自动化制造系统的智能工具管理策略

Intelligent Tool Management Strategies for Automated Manufacturing SystemsAbstractWith the increase in automation and use of computer control in machine tools, the number of cutting toolsper machining setup is on the increase. On one hand, su ch multi-tool setups offer the advantages of reduceddown-time and cost of production and require less space and in-process inventory, and on the other hand,require proper tool management for economic operations. A number of strategies have been devised to solvethe tool selection problems and a number of tool replace ment policies have been pr oposed in the past. Thesestrategies have been solved in isolation, whereas, a comprehensive algorithm for proper selection of tools outof those available in the tool magazine for performi ng operation and for replacement of tools on failure/wearis necessary. In this paper, taking cue from the computer memory management policies, four tool selectionstrategies have been presented and their performance in tandem with various tool replacement policies hasbeen studied. The effect of important parameters su ch as reduction of tool life due to regrinding, limitednumber of regrindings, catastrophic failures etc. have been considered. Cost has been computed for eachcombination of tool selection and replacement polic y. Also, the number of machine stoppages has beenworked out in each case. The results indicate that the combination of various selection strategies with suit-able replacement policies affects the overall cost. Keywords: Tool Management, Tool Selection, Tool Replacement, Multi-Tool Setup 1. IntroductionThe present manufacturing industry requires producing alarge variety of complex components in small batch sizesand in a cost effective manner. At the same time the num-ber of tools to be magazined has increased, particularly inautomated machining centres. It is not unusual for there tobe over 300 different tools in a direct storage unit at amachining centre. Such multi-tool setups offer the advan-tages of reduced time and cost of production and requireless space and in-process inventory. With the above de-velopments, the toolmanagement task has become sub-stantially more demanding because the increase in numberof tools per setup requires a better tool reliability man-agement, as wearing/failure of any one tool renders thewhole system non-operational. As the number of tools inthe machining setup increas es, the frequency of replace-ment due to wearing out of tool and due to failure of toolalso increases. A tool management system adopted shouldbe such as to make the right tool available at the right timein proper sequence for processing a job while keepingcosts down to a reasonable limit. Hence, it becomes nec-essary to devise strategies which would select the righttool out of those available in the tool magazine of the ma-chining centre [1-3].In a traditional machining environment, a skilled op-erator, in liaison with the tool store personnel monitors,maintains, and replenishes the small number of tools as-sociated with each machine tool. The expertise of the ma-chine operator is the determinant factor in ensuring thatcorrect tools are used for each operation and the tools arereplaced before they complete ly wear out and damage thework piece. This working pattern is no longer acceptablein a modern machine shop equipped with the CNC ma--chines, as a greater variety, and a large number of toolsare used to machine components on these machines andthe task of determining the correct tool out of the alternatetools available, and task of replacing the tool on failure orbefore failure, is too complicated to be left to the machineoperator. Also, a large number of machine stoppages maybe caused due to incorrect tool selection and replacement.The tasks of tool selection for performing the operations,D. G. RAO ET AL. 406and tool replacement decisions are complex and requiredevelopment of sophisticated programs which should besupported by databases containing information on themanufacturing resources and company-specific machin-ing practices [4].2. Literature ReviewTool management problems have been attempted by re-searchers since the turn of the century. The detailed sur-vey of literature is as follows.2.1. Studies on Tool ReplacementA thorough review of the solution methodologies relatedto tool replacement can be found in McCullough [5],Armarego and Brown [6], Pa’sko [7], Batra and Barash[8], Bao [9], LaCommare et al. [10], Sharit and Elhence[11], Zhou et al . [12] and Tak [13]. As has been men-tioned by Zhou et al. [12], the various authors have clas-sified tool replacement strate gies in different ways. Mostprobabilistic optimization models are based on tool lifedistribution [10,14]. A major development in the processof computer integration of automated manufacturingsystems has been the implementation of automated toolreplacement [15,16]. A complete Tool Replacement Stra-tegy specifies a tool change schedule based upon theeconomic service lives of