An integrative model for automatic warehousing systems

Download An integrative model for automatic warehousing systems

Post on 22-Feb-2017

220 views

Category:

Documents

0 download

Embed Size (px)

TRANSCRIPT

<ul><li><p>This article was downloaded by: [Carnegie Mellon University]On: 20 October 2014, At: 01:48Publisher: Taylor &amp; FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office:Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK</p><p>International Journal of ComputerIntegrated ManufacturingPublication details, including instructions for authors and subscriptioninformation:http://www.tandfonline.com/loi/tcim20</p><p>An integrative model for automaticwarehousing systemsMoshe Eben-ChaimePublished online: 08 Nov 2010.</p><p>To cite this article: Moshe Eben-Chaime (1996) An integrative model for automatic warehousing systems,International Journal of Computer Integrated Manufacturing, 9:4, 286-292, DOI: 10.1080/095119296131580</p><p>To link to this article: http://dx.doi.org/10.1080/095119296131580</p><p>PLEASE SCROLL DOWN FOR ARTICLE</p><p>Taylor &amp; Francis makes every effort to ensure the accuracy of all the information (theContent) contained in the publications on our platform. However, Taylor &amp; Francis, our agents,and our licensors make no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Any opinions and views expressedin this publication are the opinions and views of the authors, and are not the views of orendorsed by Taylor &amp; Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shallnot be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.</p><p>This article may be used for research, teaching, and private study purposes. Any substantialor systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, ordistribution in any form to anyone is expressly forbidden. Terms &amp; Conditions of access and usecan be found at http://www.tandfonline.com/page/terms-and-conditions</p><p>http://www.tandfonline.com/loi/tcim20http://www.tandfonline.com/action/showCitFormats?doi=10.1080/095119296131580http://dx.doi.org/10.1080/095119296131580http://www.tandfonline.com/page/terms-and-conditions</p></li><li><p>An integrative model for automaticwarehousing systems</p><p>MOSHE EBEN-CHAIME and NAVA PLISKIN</p><p>Abstract. Automatic warehouses, for the most part, have beenstudied in isolation. This paper departs from this approachby proposing an integrative model in which warehouseactivities interact with other functions of the total system.The model has been simulated under three modes of opera-tion: single command, dual command (DC), and hybrid.Simulation results con rm that the length of stay of unitloads outside the warehouse affects performance and alsosuggests that, under the DC mode, the warehouse may losestability. Hence, the hybrid mode is proposed as an alter-native. The hybrid mode outperforms the DC mode on mostperformance measures, except that the length of the storagequeue at high throughput levels is longer. Finally, savingopportunities, via a reduction in the number of storage/retrieval machines, are discussed.</p><p>1. Introduction</p><p>An automatic warehousing (AW) system consists ofracks, storage/retrieval (S/R) machines, input/output(I/O) station(s), and computerized control devices.The racks are paired back to back with aisles betweenthe pairs. The S/R machines are cranes that travel inthe aisles and move objects between the racks and theI/O stations. Each S/R machine can move horizontallyand vertically at the same time and can access the front(pick) face of the racks on both sides of the aisle. An S/R machine is either dedicated to a single aisle, or canmove between aisles. The cranes operate under thecontrol of a computerized system in one of two modesof operation. These operation modes are either basedon a single command (SC) cycle, during which a singlestorage or retrieval operation is performed, or on a dualcommand (DC) cycle, during which both a storageoperation and a retrieval operation are performedbetween two consecutive visits to I/O stations.</p><p>Operations management is concerned with thesequencing of storage and retrieval requests, and thematching of both types of requests in DC cycles. Bozerand White (1984) proposed the standardization andapproximation approach, while assuming randomizedstorage and FIFO sequencing. They offered to stan-dardize (or normalize) the pick face of the racks and touse continuous approximations for developing generalexpressions of the expected travel times for single anddual command cycles in AS/RS. This work inspired theinvestigation of alternative sequencing policies by Hanet al. (1987), who developed the ef cient `nearestneighbour (NN) heuristic rule.