research article integrated location-production-distribution...
TRANSCRIPT
-
Research ArticleIntegrated Location-Production-Distribution Planning in aMultiproducts Supply Chain Network Design Model
Vincent F. Yu,1 Nur Mayke Eka Normasari,1 and Huynh Trung Luong2
1Department of Industrial Management, National Taiwan University of Science and Technology, No. 43, Section 4,Keelung Road, Taipei 10607, Taiwan2Industrial and Manufacturing Engineering, Asian Institute of Technology, 58 Moo 9, Paholyothin Highway Klong Luang,Pathumthani 12120, Thailand
Correspondence should be addressed to Nur Mayke Eka Normasari; mayke [email protected]
Received 31 October 2014; Revised 7 February 2015; Accepted 23 February 2015
Academic Editor: Xuefeng Chen
Copyright Β© 2015 Vincent F. Yu et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This paper proposes integrated location, production, and distribution planning for the supply chain network design which focuseson selecting the appropriate locations to build a new plant and distribution center while deciding the production and distributionof the product. We examine a multiechelon supply chain that includes suppliers, plants, and distribution centers and develop amathematical model that aims at minimizing the total cost of the supply chain. In particular, the mathematical model considers thedecision of how many plants and distribution centers to open and where to open them, as well as the allocation in each echelon.The LINGO software is used to solve the model for some problem cases. The study conducts various numerical experiments toillustrate the applicability of the developed model. Results show that, in small and medium size of problem, the optimal solutioncan be found using this solver. Sensitivity analysis is also conducted and shows that customer demand parameter has the greatestimpact on the optimal solution.
1. Introduction
A supply chain is a network that consists of a set of geographi-cal facilities (suppliers, plants, andwarehouses or distributioncenter). Through those facilities, there is material flow fromsupplier, plant, warehouse, and end in the customer. It aimsat bringing the right amount of the right product to the rightplace at the right time [1]. Moreover, a supply chain networkdesign is a strategic decision that has high risk and long-termimpact in the supply chain system. The impact of efficiencysupply chain has become more important on the businesscompetitiveness [2]. The topic has triggered both researchersand practitioners to pay more attention to the supply chainnetwork design. Many studies have been conducted to helpthe practitioner inmaking the best decision on a supply chainnetwork. Indeed, determining the best supply chain networkis a challenge, starting with problem identification, problemformulation, and its final solution and decision.
Todayβs competition among companies and marketβsglobalization have resulted in firms developing a supply chain
that can respond quickly to customersβ need. In the currentbusiness environment, a company has to reduce costs whileimproving its customer service level to remain competitive[3], which also helps maintain profit margins. In order toachieve these goals, a company should appropriately selectthe location of the factory and the distribution center [4β6]. According to Altiparmak et al. [7], an optimal, efficient,and effective supply chain platform is provided by supplychain network (SCN) design, which also helps to improvesupply chain performance. Moreover, Ballou [8] noted thatthe SCN design goal is to maximize the financial ratio, whichis relevant to the objective of gaining the maximum return ofinvestment at the minimum cost.
Supply chainmanagement is divided into two levels: stra-tegic and operational. The strategic level primarily is aboutthe cost-effective location of facilities (plants and distributioncenters), the flow of products throughout the entire supplychain system, and the assignment in each echelon [9β12].The operational level is about the safety stock of eachproduct in each facility, the replenishment size, frequency,
Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2015, Article ID 473172, 13 pageshttp://dx.doi.org/10.1155/2015/473172
-
2 Mathematical Problems in Engineering
transportation, and lead time, and the customer service level.According to Beamon [13], determining an effective supplychain is an important component in supply chain design.In addition, the decisions regarding in which facilities theproduct should be made and how to serve customers are verycritical [14].
This paper provides a system optimization perspectivein strategic planning for a supply chain network designthat allows simultaneously determining the best location offacilities, raw material flow, and product flow on variousechelons. Previous research on strategic planning for supplychain starts by considering the basic problems that haveseveral characteristic, namely, single-period, single-product,single-echelon, and deterministic [15β25]. However, this isnot sufficient to cope with the realistic problem. Therefore,many extensions to the basic problem are needed to makethe problem more realistic. In this case, our paper considersmultiperiod, multiproduct, and multiechelon which are stillin deterministic situation tomake the basic strategic planningproblem more reasonable. A supply chain network designmodel helps managers conduct strategic planning for theircompany by selecting the best facility location thatminimizesthe total cost of the supply chain.The proposed multiproductsupply chain network design model herein helps in choosingthe appropriate location of a new plant and distributioncenter as well as the distribution of the product and rawmaterials when the demand varies during the differenttime period. Moreover, multiechelon which represents themultitype of facility is the crucial aspect to be considered instrategic planning.
The paper is organized as follows. Section 2 presentsprevious research to find the gap between this study andearlier related research. Section 3 describes the problemdefinition and the proposed mathematical model. Section 4contains the numerical experiments for the small andmedium cases. Section 5 offers a sensitivity analysis resultof the proposed model. Finally, Section 6 consists of theconclusion and suggestions for future research.
2. Literature Review
Several research integrated supply chain network designshave been developed to help practitioners solve their supplychain planning. Syarif et al. [26] studied a multiechelon,single-product logistic chain network model and proposed anovel technique as the solution method, called the spanningtree-based genetic algorithm (st-GA). The model is formu-lated by using a mixed integer liner programming (MILP)model. Their model only considers a single-product. Todemonstrate the effectiveness and efficiency of their proposedmethod, it is compared to the traditional matrix-basedgenetic algorithm (m-GA).The experiment result shows thatthe proposed method presents a better solution almost alltime and also performs better in computational time andmemory for computation.
Jakeman et al. [27] considered the strategic and opera-tional planning level decision in their research by developinga static model for a multiechelon, multiproduct supply chainnetwork design.They examined the single source distribution
system. For their solution, they used Lagrangian relaxationand a heuristic algorithm that utilizes the Lagrangian solu-tion. The result of their computation shows that the solutionmethod is both efficient and effective.
