Chapter 3 Linear Programming: Computer Solution and Sensitivity Analysis

Computer Solution Early linear programming used lengthy manual mathematical solution procedure called the Simplex Method (See CD- !M Module A"# Steps o$ the Simplex Method have %een programmed in so$t&are pac'ages designed $or linear programming pro%lems# Many such pac'ages availa%le currently# (sed extensively in %usiness and government# )ext $ocuses on Excel Spreadsheets and *M $or +indo&s# Linear Programming Problem: Standard Form Standard $orm re,uires all varia%les in the constraint e,uations to appear on the le$t o$ the ine,uality (or e,uality" and all numeric values to %e on the right-hand side# Examples: x- x. / x0 must %e converted to x- - x. - x0 1 x.2(x0 / x-" 0 %ecomes x. 0 (x0 / x-" and then x. - 0x0 - 0x- 1 Sensitivity analysis (or post-optimality analysis" is used to determine ho& the optimal solution is a$$ected %y changes3 &ithin speci$ied ranges3 in: the o%4ective $unction coe$$icients the right-hand side ( 5S" values Sensitivity analysis is important to the manager &ho must operate in a dynamic environment &ith imprecise estimates o$ the coe$$icients# Sensitivity analysis allo&s him to as' certain &hat-i$ ,uestions a%out the pro%lem# Objective Function Coefficients Let us consider ho& changes in the o%4ective $unction coe$$icients might a$$ect the optimal solution# )he range o$ optimality $or each coe$$icient provides the range o$ values over &hich the current solution &ill remain optimal# Managers should $ocus on those o%4ective coe$$icients that have a narro& range o$ optimality and coe$$icients near the endpoints o$ the range# Range of Optimality 6raphically3 the limits o$ a range o$ optimality are $ound %y changing the slope o$ the o%4ective $unction line &ithin the limits o$ the slopes o$ the %inding constraint lines# )he slope o$ an o%4ective $unction line3 Max c.x. / c0x03 is -c.2c03 and the slope o$ a constraint3 a.x. / a0x0 7 b3 is -a.2a0#

Standard $orm re,uires all varia%les in the constraint e,uations to appear on the le$t o$ the ine,uality (or e,uality" and all numeric values to %e on the right-hand side# Examples: x- x. / x0 must %e converted to x- - x. - x0 1 x.2(x0 / x-" 0 %ecomes x. 0 (x0 / x-" and then x. - 0x0 - 0x- 1

<

.1

..

eaver Cree! Pottery "#ample Sensitivity $nalysis

Sensitivity analysis determines the effect on the optimal solution of changes in parameter values of the objective function and constraint e%uations& Changes may be reactions to anticipated uncertainties in the parameters or to ne' or changed information concerning the model&

.0

.-

.8

Objective Function Coefficient Sensitivity Range

)he sensitivity range $or an o%4ective $unction coe$$icient is the range o$ values over &hich the current optimal solution point &ill remain optimal# )he sensitivity range $or the xi coe$$icient is designated as ci#

.9

.:

.;

.<

Changes in Constraint (uantity )alues Sensitivity Range

)he sensitivity range $or a right-hand-side value is the range o$ values over &hich the ,uantity>s value can change &ithout changing the solution varia%le mix3 including the slac' varia%les#

.=

01

0.

Constraint (uantity )alue Ranges by Computer "#cel Sensitivity Range for Constraints

00

Shado' Prices *+ual )ariable )alues, De$ined as the marginal value o$ one additional unit o$ resource# )he sensitivity range $or a constraint ,uantity value is also the range over &hich the shado& price is valid#

Maximize Z = $40x1 + $50x2 subject to: x1 + 2x2 40 hr of labor 4x1 + 3x2 120 lb of clay x1 x2 0

0-

Problem "#ample )&o airplane parts: no#. and no# 0# )hree manu$acturing stages: stamping3 drilling3 milling# Decision varia%les: x. (num%er o$ part no#. to produce" x0 (num%er o$ part no#0 to produce" Model: Maximi?e @ 7 A:91x. / =.1x0 su%4ect to: 8x. / ;#9x0 .19 (stamping3hr" :#0x. / 8#=x0 =1 (drilling3 hr" =#.x. / 8#.x0 ..1 ($inishing3 hr" x.3 x0 1

08

09

"#ample !lympic Bi'e is introducing t&o ne& light&eight %icycle $rames3 the Deluxe and the Pro$essional3 to %e made $rom special aluminum and steel alloys# )he anticipated unit pro$its are A.1 $or the Deluxe and A.9 $or the Pro$essional# )he num%er o$ pounds o$ each alloy needed per $rame is summari?ed %elo&# A supplier delivers .11 pounds o$ the aluminum alloy and <1 pounds o$ the steel alloy &ee'ly# Deluxe Pro$essional Aluminum Alloy 0 8 Steel Alloy 0

