), we have
Haw italy rete Blo eri pp ead Sa 5 aa jaca
in the dual is an equation. Also x, = 0 implies y,,,=0and 3 a,y, ><,
7 nth Rowers
Silty foreach > 0,46 12s...) implies, =Oand £ ays,
iF
1» ke. fh constraint in
the primal is an equation. Also y, 0 implies x,,,>0 and ¥ ayx, mix
(Gf no ratio is positive)}
iginal value, ey
Per unit improvernent value
Input — Output coefficient in the variablerow ay
where Improvement ratio =
Remark While performing sensitivity analysis, the artificial variable columns in the simplex table are
ignored. Any exchange of quantities between basic variables and an artificial variable makes no sense,
because an artificial variable has no economic interpretation. Thus, improvement ratios using coefficients
corresponding to artificial variables should not be considered.
Gase lit Change in coefficient of a non-basic variable in a cost minimization problem ‘The
procedure for calculating sensitivity limi of a cost minimization LP problem, where the objective function
coefficients are unit costs is identical to the Case I discussed above. In this case, the unit cost coefficientBEL Operations Research: Theory and Applications
‘can be increased to any arbitrary level but it cannot be decreased by more than the per unit improvement value,
‘without making it eligible so that a non-basic variable can be entered into the new solution mix. The sensitivity
limits can be calculated as:
riginal value c, - Absolute value of per unit improvement value (c) ~z))
Infinity (+ =)
Alternative method Ifwe add Ac, to the objective function coefficient ¢, of the basic variable xj, then
its coefficient cg, will become cj, = Cp, + Acgy- The new ¢, ~ 2% can be calculated as follows:
{3 ay + (Can + Sean) Ye,
isk
Leg y =
i
me bey + Meme yy) = (6 ~ 2) ~ Bee ry
For a carrent basic feasible solution to remain optimal for a maximization LP problem, we must have
$0. That is,
G- FG 2)~ Mery $0 oF 6-35 Dew ry
(2S Aen when yy > 0
(6-2)i4j 2 cy when yy <0
Hence, the value of Acpy that satisfies the optimality erierion can be determined by solving the
following system of linear inequalities:
Min] 27 b> deg, 2 Max — 5
is a " at 2
Here it may be noted that y, sare entries in the non-basic variable columns (i.e. variables) of the optimal
simplex table. The new value of the objective function becomes:
Denko + (ae +m a4 = 2 coetsi + Seo tee = Z+ Sep, Soy
Zz
Hence if dcp, satisfies inequality (1), then the optimal solution will remain unchanged but the value
‘of Z will improve by an amount Aci, Xgy
Example6.1. Solve the following LP problem:
Maximize Z=3r, +51,
subject to the constraints
@ 3yt2_Z18, GD mZ4, Gi) Zw
and xyty20.
(@)_ Determine an optimal sclution to the LP problem.
(b) Discuss the change in con the optimality of the optimal basic feasible solution.
‘Solution (a)The optimal solution ofthe given LP problem is shown in Table 6.1. From Table6. 1, the optimal
solution can beread as: and Max 2 = 38,
(b) Here c= (cy, ¢>, ¢s Cys Ce), and the cost coefficients associated with basic variables x,, x», and, are
4" (61, ¢p.¢4) = G, 5,0). The ckanges inc, can be classified as under:
() Changes inthe coeffictents (he. and es of non-basic variables, and sy Lets adc, and Acy
tothe objective function coeficents and, Thenthe new objective function coeficient will become c3 =
+ Aes and cs = c,+ Ac,, Since c, =0 and c5=0, cf = Ac, and cj = Acs. Thus, the new values ofc —z5 and
z,will become Ac, ~ I and Ac, ~3, respectively. Now in orderto maintain optimality, we musthave
‘Bcy—150 and Ac,—350(Maximizationcase) or cy and Acg<3Sensitivty Analysis in Linear Progamming BRAg
@ Change in he coeffciems cj (.e.¢y cand, of base variables x, and sy: The value of additional
increment cy, t= 1,2, 4in the coefHcients, ¢,andc, which satisfy the optimality condition can be determined
by solvingthe following system oflinearinequaties refer to Eqn, (1)]-
74
we
5° 0 0
2 x 2 1 0 18 0 23
° a 0 0 o 23 1 a3
8 % 6 0 1 o 0 1 Table 64
Z=36 o=% 0 0 a 0 Optimal Solution
Fork=1 (je. basic variable x, in row 1), we have
=3 a1
win { $3 fen = man (=, G35 or 922hq23
Here it may be noted that y,,= (43 = 1/3, y,5=—2/3) are only for those columns corresponding to which
‘variables are not in the optimal basis (ie. non-basic variables}. The current optimal solution will not change as
long as
9 15
(2#2}ae20-3 ot 2e,20
Fork=3 (ie. basic variable x, in row 3), we have
= 3
sacl neal flocs wean
Hence, the current solution will not change as long
Example 6.2 A company wants to produce three products: A, B and C. The per unit profit on these
products is Rs 4, Rs 6 and Rs 2, respectively. These products require two types of resources, manpower
and raw material. The LP model formulated for determining the optimal product mix is as follows:
Maximize Z = 4x, + 6x, + 2r,
subject to the constraints
© x, +x, +4, 53 (Manpower required), (ii) x, + 4x) + 7x, $ 9 (Raw material available)
and Spay 0
where x) x, and x, = number of units of products A, B and C, respectively, to be produced.
