• 検索結果がありません。

T ermination

ドキュメント内 JAIST Repository (ページ 36-47)

The algorithm terminates either by delivering an optimal solution or by exceeding

pre-scribed maximal time. The former case is indicated by the emptiness of the active list.

The incumb entof permutationscheduleis thenoptimaland the incumb entupperbound

is the optimalvalue.

Computational experiments

Wehavetested the performance of several variantsof the proposed branch-and-cut

tech-nique using instances where the processing times on the second machine were generated

randomly. The objective of these experiments is to evaluate the performance of the

de-velop ed heuristics. Herewereport the resultswhere:

numb er of jobs rangesfrom 9to 13;

processing time onthe rst machine ranges from31 to59;

processingtimes b

1 and b

n

onthe second machineare xed atb

1

=30and b

n

=59,

resp ectively, and remaining pro cessing times on the second machine are generated

froma discrete uniformdistribution with a range30 to59.

We have tested the variants of the proposed lowerbounds and dominance rules. As

describedinsection5.2,wehaveproposedthreekindsoflowerbounds,inshort,simple(B),

improved(I) and LP(L) lowerb ounds. Moreover insection 5.4, wehave designed agroup

of dominance rule(D

ij ).

Forease of reference toa particular algorithm, weuse the following notation:

A B

, A I

and A L

denote the algorithms which are based onlyon the simple bounds,

improved b ounds and LP-bounds,respectively. Nodominance rule is used.

A

where 2 fB;I;Lg, 1 4 and 0 3; denotes the algorithm based

onbounds and dominancerules D

ij

with 1i and 0j .

Our exp erimentsconsist of ve parts. The purposeof eachexperimentare asfollows:

1. the evaluation of the inuence of pro cessing time a on the rst machine on the

diculty of problem F2jp

1j

=a;nmitj P

C

j

(see Figure 6.1, 6.2);

WeshowtheresultofalgorithmA B

only. TheresultsofA I

andA L

are similar.

We adopt the most simple algorithmsuchthat wehave develop ed.

2. the comparison of eciency ofthe lowerb ounds algorithm(see Figure6.3, 6.4);

We show the result of algorithm A

2

( = B;I;Land 1 3) only. The

resultsof A

( =0;1;3;4) are similarand 2-gap dominance rule isthe most

eectivegap rule.

i0

We showthe resultof algorithm A I

0

(0 4)only. The results of A B

and

A L

(1)aresimilarandthe algorithmwithimprovedboundsisthe fastest.

4. the evaluation ofnonzero-gap rules D

ij

(see Figure6.7, 6.8);

We showthe result of algorithm A I

(0 ; 3) only. The results of A B

andA L

are similarandthe algorithmwithimprovedb oundsisthefastesttoo.

5. Finally, in order to make it possible to compare the prop osed technique with

var-ious mixed integer solvers, we formulated the problem as the mixed integer linear

programming problem. Weuse the mixed integer formulationof the problem given

in the subsection 5.2.3. The results of comparison with the mixed integer solverof

the CPLEX system are summarized inFigure 6.9.

0 5 10 15 20 25 30

30 35 40 45 50 55 60

Searched Nodes(%)

10 instances n=9 jobs Algorithm A B

Figure6.1:

In all variants we used the least-b ound strategy for selecting a node from the active

list. The percentage of searched nodes refers to the value of 100m=n! where m denotes

the numb er of searched nodes and n is the numb er of jobs. The experiments were done

onSparc Station 20with 128Mbytememory.

0 1 2 3 4 5 6 7

30 35 40 45 50 55 60

Solution Time(sec.)

10 instances n=9 jobs

A B

Algorithm

Figure6.2:

2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 4.2

Searched Nodes(%)

n=10 jobs No. of jobset=1

a=45(most difficult case)

LP bounds Improved bounds

Simple bounds

A 12

A B

12

I

A 12 L

A B

22

A B

32

A 22 I

A 32 I

A 22

A L 32 L

Figure6.3:

0 200 400 600 800 1000 1200

Simple bounds Improved bounds LP bounds

Solution time(sec.)

A 22 L

A 12 L

A 32 L

A 32 I

A 22 I

A 12 I

A B A 22

B

A 12 B

32

n=10 jobs No. of jobset=1

a=45(most difficult case)

Figure6.4:

1.5 2 2.5 3 3.5 4 4.5 5 5.5

Searched Nodes(%)

n=11

No. of jobset=1

a=45(most difficult case)

A I

A 10 I

A 20

I A

30

I A

40 I

Figure6.5:

60 80 100 120 140 160 180 200 220 240

Solution time(sec.)

n=11

No. of jobset=1

a=45(most difficult case)

A 10 I

A 20

I A

30

I A

40 I

A I

Figure6.6:

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Searched Nodes(%)

n=11

No. of jobset=1

a=45(most difficult case)

A I

A I

A 12

A 13 10

11

I

I

A I

A I

A 22 A

23 20

21

I

I

A I

A I

A 32 A

33 30

31

I I

Figure6.7:

60 65 70 75 80 85 90

Solution time(sec.)

n=11

No. of jobset=1

a=45(most difficult case)

A I

A I

A 12

A 13 10

11

I

I

A I

A I

A 22

A 23 20

21

I

I

A I

A I

A 32

A 33 30

31

I

I

Figure6.8:

0 10 20 30 40 50 60 70 80

30 35 40 45 50 55 60

Solution time(sec.)

n=11

No. of jobset=1

CPLEX(mip)

Algorithm A

32 I

Figure6.9:

Concluding remarks

Throughout this dissertation we have b een concerned with deterministic owshop

prob-lems where the objective is to minimize the sum of completion times of all jobs. Since

the problem is knownto bevery hardin general,wehaverestricted our attention totwo

subcases: problemswith specialdominance relations amongmachines andproblems with

twomachines.

