Research Article A Decomposition Strategy for Optimal...
Transcript of Research Article A Decomposition Strategy for Optimal...
Research ArticleA Decomposition Strategy for Optimal Design ofa Soda Company Distribution System
J. A. Marmolejo,1 I. Soria,2 and H. A. Perez2
1Faculty of Engineering, Anahuac University, 52786 Mexico City, MEX, Mexico2CADIT, Anahuac University, 52786 Mexico City, MEX, Mexico
Correspondence should be addressed to J. A. Marmolejo; [email protected]
Received 2 May 2015; Revised 9 July 2015; Accepted 28 July 2015
Academic Editor: Pan Liu
Copyright Β© 2015 J. A. Marmolejo et al.This is an open access article distributed under the Creative CommonsAttribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This work presents a distribution problem of products of a soda bottling company. Commodities are produced at several plants withlimited capacity and the demand of distribution centers is satisfied by shipping via cross-docking warehouses. The decompositionstrategy is proposed to determine which warehouse needs to be opened to consolidate the demand and by which warehouse eachdistribution center is served exclusively.The objective is minimizing fixed costs and total transportation costs.Themodel presentedis a mixed-integer programming model with binary variables for which we propose a decomposition strategy based on Bendersalgorithm. Numerical results show that the proposed strategy can provide the optimal solution of several instances. A large-scalecase study based on a realistic company situation is analyzed. Solutions obtained by the proposed method are compared with thesolution of full scale problem in order to determine the quality bound and computational time.
1. Introduction
According to international organisms, Mexico is the secondbiggest consumer of bottled soda with an average consump-tion of 160 liters per person a year after the United States,where 94% of the population consumes these types of drinks.The main point of sale of soda in Mexico is in small storeswhere 75% of the sales are carried out, 24% is in restaurants,and the remaining are in self-service stores [1]. For thecompany, it is very important to have an effective distributionnetwork that provides the possibility to keep a high levelof service to the client at the smallest possible cost; that is,the clients must have products in the opportune moment atminimum cost.
In 2012, the bottling company offered a level of serviceabove 99.6% (delivery requested/customer request), but in2013 the first supply problems were presented due to theproductive capacity of the plants; this situation originated thespecialization of the production lines of the company plants,holding the product readiness in the market at minimumcost; see Figure 1.
For this reason, we proposed a strategy for optimal designof a soda bottling company distribution system based on [1].
The proposed distribution network is constituted by plants,cross-dock warehouses, and distribution centers. Commodi-ties are produced at several plants with limited capacity andthe demand of distribution centers is satisfied by shippingvia cross-docking warehouses. The problem is to determinewhich warehouse needs to be opened to consolidate thedemand and by which warehouse each distribution center isserved exclusively.Theobjective isminimizing fixed costs andtotal transportation costs; see Figure 2.
The idea is to establish a network of arcs which enables theflow of products in order to satisfy demand of distributioncenters. A proper design can yield better operation levelsand cost reductions. In this problem, these reductions canbe significant. To the best of our knowledge, the specificmodel we pose in this study is not addressed in the literature;however, the area of fixed charge network design is closelyrelated.
This paper is organized as follows. In Section 2, wepresent a brief review of the network design problem.Section 3 contains the specific problem formulation andthen Section 4 describes the solution methodology. Section 5presents computational results of the application of Benders
Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2015, Article ID 891204, 7 pageshttp://dx.doi.org/10.1155/2015/891204
2 Mathematical Problems in Engineering
Plants K
1
2
3
.
.
.
1
2
3
.
.
.
1
2
.
.
.
K
Cross-dockingwarehouses I
Distributioncenters J
I J
Figure 1: Proposed distribution system.
Plants K
1
2
3
.
.
.
1
2
3
.
.
.
K
Distributioncenters J
J
Figure 2: Current distribution system.
Decomposition in several instances. Finally, conclusions arein Section 6.
