This article was downloaded by: [Bhaumik, Pradip K.

] On: 18 February 2010 Access details: Access Details: [subscription number 919352111] Publisher Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 3741 Mortimer Street, London W1T 3JH, UK

International Journal of Systems Science

Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713697751

Optimal shrinking of the distribution chain: the facilities delocation decision

Pradip K. Bhaumik a a International Management Institute, New Delhi-110 016, India Online publication date: 17 February 2010

To cite this Article Bhaumik, Pradip K.(2010) 'Optimal shrinking of the distribution chain: the facilities delocation

decision', International Journal of Systems Science, 41: 3, 271 280 To link to this Article: DOI: 10.1080/00207720903326860 URL: http://dx.doi.org/10.1080/00207720903326860

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

International Journal of Systems Science Vol. 41, No. 3, March 2010, 271280

Optimal shrinking of the distribution chain: the facilities delocation decision

Pradip K. Bhaumik*
International Management Institute, B-10, Qutab Institutional Area, New Delhi 110 016, India (Received 10 July 2008; final version received 27 May 2009) Closure of facilities is quite common among both business firms and public sector institutions like hospitals and schools. Although the facilities location problem has been studied extensively in the literature, not much attention has been paid to the closure of facilities. Unlike the location problem, the existing facilities and the corresponding network impose additional constraints on the closure or elimination of facilities and to highlight the difference between the two, we have called this the facilities delocation problem. In this article, we study a firm with an existing distribution network with known retailer and distributor locations that needs to downsize or shrink its distribution chain due to other business reasons. However, it is not a reallocation of demand nodes among the retained distributors. An important condition stipulates that all demand nodes must continue to get their supplies from their respective current distributors except when the current source itself is delocated, and only such uprooted demand nodes will be supplied by a different but one of the retained suppliers. We first describe the delocation problem and discuss its characteristics. We formulate the delocation problem as an integer linear programming problem and demonstrate its formulation and solution on a small problem. Finally, we discuss the solution and its implications for the distribution network. Keywords: facilities location; distribution chain; delocation; fixed-charge problem; ILP

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

1. Introduction In many sectors of the economy, favourable location of facilities is critical to the business success of the firm itself. Facility location is concerned with the placement of one or more facilities in a way that optimises certain objectives such as minimising costs, maximising the service level/providing equitable services to customers, maximising market share captured, minimising the time taken to deliver emergency services or a combination of these. The location decision for a new facility is very important for most organisations manufacturing and service in the private as well as the public sectors. In fact, for many service facilities like bank branches and retail stores, the location decision also affects the demand for the service and its market share unlike manufacturing facilities for whom the cost is largely affected. The number and location of distributors is of paramount importance in the design of a distribution network. The distributor becomes the single point of contact for each of the many retailers and the closer the distributor, the faster the response time and consequently the higher the service level of the distribution network. In fact, quite often some measure of the distance between retailers and their respective distributors, e.g. mean or maximum, is used as a measure of the service level. The distributor may
*Email: pkbhaumik@imi.edu
ISSN 00207721 print/ISSN 14645319 online 2010 Taylor & Francis DOI: 10.1080/00207720903326860 http://www.informaworld.com

provide many services including warehousing, orderfilling, information-processing, billing and collection. An additional distributor may increase the fixed cost of operating the distributorship but decrease the cost of transportation and improve the service level. The fixedcharge problem attempts to minimise the total cost including the fixed cost and the variable cost of transportation. As the flow of products through the distribution network increases, more distributors appear in the optimum solution and the minimum cost solution also has a higher service level. Although this movement towards a greater number of distributors holds true in most cases, in practice the actual distribution network is not always an optimum one due to historical legacy, strategic and other practical considerations. However, it is easy to see that if product volumes fall then a reverse movement towards fewer distributors may have to be set in motion. Globalisation has exposed firms to heightened business risks from enhanced competition. Consequently, the business environment of many companies has undergone unimaginable changes engendering new problems requiring new methodologies and new solutions. Often firms need to downsize or shrink their distribution chain for extraneous reasons anticipated to continue over the medium or the long term. This could be due to a host of factors