In Chapter 3,we haveconsidered owshopproblems with special structuredmachine

dominancerelation. AdiriandPohoryles[1]designedtwopolynomialtimealgorithms

con-structingthebestpermutationschedulesforFmjnmit;idmj P

C

j

andFmjnmit;ddmj P

C

j

respectively. Wehaveprovedthat allofthe ab ovealgorithmsconstruct not onlythe b est

permutationschedulesforthegivenproblem,butthattheyinfactproduceoptimal

sched-ules. In fact we have proved a much more general result, namely that under the above

machine dominance constraints, the search for optimalschedule can be restricted to the

set of all permutation schedules not only for the sum of completion times criterion but

for anarbitrary regularobjectivefunction.

In Chapter 4 and 5, we have considered the case of two machines. Two theorems of

paper[1] whichdeal with the F2jnmitj P

C

j

problem contain incorrect claims. We have

showed that both claims are incorrect by constructing a counter example. Moreover we

have managed to prove another statement which has a similar \avor" as the original

(incorrect) claims. Furthermore we have describ ed several variants of the

branch-and-cut technique for solving the F2jp

1j

= a;nmitj P

C

j

problem and reported results of

computational exp eriments. Their analysis leads tothe following conclusions.

Figures 6.1 and 6.2 suggest that the instances in which pro cessing time a on the

rst machineisaroundthe middleof therange ofpro cessingtimes b

i

onthesecond

machinearethemostdicult. Thereasonisthatwhenaisaroundthemiddleofthe

rangeof pro cessingtimesonthesecondmachine, many candidatejobswhichmight

b eassignedeitherafterorb eforeblockingpositionexistand,therefore,subproblem

P k

has many feasible solutions.

Figure6.3 showsthatthe improvedboundsand LPb ounds decreasethe numb erof

searchednodes. Figure6.4showsthatthe improvedboundsdecreasesolutiontimes

to o. In the moredetailed observation,wehaverealizedthat LPboundsisthe most

eective aroundthe rootof tree.

Figure6.5 and6.6 showthatD

10

-rules and D

20

-rulesincreasethe eciencyof

algo-rithms, but there is no improvement when dominance is based on the interchange

Similarly gures 6.7 and 6.8 show that D

i1

-rules and D

i2

-rules are eective in our

algorithms,but noimprovementisachievedwhengapisbasedonmorethan 2jobs.

Moreover these gures imply that D

ij

-rules (i = 1;2 and 0 j 2) are the most

eective aroundthe rootof the tree.

A comparisonwith the mixed integer solver ofCPLEXindicates thatthe proposed

branch-and-cuttechniqueis fasterthan CPLEX (see Figure6.9).

[1] Adiri, I. and Pohoryles,D. Flowshop/No-idle or No-wait scheduling to minimize the

sum of completion times, NavalResearchLogistics Quartely29, 495-504(1982)

[2] Baker,K. R.Introduction to sequencing and scheduling, John Wiley & Sons,(1974)

[3] Conway, R. W., Maxwell, W. L., and Miller, L. W. Theory of Scheduling,

Addison-Wesley, Reading,Mass. (1967)

[4] Della Croce, F., Narayan, V., and Tadei,R. The two-machine total completion time

ow shop problem. , European Journalof Operation Research,90, 227-237(1996)

[5] Garey,M. R.,Johnson,D. S.,and Sethi,R. Thecomplexityof owshop and jobshop

scheduling,Mathematics of Operations Research 1,117-129 (1976)

[6] Graham,R.L.,Lawler,E.L.,Lenstra,J.K.andRinnooyKan,A.H.G.Optimization

and approximation in deterministic sequencing and scheduling, a survey. Annals of

Discrete Mathematics 5, 287-326(1979)

[7] Ho, J.C. and Gupta,J. N. D. Flowshop Scheduling with Dominant machines,

Com-puters Ops. Res., 22, No. 2, 237-246(1995)

[8] Hoogeveen, J. A., and Kawaguchi, T. Minimizing total completion time in a

two-machine owshop: analysisof special cases.

[9] Ignall, E.and Schrage, L. Application of the Branch and Bound Technique to Some

Flow-Shop Scheduling Problems, Oper.Res., 13, 400-412(1965)

[10] Johnson,S.M.OptimalTwo-andThree-StageProductionScheduleswithSetupTimes

Included, NavalResearch Quarterly, Vol.1,No. 1 (March,1954)

[11] Kohler, W. H., and Steiglitz, K. Exact, approximate and guaranteed accuracy

algo-rithms for the owshop problem n/2/F/

F., ACM,Vol.22, No. 1,106-114(1975)

[12] Lawler, E. L., Lenstra, J. K., Rinnooy Kan, A. H. G. and Shmoys, D. B.

Sequenc-ing and Scheduling : Algorithms and Complexity, Handbooks in OR& MS, Elsevir

SciencePublishers B. V., Vol.4445-522(1993)

[13] Lomnicki, Z. A Branch-and-Bound Algorithm for the Exact Solution of the

Three-Machine SchedulingProblem,Operational ResearchQuarterly,Vol.16,No.1(March,

1965)

Special Cases of the Permutation Flow-Shop Problem, RAIRO, Vol. 17, No. 2

105-119(1983)

[15] Van de Velde, S. L. Minimizing the sum of the job completion times in the

two-machine owshop by Lagrangean relaxation., Annals of Operations Research, 26,

257-268(1990)

ドキュメント内 JAIST Repository (ページ 36-47)

関連したドキュメント