tools and a control policy re-garding unscheduled tool changes following breakage.The most realistic replacement strategies have consid-ered the distributed nature of tool lives under actual ma-chining parameters, as well as the option to change sev-eral tools once one fails [9,10], rather than consideringonly expected lives and single tool replacement [5,6]. Allof these tool replacement studies have considered onemachine in isolation. Sharit and Elhence [11] have gonebeyond the single machine model to examine tool re-placement strategy at the system level. Rather than pro-posing an automated, optimizing strategy, their study emphasizes the limitations of both human and computerat making the tradeoff betw een economic tool replace-ment costs and system throughput in a real-time, dy-namic environment. Zhou et al. [12] have proposed anoptimization model for tool replacement based on toolwear status. The model is capable of utilizing tool wearstatus in determining an optimal replacement policy.They have mentioned three types of tool replacementstrategies employed in the industry: 1) scheduled toolreplacement, 2) preventive planned tool replacement, and3) failure tool replacement. Under the first strategy, thecutting tool is replaced either at pre-scheduled time orupon failure whichever is earlier. The second is alsosimilar, except that it is based on the number of finishedworkpieces. The third strategy is to use the tool untilfailure. Research has been done to find the optimumpre-established time or lo t size for replacement and tocompare the different strategies [10].Two broad categories of the tool replacement policies,viz. unscheduled tool replacement schemes and sched-uled tool replacement schemes have been mentioned inthe literature. Under unscheduled tool replacement sche-me, Series Tool Replacement and Parallel Tool Replace-ment have been suggested [8]. Under scheduled tool re-placement schemes, Individual Tool Replacement [7],Group Tool Replacement [8], Skip Schedule Group ToolReplacement and Sub Group Tool Replacement [13]have been suggested. Tak [13] has suggested the Dy-namic Tool Replacement policy, under which every toolfailure, which essentially results in machine interruption,is utilized as an opportunity for evaluating every tool ofthe setup for its effective utilization and reliability. Un-der this strategy, whenever a tool fails, all the tools in themagazine are evaluated for the useful lives that they havelived and the tools exceeding their warning limits arealso replaced along with the tool which has failed. Hedinet al. [17] have investigated static and dynamic toolingpolicies and presented a comparison between the two incontext of a general flexible manufacturing system. Cur-rently, many tool replacement models are deficient inthat they ignore the relationship between the processingrates and the tool replacement policy.2.2. Studies on Tool SelectionMajor research efforts in the area of Tool Selectionstarted in the early 80’s. A review of solution method-ologies for optimum tool selection can be found in Ced-erqvist [18], Giusti et al. [19], Chen et al. [20], Syan [21],Domazet [22], Maropoulos [23], Hinduja and Barrow[24], Eversheim et al . [2], Zhang and Hinduja [4], Hedinet al. [17]. The task of selecting cutting tools which arenot only functionally correct but also optimum, is a com-plex one. Cederqvist [18] has suggested that when a newbatch is planned, the number of cutting edges requiredfor each tool in the setup can be calculated and a com-plete set of block tool heads can then be prepared andsent out to the machine. Some researchers viz. Giusti etal. [19], Syan [21] have developed expert systemswherein the technological knowledge is represented asproduction rules which are consulted when selecting atool for a given operation. Chen et al . [20] have adopteda heuristic-deterministic appr oach to reduce the comput-ing time to determine the optimum tooling for roughturning operations. Domazet [2 2] has proposed a hybridapproach to automatically sel ect turning tools; the selec-tion is done in stages and for those stages which requireCopyright ©2011 SciRes. ICAD. G. RAO ET AL. 407special knowledge and exper tise, a non-algorithm ic me-thod is followed. Maropoulos [23] has used an algorithmapproach to automatically de termine tools for rough andfinish turning operations. Hinduja and Barrow [24] havesuggested an automatic interactive system. In the interac-tive part, the user is guided by the system towards theparameters of the optimum tool. Eversheim et al . [2]have reiterated the importance of tool selection in amodern manufacturing environment. They have proposedan integrated method for tool selection on the basis ofmanufacturing features. Zhang and Hinduja [4] have pro-posed automatic generation of a tool set for a given batchof components, the optimization criterion being eitherthe minimum machining cost or minimum number ofmachine stoppages or a combination of both. Hong-BaeJun et al . [25] have considered a tool