</p><p>Expected travel time was the only concern in thesestudies, which applied the NN rule to block sequencing.The travel time is the major component of the servicetime in AW systems. The other components are the pickand deposit (P/D) times which are constant and inde-pendent of the operation mode, sequencing rule, andthe like. Travel time was considered because often, inqueuing systems, the mean service time is highly cor-related with the mean response time which measuresthe service level of the system. The response time is thetime elapsed since a request for service is issued until itis completed, including both the waiting time in thequeue and service time. Eben-Chaime (1992) showedthat block sequencing can be hazardous in terms ofresponse times and queue lengths. As an alternative heproposed a dynamic application of the NN rule fordispatching, an alternative which was shown to main-tain response times and queue lengths at the levels ofFIFO sequencing, while signi cantly reducing cranetravel times.</p><p>Previous operations management studies con-sidered the storage function in isolation, ignoring itsrelationships with, and dependencies on, the otherfunctions of the total system. While this might beacceptable for the distribution centre studied by Seid-mann (1988), the isolation premise does not hold</p><p>0951-192X /96 $12.00 1996 Taylor &amp; Francis Ltd</p><p>INT. J. COMPUTER INTEGRATED MANUFACTURING, 1996, VOL. 9, NO. 4, 286 292</p><p>Authors : Moshe Eben-Chaime and Nava Pliskin, Department of Industrial</p><p>Engineering and Management, Ben-Gurion University of the Negev, P.O. Box</p><p>653, 84105 Beer Sheva, Israel.</p><p>Dow</p><p>nloa</p><p>ded </p><p>by [</p><p>Car</p><p>negi</p><p>e M</p><p>ello</p><p>n U</p><p>nive</p><p>rsity</p><p>] at</p><p> 01:</p><p>48 2</p><p>0 O</p><p>ctob</p><p>er 2</p><p>014 </p></li><li><p>for the more general case, where storage sub-systemsinteract with various other subsystems such as manu-facturing. Some studies assumed FIFO service and didnot account for the economic bene ts which may resultfrom improved operation management policies, suchas fewer cranes and a smaller storage area. Recently, itwas noted that `what makes AS/RS most justi able iswhen you integrate it into an entire manufacturingsystem scheme (Knill et al. 1993).</p><p>This paper presents an integrative systematic modelthat considers the relationships between storage andother functions of the total system. The proposedmodel, in Section 2, provides a conceptual frameworkfor investigating, by simulation, the operational charac-teristics of the system under different circumstances.The simulation environment is also described in Section2. Queue behaviour and the effects of sequencing rulesare studied in the third section while Section 4 isdevoted to a comparison of various operations modes.It is noteworthy that although the paper focuses onoperations management, the model also lends itself tothe study of storage policies and assignment and systemdesign.</p><p>2. The model and simulation environment</p><p>Many automated warehouses (AWs) function on thebasis of unit loads (UL) such as bins, drawers, or pallets,which are used to store items, rather than on the basisof single items. When performing a retrieval operation,an S/R machine pulls a UL from a rack and</p><p>delivers it to an I/O station. When performing a storageoperation, a UL is taken from an I/O station and isdeposited in an empty rack slot. A retrieval (storage) ofa UL does not necessarily imply that items are retrieved(stored). A UL, for instance, may be retrieved in orderto add to its contents , i.e. store item(s). Similarly, ULsare often re-stored after some items are removed, i.e.retrieved from them. Hence, storage requests are notgenerated independently but as a result of prior retrie-vals. Further, warehouse activities are triggered by othersystem functions that generate UL retrieval requests.These observations imply that there are two separatequeue types, one for storage and one for retrieval, andmore than a single queue of each type may exist. Thereare two principle differences between the two queuetypes. First, the queues of retrieval requests are listsstored in the computer memory while ULs are physi-cally waiting to be stored in the storage queues. Second,an exact location of a UL is speci ed for each retrieval,while under the assumption of randomized storage,ULs can be stored in any empty slot in the warehouse.The addresses of the empty slots are stored in anotherlist in the computer memory.</p><p>The integrative model for AWs proposed belowconsists of the system, the racks, the S/R machines,the storage and retrieval queues, and the list of emptyslot addresses. The operation cycles are illustratedgraphically in Figure 1 where the cause-and-effectrelationships between both service types are clari edby the directed arrows. The length of time a UL spendsout of the warehouse is determined by activities</p><p>Integrative model for automatic warehousing 287</p><p>Figure 1. The integrative model for automatic warehouses.</p><p>Dow</p><p>nloa</p><p>ded </p><p>by [</p><p>Car</p><p>negi</p><p>e M</p><p>ello</p><p>n U</p><p>nive</p><p>rsity</p><p>] at</p><p> 01:</p><p>48 2</p><p>0 O</p><p>ctob</p><p>er 2</p><p>014 </p></li><li><p>performed on the items in the UL by other systemfunctions. Since this time segment may in uence thewarehouse operation, it must be incorporated into theanalysis.