5o& many Deluxe and Pro$essional $rames should !lympic produce each &ee'C
Model Dormulation Eer%al Statement o$ the !%4ective Dunction Maximi?e total &ee'ly pro$it# Eer%al Statement o$ the Constraints )otal &ee'ly usage o$ aluminum alloy F .11 pounds# )otal &ee'ly usage o$ steel alloy F <1 pounds# De$inition o$ the Decision Earia%les x. 7 num%er o$ Deluxe $rames produced &ee'ly# x0 7 num%er o$ Pro$essional $rames produced &ee'ly# Max .1x. / .9x0 s#t# ()otal +ee'ly Pro$it"

0x. / 8x0 F .11 (Aluminum Availa%le" -x. / 0x0 F <1 (Steel Availa%le" x.3 x0 G 1

A 1 2 3 4 Material !lumi"um #teel

B C Material Requirements Deluxe Profess. 2 4 3 2

D Amount Available 100 $0

A 6 7

Bi!es Ma"e # 1$ Maximi(e" )otal Profit 11 12 Constraints Amount %se" 13 !lumi"um 100 14 #teel $0

B C De&ision 'ariables Deluxe Professional 15 1%&500 412&500

'= '=

Amount Avail. 100 $0

0:

 !ptimal Solution According to the output: x. (Deluxe $rames" 7 .9 x0 (Pro$essional $rames" 7 .;#9 !%4ective $unction value 7 A8.0#91 ange o$ !ptimality (uestion: Suppose the pro$it on deluxe $rames is increased to A01# Hs the a%ove solution still optimalC +hat is the value o$ the o%4ective $unction &hen this unit pro$it is increased to A01C $ns'er: )he output states that the solution remains optimal as long as the o%4ective $unction coe$$icient o$ x. is %et&een ;#9 and 00#9# Since 01 is &ithin this range3 the optimal solution &ill not change# )he optimal pro$it &ill change: 01x. / .9x0 7 01(.9" / .9(.;#9" 7 A9:0#91# (uestion: H$ the unit pro$it on deluxe $rames &ere A: instead o$ A.13 &ould the optimal solution changeC $ns'er: )he output states that the solution remains optimal as long as the o%4ective $unction coe$$icient o$ x. is %et&een ;#9 and 00#9# Since : is outside this range3 the optimal solution &ould change#

.//0 Rule
)he .11I rule states that simultaneous changes in o%4ective $unction coe$$icients &ill not change the optimal solution as long as the sum o$ the percentages o$ the change divided %y the corresponding maximum allo&a%le change in the range o$ optimality $or each coe$$icient does not exceed .11I# ange o$ !ptimality and .11I ule (uestion H$ simultaneously the pro$it on Deluxe $rames &as raised to A.: and the pro$it on Pro$essional $rames &as raised to A.;3 &ould the current solution %e optimalC $ns'er H$ c. 7 .:3 the amount c. changed is .: - .1 7 : # )he maximum allo&a%le increase is 00#9 - .1 7 .0#93 so this is a :2.0#9 7 8<I change# H$ c0 7 .;3 the amount that c0 changed is .; - .9 7 0# )he maximum allo&a%le increase is 01 - .9 7 9 so this is a 029 7 81I change# )he sum o$ the change percentages is <<I# Since this does not exceed .11I3 the optimal solution &ould not change# )he .11I rule states that simultaneous changes in right-hand sides &ill not change the dual prices as long as the sum o$ the percentages o$ the changes divided %y the corresponding maximum allo&a%le change in the range o$ $easi%ility $or each right-hand side does not exceed .11I# ange o$ Deasi%ility and Sun' Costs (uestion 6iven that aluminum is a sun' cost3 &hat is the maximum amount the company should pay $or 91 extra pounds o$ aluminumC
!(justable )ells )ell $3$$ $)$$ /ame 2eluxe 4rofess& *i"al +e(uce( ,bjecti-e !llo.able 0alue )ost )oefficie"t 1"crease 15 0 10 12&5 1%&500 0&000 15 5 !llo.able 2ecrease 2&5 $&33333333

)o"strai"ts )ell /ame $3$13 !lumi"um $3$14 #teel *i"al #ha(o. )o"strai"t !llo.able !llo.able 0alue 4rice +&5& #i(e 1"crease 2ecrease 100 3&125 100 60 46&6666666% $0 1&25 $0 %0 30

ange o$ Deasi%ility and Sun' Costs $ns'er: Since the cost $or aluminum is a sun' cost3 the shado& price provides the value o$ extra aluminum# )he shado& price $or aluminum is the same as its dual price ($or a maximi?ation pro%lem"# )he shado& price $or aluminum is A-#.09 per pound and the 0;

maximum allo&a%le increase is :1 pounds# Since 91 is in this range3 then the A-#.09 is valid# )hus3 the value o$ 91 additional pounds is 7 91(A-#.09" 7 A.9:#09#

0<