provisioning prob-lem in a flexible manufacturing system (FMS) with anautomatic tool transporter. Their study determines thenumber of copies of each tool type for a limited budgetwith the objective of minimizing makespan. Two heuris-tic algorithms have been proposed. One is a compositesearch algorithm based on two greedy search methods,and the other is a search algorithm in which numbers oftool copies are determined based on tool groupings. Inboth algorithms, simulation results are used to findsearch directions. Mözbayrak et al . [26] have addressedthe design of an integrated tool management system forflexible machining facilities (FMFs). Modules with func-tions ranging from issuing tools according to a toolingstrategy to diagnosing system operation have been de-veloped and integrated aro und a centralized manufactur-ing database to guarantee streamlined manufacture.Selim Akturk M. et al . [27] have shown that there is acritical interface between the lot sizing and tool man-agement decisions, and these two problems cannot beviewed in isolation. They have proposed the alternativealgorithms to solve lot sizing, tool allocation and ma-chining conditions optimization problems simultane-ously. Svinjarevi d G. et al. [28] have studied the con-trolled testing and analysis in every phase of tool man-agement in departments and otherservices which aredirectly involved in the tool management system toreduce stock and costs. They have identified some dis-advantages and given a few suggestions for the im-provement in the tool management system. Haslina Ar-shad et al. [29] have introduced a virtual cutting toolmanagement system to reduce or eventually solve manyof the tool management problems. It has the capabilityof choosing the right cutting tool from a virtual cuttingtool catalogue. Their system provides the virtual selec-tion process for cutting tool and a virtual milling proc-ess.3. Problem DefinitionThe above survey of literature reveals that several re-searchers have attempted different tool selection schemesand tool replacement strategies, but most of the abovemodels, except for a few, developed for optimizing toolreplacement, do not consider the actual status of the cut-ting tool. From the above review , it is clear that researchefforts have been mainly di rected towards the selectionof optimum tooling for a single machining operation.Also, the possibility of adjusting the wear rates of indi-vidual tools and synchronizing tool changes due towear/failure in order to re duce the number of machinestoppages has been suggested. Also, investigations havebeen carried out in the past using simulation strategies tofind out the best strategy under different operating situa-tions. However, it is necessary to investigate how eachselection policy performs in tandem with different re-placement policies. The reason is that some tool selectionstrategy might be better in combination with a particularreplacement strategy but may prove to be worse withother replacement strategies.The focus of this paper is to evaluate the performanceof identified tool selection policies in tandem with thedifferent tool replacement strategies. In developing arealistic simulation model for evaluating various toolselection and replacement policies, it is very important toconsider various factors such as remaining tool life ofpartially used tools, catastrophic failure of the tools,number of times the tool has been reground, reduction intool life due to regrinding, applicability of tools for vari-ous operations, operations-pro file etc. In the presentwork, a simulation model has been developed to investi-gate the most appropriate tool selection policy with eachreplacement strategy considering the above mentionedfactors.4. Adaptations for Tool Selection andReplacementThe strategies considered for tool selection and replace-ment for developing the simulation model have beenexplained in the following sections.4.1. Tool Selection StrategiesThe goal of any tool management system is to make theright tool available at the right time in the right sequencefor processing a job. To achieve this goal, the selectionof the right tool is very important as there may be anumber of similar tools or different tool types in themagazine with different remaining tool lives, capable ofperforming the operation. If proper selection of the toolCopyright ©2011 SciRes. ICAD. G. RAO ET AL. 408is not done then it is possible that a certain tool of onetype may not be used at all or may be lying unused for along time in the tool magazine, even if it has a goodamount of remaining tool life. The tool selection logic isvery simple to describe. The system must maintain arecord of all the tools with different tool lives. Sometools may be having a small value of tool life whereassome tools may have a large value of tool life. When atool request is made, the tool magazine should besearched for the type of tool required and the tool liferequired. Since the magazine has tools with different toollives, a procedure for