272

P.K. Bhaumik its earliest roots to present maturity. An extensive literature review of facility location models in the context of supply chain management is given in Melo, Nickel, and Saldanha-da-Gama (2009). Different authors have formulated and suggested solution procedures for the problem of locating facility(ies) on a plane, on a network or on discrete points (Liao and Guo 2008; Meyerson, Munagala, and Plotkin 2008; Nadirler and Karasakal 2008; Batanovic, Petrovic, and Petrovic 2009; Bischoff, Fleischmann, and Klamroth 2009; Boccia, Sforza, and Sterle 2009; Marin, Nickel, Puerto, and Velten 2009; Meng, Huang, and Cheu 2009; Puerto, Ricca and Scozzari 2009). In planar location problems, demand occurs and new facilities can be located anywhere on a plane. In network location models, demand occurs and new facilities can be located only on the nodes or links of the network. In both of these, the distance between any two points can be expressed as a function of the coordinates of the two points. Discrete location models allow for the use of arbitrary distances or costs between nodes and as such, the structure of the underlying network is lost. Although the location problem has been studied in detail even from different disciplinary areas of geography, operations research, economics, management and mathematics, it is surprising to note the absence of attention paid to the delocation problem in the literature. In economics and in economic geography, the term delocation is often used to mean the first half of relocation, i.e. closing a facility in its existing location so as to start operations from a new location (e.g. Midelfart-Knarvik and Overman 2002; Baldwin, Forslid, Martin, Ottavino, and Robert-Nicoud 2003). However, we use the term as an antonym of location; while in a location decision we decide where to commence operations from a new facility, a delocation decision involves deciding which of the existing facilities should cease operations. As shown by ReVelle, Murray, and Serra (2007), closure of facilities is quite common among business firms as well as health and educational institutions in the public sector. Modelling the school system consolidation under declining enrolments, Bruno and Anderson (1982) developed an early model providing for the closure of some schools. Diamond and Wright (1987) and Church and Murray (1993) extended these models with the concepts of balancing utilisation. Murray and Wu (2003) modelled urban public transport networks where some stops would be eliminated, whereas some others would be relocated due to the changes in population density and network changes. Klincewicz, Luss, and Yu (1988) developed a largescale multilocation capacity planning model providing

related to competitive, technological, environmental or lifestyle changes, and as the demand for its products or services reduce, the firm may respond by shrinking its distribution network along with other responses. For example, Coca-Cola India faced a similar situation a few years ago when the Delhi-based Centre for Science and Environment, a green think tank, tested its soft drinks and reported that all contained unsafe levels of pesticide in them. Lifestyle changes among consumers in many countries brought about by the current global recession may require similar response from many firms. Although closure of facilities is quite common, surprisingly, not much attention has been paid to this problem in the literature and in the absence of any formal approach to guide them, the decision maker is forced to use methods that are mostly ad hoc and intuitive in nature. In this article, we focus on a firm that wishes to downsize its distribution network. As the flow of products in the distribution network reduces, the number of distributors needed may have to be reduced. Given an existing distribution network with known retailer and distributor locations, it becomes very important to decide which distributors to eliminate as the decision affects both the cost and the service level of the modified distribution chain. It should be understood that while in the growth phase the company has to decide where to locate the distributors (the location problem), in the downsizing phase it has to decide which distributors to eliminate or delocate and to highlight the difference, we have called this the delocation problem. After the short introduction, a brief review of delocation problems discussed in the literature is presented in Section 2. The delocation problem is then described and formulated as a mathematical programming problem in Section 3. We next present a small application and discuss the results in Section 4. Finally, conclusions are drawn in Section 5.

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

2. The delocation problem Many variations of the location problem have been studied in the literature. An excellent review of location problems their formulation, solution procedures and applications is presented by Brandeau and Chiu (1989). They also refer to some earlier review articles published in the literature. Two books one by Daskin (1995) and the other by Drezner (1995) also provide an overview of different location problems and a detailed bibliography on the subject. A recent volume edited by Drezner and Hamacher (2001) provides current development in this area, while Smith, Laporte, and Harper (2009) charts the progress of locational analysis from