2. Literature Review
Nowadays, companies still pay attention to efficiency becauseof tactical and operational impact at the executive level;these aspects remain relevant when a new framework isproposed, because all decisions will depend on it to continuethe pursuit of all the company objectives. This issue isreferred to as supply chain network design problems by mostresearchers and frequently begins from real-life applications.In this sense, a traditional paper is [2]. They present anew method for the solution of the problem addressing theoptimal location of distribution centers between plants andcustomers. A particular interest involves a study in facilitylocation that is reviewed in [3], where the authors explain
some important characteristics, methods, and application inthe context of supply chain management. Furthermore, theymentioned that the discrete facility location problems couldbe categorized as supply chain network design problems.Despite depending on situations, many techniques used inthe design of a distribution network are from the researchoperations area, especially those presented in [4]; somecontributions are made to the state of the art that followsthis subject for facility locations models. However, severalindustries are still facing problems related to design of thedistributionnetwork because it is a decision linked to improv-ing the level of service, and many other important metrics.Along the same line, in [5], they use a case study relatedwith a model for the distribution network of a region ofconsumer goods company taking into account several tacticaldecisions like lead time, credit performance, power, anddistributionβs reputation. Among various types of problemsdiscussed in supply chain management, facility location hasbeen studied for a long time by some researchers [6]. In [7],the authors follow an objective to propose a novel scheme ofdistribution. In these investigations, they include a classicalfacility location problem, and they presented cases of studiessolved by appropriate techniques.
Typically, these problems are presented as a mixed-integer programming (MIP) formulation [8, 9]. In [2], theauthors develop an algorithm based on Benders Decom-position for solving multicommodity distribution networkdesign problem. Another classical model is presented in [10].They consider a model to solve a minimization functionwhich includes fixed cost in warehouses and distributioncenters and transportation cost for multicommodities fromplants to warehouses and finally to customers.Moreover, theyconsider a tri-echelon, multicommodity system concerningproduction, distribution, and transportation planning. Theauthors use a Lagrangian relaxation-based heuristic to pro-vide an effective feasible solution for the problem. Also, theyreconsider other different characteristics for solving themainproblem of integrated logistics model [8, 11, 12]. In [13], theyconsider an integrated distribution network design and siteselection problem arising in the context of transportationplanning faced by the freight-forwarding industry; in thissense, they consider a strategic level multicommodity net-work design where each commodity is defined by a uniquepair of origin and destination points and known requiredflow amount and other considerations proper to the realproblem but using Benders Decomposition for its solution.Similarly, they illustrate the efficiency and the effectivenessof this approach. As in [14, 15] also a Benders Decompositionapproach is used, first in combinationwith an intelligent algo-rithm to improve the time solution for the master problemand then in modified version to exploit the mathematicalformulation of the problem in deterministic, multicommod-ity, single-period contexts, respectively. In [16], the authorsexposed that while these production-distribution problemsfocused to pursue exact solutions efficiently by using opti-mization software (only for small instances and small dimen-sions), the main reason to avoid it is because they containa large number of constraints and variables. Besides, they
Mathematical Problems in Engineering 3
propose heuristics for this bilevel mathematical problemusing Stackelbergβs equilibrium.
3. Mathematical Formulation
Let πΎ be the set of manufacturing plants. An element π β πΎ
identifies a specific plant of the company. Let πΌ be the set ofthe potential cross-dock warehouses. An element π β πΌ isa specific cross-dock warehouse. Finally, let π½ be the set ofdistribution centers; a specific distribution center is any π β π½.Let Z denote the set of integers {0, 1}.
Parameters are as follows:
ππ: capacity of plant π.πΎπ: capacity of cross-dock warehouse π.
πΉπ: fixed costs of opening cross-dock warehouse inlocation π.
πΊππ: transportation cost per unit of the product fromthe factories π to cross-dock warehouse π.
πΆππ: cost of shipping the product from the cross-dockπ to distribution center π.
ππ: demand of the distribution center π.
Decision Variables. We have the following sets of binaryvariables to make the decisions about the opening of thecross-dock warehouse, and the distribution for the cross-dock warehouse to the distribution center:
ππ =
{
{
{
1 if location π is used as a cross-dock warehouse,
0 otherwise,
πππ =
{
{
{
1 if cross-dock π supplies the demand of distribution center π,
0 otherwise.
(1)
πππ is the amount of product sent from factory π to cross-dock π and is represented by continuous variables.