</p><p>The integrative model has been simulated on an80486 platform to study an AS/RS consisting of a pair ofracks, an S/R machine, and a single I/O station. Thesimulation was programmed using both Turbo-Pascal6.0</p><p>TMand Paradox</p><p>TMin a complementary manner. For</p><p>the sake of generality, the continuous approximationand rack standardization of Bozer and White (1984) isadopted in the simulation. This approach allows analy-sis in terms of the standard time unit, T, and theshape factor, b, of the rack, disregarding physicalattributes, size and structure of the rack and speeds ofthe S/R machines. Our simulation, of a pair of squaredin time racks and an S/R machine dedicated to servethem, assumes randomized storage and a single I/Ostation at the lower-left corner of the racks. The inputsto the simulator include the following: (1) the distribu-tion of the inter-arrival times of retrieval requests; (2)the distribution of the length of stay of ULs outside thewarehouse; (3) the standard time unit T (squared intime racks, b = 1, were assumed); (4) the number ofempty slots; (5) the P/D times; (6) the length of thesimulation in terms of completed service cycles; and (7)the mode of operation.</p><p>3. The behaviour of the queues</p><p>The behaviour of the two queues in terms of waitingtime, as a function of mean item-processing time, isshown in Figure 2, for three different load sizes. Thepattern is the same for all loads. Waiting times in bothqueues decrease and converge to a certain value whichis, of course, higher for heavier loads. For each loadsize, both storage and retrieval time curves, thoughclose, do not coincide and intersect once.</p><p>The in uence of the sequencing rule on warehouseperformance is studied via comparison between FIFOand SPT. The simulation con rms that throughputrates are equal under both rules for the same loadsize. Figure 3 displays the effect of the load size onwaiting time in the storage queue and the retrievalqueue, the retrieval queue is hardly affected by thesequencing rule, in contrast to the remarkable effect onthe storage queue.</p><p>Next, DC cycles are performed. The use of DCcycles increases the dependency between both servicetypes. The patterns for FIFO sequencing are presentedin Figure 4. Similar patterns are observed under NNdispatching except that, as expected, the means of thewaiting time are shorter under the NN rule. The lengthof stay outside the warehouse, the item processingtime, seems to have a vital in uence on both queues,as can be observed in part (a) of Figure 4. For longer</p><p>M. Eben-Chaime and N. Pliskin288</p><p>Figure 2. Queue performance SC model.</p><p>Dow</p><p>nloa</p><p>ded </p><p>by [</p><p>Car</p><p>negi</p><p>e M</p><p>ello</p><p>n U</p><p>nive</p><p>rsity</p><p>] at</p><p> 01:</p><p>48 2</p><p>0 O</p><p>ctob</p><p>er 2</p><p>014 </p></li><li><p>mean processing time the waiting time in the storagequeue decreases in parallel to a growth of the waitingtimes in the retrieval queue. The load, controlledprimarily by the arrival rate, has a similar effect, part(b) of Figure 4. It is noteworthy and disturbing that forheavy loads and long processing time (e.g. 80% loadand 8 time units) the system collapses and the simula-tion is aborted. The trends are switched with respect to</p><p>the number of empty slots, as shown in Figure 4(c). Inorder to apply the NN rule, there should be a numberof empty slots in the racks. This number remains xedsince, during each cycle, one slot is lled up by storing aUL in it, while another slot is emptied when its UL isretrieved. The larger the number of empty slots, thelarge is the cycle time reduction and the consequentthroughput increase obtained under the NN rule.</p><p>Integrative model for automatic warehousing 289</p><p>Figure 3. Sequencing rules comparison SC mode.</p><p>Dow</p><p>nloa</p><p>ded </p><p>by [</p><p>Car</p><p>negi</p><p>e M</p><p>ello</p><p>n U</p><p>nive</p><p>rsity</p><p>] at</p><p> 01:</p><p>48 2</p><p>0 O</p><p>ctob</p><p>er 2</p><p>014 </p></li><li><p>Clearly, the cycle time is reduced at the expense ofmuch longer waiting time in the storage queue. Longerwaiting time implies longer queues, creating a need fora larger space to hold the actual ULs that are waiting inthe storage queue.</p><p>The advantage of the DC mode over the SC mode isthe resultant reduction in travel time per transaction, inthe order of 40%! However, the sensitivity of warehouseperformance (in terms of waiting time) under pure DCoperation mode and the resulting system collapsemotivated us to search for a better alternative. Ahybrid mode of operation is proposed in an attemptto maintain stability while taking advantage of the shorttravel times under the DC mode. Performance levelsunder the proposed hybrid mode of operation areanalysed and compared with the DC mode, in thenext section.</p><p>4. Comparative analysis of the hybrid mode</p><p>Under the hybrid mode, DC cycles are per-formed whenever possib le. Otherwise, SC cyclesare performed. The S/R mach ine halts on ly whenboth queues are empty. Both the FIFO and the NNrules can be used to sequence operations. Themotivation to th is mode is to achieve stability inthe warehouse queuing system . The average waitingtime in both retrieval and storage queues is plottedin Figure 5 again st mean item processing time, forload size of 50%. Clearly, the performance under theh...</p></li></ul>

Recommended

View more >