International Journal of Systems Science for plant closures in some periods in a multiperiod setting. In a similar multiperiod framework, Melachrinoudis and Min (2000) formulated a multiple objective model involving the dynamic relocation and phase-out of a hybrid, two-echelon plant/warehousing facility. A budget-constrained location problem was formulated by Wang, Batta, Badhury, and Rump (2003) that simultaneously considered opening some new branches and closing some existing branches in the branch network of a commercial bank. Almost all of these papers consider relocation problems where some existing facilities would be closed while some new facilities could be opened at other locations to accommodate demand changes and other factors. The only paper that focuses on the facility closure was published as late as 2007 and models firms both with and without competition (ReVelle et al. 2007). For the firm facing competition, they model a market served by the firm as well as by the competition. They assume that demand from each demand node will be fully supplied from the nearest supplier either belonging to the firm or to its competitor and the formulation minimises the demand lost to competition. In the second model for a firm without competition, for example in the public sector providing healthcare or education services, they minimise the number of people made worse-off when some of the facilities are closed to reduce costs. 3.1. A simple fixed-charge formulation of the facilities closure problem

273

The problem appears to have all the characteristics of a simple fixed-charge problem and amenable to such a formulation. It can then be formulated as an integer linear programming (ILP) after defining the parameters and the decision variables as shown below. Parameters: i index of servers, distributors or supply nodes, in the existing distribution network m total number of existing supply nodes or distributors r number of supply nodes to be retained (r 5 m) j index of demand nodes or retailers in both the existing and the new set up n total number of demand nodes or retailers fi cost of operating the ith supply node per period Mi the maximum number of demand nodes that can be served from the ith supply node cij cost of supplying the jth demand node from the ith supply node per period Decision variables:  1, if the ith existing supply node is retained xi , 0, if not i 1, 2, . . . , m 8 > 1, if demand node j is served from supply < node i in new set up , yij > : 0, if not i 1, 2, . . . , m; j 1, 2, . . . , n
m X i1 m n XX i1 j1

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

3. Problem formulation We consider the general case of a firm serving a spatial market that is represented by discrete points in a connected network. Each node in the network represents a demand area, local market or a retailer location. Customers get their respective demand met by the retailers. Some of the nodes in the network also represent the current location of servers or distributors that service the demand nodes or retailers. The network under consideration has n demand nodes currently served by m servers. In the changed scenario as the demand for its products has reduced (although not uniformly across all demand nodes), the firm wants to reduce the number of servers to r (r 5 m) by delocating or eliminating [(m r) p] distributors. We assume that this reduction will not affect the demand at any of the n demand nodes. However, the closure of some distributors is expected to reduce both the cost and the service level. We further assume that the cost effect can be measured through a fixed cost of operating the distributorships and a variable cost of transportation that varies with the product flow to each retailer.

Minimise z subject to:

m X i1 n X j1

fi xi

cij yij

yij 1,

j 1, 2, . . . , n

yij

Mi xi ,

i 1, 2, . . . , m

n X j1

xi r, xi 2 f0, 1g yij 2 f0, 1g

274

P.K. Bhaumik demand nodes ( j 1, 2, . . . , n) such that each demand node ( j) is supplied from a unique supply node. The firm wishes to close p of these existing m supply nodes. With the closure of some servers, the service level would deteriorate but that is expected to be compensated by a reduction in cost. Also, all the demand nodes must continue to get their supplies from their respective existing supplier, and exceptions will be made only if the existing supplier is eliminated. The pure delocation problem can thus be formulated as shown below. Additional parameters: p number of supply nodes to be eliminated sj index of the supply node for the jth demand node Di Set of indices of all the demand nodes for the ith supply node Ri the maximum number of additional demand nodes that can be served from the ith supply node Decision variables: 8 > 1, if the ith existing supply node is eliminated > > > < in the new set up , xi > 0, if not, i:e: if the ith existing supply node > > > : is retained in the new set up i 1, 2, ... ,m 8 > 1, if demand node j is served from a different > > > > retained supply node i in new set up because > > > > > > the current supply node is eliminated > < , yij 0, if not, i:e: if demand node j is continued > > > > to be served from supply node i both in > > > > > the current and the new set up or if supply > > > : node i is eliminated i 1, 2, ...,m; j 1, 2, ...,n
m n XX i1 j1