We can now state the mathematical model as a (P)problem
minπππ,ππ,πππ
π = β
πβπΎ
β
πβπΌ
πΊπππππ +β
πβπΌ
πΉπππ +β
πβπΌ
β
πβπ½
πΆπππππππ (2)
subject to the following constraints:Capacity of the plant:
β
πβπΌ
πππ β€ ππ, βπ β πΎ. (3)
Balance of product:
β
πβπ½
πππππ = β
πβπΎ
πππ, βπ β πΌ. (4)
Single cross-dock warehouse to distribution center:
β
πβπΌ
πππ = 1, βπ β π½. (5)
Cross-dock warehouse capacity:
β
πβπ½
πππππ β€ πΎπππ, βπ β πΌ. (6)
Demand of items:πππ β€ πππ, βπ β πΌ, βπ β πΎ, (7)
π = min {ππ} , (8)πππ β₯ 0, βπ β πΌ, βπ β πΎ, (9)
ππ β Z, βπ β πΌ, (10)
πππ β Z, βπ β πΌ, βπ β π½. (11)
Objective function (2) considers in the first term the costof shipping the product from the plants π to cross-dockwarehouse π. The second term contains the fixed cost toopen and operate the cross-dock warehouse π. The last termincorporates the cost to supply the demand of the distributioncenter π. Constraints (3) imply that all that is produced inplant π does not violate the capacity of plant π. Balanceconstraints (4) ensure that the amount of products that arriveat a distribution center π is the same as that sent to the plantsπ. The demand of each distribution center π will be satisfiedby a single cross-dock warehouse π, which is achieved byconstraints (5). Constraints (6) bound the amount of prod-ucts that can be sent to a distribution center π from a cross-dock warehouse π that has been open. Finally, constraints (7)guarantee that any opened cross-dock warehouse π receivesat least the minimum amount of demand for distributioncenters π. The demand is satisfied by shipping via cross-dock warehouse with each distribution center being assignedexclusively to a single cross-dock warehouse. The possiblelocations for the cross-dock warehouses are given, but theparticular facilities to be used will be selected as a result ofthe minimum total distribution cost. As a result, efficientload consolidation arises as an important opportunity forprofitability in this work. To the best of our knowledge, thespecific model we pose in this study is not addressed in theliterature; however, the area of fixed charge network designis closely related. Our formulation facilitates the use of aBenders Decomposition framework for its solution.
4. Solution Methodology
Because of the economic importance of the network designproblems and their combinatorial nature, several solution
4 Mathematical Problems in Engineering
.
.
.
.
.
.
1
2
.
.
.
Cross-dockingwarehouses I
Distributioncenters J
I
J
Figure 3: One cross-dock warehouse to one distribution center.
methodologies have been developed. In this paper, wedescribe a decomposition approach for solving the sodabottling company problem.The approach is based onBendersDecomposition, which is one of the most successful solutionapproaches. Model (2)β(11) presents a structure well-suitedfor a primal decomposition approach (Benders Decomposi-tion).
This solution method is based on the situation thatwe can decompose the original problem obtaining severalsmaller and thus easier to solve subproblems. In this case,Benders Decomposition is used to meet the optimal locationof intermediate warehouses between plants and distributioncenters. As in [2], this model proposes that no customer zoneis allowed to deal with more than one cross-dock warehouse,since πππ must be 0 or 1 and not fractional. Additionally, wecan mention that the real case presented has a large numberof binary variables compared to the number of constraints.Thus, the Benders Decomposition is the better way to solvethe optimal design of the distribution system, because itexploits the special structure of the original problem. InFigure 1 a graphical representation of the proposed distribu-tion system is shown. This graphical illustration is differentfrom the one presented in Figure 2. The difference betweenthe current system and the proposed is because the demandis satisfied by shipping via cross-dock warehouse with eachdistribution center, which is assigned exclusively to a singlecross-dock warehouse; see Figure 3.
4.1. Benders Decomposition. The optimal design of a sodabottling company distribution system is a problem thathas a large number of complicating binary variables andthus involves a large CPU time to find an optimal integersolution; for this reason, we applied Benders Decompositionframework [17]. This method projects problem (2)β(11) ontothe space defined by the binary variables πππ and ππ. Theoriginal problem P is decomposed into two different prob-lems: a restricted master problem and Benders subproblem.
The subproblem (SP) is a dual lineal problem (LP) of P andis obtained by fixing the variables to either 0 or 1. Benderssubproblem provides an upper bound of the original prob-lem. Let (SP) be the dual problem of P and the dual variablesassociated with constraints (3), (4), and (7), respectively. Wehave the following:
Subproblem (SP):
maxππ,πΌπ ,π½ππ
Ξ¦ = β
πβπΎ
ππππ +β
πβπΌ
β
πβπ½
(πππππ) πΌπ + β
πβπΎ
β
πβπΌ
πππ½ππ,
ππ +πΌπ +π½ππ β€ πΊππ, βπ β πΌ, βπ β πΎ,
ππ β€ 0,
πΌπ unrestricted,
π½ππ β₯ 0.