Shen [20] proposed a supply chain network designmodelwith profit maximization as the objective function, but itconsiders only a single-product. In addition, the companymay lose the customer if the productβs price is higherthan the customer reserve price. Altiparmak et al. [28]studied a single-product, multiechelon, and multiobjectiveSCN design. They set up a solution procedure based onthe genetic algorithm (GA) to find the optimal solutionto their problem. The multiobjective optimization problemconsists of many optimal solutions, called Pareto-optimalsolutions. The problem is formulated as a multiobjectivemixed integer nonlinear programmingmodel.The objectivesare to minimize total cost, maximize customer service, andmaximize utilization of the distribution centers (DCs).
Altiparmak et al. [7] presented a solution procedure fora multiproduct supply chain network (SCN) design based onthe steady-state genetic algorithm (ssGA) with a new encod-ing structure. They considered a single source, multiproduct,and multiechelon supply chain network design in which thenumber of customers and their demands are assumed tobe known. The problem, which is the NP-hard problem,is provided in mixed integer programming formulation. Inorder to investigate the effectiveness of the ssGA, threeother heuristic approaches are also used: Lagrangian heuris-tic (LH), hybrid genetic algorithm (hGA), and simulatedannealing. The experimentβs results show that ssGA has abetter solution than the other heuristic approaches used.Ying-Hua [29] adopts the model developed by Altiparmaket al. [7], which considers a single source, multiproduct,and multiechelon supply chain network design, but themodel only has multisources instead of a single source.Additionally, the plants and DCs that are open are known.To verify the efficiency of his proposedmethod, he comparedit to other algorithms, such as mathematical programming,the simple genetic algorithm, the coevolutionary geneticalgorithm, and the constraint-satisfaction genetic algorithm.The experimental result in Taiwanβs textile industry showsthat the proposed method of Ying-Hua [29] performs betterthan other researchersβ methods.
Bhutta et al. [30] developed an integrated location, pro-duction, distribution, and investment mixed integer linearprogramming (MILP)model in a two-echelon,multiproduct,multiperiod, and flexible facility capacitated with maximumprofit as the objective function. CoΜccola et al. [31] setup an integrated production and distribution MILP modelin a multiechelon, multiproduct, and single-period settingwith minimum total cost as the objective function. Theyconducted an empirical numerical experiment on six Euro-pean countries. Fahimnia et al. [32] presented an integratedproduction and distribution planning MILP model for atwo-echelon SC that considers several real world variablesand constraints. They used GA to optimize the model andsolved the medium-size case problem in their numericalexperiment. In addition, Bashiri et al. [33] and Badri etal. [34] developed a multiple-echelon, multiple-commodity
-
Mathematical Problems in Engineering 3
mathematical model for strategic and tactical planning. Themodel is developed as a MILP model in four echelons, butthey did not consider satisfying the demand constraint.
Many papers have developed a supply chain networkdesign through a mixed integer programming (MIP) model[35, 36]. However, in fact, the quantity of the commodity isusually an integer. Our paper considers location, production,and distribution planning in the supply chain network designproblem with multiechelon, multiproduct, and multiperiodcharacteristics in which the proposed model is pure integerlinear programming (PILP) model, having four echelons,multiproduct, and multiperiod demand and satisfying ademand constraint. Consideration of using PILP is intendedfor providing quality guarantees of optimality [37].Moreover,its application can be used for low volume discrete manufac-turing company of large equipment. In terms of multiplicity,our paper considers the most complex model in the area ofintegrated production and distribution planning.
3. Problem Definition and Model Formulation
Development of an efficient and effective supply chain isvery critical to achieving good performance. Therefore, in-depth analysis is needed when opening a new plant andnew distribution center in the appropriate location. Asidefrom that, multiple products instead of a single-product needto be considered in the problem of supply chain networkdesign and taking into account that the integer quantity inthe supply chain network design is more applicable. To dealwith this problem, this paper develops a pure integer linearprogramming (PILP) model that focuses on determining thelocations of the plants and distribution centers, as well as thenumber of those facilities, so that customer needs are satisfiedat a minimum total cost during the planning horizon.
This research focuses on the supply chain design problemwith the following characteristics.
(1) The distribution network under consideration is amultiechelon and multiproduct supply chain net-work.
(2) Demand in each time period (yearly) is deterministicand known.
(3) The plant or DC does not need to be opened at thebeginning of the planning horizon, and when one isopened, it will not be closed.
(4) Customers can receive the product from multipleDCs.
This research develops amathematicalmodel that helps todetermine the number and locations of plants and distribu-tion centers in a supply network and the assignment-relateddemand allocation in each echelon. Figure 1 depicts the sys-tem considered in this research. According to Jayaraman andPirkul [38], the key components of supply chain modelingthat should be considered by the model builder are supplychain drivers, supply chain constraints, and supply chaindecision variables of the model. Supply chain drivers repre-sent the goal setting of the model, supply chain constraintsrepresent the limitations on the range of decision alternatives,
and supply chain decision variables are the components thatset limits on the range of decision outcomes.
The objective function of this model is to minimize thetotal cost of the system. According to Fahimnia et al. [39],the total cost in the production and distribution networknaturally consists of the production cost and distributioncost. Production cost is the sum of the fixed opening costand the variable production cost, while the distribution costis the sum of the fixed cost of opening the distribution centercost, the variable inventory cost, and transportation cost.Therefore, the total cost in this model consists of cost to openthe plant, cost to purchase and transport raw material fromsupplier to plant, cost to manufacture the products, cost toopen the DCs, cost to transport the product from plant toDCs, inventory holding cost of each product in DCs, and costto transport the product from DCs to customer.
The decision variables in production and distributionplanning consist of the supplier stage and distribution stage.In the supplier stage, the decision variables consist of howmany suppliers should be there, how many quantity andfrequency of shipment from each supplier, what is theconfiguration of the supplier-plant distribution network, andwhere are the selected locations of the suppliers and plants.The decision variables in the distribution stage consist ofhow many distribution centers to operate, where shouldthey be located, and inventory in the distribution center[40, 41]. Therefore, we decide that the decision variables inthis research encompass determining where the plant anddistribution center will be opened, their distribution, and theproduction of the plant when it is opened.