(® Find the optimal product mix and the corresponding profit of the company.
() Find the range of the profit contribution of product C (ie. coefficient c, of variable x,) in the objective
function, such that the current optimal product mix remains unchanged.
(© What shall be the new optimal product mix when per unit profit ftom product C is increased from
Rs 2 to Rs 10?
(© Find the range of the profit contribution of product A (i. coefficient ¢, of variable x,) in the objective
function such that the current optimal product mix remains unchanged.
Solution (a) Convert the given LP mode! into the standard form by introducing the slack variables 5, ands.
Maximize Z = 4x, + 6x, + 2r, +05,+ 0s,
subject to the constraints
Ox tmtytsy—3 Gx +4yt Byte
and Xp Xp Ty Sy 520
‘The optimal solution obtained by applying the simplex method is shown in Table 6.2
5: (S +09) 2 2 (5-3) ore, 22,"> 4 6 2 o o
4 7 1 1 0 =I 43 13
6 % 2 0 1 2 sia 1B
Table 6.2 z= 16 4 6 8 108 Ey
Optimal Solution 9, 0 ° =6 103-28.
‘The optimal sotution is: x, = 1, 1) = 2 and Max Z= Rs 16,
(b) Effect of change in the coefficient c, of non-basie variable x, for product C In optimal simplex Table
6.2, the variable x, is non-basic and its coefficient c, ~ 2 is not listed in the eg column of the table, This
‘means a further decrease in its profit contribution c, (= Rs 2) per unit will have no effect on the current
optimal product mix. But, if cy is increased beyond a certain value, the product may become profitable 10
produce. Hence, there is only an upper limit on c, for which the current optimal product mix will be effected.
97 4 6 2746 ° °
ay a 1 -l Taga
6 a 2) 0 1 2 “13
Table 6.3
Table with a Acy y 4 6 103
9-5 6 o 108
Table 6.4
‘As wealready know, a change Ac, inc will cause a change in z, and c3 ~ 3 values in the x-column,
‘The change in cy results in the modified simplex Table 6.3.
For an optimal solution shown in Table 6.3 to remain unchanged, we must have: Ac, ~ 6 $0
or ey 5 6.
a dep ils Regret abi il con't emi
wwe get c; - 2 £6 of cy <8. This implies that as long as the profit contribution per unit of product C is
less than Rs 8 (i.e. change should not be more than Rs 8) it is not profitable to produce it and therefore
the current optimal solution will remain unchanged,
(©) If the value of ¢, is increased from Rs 2 10 Rs 10, the new value of ¢, ~ 23 = (€,~ €s) will be
=i
gos-10-6 6[ 4]or-c4e1y-260
‘Thus, if coefficient of variable x; is increased from Rs 2to Rs 10, the value of ¢, ~2, will become positive
as shown in Table 6.4. The variable x, (corresponding to product C) becomes eligible to enter into the basis.
Hence, the solution cannot remain optimal any more.
57 4 6 wo 0 o
an
low 2] 0
Copyrighte| Sensitivity Analysis in Linear Programming BRED
Applying the following row operations to enter variable x, into the new solution mix
R, (new) —> R, (Old)/2 (Key elememi) ; Ry (new) > R, (old) + R, (new)
The new optimal solution after entering non-basic variable x, into the solution is shown in Table 6.5.
1
2-18 y
g=% Table 65
Since all ¢/~ 4; $0 in Table 6.5, the new optimal solution obtained is: x ~ 2, xy ~ Oand.x3 ~ 1 and
Max Z=Rs 18.
(a) Effect of change in the coefficient c, of basic variable x, for product A_ In the optimel simplex Table
62, the variable x appears in ‘Basis B’. This means a further decrease in its profit contribution c,
Rs 4) will make it less profitable to produce and therefore the current optimal product mix willbe affected.
