Scribd Logo
Upload

Welcome to Scribd!

  • Chapter 13 Facility Location Decisions

    Uploaded by

    Anirban Mondal
    283 views68 pages
    facility location

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    facility location

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    283 views68 pages

    Chapter 13 Facility Location Decisions

    Uploaded by

    Anirban Mondal
    facility location

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 68
    LOCATION STRATEGY Wernied by what they Ae Singlest and rg accupatic {by slow graduations. ALECANDER HAMILTON, 1797 along with their associated costs and investment levels, used to operate Inventory Stratogy + Foreeasting | inventory decisions Transport Strategy * Purchasing and supply Transport fundamentals ‘scheouiling decisions Customer “Transport decisions + Storage funcamertaie _//Service goals + Storage decisions 1 The product systems Location Strategy + Location decisions : planning process CLASSIFICATION oF Location PRosLems Teed ncussing location methods, Ws uschl fo assy Tecan problems into a Co deceetmtber of categories, namely, by (1) driving force tienes facilities, (2) dscreteness of the choices, (8) degre of data aggregation oxtane horizon, Driving Force Kaauty location is often determined by one critical factor. nthe case of plant and achouse location, economic factors usually dominate In cere location, revenue Bererated by a location is often the determining factor, with ene costs subtracted a from revenuies to determine profitability. Where ert operation thospital, axto- inaved bank teller charity collection center, or maintemanes ty) is to be located, SES RIN te the site may be the primary location factor especially when revenue and costs are not easily determined Number of Facilities {Locating one facility i considerably diferent problem from locating many facilities dicen Single faciity location avoids the need to cones competitive forces, division of demand among facilities, inventory consolidation effects, and facility Toection anspertation costs are typically the primary consideration, Single facility 'ecaton isthe simpler of the two probiem type Discreteness of the Choices S Ste methods will explore every possible location along a space continuum and St the Best one, These we refer to as continoas location mode Alternatively, ethoxl see Margate Brandes and Smut cation Reseszch Nirnagonant ra tNew hy Facility Location Decisions 351 Se eee A Historica Perspec ON Location? Jocation methods may salevt from a list of possible choices that have Ss for their reasonableness, These a commonly used in practs fisccele location methods. The lortcr anwme or inuttiple facility location, Degree of Data Aggregation Location problems typically in: lve the evahnation of an exceeding of network design configaratinns. livmanage problem size and obtery 2a: generally necessary to us aggregate data relationships when salvin location problem. This rests methods whose accuracy limits Ineatians to aah geographic areas such os enti vities, On the other hand, methods sii sitie aggregation, especially those fcr sits selection, can differentiate between locations seperated only by a city stieet, The latter is particularly needed for retsi! location intracity locations, and woking: fina! slant and warehouse site selections Time Horizon ‘The time nature of location » hod is to be static or dynamic. That is, static methods Find locations based on slain for a single period, such as one year. Location plans may cover many years at orice, however, especially if facilities represent a fixes invest- ment and the costs af relocating from one location to another are high. Methods that handle multiperiod location plowing. are referred to as dynamic: VE Much of the early theories about IovaFon were postulated by land economists and regional geographers such as Johann von Thiinen, Alfred Weber,’ T. Palander® August Lésch,* Edgar Hoover,” Mefvin Greenhut, and Walter Isard.” A common theme throughout all ofthese eanly works was the importance of transportatio in determining location, Although much of the work was conducted in an and early industrial society, 9 number of the concepts that they suggested applicable today. Consider » vie «ntinw of just a few of these, a railee OF ehiston vf gation yo |New Yock Spi vs see T Pan Sa Jodha: Heirsich von Tin, Soci mh nd ® Bed ud Betis Schumacher Zari 157 “Alla Weber Shi de tintsren (Abo, Tabi, 19), eae by C hes i v= Chiao Liner of Chcage Pres ‘Jena: Gustav Fschoe Verlag 1940; ige MA Flava Maltin’ Green Practice (Chapel Bl, NC Unive oF North Castine ‘Walter ti iesce (Neve York: fo Wie Spare Fron vay (Cambridge, MA MIT Dress, 858 Vocation Strate i | 1 | | | RSet we Figure 13-1 Thunen’s Rene Curve for Land Prices ~ Transport cost Profit ~ Land rent Vegerabla an Dairies , Fruit farms , Wheat andooT Distance fvom market Bid-Rent Curves ZThonen recognized that the maximum rent, or profit, that any economic develop- srerteatg Pay {OF land was the difference between the price os the g00ds int the ‘marketplace and the cost of transporting the goods fo the marketplace. He visualized an Solated city-state (marketplace) sitwated on a plane of equal fertility. Economic land. In an agricultural economy, agricultural activity might locate out from the sartetplace, as shown in Figure 13-1. Today, this idea stil wore e hold as we ghserve the pattern of retail, residential, manufacturing, and agricultural locations iat sing the city center. Those activities that can pay theses fat iosn will be located nearest the city center and along major transportation links Weber's Classification of Industries ‘Mis Weber recognized the role that raw materials playin the production process ang Mow they affect location. He observed that some procereshons weight losing, the weight ot ming That the sum ofthe weight of aw mates oe ae prodockpy the fished prot. Weight fs lot in processing tee eer by- BRUES. Therelore, to avoid shipping by-products to te marketplace, such Fovenion ceeataten toward their raw material sources in order to wee et Portation costs (see Figure 13-2) he other hand, processes may be weight gaining, This commonly occurs ra ete eer ito the process. Acconing to Weber ubisories ye the Managua available everywhere suchas ait nd waten Thentioae minimize “Sesportation costs by shipping ubiquities the shortest poe distance, such Chepter 13 Facility Location Decisions $83 554 Material woight in the production process Location to draw Type of production Weight before Weightafter Sources of process processing processing raw materials, weight losing Weight saining Neither weight losing nor weight gaining Figure 13-2 Effect on Process Location of Product Weights Before and After Processing processes should be located as close to markets as possible (see Figure 13-2). An ‘example of an industry that locates its plants in this manner is soft drink bottling. = Syrups are shipped into the bottling plants and mixed with water. These plants are. 2 typically located in the general region of the markets for the products. Finally, there are processes where there is no change in weight between raw materials and finished product. Assembly operations are representative of this cate- gory, where the finished product is a sum of the weight ofthe parts and components assembled into it. Such processes, according to Weber, are bound neither to the sources of the raw materials nor to the markets (see Figure 13-2). That is, the total of inbound and outbound transportation costs is the same at any location between source and market points Hoover's Tapered Transportation Rates Hoover observed that transportation rates are tapered with distance. To minimize inbound plus outbound transportation costs where they are the dominant location fore, a facility located between a raw material source anda market point wil have & Iinimum transportation cost at one of these two points. As show in Figure 133 location between these points is economically unstabie. Since Y is lower than X of the cost curves, location shoul Part V Location Strategy eS comune eee ne ARES Figure 13-3 Tapered Transportation Rates Force Locatian to the Source of Materials fr to the Market Outbound transportation cast ‘Inbound transportation Source of materials SINGLE Facitity Location [ct us now turn to contemporary w Last of applied mathematics and computer, jalure rather than conceptual. We begin with Popular model th ey Hable Plant, terminal, warehouse, or rial ee ne iat is used for locat- ce point. It has been vari. ‘dian, the grid method, the point volume are the TC = total transportation cost ¥.