The mathematical model of this research is developedbased on the model of Altiparmak et al. [28]. They set upa mathematical model that considers a single-product, fourechelons, a single source, and static demand. Our paperβsmathematical model has multiproducts, 4 echelons, multi-sources, and multiperiod demand characteristics.
To meet fluctuating customer demand, the end prod-ucts and the information exchange are conducted regularlythrough plants and distribution centers within a given pro-duction and service network. Indicators of supply chainperformance such as fill rate, customer service level, associ-ated cost, and capability of response can be obtained underdifferent network configurations through an evaluation of thesupply chain network configuration itself. Different networkconfigurations involve different stock levels of raw materials,subassemblies and end products, distribution center loca-tions, production policy (make-to-stock or make-to-order),production capacity (amount and flexibility), allocation rulefor limited suppliers, and transportation modes.
The common multiechelon supply chain network(MSCN) problem searches for a network configuration at aminimum cost. This is a NP-hard (nondeterministic polyno-mial-time hard) problem that employs a mathematicalprogramming formulation as a natural way to build anNP-hard problem, although it is not an efficient procedure.In Yeh [42], some parameters are known in advance,namely, the numbers and capacities (demand) of suppliers,plants, distribution centers (DCs) and customers, the unittransportation cost between suppliers and plants, plants
-
4 Mathematical Problems in Engineering
s = 1 Β· Β· Β· S
supplier(fixed location)
p = 1 Β· Β· Β· P
plant(location to be selected)
j = 1 Β· Β· Β· J
distribution centre(location to be selected)
i = 1 Β· Β· Β· l
customer zone(fixed location)
Dp,Op
brpjt
Wj, Gj, cji
mrjitdirt
QοΏ½spt
Figure 1: The supply chain network under consideration.
and DCs, and DCs and customers, as well as the fixed costfor operating plants and DCs. The goal of his research is toidentify the locations of plants and DCs and the quantitiesshipped between the various points that minimize total costand transportation costs. The problem in his research isformulated using a pure integer programming (PILP) model.
The proposed model uses the notations shown in Nota-tions section.
The problem is formulated as follows:
Min π = βπ
{πΈπππ1 +
π‘=π
β
π‘=2
πΈπ (πππ‘ β ππ(π‘β1))}
+
π‘=π
β
π‘=1
β
π
β
π
β
Vπ΅Vπ ππVπ ππ‘ +
π‘=π
β
π‘=1
β
π
β
π
π΄ππππππ‘
+β
π
{πΉππΊπ1 +
π‘=π
β
π‘=2
πΉπ (πΊππ‘ β πΊπ(π‘β1))}
+
π‘=π
β
π‘=1
β
π
β
π
β
π
πΎππππππππ‘
+
π‘=π
β
π‘=1
β
π
β
π
βπππ‘πππ +
π‘=π
β
π‘=1
β
π
β
π
β
π
πΏππππππππ‘,
(1)
subject to
β
π
πππππ‘ β₯ ππππ‘ βπ, π, π‘, (2)
β
π
β
π
πππππ‘ππ +β
π
βππ(π‘β1)ππ β€ πππΊππ‘ βπ, π‘, (3)
β
π
1
PRπππππ‘ β€ π·ππππ‘ βπ, π‘, (4)
β
π
πππππ‘ β€ ππππ‘ βπ, π, π‘, (5)
β
π
πVπππππ‘ β€ βπ
πVπ ππ‘ βV, π, π‘, (6)
β
π
πVπ ππ‘ β€ πΆπ V βV, π , π‘, (7)
βπππ‘ = β
π
πππππ‘ + βππ(π‘β1) ββ
π
πππππ‘ βπ, π, π‘, (8)
πππ‘ β₯ ππ(π‘β1) βπ‘ = 2, . . . , π, (9)
πΊππ‘ β₯ πΊπ(π‘β1) βπ‘ = 2, . . . , π, (10)
πππ‘ = {0, 1} , (11)
πΊππ‘ = {0, 1} , (12)
πVπ ππ‘ β₯ 0, and integer, (13)
ππππ‘ β₯ 0, and integer, (14)
πππππ‘ β₯ 0, and integer, (15)
πππππ‘ β₯ 0, and integer, (16)
βπππ‘ β₯ 0, and integer, (17)
βππ0 = 0 βπ, π. (18)Equation (1) shows the objective function of the model.
Equation (2) is the constraint for satisfying customerdemand. Equation (3) is the capacity constraint for DC π.Equation (4) is the capacity constraint for plant π. Equation(5) is the limitation of the product that is transported fromplant π to all DCs. Equation (6) is the requirement ofraw material V for production. Equation (7) is the capacityconstraint for the supplier. Equation (8) is the inventory
-
Mathematical Problems in Engineering 5
Table 1: Data for instances in the numerical experiment.
Single-period Multiperiod Multiperiodsmall-sized problem small-sized problem medium-sized problem
Number of customers 2 2 4Number of locations for distribution centers 2 2 4Number of locations for plants 2 2 4Number of suppliers 2 2 4Number of products 2 2 4Number of raw materials 2 2 4Length of planning horizon 1 2, 5, 10 4
Table 2: Parameter values for the numerical experiment.
NumberParameter Generated using1 Cost to open the plant π(πΈπ) Integer uniform distribution π(25000, 30000)2 Maximum capacity of plant π(π·π) Integer π(18, 22)3 Production rate of manufacturing product π(PRπ) Integer π(10, 15)4 Cost of transporting and purchasing raw material V from supplier π to plant π(π΅Vπ π) π(10, 15)5 Unit manufacturing cost of product π at plant π(π΄ππ) π(8, 10)6 Cost of transporting product π from plant π to DC π(πΎπππ) π(4, 8)7 Capacity of supplier s for raw material V(πΆπ V) Integer π(1250, 1500)8 Utilization rate of raw material v per unit of finished product π(πVπ) Integer π(1, 5)9 Cost to open DC π(πΉπ) Integer π(20000, 30000)10 Capacity of DC π(ππ) Integer π(250, 350)11 Space requirement rate of product π on a DC(ππ) π(1, 2)12 Demand at customer zone π for product π in time period π‘(ππππ‘) Integer π(30, 50)13 Unit inventory holding cost of product π in DC π(πππ) π(5, 10)14 Cost of transporting product r from DC j to customer π (πΏ πππ) π(8, 12)
balance equation of product π in DC π at time period π‘.Equation (9) ensures that the plant only opens once. Equation(10) ensures that the DC only opens once. Equations (11)-(12)are binary constraints for the decision variables. Equations(13)β(17) give the requirement of nonnegativity. Equation (18)shows the initial inventory in DC at the beginning of theplanning horizon.