selecting right tool is needed forthe optimum utilization of the tools.In computer systems, memory management is per-formed in which a system maintains a list of all theblocks of memory. Some of these blocks are free at anytime and some currently allocated to a user. When anallocation request is made, the system must locate a freeblock of memory of sufficient size and allocate all or partof it. In case only a portion of a free block is to be allo-cated, the allocation is made from the bottom of theblock. When a re-allocation request is made, the systemmust recover the re-allocated block of memory. In addi-tion, the system should be able to find adjacent freeblocks of memory and combine them into a single largeblock, to maximize the probability of being able to sat-isfy a large allocation request. For this reason, a numberof memory management systems have been devised fordifferent applications. Some of the most common areknown by the name First Fit, Next Fit, Best Fit andWorst Fit. These policies have been adapted here forselecting the tools from the tool magazine for performingvarious operations on a mach ining centre. The adaptationof the policies in context with the tools selection is ex-plained in the following paragraphs.1) First FitIt is a very simple scheme. The available tools areplaced in the magazine in a random manner irrespectiveof their remaining lives. When a tool request is received,the magazine is searched for the first tool with remaininglife, large enough to satisfy the request, and the tool isused for processing the requesting job.2) Next FitIn the First Fit scheme, the search for the tool havingsufficient remaining life to serve the request always be-gins from the starting position of the magazine. Conse-quently, it requires longer time for search and also, allthe tools with smaller tool life tend to collect at the be-ginning. Hence, it is necessary to examine several toolsbefore allocating a tool for processing. A modification tothe first fit strategy is to start a search for a suitable toolat the position where the previous search ended. Thisapproach causes the decrease in search time and also,tends to distribute tools with lesser tool life uniformlyover the entire magazine rather than concentrating themnear the front. This approach is named Next Fit scheme.3) Best FitThe Best Fit approach is to search the entire magazinefor the tool with the smallest tool life which satisfies therequest. This approach tends to save the tools havinglarger tool lives until they are needed to satisfy a largerrequest.4) Worst FitIn this approach, the entire magazine is searched forthe tool with largest tool life that satisfies the request.The idea is that the tool life remaining after processingthe current request is large enough to process anotherrequest. However, this approach tends to generate largenumber of tools with very small tool lives that are in-adequate to satisfy most subsequent requests.4.2. Tool Replacement StrategiesA number of tool replacement schemes have been sug-gested in the past. Tool replacement has its own impor-tance in the field of manufacturing. If a tool is replacedtoo early, the remaining tool capacity is lost and too fre-quent changes take place. On the other hand, if a tool isreplaced too late, the probability of tool failure goes high.Since the proposed tool selection schemes are based onthe remaining tool life of the tool, the same factor will beconsidered for tool replacement also. Out of the availableschemes in the literature, the following have beenadapted in the present work:1) Single Tool ReplacementIn this replacement scheme, tools are continuouslymonitored and as and when a tool exceeds its warninglimit or fails due to some other reason, it is replaced.This results in full utilization of tool life but results inhigher down-time cost.2) Multi Tool ReplacementUnder this strategy, whenever a tool in the setup ex-ceeds its warning limit or becomes unusable due to fail-ure, all the tools of the setup are replaced. Such a Strat-egy may result in lower down-time cost but the tools arenot utilized to their capacity and hence tooling cost goesmuch higher.3) Dynamic Tool ReplacementUnder this scheme, whenever a tool in the setup ex-ceeds its warning limit or becomes unusable due to fail-ure, every other tool of the setup is evaluated for usefullife which it has lived. If this life exceeds the critical lifeof the tool, this tool is also replaced along with the toolwhich has failed.5. MethodologyA simulation model has been developed to investigateCopyright ©2011 SciRes. ICAD. G. RAO ET AL. 409the most appropriate tool replacement policy with severalto ol selection schemes. All the relevant informationabout the different tools such as, tool material, differentoperations that can be performed by the tool, maximumtool life, cost of the tool, cost of regrinding, warninglimit of tool etc. has been stored in a tool database. Also,the cutting parameters like depth of cut, cutting speed,feed, etc. for different comb