= volume at point j A= transportation rate to point / distance to point from tne facility tobe located Chapter 13. Facility Location Decisions $55 356 [Et faclity location is found by solving, two equations forthe coms ts of location.) These exact center-of-gravity coordinates are varia, je sees 022 Lvava, and Dvayva, i133) Lava, where X, ¥ = coordinate points ofthe located facility Xi. Y= coordinate poims of source and demand points The distance d,is estimated by Kx, Xu, -¥F 13.4 where K represents a scaling factor to convert one unit of a coordinate point io 9 more common distance measure, such as miles or kilometers The solution process involves several steps, which are outlined as folloxs: 1 Determine the X-Y coordinate points for each source and demand point, long ‘with point volumes and linear transportation rates, 2. Approximate the initial location from the centerof-gravity formulas by omte tung the distance term d, as follows: 113) and (3-6) "These eaisations are derived fom Eguatons (13-1) and with rapeet to X and 1, setting tem ee ng the pacial der Past V Location Strat 3. Using the solution for X, F from step 2, calculated according to Equation EES). Me scaling factor K need not be used at this point i 4 Substituted; into Equations (13-2) and (13-3), and solve for the revised RY coordinates. ef 5: Recalculate d, based on the revised ¥, ¥ coordinates & Repeat steps 4 and 5 until the X, 7 coordinates do not change for successive Feanens OF they change so lite that continuing the calculations ae fruitful 7, Firally, calculate the total cost for the best location, i desi by using Equation (1341). Seger rear ese ee eet Example Consider the problem of Limited Distributors, Inc, with two plants supplying the these, thich. te tin, supplies thee demand centers, The sential atrangement of ihe Plans and market points is shown in Figure 134%. Weswh the oe for the sin- she warehouse that will minimize transportation costs. A grid overlay ona highway an ema 3 Convenience in establishing the relative point locations, Eee plant and demand center location is expressed as a geometric coun Point, Product A ig Ceara. and product B from P~ These products arereshipyed wt ‘markets, Coordinate points, volumes, andi transportation rates ae sumanestad Table 13-1, Figure 13-4 Location Map of Plants P, and P, and Markets M,, My, and ‘M, and Suggested farehouse Location Limited aistnbutors) Vertical coordinate 5 ‘Determines by dvsing repsentativequted rate Set detrei) courbes Table 13-1 Volume, Transportation Rates, and Coordinates for Market and Supply Points Using Equations (13-5) and (13-6), we can find an initial, or tion for the warehouse. Calculations are easy if we solve the equations in a tabular form. That is, ¥, R VE WRX, 2000 0050 00 300.00 3000 0050 15000 1200.00 2500 007518750 375.00 1,000 500 450.00 1,500 1 300.00 32500 Now, we have X = 3225.00/625.00 = 5.16 and ¥ = 3237.50/625.00 = 6.18 ‘These coordinates define the warehouse location, as shown in Figure 13-4. The total transportation cost associated with ths locaton fs determined in Table 192. — aie The previous example was terminated at step 2of the solution process. This is an approximate solution. In many applications, it will provide a location that is reason ably close to the optimum. It will provide a first approximation to the least cost solt tion and will give an optimum when there is perfect symmetry in location, volume, and costs associated with the points. When these conditions are not fully met research has shown that the potential error still can be quite small, if the volume associated with one or a few points is not substantially larger than the rest; there is a 358 Part V_Location Strategy Table 13-2 Calculation of Transportation Cost for Limited Distributors’ Warehouse Location lange number of demand or supply points in the problem; and the rates are linear, or nearly linear, with involving 50 demand points with randomly dispersed locations, volumes ave transportation rates had an average error of L6 percent fr this method. Of course, this error can increase subst demand points is decreased. Finding a more exact center-of-gravity sol Met eon blition process. We are not able to find the solution directly and ee mor to an iterating procedure. A rather simple and straightforward meth net of successive approximations, Although there are others, this procedure sencas se yell in this application. It can be time-consuming for hand caleulation but tones itself nicely to computer solution, transportation distance." For example, a modest problem fom optisnum when using, fantially as the number of tution requires completing the remain- Se Example wuing with the Limited Distributors’ proble: ity solution 'm, we woulkl now use the centerof- as the starting solution in Equations (13-1) and (13-2) to find the fact location. The location coordinates for the fist iteration can be found by solving Heequations in the following tabular form, using results from the previ is example, eae 3000) 14.48 90.89 9A adaR Table 13-3 (One Hundred Computations! Cycles of Location Coordinates and Total Transportation Costs As Generated from the COG Software Module = Brac salution ‘The revised location coordinate points can be calculated as, X = 102.009/20.249 = 5.038 and 102,388/20.249 = 5.057 with a total cost of $21,431, Using the computer software module in LOGWARE known as COG, we can complete 100 iterations of this procedure. The results are given in Table 13.3 In this Problem, note thatthe total cost does not decline further after the eleventh iteration, and there i litle change in the location coordinates, Ths i the nature of this parte. ular problem, but other problems may show dramatic differences esses leaes tee eT Extensions to the Single Facility Location Model The continuous location nature and simplicity ofthe exact centerof gravity approach Bivenits appeal either as a location model unto itself or as a submodel in more cabo. rate methods, has encouraged researchers to extend its capability. Primary among these extensions are to include customer service and revenues,” to handle multiple locations, * and to represent nonlinear transportation costs. See Donald Bowersox, “An Analytical Approsch wo Warchowse| ling & Shipping, (February 1962) pp. 17-26; and Rona i ics Maralgrment nod. (Upper So River, Ni Pronace Hal, 98 E-Hill, “Program Finds New Ses in Multi-Facilty Location Problem,” Inns 3Fe Ballo, Business Log nslon ofthe Generalized We UD pp I-19? Part V_ Location Strategy SHU - 0k Appraisal of Single Facility Location 1h gldition to the center-of-gravity model, other single facility lation approaches wate Bephical techniques and approximating methods. Allvary in te degree of ealism that they portray, in theic speed and ease of computation, and in thelt cok iP to suarantee an optimum solution. Clearly, no single model i likely to have all of the features desired tor a particular location problem so that the solution wil lead shrectly to a final decision and management can merely delegate location devisinns {oan analyst. Therefore, these models can only provide solution guidelines and thee cifective use requires a good understanding of their strengths as well as their shore, comings. _[he benefit of these single location models is quite clear—they aid the search for ar solution to a location problem, and they capture enough ofthe reality of the Actual problem so that the solution is meaningful fo management. The shortcomings may not be so obvious, and they need to be noted. Although any model will exhitee some shortcomings when applied to a real problem, this does not mean that the model is not useful. What is important is the sensitivity of the location model's resuls toa poor representation of reality. i a simplifying assumption, such as linear, ea aeamsportation rates, has litle or no efect on a models suggestion fora facility ‘Ceatjon,a simpler model may well prove to be more effective than elegant ones Some of the simplitying assumptions in single location models are listed next { Demand volumes are frequently assumed concentrated at one point, when in {act they ae generated from a numberof customer points that ae dispersed over a wide area. The market center of gravity is often used as the demand cluster, but this is subject to some error in calculating transportation costs to the demand cluster instead of to individual demand points, ingle focility location models typically find a location based on variable costs They make no distinction between the differences in capital cost required for establishing a warehouse at various locations and other costs such labor ventory carrying costs, and utilities associated with operating a facility a dif ferent locations, - Total transportation costs usually are assumed to increase proportionately with Gistance; however, most transport rates are composed ofa fixed component and 2 Variable component that varies with distance. Rate minimums and sate blan keting may further distort their linearity “sighting routes are commonly assumed between the facility and other points. This is rarely true, since travel is over a defined road network, Shed rail system, or dhrough a rectilinear city street network. A propre ty factor can be included in the model to convert straight-line distances to ‘oximate highway miles, rail miles, or whatever. This conversion factor “Hedda circuity factor, varies by location. For US. intercity transport, calculated ‘sight-line miles should be increased by 20 percent to get highway dirvet- Stes aed Distance hoe Pa route miles and by 24 percent to get rail short-line miles, For city streets, a fetog ‘of #1 to 44 percent cam be used, A table of circuity factors for truck travel in if, ferent countries is given in Chapler 14 5. There is some concern that location madels such as thes: is, they do not find a solution that reflects future changes in revenues and costs Applications _ * Leaseway Transportation Corporation was able to make use of the exact centers ogravity model to locate a truck maintenance facility in Boston. The company. leased varying numbers of trucks to many accounts throughout the Boston mek ropolitan area, Te tuck maintenance facility was tobe locoted for maximum convenience of all accounts. The location ofeach account andthe number of trucks leased was known. The transportation rate was the same throughout the region. The center-of-gravity model gave the general location within which a specific site could be selected *+ Anil company used the center-of gravity method to locate il collection plat) forms in the Gulf of Mexico. Many weilheads were