4. Numerical Experiment
This paper examines both small-sized and medium-sizedproblems. We conducted the experiment mainly to show thecost savings advantage of the proposed integratedmodel overprevious related published models. Tables 1 and 2 present theindices and parameters of the model, respectively.
Table 1 gives the data of the test instances. There arefive test instances: one single-period small-sized problem(Instance 1), three multiperiod small-size problems (Instan-ces 2, 3, and 4), and one multiperiod medium-size problem(Instance 5). Instance 1 will be used as our base for thecomparative analysis to show the advantages of integratingmultiple period planning. In Instance 1, the numbers ofcustomer, DC, plant, supplier, product, and rawmaterials areall set to be 2, except for the planning horizon which is set tobe 1 year.We adopt the same data for Instances 2, 3, and 4with
different planning horizons of 2, 5, and 10 years, correspond-ingly. Lastly, in Instance 5, all parameters are set to be 4.
Table 2 gives the corresponding parameter value of themodel. The model consists of 14 parameters which are allgenerated using uniform distribution, some in integers andsome in real numbers.
Our proposed model is exactly solved using LINGO.The model formulation using the LINGO framework con-sists of three sections: (1) sets of variables and parameters;(2) corresponding data sets; (3) mathematical model. TheLINGO solver used branch and bound method to solve theproblem. This method is an intelligent enumeration processseeking a sequence of better and better solutions until the bestsolution is found. In the process of finding the best solution,the memory is updated with the best objective functionvalue found so far. This process continues until no furtherimprovement can be found.
The results of location and assignment planning for thesmall-size numerical experiment can be seen in Tables 3β6. The results of location and assignment planning for themedium-size one can be seen in Table 7.
Table 3 shows the network design results of single-periodsmall-sized problem.This implies that having plant 2 openedmaterials 1 and 2 only come from supplier 2. Plant 2 deliversall products 1 and 2 to DC 1; then DC 1 delivers all products
-
6 Mathematical Problems in Engineering
Table 3: Location and assignment planning for the single-period small-sized problem.
Optimal solution(objective function) $81,105.48Period Supplier(Plantrawmaterial) Plant(DCproduct) DC(Cusproduct)1 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)
Table 4: Results for multiperiod small-size problem, π = 2 years.
Optimal solution(objective function) $94,766.42Period Supplier(Plantrawmaterial) Plant(DC product) DC(Cusproduct)1 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)2 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)
Table 5: Network design result for multiperiod small-size problem, π = 5.
Optimal solution(objective function) $137,372.6Period Supplier(Plantrawmaterial) Plant(DCproduct) DC(Cusproduct)1 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)2 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)3 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)4 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)5 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)
Table 6: Network design results for multiperiod small-size problem, π = 10.
Optimal solution(objective function) $207,537.2Period Supplier(Plantrawmaterial) Plant(DCproduct) DC(Cusproduct)1 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)2 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)3 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)4 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)5 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)6 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)7 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)8 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)9 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)10 2(21, 22) 2(11, 12) 1(11, 12, 21, 22)
to all customers. The same way of explanation uses for thesolution configurations in the other columns and for thesucceeding tables until Table 7. Plant 2 delivers all products1 and 2 to DC 2; then DC 2 distributes these products to allcustomers.This gives us theminimum total cost of $81,105.48.
Table 4 shows the optimal network design for Instance 2(π = 2). The same optimal solution as in Table 3 is obtainedfor this instance for all periods. However, the optimal totalcost is different with the value of $94,766.42. Consequently,we observe a 17% increase in total cost by doubling π to 2years.
Tables 5 and 6 show the optimal network design forInstances 3 and 4 with π = 5 and π = 10, respectively. Thesame solution as in Table 3 is obtained for all periods in bothinstances, except that they have different optimal total costs.We have total cost of $137,372.6 for Instance 3 (π = 5) and$207,537.2 for Instance 4 (π = 10). Here, we observe a 51%
increase in total cost by doubling the planning horizon to 10years.
Table 7 shows the network design results of medium-sized problem. Only plants 1, 3, and 4 are opened. Supplier1 delivers raw materials 2 and 3 to plant 1, raw materials 2and 4 to plant 3, and raw material 3 to plant 4. Supplier 2delivers raw materials 1 and 4 to plant 1. Supplier 3 deliversraw material 2 to plant 1, raw material 3 to plant 3, and rawmaterial 4 to plant 4. Supplier 4 delivers raw material 1 toplant 2 and rawmaterials 1 and 2 to plant 4.The same solutionis obtained for all periods, except in period 3. In period 3,supplier 2 only delivers product to plants 3 and 4. All DCs 1,2, 3, and 4 are opened for all periods. All period has differentplant-DC solution mix. Furthermore, each DC delivers theircorresponding product to all customers with different DC-customer solution mix in period 4.
-
Mathematical Problems in Engineering 7
Table 7: Results for medium-size problem with π = 4.