Also, an increase in the value of c,, beyond a certain limit, will make the product A much more profitable
and may force the decision-maker to decide to only produce product A. Thus, in either case, the current
optimal product mix will be affected and hence we need to know both the lower as well asthe upper litt,
‘on the value of ¢,, within which the optimal solution will not be affected.
Referring again to Table 6.2, the range of change in the value of ¢, (and/or also in ¢,) that does not
affect the current optimal product mix cen be determined once again by calculating the values of ¢; ~ zj,
‘values that correspond to non-basic variables x, s, and 3» respectively. For this reproduce Table 6.1 once
again with unknown velue of c, as shown in Table 66.
go thse, 6 2 ° o
athe, 1 1 0 1 43 -13
6 a 2 o 1 2 “13 re
Table 66
Z=12+@ + de) = | ata 6) BO, «WAT A4Ac 3 23 —Ac/s Table with a Acy
ons | 0 0 By 6 103 4ey3__bey/3—-23_ change in c
For the solution shown in Table 6.6 to remain optimal we must have all ¢, ~ z; $ 0. That is
Ae 6 £0 gives Ao, <6
acy
-Z- Eso gives Ac 2-52
dy 2 , 3
SF <0 ives dey 52
Thus, the range of values within which c, may change without affecting the current optimal solution is:
3 3
2.8 be
5 She, $2 or 4-5
From these calculations, it may be concluded that in the case of competitive pressure or in the absence
‘of any competition, the decision-maker knows to what extent the prices can be adjusted without changing,
the optimal product mix.
3
S442, ie. 5 $056
Copyright erial4 Opsrations Research: Theory and Applications
Example 6.3 Given the following LP problem
Maximize Z= x, +2e)—x,
subject to the constraints
@3x4x)-x,S10, (i) —1, +4, 4,26, Gill) xy+x, $4
and By kyxy20.
Determine the effect of discrete changesin c)(j=1, 2,.. 6) onthe optimal basic feasible solution shown in
Table67,
6° 2 4 0 0 0 -~M
° 4 6 Bc qi renege) Saji 20 Fo Memes oe
2 xy 4 Ue NE ee te age Sat 0
o 3 10 Libeat Caestinaedane tena Ae onae
Table 6.7 Z8 2 z Or seieme eat TO aren ON
‘Optimal Solution “4 SEO Ca, 4s eal
Solution In Table6.7, variable, isa non-basic with c= -1. If, further decreases, x, will not enter the basis,
and hence will nat afect the optimality ofthe solution. Thus there is no lower limit onthe valueofc,. However,
if, inereases and exceeds a certain value, it may become profitable to have.x, in the basis. Thus, there should
‘bean upper limit on the value ofc.
Solution shown in Table 6.7 will remain optimal provided largest value of @ <0, ic.
3
€,-.2,0)]0/<0 or eS0.
1
‘The variablex, in Table 6.7is abasic variable, the solution will remain optimal so long as nor-basic variables
remain non-positive, ie,
3
=10,6,,0)/0|S0 or <0
1
] 2
€5=-1-@.c,,0)] 1 |s0 or 2-1
3
=I
= 0-(,6,,0)| 1]s0 ot e520
4
0
M~(0,¢,,0)| 0|<0 or MZ0
-1
is enon-asie variable, and to find its upper limit, we musthave c, $0, i.
a
€5-,2,0)| 1]50 or ey-250 or e552,
3
‘The variables; in Table 6.7 is abasic variable andi limiting value is obtained from largest value ©),23,%
and &; of cy, ¢4,¢4 and c, respectively as follows:
‘The variablex, in Table 6.
;
7‘Sensitivity Analysis in Linear Programming
:
-1-¢40)| 1 |s0 1425-28005?
[s
7
ceyain | aleumene aeons
3
o
3 =-M-~(cy.2,0) | 0/0 or - M0, which is tue,
ft
Hence, limits for ¢, are:
‘The variable s, in Table 6.7 isa basic variable and its limiting value is obtained from largest value Z, 2), 2%
1B $e $32
and & of ¢, cy and c, respectively as follows:
F=-1- 0,2) |0]s0 or $0 or c2-1,
li
2
B=-1-0,2c)| 1)s0 or -1-(2+3c,) <0 or cg2-1,
3
% =0-(0.2,6)| 1]s0 or 0-@+4e,)<0 oF
% =-M-(,2,¢) | 0/s0 or -M+e5<0 or cg SM.
Hence, limits for cg are:~1/2-