located throughout the floor of the Gulf. A group of these was connected using pipe that moved the oil loa”) = collestion platform on the surface. The center-of gravity method was appropri> > ate for finding the collection platform location that would minimize the total length of pipe needed oe Muttiece Facitity Location Oo The more complex, yet more realistic, location problem for most firms occurs when two or more facilities must be located simultaneously, ot additional facilities are to be located when at least one already exists. This problem is common because all bbut the smallest companies have more than one facility in their logistics systems. It is complex because these facilities cannot reasonably be treated as economically independent, and the number of possible location configurations becomes enormous. esr EE Observation A few years ago, a company producing industrial cleaning compounds so its products in approximately 2,000 U.S. counties, used 108 public warehouses, and ‘manufactured its products in four plants, There were mote them 800,000 possible plant-varehouse-customer combinations to be considered emong only the existing locations. Finding an optimum warehouse configuration was further complicated by the several hundred product items sold and several modes of transportation used 562 Part V Location Su Let us direct our attention to the Since it a common problem exper characterized by several by Warehouse location problem as ienced by many kinds of b asic planning questions. 1. How many warehouses should there be in # lange should they be, and 2 Which demand points sh should be assigned to e 3. Which products shoul be shipped directly ‘number of location methods have! all of these questions. Several of ae presented here to show the location methods may be catego 2 general class, uisinesses, It can be he supply ch, where should they be located? would be assigned to.a warehouse? Whi warehouses ach plant, vendor, or port? id be stocked in each w. from plants, vendors, fain network? How ‘arehouse? Which products should r ports to customers? been developed that aid in answering some or these, although by no means an exh tive selection, variety and power of the approaches, Mathematical rized as exact, simulation, and heuristic Exact Methods Exact methods refer to those Liathematically optimum solution to the location Problem, Known accuracy. In many respects, this is Fro aLates with the capability to guarantee either a or at least a solution of {an ideal approach to the problem of locas Hon; however, the approach can result in long computer running times, huge men probes nents and a compromised problem detimitens when applied to practical Problems. Calculus!” and mathematical Programming models are examples of this Spproach, and both will be illustrated Multiple Center-of-Gravity Approach The nature of the multiple facility locator Problem is seen if we malt abbroach in a multilacation format, Reco that this ‘model that finds the minimum trans ‘iy located among origin and located, then it is necessary to locations, This forms clusters of, Then, an exact centerof tise the exact center is a caleulus-based portation cost solution for an intermediate land destination points. If more than one facing x tobe “sign the origin and destination points to arbitrary Focal the numberof facilites being located Scan be enti0n is found for each of the clusters Tres ae gan be made in many ways, especially when considering 2 ri and destination points in the prokt jpbreach sto form the clusters by grouping the Points that are the closest ved fo ten Met the conter-of-gravity locations found, the points are reas. ened to locations. New eenter-of gravity locations found for the revised The process is continued until there ino further change. This completes the “tations fora specitied numberof facilites to he located, Itcan be repeated for "ent numbers of facilities suite facilities and a | ean. One large number | Sn s, Some companies prefer using, implying a velatively small tixed investment ial operation 3 inist processing, available at each fa | and handling chary iclory may be prope Processing of the input data was handled in two parts, First q preprocessing pro- alors teat orene atomer overs that cosld be filled via a warehouse from ens ors that Were sulficiently large to be economically filled from a plant. Next, the Plants te Program computed distances from customers t0 varchonee and Plants to warehouses from a longitude and latitude coordinate system Customers sulecting el £0 warehouses by examining the five closest wamhouses sea meet Rouse te Warehouse offering the last cost in terms of delivery cose Rene wee f [onset Customer, handling and storage costs at the warehouse, oat transportation : dations toa ant %9 warchouse. The computer then performed the nececenct compu: lows thereat 2 Pacticular warehouse configuration, given the assigned poctey {lows theough the warehouse system and geographic cata ead into the test program, Sane Programming approach was used to resolve any capacity limitations at the Hove dg an, Matehouse locaiion configurations were evaluated. Figo aot a > diagram of the model’s operation, Simulation models continue to play an important role in warehouse location, Ue often weitten primarily as inventory simulators (LREPS™ wet PIPELINE JANAGER™), but others are dlevei A more directly t0 be warehouse locators. Figure 13-7 Fercnaainre see Bevlore othe Tad ina customer order {data and losations oat for Bote Preprocessing Distribution,” Harenet program Busites Reiss Vol 38 {lovember-December 1940), Geders Tied Bp through ware housing system Reade [Feadin wararouse] freight rates, Test |_| tocation warehousing Program configuration gins imenate | ——"T———"_ | "re ute Coat of warehouse location configuration = << tater at desired? desirable feature inherent in them is their ability to handle time-related aspects of inventory along with the geographical aspects of location, On the other hand sive data requirements and long computer-running times can be a problem for this methodology. Nevertheless, the precise descriptions of reality are the primar sons for their appeal A major problem with location simulators is that the user may nat know how close the chosen warehouse configurations are to the optimum. Of course, we do know that the total cost curve for the location problem usually has a “flat bottom.” | Therefore, costs between closely ranked alternatives change litle in the region of the | ‘optimum. As long as a reasonable number of prudently selected configurations have been evaluated, we can have a high degree of confidence that atleast a satisfactory solution has been found 572 Part V Location Strategy Heuristic Methods Heuristics can be referred to as any principles or concepts that contribute to reducing the average fime to search for a solution, They are sometimes referred to as rues of thumb that guide problem solving. When applied to problems of location, such rules of thumb, which are a consequence of insight into the solution process, allow good solutions to be obtained quickly from numerous alternatives, Although heuristic ‘methods do not guarantee that an optimum solution has been found, the benefits of reasonable computer running times and memory requirements, a good representa~ tion of reality, and a satisfactory solution quality are reasons to consider the heuristic approach to warehouse location. Heuristic methods have been popular as a methodology for warehouse location. A classic yet still useful heuristic approach to the warehouse location problem was developed by Kuehn and Hamburger. Other examples abound," To help under. stand a heuristic model type for realistic problems, consider the nature of the loca tion problem typically encountered in practice. The location problem is one of trading off the costs relevant to location, which include * Production and purchase costs, + Warehouse storage and handling costs + Warehouse fixed costs + Cost for carrying inventory + ‘Stock order and customer order processing costs * Warehouse inbound and outbound transportation costs ach of these cost categories should reflect geographic differences, volume and ship- ping characteristics, policy variations, and economies of scale. The nature of the cost trade-offs is graphically shown in Figure 138, Inventory, storage, and fixed costs are in direct trade-off with inbound and outbound trans, Portation costs. Production and order processing costs also enter into the cost trade. off, but they cannot be adequately shown in this particular figure. The task of a location model is to seek out the warehouse/plant configuration that results in the ‘minimum total relevant cost, subject to customer service and other practical restric. tions placed on the problem, Figure 13.8 shows that transportation costs decline with the aumber of ware- houses in the distribution system. This is generally true because inbound shipments te a warehouse are made in larger quantities, and at lower rates, than outbound 4.8 Rodn and M, J, Hembusges, "A Heucstc Program tor Loceting Warehouses,” Muigenent se, Vo 10,Guly 863), pp. of 566, “4 Chiu, “An Ovetvaw of Representative Problems in Loeation HB) tat ames M. Master, "Committ Sofa or the Bsns Logit, Wl. 14, Ne Chapter 13. Facitty Location Decisions 573 574 inte Figure 13-8 Gonoralized Cost Trade-Offs in the Facity Location Problem | | \ Total cost ae \ wegen I fod, Cost Tnweniary eoeryina and warehousing ai Production/porchase | a and order orocessing Inbound an |" fesbound renaperisoon Number of wardhowses shipments. As more warehouses are put into the