Optimal solution(objective function) $510,608.30Period Supplier(Plantrawmaterial) Plant(DCproduct) DC(Cusproduct)
11(12, 13, 32, 34, 43);
2(11, 14);3(12, 33, 44);4(31, 41, 42)
1(24, 33, 34, 42, 44);3(11, 13, 21, 32);
4(12, 32, 33, 42, 43)
1(12, 23, 31, 41, 42, 43);2(11, 21, 34, 41, 44);3(14, 22, 32, 43);
4(13, 24, 33, 34, 42)
21(12, 13, 32, 34, 43);
2(11, 14);3(12, 33, 44);4(31, 41, 42)
1(24, 33, 34, 42, 44);3(11, 13, 21, 32);4(12, 32, 33, 43)
1(12, 23, 31, 41, 42, 43);2(11, 21, 34, 41, 44);3(14, 22, 32, 43);
4(13, 24, 33, 34, 42)
31(12, 13, 32, 34, 43);
2(11, 14);3(33, 44);4(31, 41, 42)
1(24, 33, 34, 42, 44);3(11, 13, 21, 32);
4(12, 32, 33, 42, 43)
1(12, 23, 31, 41, 42, 43);2(11, 21, 34, 41, 44);3(14, 22, 32, 43);
4(13, 24, 33, 34, 42)
41(12, 13, 32, 34, 43);
2(11, 14);3(12, 33, 44);4(31, 41, 42)
1(12, 24, 34, 42, 44);3(11, 13, 21, 32);4(12, 32, 33, 43)
1(12, 23, 31, 41, 43);2(11, 21, 34, 41, 44);3(14, 22, 32, 43);
4(13, 24, 33, 34, 42)
Table 8: Network design results for medium-size problem with π = 4.
Year Annual cost ($) Total1 2 3 4 5 6 7 8 9 10
Single-period 81105 81105 81105 81105 81105 81105 81105 81105 81105 81105 811055Multiperiod (π = 10) 20754 20754 20754 20754 20754 20754 20754 20754 20754 20754 207537Cost saving 60352 60352 60352 60352 60352 60352 60352 60352 60352 60352 603518
Table 8 shows the potential cost savings for multiple-period plan versus single-period plan. For example, given theoptimal cost, we have for Instance 1 a total of $811055 which isrequired for a plan done individually at the beginning of eachyear. On the other hand, given the optimal cost of $207,537.2for Instance 4, we can roughly estimate average annual costfor the entire planning horizon by dividing total cost by 10years. The potential estimated total saving is $603,518.
It should be noted that we intend to generate a fairlyconcentrated demand across periods. In this way, there willbe no much effect to the optimal network design across theplanning horizon. With this, the results imply that the costincreases with the increases in planning horizon but notlinearly as opposed to single-year planning.Thefixed cost ele-ment of the total cost is distributed over the number of peri-ods (years) included in the plan. As this number increases,this fixed cost will be stretched over the years. Thus, it ulti-mately gives us annual savings compared to single-year plans.
Tables 9, 10, and 11 show the production-distribution planfor Instances 1, 2, and 5, respectively. In these tables, thefollowing optimal values are indicated, namely, productionquantities needed for plants tomanufacture to fully fulfill cus-tomer demand, raw materials requirements from suppliers,finished product quantities to be transferred from plant toDC, and finished product quantities to be delivered fromDCto customers.
Table 9: Production-distribution plan for the single-period small-sized problem.
Origin Destination Period 11 2
Supplier 2 Plant 2 521 445Plant 2 DC1 86 91
DC1 Customer 1 44 45Customer 2 42 46
Table 10: Production-distribution plan for the multiperiod small-sized problem, π = 2.
Origin Destination Period 1 Period 21 2 1 2
Supplier 2 Plant 2 521 445 423 398Plant 2 DC1 86 91 67 88
DC1 Customer 1 44 45 31 46Customer 2 42 46 36 42
5. Sensitivity Analysis
Jakeman et al. [27] noted that sensitivity analysis is onestep in developing a model. Sensitivity analysis can alsoassist in executing the model [40]. Sensitivity analysis looks
-
8 Mathematical Problems in Engineering
Table 11: Production-distribution plan for the multiperiod medium-sized problem.
Origin Destination Period 1 Period 2 Period 3 Period 41 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
Supplier
1
Plant 1 190 327 219 342 202 311 159 339Plant 2Plant 3 1195 323 1166 334 1182 318 1226 274Plant 4 422 450 437 450
2 Plant 1 536 818 583 853 498 758 645 867
3Plant 1 12 14 108Plant 3 543 510 518 540Plant 4 434 414 423 414
4 Plant 3 393 418 386 298Plant 4 654 850 612 882 631 867 612 882
Plant
1
DC1 7DC2 41 49 35 44DC3 1 46 2 34 4 44 35DC4 14 68 22 74 13 64 32 71
3
DC1 39 72 55 52 50 52 74 52DC2 107 101 114 112DC3 35 42 34 12DC4
4DC1 47 59 63 37DC3 46 24 40 33 26 26 62 40DC4 17 80 84 15 85 77
DC
1
Cust. 1 45 36 42 44Cust. 2 48 48 38 48Cust. 3 38 33 35 45Cust. 4 1 2 24 22 23 4 15 21 14 29 4
2
Cust. 1 37 36 44 48Cust. 2 31 48 48 46Cust. 3 1 16 3 5Cust. 4 39 40 17 33 22 32 18 39
3
Cust. 1 46 34 44 35Cust. 2 49 48 30 39Cust. 3 32 34 30 35Cust. 4 25 35 30 40
4
Cust. 1 30 46 48 33Cust. 2 38 46 37 46Cust. 3 50 30 38 28 37 27 44 25Cust. 4 31 22 28 32
at the influence of parameters changes upon the objectivefunction. In addition, sensitivity analysis can characterize theuncertainty in the parameter [12].
We therefore conduct sensitivity analysis to analyze thechanges in the decision variables and the objective functionwhen the parameter values are changed. This section inves-tigates the changes in decision variables, namely, πππ‘, πVπ ππ‘,πππππ‘,πΊππ‘, andπππππ‘, the changes in the network configuration,
and the changes in objective function. We use Instance 2 toconduct the sensitivity analysis.
We performed the one-at-a-time (OAT) method in thisanalysis. In this method, we commonly used one param-eter to change at a time while leaving all others at theirbaseline values. This method of sensitivity analysis considersthe parameterβs variability and its associated influence inthe output model [43]. Table 12 presents the scenarios for
-
Mathematical Problems in Engineering 9
Table12:Scenario
susedin
sensitivityanalysis.