The problem is formulated to minimise the total cost per period comprising of a fixed cost of operating the retained servers and another cost of supplying from each retained server to each of the n demand nodes. Very high or infinite cost can be assigned to those routes which are infeasible or undesirable. Constraint set (2) ensures that each demand node is supplied from a unique server, while constraint set (3) mandates that only retained servers can supply to any demand node and the total number of demand nodes supplied from a retained server i is within its capacity Mi. As described earlier, Mi is the maximum number of demand nodes that can be served from the ith current supplier with existing supply infrastructure that costs fi per period. The motivation for the closure of some distributors being the reduction of demand for its products, the firm may find that for many supply nodes Mi is higher than the number of demand nodes actually supplied from the ith supply node. It is obvious that the above simple fixed charge formulation is independent of the existing distribution network, but for the fact that the candidate server nodes that can be retained are from among the existing servers that is, only r out of the existing m server nodes are to be retained. The existing network may have been optimal at some previous period but may not be optimal at present when the product flow in the network has reduced in general, but the demand at each retailer may not have reduced by the same percentage. Constraint (4) has been added so that the exact number of distributors can be ensured in the solution. The pure delocation problem cannot be viewed as a simple fixed charge location problem to locate r (r 5 m) distributors. It is not difficult to see that the solution to the simple fixed charge formulation above may reallocate some of the demand nodes to new supply nodes to reduce costs. This may not be acceptable due to operational and practical considerations. This is because of the legacy of the existing distributors who may or may not be retained. The firm is not considering starting afresh and finding an optimum r distributor location solution due to the heavy cost of disruption of services and loss of trust and faith among current distributors. In other words, each demand node must continue to get its supplies from its current supplier except when the current supplier is eliminated. The formulation of the delocation problem becomes more complex if these restrictions are taken into account and is presented next.

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

" cij yij

Minimise z

m X i1

fi xi

m X X i1 j2Di

! # cij xi 5

subject to:
m X i1

yij xsj ,

j 1, 2, . . . , n

3.2. Formulation of the pure delocation problem The pure delocation problem begins with a given network with known supply nodes (i 1, 2, . . . , m) and

n X j1

yij

Ri 1 xi ,

i 1, 2, . . . , m

International Journal of Systems Science

275

j=2 j=1 j=5 j=4 j=3 j=6 j = 10 i=1 j=8 j = 13 j = 11 j = 12 j = 16 j = 14 i=4 j = 15 j=7 i=3 j=9

i=5 i=2 j = 19 j = 18

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

j = 17

Figure 1. Existing distribution network.

m X i1

xi p, xi 2 f0, 1g yij 2 f0, 1g

The objective is to minimise the cost increase consequent to the closure of some facilities. This is the increase due to the cost of the new links created net of the savings due to closure of some existing servers and links therefrom which would not be used in the new set up. Constraint set (6) defines that for every demand node j, yij 0 if the source node (sj) is retained (i.e. xsj 0) and if the source node is eliminated, then this uprooted demand node has to be linked to one new source node. Similarly, the constraint set (7) requires that corresponding to every eliminated supply node (i.e. with xi 1), all the yijs ( j 1, 2, . . . , n) should be equal to zero as new links cannot be established from eliminated supply nodes. The same constraints also ensure that the number of additional demand nodes connected to the ith retained supply node (after the closure of some other supply nodes) does not exceed its unutilised capacity Ri. Constraint (8) specifies the number of supply nodes that need to be eliminated. 4. Application and results We present a small application with 19 demand nodes ( j 1, 2, . . . , 19) supplied from 5 servers or

distributors (i 1, 2, 3, 4, 5) in the existing set up. The existing distribution network is shown in Figure 1. Table 1 provides the parameter values for the model formulation. The cost figures per period of operating the ith supply node ( fi) as well as of using the link from supply node i to demand node j for each value of i and j, (cij), are shown. Similarly, Table 1 also lists the values for sj and Di for each demand node and each supply node, respectively. The cost of having a distributor at the ith supply node ( fi) is actually a fixed cost and incorporates the establishment costs, salaries, utilities, storage costs and other similar expenses incurred per period. Similarly, cij represents the cost per period of supplying demand node j from the ith distributor and incorporates the transportation and distribution cost. In fact, it can be computed if the demand per period from the jth node is known and also known is the unit cost of transportation on the route from the ith distributor to the jth demand node. With the given data, the delocation problem can be formulated as shown below: Minimise z
5 19 XX i1 j1