system, the warehouses 2-0 closer to customers, stich that the inbound cost is increased! but the outbound cost i= dispre- portionately reduced. The transportation cost curve continues to decline in U's man ner until so many warehouses are used in the network that it is no longer p-actical to ‘maintain full vehicle-load shipments to all warehouses. The tfansportetion curve ‘would turn up at this point The inventory-carrying and warehousing cost curve is shown to increase at a decreasing rate as the number of warehouses in the network increases. T's is pti marily a result of the inventory policy of a firm, as well as how that poli executed, and the increasing amount of fixed cost in the network, With more ware houses, there is a proliferation of the amount of safety stock in the network, ithe firm is controlling inventories by means of economical order quantity proveclures, & tapered average inventory level and inventory-carrying cost curve results. Other policies may result in somewhat different inventory-careving and warehousing cost curves, ranging from linear to tapered. If the warehouses ate privately owned ot leased, there will be an annual fixed change per warchouse. Total fixed costs ‘7 the network then will increase with the number of warehouses, Selective Evaluation A heuristic procedure can be developed from a method that has already beer °° sented in this chapter, namely the multiple eenter-of-gravity method. The pro is to solve for a specified number of facilities. Since the method only acco transportation costs, additional costs such as inventory and facility fixed costs “Say be added to create a more representative total cost. By repeating the procectis? fot various numbers of facilites, the best number of facilities, and their associated © tions, can be found, Part Location sg Suppose we have data for ten markets and their corresponding transportation rates, Figure 139 ff MCOGOLDAT ofthe LOGWARE softwere the ‘markets are shown in figure 15.9. n addition, there isan annual fhe charge of $2,000,000 for each ware. Taube, AU warehouses have enough capacity to handle the entire market demand, [ig Moun of inventory” in. the loviete system is estimated from [r(S) = $6,000000 Viv, where Nis the mete of warehouses in the network Invettory-cartying costs are 25 percent pet rene Handling rates at the warehouses poate same: theretore, they do not afet the locate outcome, How many ware. houses should there be, where should they be located, and which markets should be assigned to each warehouse? Using the MULTICOG software module in LOGWARE and repeatedly solving {ot various numbers of warehouse, a spreadche can be developed as shown in gtefarehouses yield the best cst balance. As seen in Figure 13-8, warehouses Row be located in markets 3, 7,9, and 10. Markee een 4,and 5 are: Figure 13:9 Venoan ns Markets for Example eee Becem ong | 90 Pr Seliton thou |, Warehouses p 72 fe 63 | ; | re ° PEG Uae VEC EET We er Saar are coordinates ints Chapter 13 Facility Location Decisions 395 Table 13-5 Selection Evaluation Location ——— Alternatives 1409,628 576 1 5864 2806 1.368330 p00, ,00n.090 9ats327 —100K,0m awe M0000 3 T6mn0.000 4.2 zsaan 800.0% 4800.00 25,483,585 mon0900 4743416 24743416 The selection evaluation approach is heuristic for several reasons. First the mul Liple center-of-gravity method includes some rules that are used to determine initia = warehouse locations. This may cause sub-optimality in the results, Second, fixed” costs and inventory costs are added to transportation costs ater the warehouse loca! tions ate determined. Itis preferred that these costs combine during the process that <2 determines warehouse location for more optimal resuilts. Regardless of its shortcom: ings, the approach has value when there is a minimum of information available to solve a location problem. Its useful in generating candidate locations that may be more thoroughly evaluated using procedures that are more robust 2 Another form of selective evaluation specifies the number of warehouses to be evaluated and the particular warehouses in that number, Although the overall analy sis is similar to that just presented using the center-of-gravity method, the analyst uses human judgment, logic, cognitive skills, and results from other model types to select the warehouses for evaluation. Since models seeking the optimum cannot hope to consider all the factors necessary for finding a satisfactory network design, this type of what-if analysis becomes very useful for practical network design. Linear programming is commonly used to allocate demand throughout the specified net- work. Choosing specific warehouses for evaluation is an effective way of dealing with practical issues in network design and making sure that desired warehouse combinations are considered. Most location analyses are dominated by selective evaluation of this type. Selective evaluation may be used in solving location problems where the model Used in the analysis is not primarily of a location nature. A common problem of this type is the location of a truck depot from which trucks are dispatehed. Multiple fucks are routed to multiple stop-off points, and the route configuration is depen dent on the location of the depot in proximity to the stop-off points, Depot location is dominated by transportation costs such that solving the truck routing problem is critical to depot location. A vehicle routing model suich as ROUTER in LOGWARE can be used to form routes and minimize transportation cost. Then, selecting a par ticular depot location, solving the routing problem for the selected lacation, a Part V Location Strategy oe ce ae — oe i = 153095 Miles of aval EW. FiBELe 13-10 Current pot Location wth Stop Volumes in Cae and Truck Routes acting costs specific to the location allow each location to be evaluated. This is a ifakand-error procedure, and a satisfactory solution to the ‘epends on the quality ofthe locations selected for evaluation E. xample location. problem au Rurant supply house makes deliveries to its customers (restaurants) on a Say basis: Currently, four tracks are dispatched on toute haan depot located in thor dae chown in Figure 13-10 for atypical day’s delivery voles Considering hat the trucks are amortized at « cost of $20 per day per truck, gas and mainte. ee Fae ts {0 Operate each teuck are $0.40 per mile, and drivers are paid S11 per That for wages and benefits, the current daily cost to serve the customers is $508, ne company is considering a move to one of two cenend locations indicated in Figure 13-20 as A and B, The lacility operating cons are expected to be about the the RO the amortized one-time moving cost is estimated ve 24) per day. Using the ROUTER px vm in LOGWARE to generate routes from depots at 1 ard eat bee vised transportation costs, Comparing the threw alternerg ne Facility Location Decisions Numberoi Routing Truck Moving « Location “Trucks Cost Cost Crrent + 508 A 5 497 on 40 B 4 488 40 60H Since the savings in routing costs cannot overcome the cost of moving te. tion, the economic decision is to keep the depot at the current location, Se Guided Linear Programming When serious heuristic procedures are developed for real-world location problems, they generally will include linear programming as part of the solution methodology The appeal is that linear programming gives optimal results and can handle capaci restrictions that elude many other approaches. However, to be a truly robust prove: dure for location, fixed costs and nonlinear inventory costs need to be handled as ‘well, Heuristic procedures need to be employed with linear programming to create an effective model Consider the smal, single product problem shown in Figure 13-11. The frst step is to construct a matrix that is formatted like the transportation problem of linear programming. By giving it a special structure, two logistics network echelons can be Figure 13-11 A Single-Product Location Problem with Warehouse Fixed Costs and Inventory Costs Handling » s2iewt, Capacity = 60,000 cwr, Fixed = $100,000, be ea Plant P, Production ~ Stew. Capa iad Warshouse W, Handling ~ Sticw, Capacity ~ Unrestricted I Fixed » $400,000 Customer C> | Plant Py 50,000 owt, Produtzon ~ Sdicmt. lventory carrying gost Capacty ~ Unrestricted 100 (Throughput?? 