(12)
Relaxed Master Problem (RMP):
minππ,πππ
Ξ©, (13)
Ξ© β₯ β
πβπΌ
πΉπππ +β
πβπΌ
β
πβπ½
(πΆπππππππ) ππ +β
πβπΌ
β
πβπΎ
πππ½ππ
+β
πβπΌ
β
πβπ½
(πππππ) πΌπ,
(14)
β
πβπΌ
πΉπππ +β
πβπΌ
β
πβπ½
(πΆπππππππ) ππ +β
πβπΌ
β
πβπΎ
πππ½ππ
+β
πβπΌ
β
πβπ½
(πππππ) πΌπ β€ 0,(15)
β
πβπΌ
πππ = 1, βπ β π½, (16)
β
πβπΌ
πππ β€ 0, βπ β π½, (17)
β
πβπ½
πππππ β€ πΎπππ, βπ β πΌ, (18)
β
πβπ½
πππππ β€ 0, βπ β πΌ, (19)
ππ, πππ β {0, 1} . (20)
The master problem is an integer problem and its vari-ables are considered as complicating variables. To determinethe values of these variables, the Benders master problemmust be solved. Since the number of primal cuts is toolarge, Benders [17] proposed to solve a Relaxed MasterProblem (RMP), by taking only a subset or Benders cuts,and to generate these cuts, one by one, in each iterationof the algorithm. Relaxed Master Problem provides feasiblesolutions through Benders optimality cuts (14) because theyare based on optimality conditions of the subproblem andthrough valid constraints called feasibility cuts in everyiteration of the algorithm. Feasibility cuts (15), (16), (17), (18),and (19) enforce necessary conditions for feasibility of the
Mathematical Problems in Engineering 5
Table 1: Comparison of the instances.
Instance πΎ πΌ π½ Continuous variables Binary variables ConstraintsINST 0 2 2 2 4 6 10INST 1 4 5 17 20 90 36INST 2 4 10 17 40 180 51INST 3 6 25 40 150 1025 121INST-R CASE 44 56 254 2464 14280 466
{Initialization}ππ, πππ := Fix integer variables to given feasible integer valuesLB := ββ
UB := +β
while UB β LB > π do{solve subproblem SP}if Unbounded then
Get unbounded ray ππ, π½ππ, πΌπ
Add feasibility cut to master problem (MP)else
Get extreme point ππ, π½ππ, πΌπAdd optimality cut to master problem (MP)UB := min{UB, SP}
endif{solve master problem (MP)}LB := Ξ©
endwhile
Procedure 1: Benders Decomposition (ππ, πππ, ππ, π½ππ, πΌπ).
primal subproblem. Master problem provides a lower boundof the original problem (P).
Benders Algorithm. A brief summary of the algorithm is givenin Procedure 1 for completeness.
5. Computational Experience
The full scale model and the decomposition strategy pro-posed were implemented in GAMS [18] using the solverCPLEX [19] for MIP and LP problems (master problem andBenders subproblem). All mathematical models were carriedout on an AMD Phenom II N970 Quad-Core with a 2.2GHzprocessor and 4GB RAM. Because the major difficulty ofBenders method is the solution of master problem, andbecause [2] suggests that the Benders master problem shouldnot be solved to optimality, we set GAMS parameter OPTCRat 0.0015; that is, the relative termination tolerance is within0.15% of the best possible solution. In the first iteration ofalgorithm, we fixed all integer variables to 1. Additionally,the size of all MIP models was reduced through presolverphase of CPLEX. Benders algorithm stops when the valuesof lower bound and upper bound are equal, except for a smalltolerance π = 0.15%:
π = [(UB β LB)
UB] β 100%. (21)
16000
14000
12000
10000
8000
6000
4000
2000
0
INST0Q
uant
ityINST1 INST2 INST3 INST-RCASE
Instances
Continuous variablesBinary variablesConstraints
Figure 4: Model complexity.
To test the efficiency in terms of CPU time and qualityof our solution method, we solved instances with data havingdemand distribution characteristics that reflect real applica-tions. We generated randomly four instances with differentnumber of plants, cross-dock warehouses, and distributioncenters. Additionally, we solved a realistic case of a sodabottling company. Table 1 indicates the specific size of eachinstance.