cij yij 256x1 124x2 136x3

152x4 111x5

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

276

Table 1. Parameter values for the existing distribution network. j Max total demand nodes 1 2 3 4 5 6 7 8 9 Mi 11 7 8 9 6 sj 1 3 1 1 4 1 5 4 5 5 3 37 62 39 57 71 40 55 35 51 63 19 57 46 68 75 32 37 27 48 54 49 59 22 29 45 35 47 52 59 63 21 49 27 32 46 3 Ri 43 53 22 32 35 1 58 64 44 26 45 4 Max. addl. demand nodes 10 cij 41 45 50 63 58 1 30 42 44 49 47 2 44 36 38 39 34 3 49 48 30 27 25 4 57 45 39 18 24 5 69 63 47 29 36 4 40 29 35 41 27 5 58 28 64 64 54 2 56 22 62 57 39 2 74 48 51 48 32 5 11 12 13 14 15 16 17 18 19

Supply node

Optg cost

Supplying to demand nodes

P.K. Bhaumik

fi

Di

1 2 3 4 5

49 32 36 41 28

1,3,4,6,8,10 11,17,18 2,7,12 5,9,13,15 14,16,19

International Journal of Systems Science

Table 2. Solution of the delocation problem. Optimum solution No. of servers to be eliminated (p) 0 1 2 3 4 Cost increase 0 41 55 20 Cost 779 738 724 759 Eliminated supply nodes Nil x3 x3, x5 x2, x3, x5 New links

277

Nil y1,2, y1,7, y5,12 y1,2, y1,7, y2,12, y2,16, y2,19, y4,14 y1,2, y1,7, y1,11, y4,12, y4,14, y4,16, y1,17, y1,18, y1,19 No feasible solution

subject to: y1,1 y2,1 y3,1 y4,1 y5,1 x1 y1,2 y2,2 y3,2 y4,2 y5,2 x3 y1,3 y2,3 y3,3 y4,3 y5,3 x1
Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

y4,15 y4,16 y4,17 y4,18 y4,19 5x4 5 y5,1 y5,2 y5,3 y5,4 y5,5 y5,6 y5,7 y5,8 y5,9 y5,10 y5,11 y5,12 y5,13 y5,14 y5,15 y5,16 y5,17 y5,18 y5,19 3x5 3 x1 x2 x3 x4 x5 p x1 , x2 , x3 , x4 , x5 2 f0, 1g y1,1 , y1,2 , . . . , y5,19 2 f0, 1g When solved for different values of p, using a standard package for ILP, the solution obtained is as given in Table 2. If one server is to be closed, the cost per period actually decreases and the solution suggests that x3 should be eliminated. In the new set up, each of the uprooted demand nodes, namely [D3 {2, 7, 12}], will have to be supplied from some operating server. In the optimum solution, demand nodes 2 and 7 are being supplied from supply node 1 (using new links y1,2 and y1,7) while demand node 12 is being supplied from supply node 5 (using new link y5,12). We also notice that with a larger number of closure of servers, the cost first decreases and then increases. If the motivation for closure of servers is cost reduction as the service level is likely to worsen with fewer servers then it works up to a limit (p 2) and after the limit, both cost and service level may worsen. If the simple fixed charge formulation is used then the solutions for different number of supply nodes retained is as shown in Table 3. Here again we notice that the cost first decreases and then after a point starts rising. For our problem this happens when three supply nodes are retained. The difference between the delocation problem formulation and the simple fixed charge formulation can be seen more clearly from the two diagrams shown in Figures 2 and 3, respectively. The simple fixed charge solution of Figure 3 requires demand node 7 to be supplied from supply node 1, whereas it is being supplied from supply node 3 in the current set up. While supply node 3 is retained in the new set up, it does not supply to its earlier demand nodes 7 and 12. This may not be acceptable in the long-term strategic