Parameter
Distrib
utionto
generate
rand
omvaria
ble
Scenarios
12
34
56
78
910
πΈπ
Integeru
niform
distr
ibution
1500
0β1700
01700
0β1900
01900
0β2100
02100
0β2300
02300
0β2500
03000
0β3200
03200
0β3400
03400
0β3600
03600
0β3800
03800
0β40
000
π·π
Integeru
niform
distr
ibution
4β6
7β9
10β12
13β15
16β18
22β24
25β27
28β30
31β33
34β36
PRπ
Integeru
niform
distr
ibution
1-22β4
4β6
6β8
8β10
15β17
18β20
21β23
24β26
27β30
π΅Vπ π
Uniform
distrib
ution
1-22β4
4β6
6β8
8β10
15β17
17β19
19β21
21β23
23β25
π΄ππ
Uniform
distrib
ution
3-4
4-5
5-6
6-7
7-8
10β12
12β14
14β16
16β18
18β20
πΎπππ
Uniform
distrib
ution
1-22-3
3-4
8β12
12β16
16β20
20β24
24β28
πΆπ V
Integeru
niform
distr
ibution
50β200
200β
500
500β
700
700β
1000
1000β1250
1500β160
01600β1700
1700β1800
1800β1900
1900β200
0
πVπ
Integeru
niform
distr
ibution
1-22-3
3-4
4-5
5β10
10β15
15β20
20β25
25β30
πΉπ
Integeru
niform
distr
ibution
1000
0β1200
01200
0β1400
01400
0β1600
01600
0β1800
01800
0β2000
03000
0β3200
03200
0β3400
03400
0β3600
03600
0β3800
03800
0β40
000
ππ
Integeru
niform
distr
ibution
1β50
50β100
100β
150
150β
200
200β
250
350β
400
400β
450
450β
500
500β
550
550β
600
ππ
Uniform
distrib
ution
2-3
3-4
4-5
5-6
6-7
ππππ‘
Integeru
niform
distr
ibution
5β10
10β15
15β20
20β25
25β30
50β6
060β70
70β80
80β9
090β100
πππ
Uniform
distrib
ution
1-22-3
3-4
4-5
10β12
12β14
14β16
16β18
18β20
πΏπππ
Uniform
distrib
ution
1-22β4
4β6
6-7
7-8
12β17
17β22
22β27
27β32
32β37
-
10 Mathematical Problems in Engineering
Table 13: Result of sensitivity analysis.
Number Parameter Impactπ πππ‘ πΊππ‘ πVπ ππ‘ πππππ‘ πππππ‘
1 Cost to open the plant π(πΈπ) β β β β β β2 Maximum capacity of plant π(π·π) β β β β β β3 Production rate of manufacturing product π(PRπ) β β β β β β4 Cost of transporting and purchasing raw material v from supplier s to plant π(π΅Vπ π) β β β β β β5 Unit manufacturing cost of product π at plant π(π΄ππ) β β β β β β6 Cost of transporting product r from plant p to DC π(πΎπππ) β β β β β β7 Capacity of supplier s for raw material V(πΆπ V) β β β β β β8 Utilization rate of raw material v per unit of finished product π(πVπ) β β β β β β9 Cost to open DC π(πΉπ) β β β β β β10 Capacity of DC π(ππ) β β β β β11 Space requirement rate of product π in a DC(ππ) β β β β β β12 Demand at customer zone π for product π in time period π‘(ππππ‘) β β β β β β13 Unit inventory holding cost of product π in DC π(πππ) β β β β β β14 Cost of transporting product π from DC π to customer π (πΏ πππ) β β β β β β
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Lrji
Yjr
dirt
ar
Wj
Fj
UοΏ½r
CsοΏ½
Krpj
Arp
BοΏ½sp
PRrDp
Ep
Figure 2: Percentage impact of each parameter.
sensitivity analysis. The ranges for all parameters values aregiven using the indicated distribution.
Table 13 shows the results of the sensitivity analysis. Theimpact for each parameter change is indicated with checkmarks under the corresponding columns for the objectivefunction and decision variables. Changes in inventory hold-ing cost show no impact to objective function and any ofthe decision variables. Parameters with the lowest impact areopening plant cost, manufacturing cost, transportation costto DC, opening DC cost, space requirement rate in DC, andtransportation cost to customer. Parameterswith themediumimpact are plant capacity, production rate, transportationcost to plant, supplier capacity, raw material utilization rate,and DC capacity. Finally, customer demand gives the highestimpact.
In addition, the impact of the variation parameter in theobjective function and decision variable are calculated as theweight of the influence, whereby the higher the value, thegreater impact to the decision variables. It is calculated bydividing the number of check marks (impact) in objectivefunction and decision variables in each parameter by total
impact. The total weight of all parameters is 1. The lowest,medium, and highest impact weights are 0.03, 0.11, and 0.17,respectively.
The impact of each parameter to the decision variable issignificantly different. It is depicted in Figure 2. Clearly, cus-tomer demand gives the greatest impact on the modelβs solu-tion. Moreover, significant impact to the decision variablessuch as those under distribution network is also revealed.Themanagerial implication from this result is that the networkconfiguration that the long-term plan may provide is animportant decision-making input.
6. Conclusions and Recommendations
We developed an integrated model for the problem oflocation, production, and distribution of multiproduct, four-echelon, and multiperiod supply chain network design. Themodel is coded using LINGO program and implemented itfor the small-sized and medium-sized problems.
The numerical experiment illustrates the applicability ofthe proposedmodel.With up to four echelons included in the
-
Mathematical Problems in Engineering 11
supply chain and asmuch as ten years of planning horizon, wedemonstrate cost savings advantage for the proposed model.Lastly, we determine the impact of all parameters involved.Customer demand gives the greatest impact on the modelβssolution.