The model complexity of each instance can be seen inFigure 4. This figure shows the increase in problem size interms of the number of constraints, continuous variables, andbinary variables.
5.1. Results. Benders Decomposition described in the pre-vious section was used to design the optimal distributionsystem for the soda bottling company. Because the amountof memory and the computational effort needed to solve theinstances grow significantly with the number of variablesand constraints, we propose a decomposition strategy basedon Benders Decomposition. The comparison results demon-strate that the π-optimal solution is very close to the optimalsolution (GAMS-CPLEX Solution).
Table 2 illustrates that the comparison results demon-strate that the CPU time of proposed decomposition strategyis less than direct solution with GAMS-CPLEX. The GAP isless than 0.2%:GAP
= [(GAMS Solution β Benders Decomposition Solution)
GAMS Solution]
β 100%.
(22)
6 Mathematical Problems in Engineering
Table 2: Performance of Benders Decomposition and direct solution of full scale problem.
Instance Opt. value CPU time LB UB GAP CPU timeGAMS-CPLEX (sec.) (%) (sec.)
INST 0 1270 1.5 1270 1270 <0.2 0.5INST 1 179170 2.5 179168 179168 <0.2 1.5INST 2 485140 3.5 485142 485142 <0.2 2.5INST 3 1.00E + 07 930 1.00E + 07 1.00E + 07 <0.2 630INST-R CASE 8.45E + 10 4160 8.43E + 10 8.43E + 10 0.2 2160
7
6
5
4
3
2
1
0
β1
Γ106
Obj
ectiv
e fun
ctio
n
Lower boundUpper boundCPLEX
INST1
1 2 3 4 5
Iterations
Figure 5: Benders versus full scale solution.
3
2.5
2
1.5
1
0.5
0
Γ106
Obj
ectiv
e fun
ctio
n
Lower boundUpper boundCPLEX
INST2
1 2 3 4 5
Iterations
Figure 6: Benders versus full scale solution.
The progression of upper bound and lower bound valuesover the iterations around the optimum in all instances isplotted in Figures 5, 6, 7, and 8. Clearly, the use of feasibilitycuts is very effective in speeding up convergence and theyaffect the quality of both the upper and the lower bounds. Inour numerical studies, we observe similar high performancein convergence with feasibility cuts in other instances. Forall the instances solved, tight lower and upper bounds were
1.2
1
0.8
0.6
0.4
0.2
0
Γ109
Obj
ectiv
e fun
ctio
n
Lower boundUpper boundCPLEX
INST3
1 2 3
Iterations
Figure 7: Benders versus full scale solution.
9
8
7
6
5
4
3
2
1
0
Γ1012
Obj
ectiv
e fun
ctio
n
Lower boundUpper boundCPLEX
INST-REAL CASE
1 2 43
Iterations
β1
Figure 8: Benders versus full scale solution.
obtained with a small GAP in just a few iterations of theproposed algorithm.
These measures reported show that the performance ofthe Benders Decomposition seems to be indifferent to itssize when feasibility cuts are implemented. The results ofthe experiments reported in Table 2 show that the proposedBenders procedure produces very good feasible solutionscompared to the optimal/best available ones generated byCPLEX in significantly less CPU time.
Mathematical Problems in Engineering 7
6. Conclusions
In this paper, we present an optimal design of a sodabottling company distribution system problem.We proposeda decomposition strategy that shows acceptable convergenceproperties. In all instances, the number of iterations requiredto get convergence is less than five. The feasibility cuts allowreaching an π-optimal solution. The total CPU time requiredto solve the large-scale case study based on the realistic com-pany situation was less than 2200 seconds. This instance has14280 binary variables, 2464 continuous variables, and 466constraints.The proposed decomposition strategywas shownto be very efficient for the case study. For this class of prob-lems, global optimality is not guaranteed in a reasonable time,but the solution through Benders Decomposition providesgood feasible solutions and good bounds to the optimum.
Moreover, it is important to mention that the mainobjective was to compare the CPU effort and the qualitybounds of the full scale solution through a commercial solverand the proposed decomposition strategy. We can finallyconclude that we could propose a strategy for designing thedistribution system of a soda bottling company in competi-tive CPU times. The solutions obtained are close to the opti-mal values reported by commercial optimizers. This ensureshaving a flexible and scalable solution tool when the companydecides to increase the number of its facilities.