y1,4 y2,4 y3,4 y4,4 y5,4 x1 y1,5 y2,5 y3,5 y4,5 y5,5 x4 y1,6 y2,6 y3,6 y4,6 y5,6 x1 y1,7 y2,7 y3,7 y4,7 y5,7 x3 y1,8 y2,8 y3,8 y4,8 y5,8 x1 y1,9 y2,9 y3,9 y4,9 y5,9 x4 y1,10 y2,10 y3,10 y4,10 y5,10 x1 y1,11 y2,11 y3,11 y4,11 y5,11 x2 y1,12 y2,12 y3,12 y4,12 y5,12 x3 y1,13 y2,13 y3,13 y4,13 y5,13 x4 y1,14 y2,14 y3,14 y4,14 y5,14 x5 y1,15 y2,15 y3,15 y4,15 y5,15 x4 y1,16 y2,16 y3,16 y4,16 y5,16 x5 y1,17 y2,17 y3,17 y4,17 y5,17 x2 y1,18 y2,18 y3,18 y4,18 y5,18 x2 y1,19 y2,19 y3,19 y4,19 y5,19 x5 y1,1 y1,2 y1,3 y1,4 y1,5 y1,6 y1,7 y1,8 y1,9 y1,10 y1,1 1 y1,12 y1,13 y1,14 y1,15 y1,16 y1,17 y1,18 y1,19 5x1 5 y2,1 y2,2 y2,3 y2,4 y2,5 y2,6 y2,7 y2,8 y2,9 y2,10 y2,11 y2,12 y2,13 y2,14 y2,15 y2,16 y2,17 y2,18 y2,19 4x2 4 y3,1 y3,2 y3,3 y3,4 y3,5 y3,6 y3,7 y3,8 y3,9 y3,10 y3,11 y3,12 y3,13 y3,14 y3,15 y3,16 y3,17 y3,18 y3,19 5x3 y4,1 y4,2 y4,3 y4,4 y4,5 y4,6 y4,7 y4,8 y4,9 y4,10 y4,11 y4,12 y4,13 y4,14 5

278
Table 3. Solution of the simple fixed charge problem.

P.K. Bhaumik

j=2

Supply nodes Supply retained (r) Cost node (i) 5 716 1 2 3 4 5 1 2 3 5 1 2 4 1 4

j=1

Distributing to demand nodes ( j) 1, 3, 6, 7, 10, 11 17, 18 2, 4, 5, 8 9, 14, 15 12, 13, 16, 19 1, 3, 6, 7, 10, 11 17, 18 2, 4, 5, 8, 9 12, 13, 14, 15, 16, 19

j=4 j=3 j=6 j = 10 i=1 j=8 j=7

j=5 i=3 j=9

j = 13 j = 11 j = 12 j = 16 j = 14

j = 15

706

i=5 i=2 j = 19 j = 18

j = 17

701

1, 2, 3, 4, 6, 7, 10, 11 12, 16, 17, 18, 19 5, 8, 9, 13, 14, 15 1, 2, 3, 4, 6, 7, 10, 11, 16, 17, 18 5, 8, 9, 12, 13, 14, 15, 19 No feasible solution

2
Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

747

Figure 3. Solution of the simple fixed charge problem with 4 distributors retained (cost 706).

j=2 j=1 j=4 j=3 j=6 j = 10 i=1 j=8 j = 13 j = 11 j = 12 j = 16 j = 14 j=7 i=4 j = 15 j=9 j=5

i=5 i=2 j = 19 j = 18

j = 17

changes in demand at the different demand nodes. However, even if the existing distribution network is optimal with the existing number of supply nodes, the solutions for the two problems would be different with fewer supply nodes retained. This can be clearly seen from Table 3. If the existing distribution network had four supply nodes these would be supply nodes numbered 1, 2, 3 and 5. If only three supply nodes are to be retained, Table 3 suggests that supply nodes numbered 1, 2 and 4 should be operating. The delocation problem would not allow the creation of a new distributor (node 4) while reducing the total number of distributors. The delocation problem formulation and its solution is relevant for distribution chains going ahead with the closure of some of its distributors without annoying any of the retained distributors.

Figure 2. Solution of the delocation problem with 1 distributor delocated (cost 738).

interest of the firm although it may be cheaper than the solution given by the pure delocation formulation. The delocation problem solution, on the other hand, does not violate any of the existing linkages as shown in Figure 2. A na ve argument can be advanced that the difference in the optimal solutions of the pure delocation problem and the simple fixed charge problem is because the existing distribution chain is not optimal with five supplier nodes. In practice, even if the distribution chain is designed to be optimal at some stage, it usually deviates from optimality due to uneven

5. Conclusions Closure of facilities and shrinkage of the distribution chain is not very uncommon for business units. But this problem has not been studied much in the literature. With reduced volume of materials flowing through the distribution chain, a firm may wish to reduce the number of distributors with a minimum consequential cost increase. However, it is not a reallocation of demand nodes among the retained distributors. An important condition of the pure delocation problem stipulates that all demand nodes must continue to get their supplies from their respective current sources except when the current source itself is delocated and only such uprooted demand