The applicability of the proposed model arises, for exam-ple, mainly on manufacturing industries such as automobile,electronic, and furniture industries. We refer to the latestpublished models and build upon those to address someimportant features. However, there are some limitationsof this study that may need further attention for futuredirections. The model belongs to the static and deterministicclass with known demand. Solving this network flow probleminvolving four stages is computationally expensive usingexact methods. Thus, we are limited with the size that theLINGO program can handle. This is when heuristic algo-rithms prove to be useful. For example, a trend in nature-inspired algorithms such as genetic algorithm (GA) andsimulated annealing (SA) to name a few is known to solveNP-hard discrete combinatorial optimization problems withhigh quality at faster computation speeds.
Notations
Indices
πΌ: Set of customers (π β πΌ)π½: Set of potential locations of distribution centers (π βπ½)
π: Set of potential locations of plants (π β π)π: Set of suppliers (π β π)π : Set of products (π β π )π: Set of raw materials (V β π)π: Length of planning horizon (π years).
Notations Corresponding to the Activities in the Plant
Parameters
πΈπ: Cost to open the plant π ($)π·π: Maximum capacity of plant π (time unit)π΅Vπ π: Unit cost of transporting and purchasing rawmaterial V from supplier π to plant π ($/unit of rawmaterial)π΄ππ: Unit manufacturing cost of product π at plant π($/product)PRπ: Production rate of manufacturing product π(product/time unit)πΎπππ: Cost of transporting product π from plant π toDC π ($/product)πΆπ V: Capacity of supplier π for raw material V (rawmaterial unit)πVπ: Utilization rate of raw material V per unit offinished product π (raw material unit/product).
Variables
πππ‘: 1 if plantπ is opened in time period π‘; 0 otherwiseπVπ ππ‘: Quantity of raw material V shipped from sup-plier π to plant π in time period π‘ (raw material unit)ππππ‘: Quantity of product π produced by plant π intime period π‘ (product)πππππ‘: Quantity of product πshipped from plant π toDC π in time period π‘ (product).
Notations Corresponding to the Activities in DCs
Parameters
πΉπ: Cost to open DC π ($)ππ: Capacity of DC π (volume)ππ: Space requirement rate of product π in DC (vol-ume/product)ππππ‘: Demand at customer zone π for product π in timeperiod π‘ (product)πππ: Unit inventory holding cost of product π in DC π($/product)πΏπππ: Cost of transporting product π from DC π tocustomer π ($/product).
Variables
πΊππ‘: 1 if DC π is opened at time period t; 0 otherwiseπππππ‘: Quantity of product π shipped from DC π tocustomer π in time period π‘ (product)βπππ‘: Quantity of product π in DC π at the end of timeperiod π‘ (product).
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper.
Acknowledgments
The authors thank the editors and referees for their helpfulcomments and suggestions. This work was partially sup-ported by the National Science Council of Taiwan underGrant NSC 102-2221-E-011-082-MY3. This support is grate-fully acknowledged.
References
[1] J. Xu, Y.He, andM.Gen, βA class of random fuzzy programmingand its application to supply chain design,β Computers andIndustrial Engineering, vol. 56, no. 3, pp. 937β950, 2009.
[2] Z. Tang, M. Goetschalckx, and L. McGinnis, βModeling-baseddesign of strategic supply chain networks for aircraft manufac-turing,β Procedia Computer Science, vol. 16, pp. 611β620, 2013.
[3] D. J. Thomas and P. M. Griffin, βCoordinated supply chainmanagement,β European Journal of Operational Research, vol.94, no. 1, pp. 1β15, 1996.
-
12 Mathematical Problems in Engineering
[4] A. Amiri, βDesigning a distribution network in a supply chainsystem: formulation and efficient solution procedure,β EuropeanJournal of Operational Research, vol. 171, no. 2, pp. 567β576,2006.
[5] M. T. Melo, S. Nickel, and F. Saldanha-da-Gama, βFacilitylocation and supply chain managementβa review,β EuropeanJournal of Operational Research, vol. 196, no. 2, pp. 401β412,2009.
[6] P. N. Thanh, N. Bostel, and O. PeΜton, βA dynamic model forfacility location in the design of complex supply chains,β Inter-national Journal of Production Economics, vol. 113, no. 2, pp.678β693, 2008.
[7] F. Altiparmak, M. Gen, L. Lin, and I. Karaoglan, βA steady-state genetic algorithm for multi-product supply chain networkdesign,β Computers & Industrial Engineering, vol. 56, no. 2, pp.521β537, 2009.
[8] R. H. Ballou, βUnresolve issues in supply chain network design,βInformation Systems Frontiers, vol. 3, no. 4, pp. 417β426, 2001.
[9] M. T. Melo, S. Nickel, and F. S. D. Sa Gama, βDynamic multi-commodity capacitated facility location: amathematicalmodel-ing framework for strategic supply chain planning,β Computers& Operations Research, vol. 33, no. 1, pp. 181β208, 2006.
[10] H.Min and G. Zhou, βSupply chainmodeling: past, present andfuture,β Computers & Industrial Engineering, vol. 43, no. 1-2, pp.231β249, 2002.
[11] E. H. Sabri and B. M. Beamon, βA multi-objective approach tosimultaneous strategic and operational planning in supply chaindesign,β Omega, vol. 28, no. 5, pp. 582β598, 2000.
[12] Y. Zhan and M. Zhang, βApplication of a combined sensitivityanalysis approach on a pesticide environmental risk indicator,βEnvironmental Modelling & Software, vol. 49, pp. 129β140, 2013.
[13] B. M. Beamon, βSupply chain design and analysis: models andmethods,β International Journal of Production Economics, vol.55, no. 3, pp. 281β294, 1998.
[14] Z. M. Mohamed, βAn integrated production-distributionmodel for a multi-national company operating under varyingexchange rates,β Internationl Journal of Production Research,vol. 58, no. 1, pp. 81β92, 1999.
[15] B. Avittathur, J. Shah, and O. K. Gupta, βDistribution centrelocationmodelling for differential sales tax structure,β EuropeanJournal of Operational Research, vol. 162, no. 1, pp. 191β205, 2005.
[16] F. Barahona and D. Jensen, βPlant location with minimuminventory,βMathematical Programming, Series B, vol. 83, no. 1β3,pp. 101β111, 1998.