378 Part V Location Strategy i | | | Customers Plant an | warshouse Gf & | Gs | pact 60.000 999.999 wii oss Ton | ar ap - 80000 | s.c00_ oo ae | 72 Tae Warehous 0000 | 40000 | 50.000 || cose Warehouse T / rand | 000 999899°| soe00 | 120,000 | 000 S Usedo roprdert on nna oh ne § Used to represent unlimited capacity, “pventory carrying, wareheusing, outbound tansporttion, and fixed cates, thatis, 32+274+05"97, Faure 13-12 Matix of Cell Costs and Solution Values for the Fre {eration in the Example Problem mennee Rein the matrix of Figure 13-12. The heuristic process is guided by the rangprtatah ch he cell costs are entered into the matrix. Since the prod fon and sropoation costs Between plants and warehouses are linear, they ctor the, plant- warehouse cells directly. For example, the cell cost ‘representing the flow between Py 4 Hs the production plus transportation costs, o H/cnt © Sis eevee $8/ews, The celiblock for warehouses ‘arvhouse hanelling plus ansportation pl andiing and transportation x 1, there are no rates for inven and they must be developed, depending on each ware, so throughput: Since this throughput is not known, we must Coca starting ‘Qushputs. For Gixed costs, each warehouse is initially given the mot favorable chron ae ing that all demand flows theough it. Thus, the rate associated cece custome gee use { would be the anual warehouse fixed cost divide! by che fs aon ote demand, 0° $100,000/200,000 = $0.50/ews. For mechan *400,000/200,000 = s: andl customers combines w: inventory-carrying plus fixed costs. Hea N be read directly from Figure 13-11, However v-sareying and fixed costs, the per cwt, rate depends on the aumber « ‘he demand assigned ¢ them. Again, te *iiey Location Dec e579 580 greatest opportunity to be selected, the assumed throughput for the wa-chouses i equal, or the throughput for each warehouse is the total customer demand divided by the number of warehouses being evaluated. The "per unit” inventory-carying cost is defined as the average inventory value in a warehouse divided by the warehouse period throughput, or IC, = K(Throughput,)/Throughput. Ieitialh each warehouse, the per cwt. inventory carrying cost is wa 10041200, 000/28"Hi2o0%bo0r = $9.21 The estimated per unit fixed and inventory-carrying costs are now entered into the warehouse-customer cells of the matrix of Figure 13-12. The problem is solved in a normal manner using the TRANLP module of LOGWARE. The computational results are shown as the bold values in Figure 13-12. This now completes round one of the computations. Subsequent computational rounds utilize warehouse throughputs from its previ ‘ous round to improve upon the estimate of the per unit inventory-carrying and fixed costs for a warehouse. To make these estimates, we note from the solution that the throughput for WY, is 60,000 ewt. and for itis 140,000 ewt. (see Figure 13-12). The allocated costs for the warehouses are Warehouse Per Unit Fired Cost $/cwt Per Unit inventory Caring Cos Wy $5100,000/60,000 cwrt, = 1.67 $100(60,000 ewe j°7/60,000 ow i $400,000/140,000 cwt. = 2.86 _ $100(140,000 ewe.)?7/140,000 ew The cell costs in the matrix for warehouses to customers (see Figure 13-12) are reca: culated to be: q c S w 1138 1036 1236 w 87 272 82 20286 2852072 The remaining cells are unaltered. Now, solve the problem again. The second iteration solution shows that all production is at plant 2 and s!! demand is to be served from warehouse 2. That is, g wy a % Ww 50,000 100,000 50,000 — Produced at plant 2 Subsequent iterations repeat the second iteration solution, since the allocation of inventory and fixed costs remain unchanged. A stopping point has been reached. To Part V Location $ ates ie the cell costs of Figure 13-12, since they conta house fixed and inventory-carrying costs. Rather, rates from Figure 13-11 in the estimated values for ware- compute costs as follows using the Warehouse T Warehouse 3 Cost Ty owt 2000 ew. Broduction| 30 2ODO% T= $800,000 Inbound transportation a 200,000 2 = 400,000, Outbound transportation ° 50.0002 = 100,000 100.0001 = 100,000 500002 = 100,000, Fixed a 400,000 Inventory careying ° 200(200,000)°7 = 513,724 Handling 9 200,000 x1 = 200,000 ‘Subtotal 3% S2513714 Tota ea ~~ _— The previous example illustrates However, many practical location pi Jacluded in the computational proced {or @ warehouse are shared among throughput, the guided linear progra ‘multiple product case. a heuristic procedure for a single product. roblems require that multiple products be lure With slight modification where fixed costs the products according to their warehouse unming procedure can be extended to handle Appraisal of Multiple Facility Location Methods Large-scale, multiple facility na diseibecs hy Provide a manager. Applications have ranged irons and distribution networks involving more than 100 ants ad 200 customer demand zones to supply networke wine hundreds of {oadoss supply a master warchouse that in turn supplies cextesne Defense, retail, intematorst orchid etdusttal goods industries operating in both denen and foe teaional environments have applied models ofthis scare The primary reasons ett oF great nS Popularity are (1) they offer decision support te solving a cae Brest consequence to management; (2) they are sufttethe robust to * shide variety of logistics networks in acceptable detail fn planning pur- eae ey at inexpensive to apply such tha the benefits destvol oars thetr use S 166.790 337.600 756,100 1.486600 E 13800 3400 1526,000 BB4_ Part V_ Location Strategy oe gui ite ag 08s to find the optimum plan to 5 x 5 = 25. Dynamic programming per- mits us to recast this multi-period problem into a series of single-decision events, Syarting withthe last yeas, wre compute the profit soncnecl age being. in locas the Bogan moving to one of the other locations, The ciscounted eecton moving in the beginning of year 5 would be $100,000/11 + 0.20)! $48,225. Given the location Frokits forthe fitth year (see Table 13-6), we want toselect ihe ho strategy, assuming, that we are in location A at the beginning of the fifth year. We evaluate the following choices: Alteenative (:) “Location Profit Moving Cost Ni Raat A $1260 — 9 = B1as6000 ls 1,398,000 — 48.295 1.349.975 Je lasrson — inzas = ons | | D 1486600 - 48,295 = 1,438.37 Le 152600 - 48255 > amr = nthe warehouse is located in , we should move ito Eto manimize profit. in Proceed to make similar computations foreach location mae fitth year, The “trateRy and the associated profits are recorded in Table 13,7, When strategy calculations are made for years othec than year 5, we must erie the profits accumulated from subsequent years, Concehos the calculations Shab would be made for location D in year 3. The diceowntat ‘moving cost would be $100.000/C1 + 0.20} = $69,444, The location profits me fees in Table 13-6, The sam yilaite profits for the subsequent year (year 4) are found in Table 13-7. We now can find the best strategy. if i itive Profifor Cumulative Alwmative Location Moving Subsequent ati a) Profit "Cost Years 80) __Year3 PD) [a 2320 ~ esas + seaman TC 743000 ~ 69444 + “Dar030 $6400 ~ 69444 ~ 2400039 756100 - gs 2459609 = donna 4ineatations are cared out until Table 13.7 is complete, The optimum RRBRK location cn be taced through the able. We seach ene yeat for the ciate egemlative profit ($3,735,130), which is location C. Fron ahs point, the the fied amar Be i Sc Se Sc 8p Sy This means to intially cate mC, stay in C for Se first three years, and then switch to location D at the beginning of the fourth year Sey locaton D the remainder o he piomning horizon See wea Table 13-7 that should wwe wish initially to locate in ang of th Tum strategy given that starting location ears esecre AEE Chapter 13. Facility Location Decisions 383 W other locations, we may trace an opti- E cas 2 a § 1004 o1/ 1 ogeL 792A wlol} uous SiOld BANEINUIND Yj WOZHOH BUIULEYd 12¢h-on1d 8 18NO SOIBAIEHIG UONEIOION-VONEDCT Z+ i I i RETAIL/SERVICE Location Retail and service centers are 0 en the final stockin poinis in a physical distribution cilities as departm network. These inelude such f stores, supermarkets, branch banks, emergency medical centers, churches, ecyeling centers med tee and police Stations. Location analysis for these points often must be hig ly sensitive to revenue lity factors rather than the cost factors so important to plant ard one Rouse location. Factors such as proximity to competition, population makeup, cus- route aeate pattems, neamess to complementary outlets, parking availability, prox- many foc featsportation routes, and community attitudes age just a fever te many factors that can influence retail or service location, Therefore the previous Treihedology does not directly apply to these problems, Since the logistician is less tikely to be directly responsible for retail or service location, we will examine only a few of the more popular methodologies Weighted Checklist piten, many ofthe factors important to retail or service locaton are not quantified eas- i) or inexpensively. Judgment remains an integral part ofthe locataos decision, yet itis Sifficut to make comparisons among site unless the analysis cane quantified to some Tie ae ceudely: One possibility is to form a weighted matix of location, factors, Lux thet shown in Table 138, and score each factor for potential sie anak number, rich the sam ofthe factor weights multiplied by the factor scones fant score ‘or the site Sites with high index values are preferable to sites with low wes values, Table 13-8 An Example List of Fectors Imp Site Selection iportant to Retail or Service Number of parking spots avaiabie Vsti of sce from stat Size a! Shape of the fot Concltion oF existing building ff any) Ingress ant esis quality Number of competitors in anes Number and types of stones in ares Complementarity of neighboring stones Prowimity to commercial areas Join promotion by local merchants Soo: Ai Gosh and ea MeLatos, Ltn Soo fo Rl on Seas Fs Lrwngon s DiC teathond Company 7) p38 Table 13-8 (con) Example Suppose that a major paint manufacturer wants to site a retail outlet for its products Outside experts, as well as standard checklists, would be consulted to generate a list of factors relevant to the problem of where to put the retail store. An abbreviated lst of factors is shown in Table 13-9. Factor weights are assigned a number from 1 to 10, according to the relative importance of each factor, with 10 being most important. A particular site is scored on a scale from | to 10, with 10 representing the most favor. able status. This particular site has a total index of 391. Other sites can be scored and the total index values compared. Of course, special care must be taken to scare dif- forent sites consistently so thatthe index values can be reasonably compares. Table 13-9 ae A Hypothetical rcocpatoi" __tocanes Pao Weighted Factor [NSTI _tocaney Paton Checklist fore Reta | 8‘ Froximiy tocompetng Location Example eee 5 Spacer tease 3 consderatore