International Journal of Systems Science nodes will be supplied by a different but one of the retained suppliers. The pure delocation problem formulated honours the existing distribution network which may not be optimal due to historical legacy, strategic and other practical considerations. We have assumed that the elimination of some distributors will not affect the demand at any of the demand nodes or retailers although it is expected to reduce both the cost and the level of service to the retailers. As the customers get their demand serviced by the retailers and there is no delocation of retailers, it is reasonable to assume that the demand at any of the demand node or retailer will not be affected. Aboolian, Berman, and Krass (2007) have studied the case of new facility location where the customer demand is elastic, expanding with the utility of the service offered by the facilities increases. In this article, we focus on a firm that wishes to downsize its distribution network. As the flow of products in the distribution network reduces, the number of distributors needed may have to be reduced. Given an existing distribution network with known retailer and distributor locations, it becomes very important to decide which distributors to eliminate as the decision affects both the cost and the service level of the modified distribution chain. It should be understood that while in the growth phase the company has to decide where to locate the distributors (the location problem), in the downsizing phase it has to decide which distributors to eliminate or delocate and to highlight the difference, we have called this the delocation problem. We have specified the exact number of suppliers to be eliminated in the formulation, but it can be easily modified to specify a maximum operating cost per period after the shrinkage or even the minimum reduction in operating cost per period. Also, if there are limits on the additional quantity that can be handled by a distributor, such limits could be easily handled by modifying the constraint set (7) to accommodate additional quantities rather than the number of additional supply nodes Ri. Similarly, although our model uses the cost per period of using a link, it can easily be modified to handle unit transportation cost and demand per period from each demand node. The formulation of the delocation problem and its justification remains valid irrespective of the size of the problem. Depending upon the size of the existing distribution network, the actual problem could turn out to be much bigger with hundreds or even thousands of retailers. We have not explored the difficulties encountered in solving the delocation problem formulated. While the small problem demonstrated above could be solved exactly using a branchand-bound procedure, efficient heuristics may be

279

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

needed for very large problems. In this context, the constraint expansion approach, which has been shown to be effective in solving location problems (ReVelle 1993), can be useful for delocation problems as well. Facility location problems are generally formulated as large integer programming problems, although in some applications mixed integer programming formulations have been used (Galasso, Merce, and Grabot 2008, Wu 2008). Specialised branch and cut procedure (Marin et al. 2009) as well as the minimax regret approach (Puerto, Rodriguez-Chia, and Tamir 2009) have also been used to solve such problems. Acar, Kadipasaoglu, and Day (2009) developed a novel approach that allows iterative interaction with a simulation model. Finally, although we have presented the delocation problem in the context of the distribution chain of a business firm, the same formulation can be easily extended to the case of public sector services such as schools. Delocation of schools is becoming a serious problem in many countries facing low birth rates and demographic shifts. Our formulation will ensure that children from retained schools are not affected and only children presently going to a school being delocated will be assigned to one of the retained schools.

References
Aboolian, R., Berman, O., and Krass, D. (2007), Competitive Facility Location and Design Problem, European Journal of Operational Research, 182, 4062. Acar, Y., Kadipasaoglu, S.N., and Day, J.M. (2009), Incorporating Uncertainty in Optimal Decision Making: Integrating Mixed Integer Programming and Simulation to Solve Combinatorial Problems, Computers and Industrial Engineering, 56, 106112. Baldwin, R.E., Forslid, R., Martin, P., Ottavino, G., and Robert-Nicoud, F. (2003), Economic Geography and Public Policy, Princeton and Oxford: Princeton University Press. Batanovic, V., Petrovic, D., and Petrovic, R. (2009), Fuzzy Logic Based Algorithms for Maximum Covering Location Problems, Information Sciences, 179, 120129. Bischoff, M., Fleischmann, T., and Klamroth, K. (2009), The Multi-facility Location-allocation Problem with Polyhedral Barriers, Computers & Operations Research, 36, 137692. Boccia, M., Sforza, A., and Sterle, C. (2009), Flow Intercepting Facility Location: Problems, Models and Heuristics, Journal of Mathematical Modelling and Algorithms, 8, 3579. Brandeau, M.L., and Chiu, S.S. (1989), An Overview of Representative Problems in Location Research, Management Science, 35, 645674. Bruno, J., and Anderson, P. (1982), Analytical Methods For Planning Educational Facilities in an Era of Declining