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper.
References
[1] I. Soria, Rediseno de la cadena de abastecimiento de un grupoembotellador de bebidas [M.S. thesis], ITESM Campus Toluca,Toluca, Mexico, 2008.
[2] A. M. Geoffrion and G. W. Graves, βMulticommodity distri-bution system design by benders decomposition,βManagementScience, vol. 20, no. 5, pp. 822β844, 1974.
[3] M. T. Melo, S. Nickel, and F. S. da-Gama, βFacility locationand supply chain managementβa review,β European Journal ofOperational Research, vol. 196, no. 2, pp. 401β412, 2009.
[4] A. Klose andA. Drexl, βFacility locationmodels for distributionsystem design,β European Journal of Operational Research, vol.162, no. 1, pp. 4β29, 2005.
[5] A. Cintron, A. R. Ravindran, and J. A. Ventura, βMulti-criteriamathematical model for designing the distribution network ofa consumer goods company,β Computers & Industrial Engineer-ing, vol. 58, no. 4, pp. 584β593, 2010.
[6] M. S. Jabalameli, B. Bankian Tabrizi, and M. Javadi, βCapac-itated facility location problem with variable coverage radiusin distribution system,β International Journal of Industiral Engi-neering & Production Research, vol. 21, no. 4, pp. 231β237, 2010.
[7] P. N.Thanh, O. Peton, and N. Bostel, βA linear relaxation-basedheuristic approach for logistics network design,β Computers &Industrial Engineering, vol. 59, no. 4, pp. 964β975, 2010.
[8] V. Jayaraman andH. Pirkul, βPlanning and coordination of pro-duction and distribution facilities for multiple commodities,β
European Journal of Operational Research, vol. 133, no. 2, pp.394β408, 2001.
[9] H. Sadjady and H. Davoudpour, βTwo-echelon, multi-com-modity supply chain network design with mode selection, lead-times and inventory costs,βComputers and Operations Research,vol. 39, no. 7, pp. 1345β1354, 2012.
[10] H. Pirkul and V. Jayaraman, βProduction, transportation, anddistribution planning in a multi-commodity tri-echelon sys-tem,β Transportation Science, vol. 30, no. 4, pp. 291β302, 1996.
[11] V. Jayaraman, βAn efficient heuristic procedure for practical-sized capacitated warehouse design andmanagement,βDecisionSciences, vol. 29, no. 3, pp. 729β745, 1998.
[12] H. Pirkul and V. Jayaraman, βA multi-commodity, multi-plant,capacitated facility location problem: formulation and efficientheuristic solution,β Computers & Operations Research, vol. 25,no. 10, pp. 869β878, 1998.
[13] H. Uster and H. Agrahari, βA Benders decomposition approachfor a distribution network design problem with consolidationand capacity considerations,β Operations Research Letters, vol.39, no. 2, pp. 138β143, 2011.
[14] W. Jiang, L. Tang, and S. Xue, βA hybrid algorithm of tabusearch and benders decomposition for multi-product produc-tion distribution network design,β in Proceedings of the IEEEInternational Conference on Automation and Logistics (ICALβ09), pp. 79β84, August 2009.
[15] H. M. Bidhandi, R. M. Yusuff, M. M. H. M. Ahmad, and M. R.Abu Bakar, βDevelopment of a new approach for deterministicsupply chain network design,β European Journal of OperationalResearch, vol. 198, no. 1, pp. 121β128, 2009.
[16] J.-F. Camacho-Vallejo, R. Munoz-Sanchez, and J. L. Gonzalez-Velarde, βA heuristic algorithm for a supply chainβs production-distribution planning,β Computers & Operations Research, vol.61, pp. 110β121, 2015.
[17] J. F. Benders, βPartitioning procedures for solving mixed-variables programming problems,βComputationalManagementScience, vol. 2, no. 1, pp. 3β19, 2005.
[18] A. Brooke, D. Kendrick, andA.Meeraus,GAMS: AUserβs Guide,Boyd & Fraser Publishing Company, 1998.
[19] GAMS Development Corporation,GAMS:The Solver Manuals;CONOPT; CPLEX; DICOPT; LAMPS; MILES; MINOS; OSL;PATH; XA; ZOOM, GAMS Development Corporation, 1994.
Submit your manuscripts athttp://www.hindawi.com
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttp://www.hindawi.com
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Stochastic AnalysisInternational Journal of