Sait ange 1 7 Brainite s omplomenary stores © © Mademinyefstorespace 8 9 Coser acest s pera ieeieas 2 3 Commit service 1 © Prociy to major ; anspor eteies linden 588 Part Location Strategy ost chasse UNA eden LE : _— Observations i * When Dave Thomas, founder of Wendy's, new locations for its restaurants, was asked how his firm decided on he replied, “We see where a McDonald's restaurant is and locate as near to it as possible,” * {ht Original Mattress Factory, which was founded by a former CEO of the Sealy Mattress Company (the largest ! i mattress producer in the United States), estab- } ahaa factory and retall tore and promoted its mattresses heavily through i | aaearising on radio and television. It was not long before computing oes | ' ‘etal stores were located immediately next door or directly cross theca | | i | Spatial-Interaction Model | | One of the most popular approaches to determining the drawing power, or overall i | 1Reirability, ofa site is the gravity model. An early version was Loown se Reilly's 4 ‘The basic idea is that two competing cities attract trade from on intervening town i UE aveey Broportion fo each city’s population but in inverse proportion emer | pits hen eats between the cities and town. Although this model i ute ine RISKS ittas Been eniched through using the mass or eurety offered by a econ aa tated of population. Mass variables are square footage ofthe ators mentary ee Siferent items instock, levels of inventory maintained, or ether fester tree ae y tance oF cg Data in the original formula becomes customer driving che ee {anes of driving time to competing retail outlets and the proposed site The power if Bs to which distance or driving time is raised can be empirically determina usually a oe by scaling from a map or driving the actual routes, to better retlecs how distance or Ie Ee time repels trade He L aay oncpt as ban ed no amor pac working mosey i & Hette ls spatal-interaction model developed an empirical baie fe hen a 1a : fumers ade tractveness of alternate eal Ses wih ecole he To & expressed as al | | a . on : - 5 = 86 y - srs 03.7) rik E i ee expected demand trom population center /that willbe attracted to a retail location j ee < Knickerbocker Pres, 1951 cS Chapter 13 Pact Location Decisions 389 590 = probability of customers from population center itraveling to retail location j = customer demand at population center / S, = size of the ratail location j T= travel time between population center Fand retail location f= qumber of retail locations j 9 = empirically estimated parameter Note that size § may include all variables that attract customers toa real site (6tong = attractiveness, inventory availabilty, price, parking space, etc). The ret site may be a single outlet or a service center of a group of outlets stich as a shopping center Travel fime 7 may include all variables that repel customers (distance, trafic congas: = tion, imitations to acess, detours, ete). The purpose ofthe model is to estimate the share of the total market that will be captured by various retail and service center sites. ae The Huff model is a basic model for spatial interaction. Over the years: researchers have improved its ability to represent real problems, have reformulated itas a multiplicative model, and have suggested different definitions ofthe variables to improve its predictive performance? Example Suppose that there are two shopping centers (R, and R,) within a metropolitan zegion located in relation to one another as shown on the time map in Figure 13-14 Ris a potential site and R, is an existing site. Customers (C, C,, and C,) who are, attracted to the shopping centers are concentrated at the centroids of their neighbor- hoods. The tolal sales potentials from the three neighborhoods are $10, $5, and $7. million, respectively. Shopping center A has 500,000 square feet of selling area and B has 1 million square feet. The parameter a is estimated to be 2. The market share for each center can be approximated as shown in Table 1310. First, the travel time is computed by using the coordinate point locations. For example, the travel time between C, and Ry is Dyp = YO%, — Xp) + — Ye = [0 ~ 50? + (20 ~ 60)" = 56.6. Second, the repelling time T* is found—T 75 3,200. Third, the probability P, of a dollar of revenue flowing to a center is com ted. For example, the probability of customer 1 selecting center Bis, ae 3128 (600,000/$00) = ,000,000/8200 ” 868 Pe = 0.36 oe "Ones compare actual sales fr an ena ‘hemodel ponerse sales. Te parametric set so that the feo are ual For arother the model see Aut Ghosh and Sarat. MeLaterty, Lacatin Srategrs for Beal and S (Lecington MA: Heath, 158 For a review of these exter ermine this parameter ms see Ghosh and Metafieny, Chapter 5 Figure 13-14 ATime-Grid Map for Shopping Center Location Example a a Lie 10 20 30 40 80 60 70 60 Time iminutes) x Fourth, the probability multiplied by the s Ieighborhood is the sales contribution thet eck neighborhood makes to the Page center sales. The expected contribution er neighborhood C, to R, 9.236 * 10 = $3.6 million, Finally {he neighborhood contributions ate sumntacs togive {tal shopping center sales. In this case, the Potential site Ry should be able to gener: see amilion in sales, This expected revere ne now be compared to operating | ests, Tents, and construction costs to see i ‘ould be made, Potential of a customer shop- 's Would be f the investment sh Other Methods A variety of additional methods plays, lens. Regression an; # role in solving retail or s alysis is important ervice location prob- to forecasting # he revenues that a specific site can expect. Covering models" ate particularly useful for locating emergeny se ices sch splice and re stations. Game they hasbeen suggest! whan tion ia key factor.” Location allocation mocks suchas goal programing integer programming ean be used. Corser an example ofthe tae of noes a jgramming to locate a bank’ principal place ol businece, Example* The Ohio Trust Company wishes to locate within the 20 counties of nartheast Ohig where it does not now have a principal place of business. According to the bank laws in Ohio, if banking frm establishes a principal place of business (PPS) in ang county then branch banks can be established in that county and in any adjacent county. Ohio Trust would like to know in which counties to establish « minima number of PPB. The 20 counties of northeastern Ohio are identified in Figure 13-15. In Table 1341, © those counties that are adjacent to each ofthe counties are listed, : To salve this problem as an integer-programming problem, we define 2) = Tifa PPB is to be located in county i:0 otherwise Based on the data in Table 13-11, we mn formulate the problem as follows: Min, Ky 4 xy s+ tego ctouniy> subject to Wyte + Hing + yy > TAshtabula 5 tay Kg sey Hg 2 1 Lake Yrazhs Lake Lake County adjacent to Lake Coun egy + Beye + ye + Begg Allxs = Oor i= Variables are integer Using any appropriate integer programming cade ts solve His pation (be MIPROG in LOCHARE), we nd hat PPUsase ocd and a neato inshore Stark and Geng sunt — eT, Service facilities may also be located by the integer programming method. igand A. Ghosh, “Covering Approsches to Retail Facility Location” in AMA Ed (Chicago: Americar Marketing Amowsatior 1984. hy. "Using Game Thaory #0 Soe! Competison.” Marketing, Vol 22.0985), n, Dents Sweeney, and Thomas Wes! Publishing Co. 1988) pp [ Counties 11 Ashtabuta | 2 Lake | > cuvahoae Lake Erie [ 7. Ashland | samen | | 8. Medina } ic. Summit | 22 Geauge UE Sie Staton | 18. Trumbull foes Pennsylvania GOMES 43-18 Northeastern Onio Counsies for Possible Principal By Ohio Trust Company, ‘Reprinted by pertission fom ire, Decnis Scone ad Thomas A, Williams, As erg at SS Pa Ws Ps Tons) oes 88S by West ng Company: All rights exerted Table 13-11 Aciacent Counties 1 Each County Under Consideration for Ohio Trust Company Portage 4. Columbiana Ses SSRERRE TE RERRIERNNREENINNNS | = & Es | 504 Example MetroHealth Hospital would like to position emergency services in the surrounding communities of the large metropolitan area in which it operates. The objective i have no patient drive more than ten minutes to reach an emergency room. The esti mated patient driving times in minutes to potential locations is given as follows: ‘To Potential Emergency Room Sie From Neighborhood 6 20 What is the minimum number of emergency rooms and where should they be located? To solve this problem, we first note which emergency room sites aze within the requited ten-minute drive. The list is B Sie Neighborhood We can now write Minimize Xq + Xp + Xo + Xp + Xe + Xp subject to 21 (Site 1 constraint) {Site 2 constraint! Xo+% (Site 3 constraint Net Xp Xe 21 (Site 4 constraint! Xp + Xe Xp 21 (Site 5 constraint} Xs + Xe+ Xp (Site 6 constraint) AIL Xs = O0rt Solving this problem using the MIPROG module in LOGWARE gives Xp = 1 and Xp = Irmeaning that emergency rooms should be placed in neighborhoods B and D. Part V Location Strategy ee OTHER Location PRog.ems Pc a any lection problems occurdngin supply chal planing that isnot aera inet OrCushly discuss all of them. However, following ate a tow aldigee ne Problem that may use the solution methodology already presented or aa may OF Probe ction procedures. They are selective istration ofthe vara, of problem types faced by logisticians Hub and Spoke A location problem solution made popular by the airlines, small package delivery services (FedEx and UPS), and communication systems is the hub-and-spoke con- crocied net an Moving teattic directly from origin to destination, the talon Giscted through one or two