280

P.K. Bhaumik
Chains, European Journal of Operational Research, 196, 487499. Meyerson, A., Munagala, K., and Plotkin, S. (2008), Costdistance: Two Metric Network Design, SIAM Journal on Computing, 38, 16481659. Midelfart-Knarvik, K.H., and Overman, H.G. (2002), Delocation and European Integration: Is Structural Spending Justified?, Economic Policy, 17, 322359. Murray, A., and Wu, X. (2003), Accessibility Tradeoffs in Public Transit Planning, Journal of Geographical Systems, 5, 93107. Nadirler, D., and Karasakal, E. (2008), Mixed Integer Programming-based Solution Procedure for Single-facility Location with Maximin of Rectilinear Distance, Journal of the Operational Research Society, 59, 563570. Puerto, J., Ricca, F., and Scozzari, A. (2009), Extensive Facility Location Problems on Networks with Equity Measures, Discrete Applied Mathematics, 157, 106985. Puerto, J., Rodriguez-Chia, A.M., and Tamir, A. (2009), Minimax Regret Single Facility Ordered Median Location Problems on Networks, Informs Journal on Computing, 21, 7787. ReVelle, C. (1993), Facility Siting and Integer Friendly Programming, European Journal of Operational Research, 65, 14758. ReVelle, C., Murray, A.T., and Serra, D. (2007), Location Models for Ceding Market Share and Shrinking Services, Omega, 35, 533540. Smith, H.K., Laporte, G., and Harper, P.R. (2009), Locational Analysis: Highlights of Growth to Maturity, Journal of the Operational Research society, 60, S140148. Wang, Q., Batta, R., Badhury, J., and Rump, C. (2003), Budget Constrained Location Problem with Opening and Closing of Facilities, Computers and Operations Research, 30, 20472069. Wu, Y. (2008), A Mixed Integer Programming Model for Global Logistics Transportation Problems, International Journal of Systems Science, 39, 217228.

Downloaded By: [Bhaumik, Pradip K.] At: 10:49 18 February 2010

Enrolments, Socio-Economic Planning Services, 16, 121131. Church, R., and Murray, A. (1993), Modeling School Utilization and Consolidation, Journal of Urban Planning and Development, 119, 2338. Daskin, M.S. (1995), Network and Discrete Location: Models, Algorithms, and Applications, New York: John Wiley and Sons, Inc. Diamond, J., and Wright, J. (1987), Multiobjective Analysis of Public School Consolidation, Journal of Urban Planning and Development, 113, 118. Drezner, Z. (1995), Facility Location: A Survey of Applications and Methods, Berlin: Springer-Verlag. Springer Series in Operations Research. Drezner, Z., and Hamacher, H. (2001), Facility Location: Applications and Theory, New York: Springer, p. 458. Galasso, F., Merce, C., and Grabot, B. (2008), Decision Support for Supply Chain Planning Under Uncertainty, International Journal of Systems Science, 39, 667675. Klincewicz, J.G., Luss, H., and Yu, C.-S. (1988), A Largescale Multilocation Capacity Planning Model, European Journal of Operational Research, 34, 17890. Liao, K., and Guo, D. (2008), A Clustering-based Approach to the Capacitated Facility Location Problem, Transactions in GIS, 12, 323329. Marin, A., Nickel, S., Puerto, J., and Velten, S. (2009), A Flexible Model and Efficient Solution Strategies for Discrete Location Problems, Discrete Applied Mathematics, 157, 11281145. Melachrinoudis, E., and Min, H. (2000), The Dynamic Relocation and Phaseout of a Hybrid, Two-echelon Plant/ warehousing Facility: A Multiple Objective Approach, European Journal of Operational Research, 123, 115. Melo, M.T., Nickel, S., and Saldanha-da-Gama, F. (2009), Facility Location and Supply Chain Management, European Journal of operational Research, 196, 401412. Meng, Q., Huang, Y., and Cheu, R.L. (2009), Competitive Facility Location on Decentralized Supply