[17] A. Dasci and V. Verter, βContinuous model for production-distribution system design,β European Journal of OperationalResearch, vol. 129, no. 2, pp. 287β298, 2001.
[18] A. MarΜΔ±n and B. PelegrΜΔ±n, βThe return plant location problem:modelling and resolution,β European Journal of OperationalResearch, vol. 104, no. 2, pp. 375β392, 1998.
[19] S. Melkote and M. S. Daskin, βCapacitated facility loca-tion/network design problems,β European Journal of Opera-tional Research, vol. 129, no. 3, pp. 481β495, 2001.
[20] Z. M. Shen, βA profit-maximizing supply chain network designmodel with demand choice flexibility,β Operations ResearchLetters, vol. 34, no. 6, pp. 673β682, 2006.
[21] K. Sourirajan, L. Ozsen, and R. Uzsoy, βA single-productnetwork design model with lead time and safety stock consid-erations,β IIE Transactions, vol. 39, no. 5, pp. 411β424, 2007.
[22] C.-P. Teo and J. Shu, βWarehouse-retailer network designproblem,βOperations Research, vol. 52, no. 3, pp. 396β408, 2004.
[23] D. Tuzun and L. I. Burke, βTwo-phase tabu search approach tothe location routing problem,β European Journal of OperationalResearch, vol. 116, no. 1, pp. 87β99, 1999.
[24] Q. Wang, R. Batta, J. Bhadury, and C. M. Rump, βBudgetconstrained location problem with opening and closing offacilities,β Computers & Operations Research, vol. 30, no. 13, pp.2047β2069, 2003.
[25] T.-H. Wu, C. Low, and J.-W. Bai, βHeuristic solutions to multi-depot location-routing problems,β Computers & OperationsResearch, vol. 29, no. 10, pp. 1393β1415, 2002.
[26] A. Syarif, Y. Yun, and M. Gen, βStudy on multi-stage logis-tic chain network: a spanning tree-based genetic algorithmapproach,β Computers & Industrial Engineering, vol. 43, no. 1-2,pp. 299β314, 2002.
[27] A. J. Jakeman, R. A. Letcher, and J. P. Norton, βTen iterativesteps in development and evaluation of environmental models,βEnvironmental Modelling & Software, vol. 21, no. 5, pp. 602β614,2006.
[28] F. Altiparmak, M. Gen, L. Lin, and T. Paksoy, βA geneticalgorithm approach for multi-objective optimization of supplychain networks,β Computers and Industrial Engineering, vol. 51,no. 1, pp. 196β215, 2006.
[29] C. Ying-Hua, βAdopting co-evolution and constraint-satisfac-tion concept on genetic algorithms to solve supply chainnetwork design problems,β Expert Systems with Applications,vol. 37, no. 10, pp. 6919β6930, 2010.
[30] K. S. Bhutta, F.Huq,G. Frazier, andZ.Mohamed, βAn integratedlocation, production, distribution and investment model for amultinational corporation,β International Journal of ProductionEconomics, vol. 86, no. 3, pp. 201β216, 2003.
[31] M. E. CoΜccola, M. Zamarripa, C. A. MeΜndez, and A. EspunΜa,βToward integrated production and distribution managementin multi-echelon supply chains,β Computers and ChemicalEngineering, vol. 57, pp. 78β94, 2013.
[32] B. Fahimnia, L. Luong, and R.Marian, βGenetic algorithm opti-misation of an integrated aggregate productionβdistributionplan in supply chains,β International Journal of ProductionResearch, vol. 50, no. 1, pp. 81β96, 2012.
[33] M. Bashiri, H. Badri, and J. Talebi, βA new approach to tacticaland strategic planning in production-distribution networks,βApplied Mathematical Modelling, vol. 36, no. 4, pp. 1703β1717,2012.
[34] H. Badri, M. Bashiri, and T. H. Hejazi, βIntegrated strate-gic and tactical planning in a supply chain network designwith a heuristic solution method,β Computers and OperationsResearch, vol. 40, no. 4, pp. 1143β1154, 2013.
[35] B. Fahimnia, R. Z. Farahani, R.Marian, and L. Luong, βA reviewand critique on integrated production-distribution planningmodels and techniques,β Journal of Manufacturing Systems, vol.32, no. 1, pp. 1β19, 2013.
[36] C. J. Vidal and M. Goetschalckx, βStrategic production-distribution models: a critical review with emphasis on globalsupply chainmodels,βEuropean Journal ofOperational Research,vol. 98, no. 1, pp. 1β18, 1997.
[37] N. D. Crossman and B. A. Bryan, βSystematic landscape restora-tion using integer programming,β Biological Conservation, vol.128, no. 3, pp. 369β383, 2006.
[38] V. Jayaraman andH. Pirkul, βPlanning and coordination of pro-duction and distribution facilities for multiple commodities,βEuropean Journal of Operational Research, vol. 133, no. 2, pp.394β408, 2001.
-
Mathematical Problems in Engineering 13
[39] B. Fahimnia, L. H. S. Luong, and R. M. Marian, βOptimization/simulation modeling of the integrated production-distributionplan: an innovative survey,β WSEAS Transactions on Businessand Economics, vol. 5, no. 3, pp. 44β57, 2008.
[40] A. Saltelli, S. Tarantola, and F. Campolongo, βSensitivity analysisas an ingredient of modeling,β Statistical Science, vol. 15, no. 4,pp. 377β395, 2000.
[41] SΜ§. S. ErenguΜcΜ§, N. C. Simpson, and A. J. Vakharia, βIntegratedproduction/distribution planning in supply chains: an invitedreview,β European Journal of Operational Research, vol. 115, no.2, pp. 219β236, 1999.
[42] W.-C. Yeh, βA hybrid heuristic algorithm for themultistage sup-ply chain network problem,β International Journal of AdvancedManufacturing Technology, vol. 26, no. 5-6, pp. 675β685, 2005.
[43] D.M. Hamby, βA comparison of sensitivity analysis techniques,βHealth Physics, vol. 68, no. 2, pp. 195β204, 1995.
-
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