hubs or transfer facilites, Tae movec fey hub to Gestination points or through a high-volume interconnection to snarhen for The es problem is to minimize the transportation cost plus the coat ath, operation by () determining the numberof hubs, 2) specifying heir lcotions woe (3) routing all {ough the hubs. Since the origin and destination identities macy ta tied [g8ether the problem isnot solved in the same manner asa warehouse Levee prob: tem. Rather, to solve the problem precisely requires a specialized slgerthan Obnoxious Facilities Teaommon that location i judged on cost minimizing or profit maximizing criteria, ‘This has the tendency to place locations close to the centers of demand, which is ment rian disadvantage. Obnoxious facilities such as waste dumps, water neat stem rants chemical reclamation plants and prisons are located or’ eta that attempts to maximize the minimum distance between them and the population. The Problem can be solved in a manner similar to the Metrolieolth Hospital emergency ean Preblem discussed above. The difference is to develop a matric ot population Prabhat exceed the minimum required distance from a potential lafaten non Problem constraints are then found from this mati. Integer programming is used to solve the problem Microlocation Practical location problems often involve signiticant-size geographical areas where {lance approximations over road, rail, water, and ais Rete we reasonable, However, where location involves small demand areas such se locating newspaper clstribution terminals, ruck delivery and pickup terminals, workecane, location in| plants, and product location in warehouses, inaccuracies in estimating travel dis- semacg cannot be tolerated. Although the methodology for these micrgheeagene prob- lems isnot necessarily ditferent from those previously descrised, ie requirement for precise data can be. iss Pika and Davi. Shing, Single Allocation Hub and 096) pp SSS Chapter 19. Facility Location Decisions 595 ee BEE CONCLUDING ComMENTS. 506 Tocating facilities in the network can be considered the most important logistics $C strategic planning problem for most firms, It sets the conditions for the pooper sel tion and good management of transport services and inventory’ levels. n ‘ways, facility location is the “bones” of the supply chain, Becauise tiv many avilities to be considered at one time, along with multipie prox them, multiple sources to serve them, and multiple customers served 6 problem is often quite complex. Decision aids are typically useful ‘The purpose of this chapter was lo survey some of the more practical for the location of plants, warehouses, and retail or service facilities i» the logistics, network. We began with classifying location problems into 3 limited mum gories so that the major characteristics of location methodology could be iden Next, the history of location theory was highlighted, Single facility loeation methodology was represented by the exact centerof gravity approach. This continuous location method is useful where transportation costs are the dominant location factor and there is to be no selection of candidate locations for testing, as in the case with mathematical programming methods Multiple facility location is the more important problem to most businesses. ‘Three location approaches are typically used: (1) optimization; (2) simulation; and (3) heuristic methods. Although there are many models of each class that have been for- mulated, only one or two in each class have been used to illustrate the nature of the methodology. The extensions of both the static single and multiple facility models to deal with the problem of location over time were shown. Finally, the problem of retail and service location was discussed. Several models ‘were shown (weighted checklist and gravity models). The retail or service location problem stands in sharp contrast with the warehouse location problem p:imarily because it is revenue-based rather than cost-based, as is mast warehouse location. Questions 1. Rofer to Weber's classification of industries, Should the following processes be (1) located near their markets; (2) located near their raw material sources; or (3 not necessarily located at markets or raw material sources? a. Bottling windshield washer fluid b. Assembling VCRs. © Smelting aluminum, 6. Refining crude oi! ©. Making apple cider How do transportation costs influence your suggestion? 2, According to Hoover, what characteristic in transportation rates mal inherently unstable between markets and raw material sources? 3. What is a ubiquity? What impact does it have on location? 4. Multiple facility location models can be classified as exact, simulation, or hens tic. Explain the differences between these and cite examples of each type. Be sure Jicate why your example illustrates the type | 1 5. Why ate single location methods not very appropriate for the multiple location problem? {hat are the relevant costs fora multiple warehouse location problem? Why are these important to a proper location analysis? + nat enelit does a “flat bottom” total cost curve have in making simulation a usable methodology for multiple warehouse location? What sa heuristic method? What i an exact method? How are they useful in solving warehouse location problems? What is dynamic warehouse location? When is it most appropriate to use this approach? When is a weighted checklist a useful methodology for location? [a the location of a McDonald's restaurant, what factors would attract customers oat dean secations? What factors would repel them? How would you go about determining the relative importance of each factor? PROBLEMS: Some problems in this chapter can be solved or partially solved with the aid Of computer software. The software packages in LOGWARE thet are sine important for this chapter are COG (C), MULTICOG (4), TRANLP (7) ROUTER (R), PMED (P), and MIPROG (IP). The CD icon FG} will appear with the software package designation shere the problem analysis assisted by one of these software programs. A database may be prepared for the problem if ‘extensive data input is required. Where the problem can be solved without the aid of the computer (by hand), the hand icon 2B isshown. Ifno icon appears, hand calculation is assumed. + Ihe Plants are fo serve three market points through one or two warehouses, as shown in Figure 13-16. Volume flowing either tO or from each point, sed dha associated transportation rates, ae given as follows Pai Paint Volume, “Transportation Rate, No. Vow, RiBvewt/mi 7 3.000) Dor 2 7,000 os 3 3,500 0.095 4 3,000 0.095 M, 0 0.095, Sa Using the center-of-gravity method, find the approximate location fora single warehouse. Ring the exact centerof-gravity method, find the optimum single ware= house location. Chapter 13 Facility Location Decisions 397 ro a x | r. iS Figure 13-16 Location of Plants and Markets with Grid Overley 598 Pari V Location St 0123456769810 Horizontal grid coordinates. x . Evaluate the quality of the solutions obtained in terms of their optimality and usefulness. Evaluate the factors included or not included in the mode Explain how management might use the solution. 4d. Find the optimum locations for 0 warehouses to serve these markets Assume that each plant serves each warehouse in proportion to the market volume assigned t0 the warehouse The Care-A-Lot Hospital Group wishes to locate an outpatient clinic or clinics in a rural area of Africa. The construction costs and other considerations suggest that one or twa centers would be about right. Since traveling is difficult for patients in this part of the world, proximity to such facilites often dictates theit choice. Therefore, location is best determined based on weighted distance (sum ber of patients times distance from the facility). Figure 13-17 shows the anal number of patients likely to visit the clinic(s) and their cluster locations. !t is esti- ‘mated that it costs an average of $0.75 per kilometer (prorated on the basis of a ‘one-way trip) for a patient to travel or be transported to the elini(s). This esti- mate is based on lost productivity, direct travel costs, and indirect travel expenses paid for by others a. What is the best location for a single clinic? by two clinics are tobe located, where isthe best location for them? & Acinic costs $500,000 (US. dollars) per year to equip and staff. Ths is paid through charitable contributions and goverment subsidies. On purely eco nomic grounds, should the second clinic be built? Bottoms-Up, Inc. isa small company that produces and distributes beer under the Old Wheez label. The company is examining the possibility of penetrating the North Shore City metropolitan area market. A plant location that would serve the area is sought. A grid overlay is placed over the selling area, as SAO? in Figure 13-18. North Shore City is area £. The suburbs surrounding E are des- ignated as A tol.Ama lowing potential demand for Oid Wheez ‘et research study shows the Democranc REPUBLIC OF CONGO Kilometers Figure 13-17 Grid Overlay on Patient Concentrations in a Region of Africa Area Annual Volume (ewe) carr 70000 5,000 70,000 3000 40.000 12,000 80,200 Chapter 19 Facility Location Decisions 599 Metropolitan Area Figure 13-18 Grid Overlay of North Shore Cay UT a ot 2 8 «6 Demand comes primarily from dealers that ate scattered uniformly over the area, Transportation costs are estimated at $0.10 /ewt/ mi 2. Ifthe center-of gravity approach is used, where should the bottling plant be located? Estimate the annie] transportation bil . Ifthe exact center-of-gravity method is used to find the plant location, where should the location be? Does it make a substantial difference in transports. tion costs compared with the location found in part 2? &Flabor costs, property taxes, and site development costs vary with how would yout propose accounting for these additional costs when aecia ing the location? Recall the problem presented in Figure 13-10. Resolve it, asstuming that both warehouses are public warehouses and, therefore, no fixed costs 1ppll Summarize your answer in terms of customer, warehouse, and plant +s ments, 5. Recall the problem presented in Figure 13-10. Resolve it, assuming that house 2 can handle ony 100,000 cvvt. Warehouse ? is expanded to hand unlimited throughput. The plant capacities remain unchanged. How much of 4 cast penalty can be expected from this c Recall the problem presented in Figuze 13-10. Resolve the problem, assuming thatthe fived cost on warehouse 2 (IV, s $200,000 instead of $420,000 pox Develop a lst of factors that you think would be important in deciding th tion of 2. AGowxtwill collection cente: A Wendy's restaurant ge? V_ Losation Strategy cnc eee ee & An automobile assembly plant d. A fire station ils indicate the weight that you would assign to each factor S Inthe problem shown in Table 13-6 and Figure 1313, Suppose that the cast asso- Ting th moving trom one location to another is S300 bi aan of $100,000, What location planning stratogy would offer the mania Profits over the five- year planning horizon? Fp © Recall the Environment Plus incinerator location problem and the data for the Ad problem in the file PMEDO2 DAT of tne PMED motiule-n LOGWARE. Consider the following additional questions, + Hows many facilities will give the lowest operating cost plus transportation Foran anete should these incinerators be located? Show he nee best num- ber of incinerators has been found, >. The company currently operates four incinerators at an De ansportation cost of $35 million. Chicego, Atlanta, Presse ce ar egeate the sites. The cost to establish new incite we one-time for the company to « rat market of Los Angelos and Seattle were to grow by 10 times, would your answer to part a be different 10 Teppose that Farmers’ Bank wishes to serve the nine customer caster ag shown Foye 1219. It has proposed locating a branch (A) at eocine ce, X, = 2, P20. A competing branch (B) is located at coowtintee X= 40, Y, = 30. agents’ Bank f to be a full-service branch sth a relate sng (attractiveness) thesoah cagibe competing branch isa partiline faclty (ne Ante drive- [rough capabilites) witha size index of 07. The travel fone customers to a bank is approximated as Tthours) = D/50, whete D = diego miles. The ‘average customer generates $100 in gross annual revenue for bane en Figure 13-19 Number of Potential Customers for Sranch Banks in a Region Chapter | Fauility Location Decisions 601 Se S BLS Michigan Lae Ere Ingiona Counties Willams Fulton van Wert Lucas Allen Onawe Detiance Henry Wood Auslaia Sancusky 19, Marion Paulding 20. Shatby Logan estimate of annual operating expense for Farmers’ branch is $300,000, ond the facility will cost $650,000 (20-year life) on land worth $100,000. a. Apply Hufi’s retail gravity model to determine the branch's annual revontte Assume a= 2. , Considering the level of investment required and the operating expenses, should the branch be constructed? i a © What additional information might you like to have before making a final decision? Tp \ 1. The Ohio Trust Company wishes to expand its principal place of business oc tions to the northwestern counties of Ohio. The conditions for location were lined in a previous exampie in Figure 13-15 on page 563. For the counties icenti= fied in Figure 13-20, find the minimum number of PPBs needed and the counties in which they should be located. (Note: A database is prepared for this problem» in the module MIPROG.) Tp N12. Biogenics is a start-up company that plans to produce biological materials used in medical research, Major customers for th hospitals located in major metropolitan areas. annual sales are 2s follows, products will be the ange researsh Cusiomer location and projected % am ae es Tha S10 2 New tork P40 73060 3 Washington 700 sso. #AMtane B38 65000 = Mam > #028 35800 © Cleveland tas 5186 25009 7 Det 2x $308 2010 & Chicago as 76 mow 9 Seton B38 90.9 2op00 10 Minneapolis ur 30 1500 1 Kanses Cty 310 458 130°0 1 Philadelphia 995 9900 1% Howton 2378 alle 258 20000 1 Phoenix a8 vote 16 Denver wn 18000 1 — Seatie vss ro000 18 Parend 46 10,000 1 san oncsco S778 sa.300 —tesAngels kos 836 products will be shipped UPS at a transport rate that averages $0.05/b/ mite. Itis estimated that the annual operating costs (FOC) fora ube ‘ory plant) are given from FOC(S) = ($5,000,000) VN, where Nis the number of laboratories being operate. The vendors forthe materials used inthe proce, fon process are assumed concentrated at Chicago. The purchase weight isthe same as the sales weight. The transport rate from Chicago to the laborstors estimated at $0.02/1b/ mile. Determine the number and location ofthe laboratories to serve Biogenies” potential markets. Which customers should be served out of each site? Fvery Eustomer location is a potential laboratory site, except St. Louis, Portlanc Kansas City, Washington, Detroit, and Chicago vendors. For the problem shown in Figuze 13-5, suppose that the capacity restriction on Saehouse 2 (WV, for both products combined is 100,000 ewe. There is no apace “¢ Zestriction on warehouse IW ). (Note: Data forthe problem shown in Figure 16-5 are available in the MiPROG module of LOGWARE. Capacitos are ae axe aertion point of Cap-W1/ZWi and Cap-W2/ZW? in the problem setup.) Recalling the problem presented in Figure 13-5, how would the solution change tthe following altezations were made to the problem seep? ‘Demand fos product 1 s doubled, but remains unchanged for proxuct The manuiacturing cost for product 2 is raised to $5/cwt at plar ling cost for warehouse 2 is increased 10 $4/cwt Chapter {3 Facility Location Decisions — 603 a There Table 13-12 Times Between Zones for Glabe | frow zone Casuslty Company, in Minutes a limited capacity in plant 2 of 90,000 Plant 1 capacity js incre plant capacity changes for product 2. Customer 2 for product 2 can no longer be served from warehouse 2 wt. to produce p -ased from 60,000 cwt. to 150,000 ewt. There: Figure 13-21. Typical Daily Demand Pattern with Current and Potential Supply Yard Locations Scale: 1 grid unit = 5Klometers 24 25 Potential yard C yard 604 Pact Location Strategy oe ce Resolve separately each of the above problem scenarios using the mixed integer inet Programming approach, (Note:The problem setup for Figure In i nce able in the MIPROG module of LOGWARE)) 15. The Globe Casualty Company positions claims adjusters atound a metropolitan Shines, SabOnd quickly to insurance claims resulting from auto accidents, tines chants, and other such emergency situations. It isa competitive feature of the serch Business for an adjuster to be on-site within 30 minutes of the time na fecident is called in, so that customers feel they are being well served. Tre city 12s been divided into ten zones feom which casualty calls originate ond inci, claims adjusters may be stationed, The respon. means ate shown in Table 13-12. To meet the 30-minute response time, here any claims adjuster stations should be established, and in which zones saeclt they be located? 16, A building supply firm supplies materials to construction sites throughout the from eoulan area of Mexico City, Mexico, Daly delivery trucks are dispatched from a materials yard. A typical daily demand pattern is shown in Figure 13-21, tehete a grid is overlaid on che mettopoltan area. The map-scaling-hetor ie GRordinate unit = 3 kilometers with a circuity factor of Lif to coment straight tine distance to approximate road distance. Demand is given im kilograms of merchandise in Table 13-13. Table 13-13 Customer Demand Votume and Coordinate Locations with Yard Coordinates Y vous Son Younes 2 0 30 16 250 00 18 1 8 wo 20 a re % z os o 1 1 2 a 2 3 3 3» 13 Facility Location Decisions ee Trucks operate with a variable cast of 2.5 pesos per kilometer, are 90 pesos per day, and a truck is amortized at 200 pesos per ds are stored in open yards and in buildings at locations shown in Fagus current materials vard from which trucks are dispatched operat pesos per day. The company is considering moving its ope materials yards have an estimated operating expense, including expense from the current location, of 480 pesos per day for A, 45) 935-5 mr for B, and 420 pesos per day for C. There are ten trucks avetladie, 9 y th hauling capacity of 1,000 kg, but not all may be needed for me: tions. 7 E demand. Trucks travel over their routes at an average speed of: Aer p noon, drivers are permitted a one-hour lunch break and they ust vetur to B j the yard before needing additional breaks. Based on a company policy, soutes B | shoulld be no longer than ten hours in a day, and trucks are not to leave the yard before 8 A.M. Ifa track has a short route and returns for reloading and rerouting, 1.5 hours for loading is required, Customers have a time window for delivery of [ between § 4.06, and 5 ns. Time for unloading at a customer location is estimated! 8 15 minutes plus 0.1 times the stop volume in kilograms, Which yard location is most economically attractive? 606 Pare V Location Strategy i i E | | i | e CASE STUDIES oe oN Company perlor Medical Equipment Company sup- piles electrical equipment that is used as come posents in the assembly of MRI, CAT scanners PET scanners, and other medical diagnostic

    You might also like