inations of tool and work-piece material have been stored. The relevant data hasbeen collected from standard hand-books. Several datasets have been created and in each data set operations tobe performed, tool material, workpiece material and thetype of surface finish required are stored. In each simula-tion run, the user has to specify the data set and tool se-lection and replacement policy to be tried. For each op-eration, the software calculates the tool life required us-ing standard Taylor’s tool life equations taking feed,speed, and depth of cut into consideration. The tools areselected for the operation us ing the policy specified andthe relevant data stored. For all the cases the number ofregrindings done, total cost involved, and number oftimes the machine was down, are computed.1) Cost FunctionTotal Cost = Machining Cost + Down Time Cost+ Tool Cost + Setter CostTotal Cost = (MC + CI) + DT + TC+ (MTRA + MTSRA) * SRwhere:MC = Machining CostCI = Labour Rate + Supervision Charges+ Interest on investmentDT = Down TimeTC = Tool CostMTRA = Mean Time for tool replacement and ad-justmentMTSRA = Mean Time for setter arrivalSR = Setter Rate per HourThe following factors have been considered in thesimulation model:2) Catastrophic FailureA considerable percentage of tool inventory is com-monly lost due to sudden and unexpected failure or toolbreakage. Therefore this feature has been incorporated inthe simulation model, with the assumption that about 5%of the new tools and 15% of the reground tools fail dueto catastrophic failures.3) Tool RegrindingsTo prevent tool breakage use of excessively regroundtools is avoided. In the simulation model, the number ofregrindings of tools has been limited to a certain number,depending on the type of tool. Further, it is consideredthat after every regrinding, the tool life reduces by a cer-tain percentage. For different tools, this percentage varies.Also, the number of regrindings permissible for each toolis different.6. Computational ExperienceA realistic simulation model developed in the presentwork, evaluates the performance of identified tool selec-tion policies in combination with the different tool re-placement strategies. A large number of data sets consist-ing of the details of different types of tools have beengenerated randomly. The realis tic values of factors suchas remaining tool life of partially used tools, catastrophicfailure of the tools, number of times the tool has beenreground, reduction in tool life due to regrinding, appli-cability of tools for various operations, and operations-profile have also been generated and incorporated in themodel.In the model, the four tool selection strategies viz.First Fit, Next Fit, Best Fit and Worst Fit against eachtool replacement policy viz. Single Tool Replacement,Multi Tool Replacement and Dynamic Tool Replace-ment have been considered.For each data set, the extensive simulations have beencarried out and the total cost involved and the number oftimes the machine was down due to change of tool havebeen computed. The results have been tabulated in Ta-bles 1 and 2 . The results indicate the combination ofmost appropriate 'Replacement Policy’ with the appro-priate “selection policy” for obtaining minimum cost foreach data set. Also, the number of stoppages of machineis obtained for each combination of tool replacement andselection strategy. Figures 1-3 show a comparison ofresults in the form of bar charts clearly indicating thetrend for some of the data sets used in the simulation.7. ConclusionsThe number of stoppages of machine due to wearing orfailure of tool plays an important role in increasing thecost of production. In the present work, the tool selectionand replacement strategies have been identified. Thesimulation model developed selects the best combinationof these strategies for minimizing stoppages and the costof production. It takes into consideration the effect ofimportant parameters such as reduction of tool life due toregrinding, limited number of regrindings, a catastrophicfailure etc.From the results obtained, it can be concluded that thecombination of various selections policies and replace-ment policies affects the overall cost. The investigationsreveal that the total cost involved is lowest in the case ofDynamic Replacement Policy in tandem with the worstfit approach. However, considering the limitations of theCopyright ©2011 SciRes. ICAD. G. RAO ET AL.Copyright ©2011 SciRes. ICA410Table 1. Total cost with different strategies for each dataset.DATA SETSReplacement PolicySelectionPolicy I II III IV V VI VIIFirst Fit 12,126 5166 7790 7534 4162 3255 7894Next Fit 11,651 4791 8132 6829 4172 3631 8581Best Fit 13,246 8816 10,702 10,834 7472 5101 10,406 Series Tool Replacement Policy Worst Fit 11,501 4466 7432 6829 3451 2546 7111 First Fit 13,031 4861 9757 8269 3887 3406 8901Next Fit 12,691 5036 10,742 9724 4062 3756 9776Best Fit 13,911 9216 9727 11,084 9632 5231 12,236Parallel ToolReplacement PolicyWorst Fit 12,261 4531 8767 8369 3457 2546 8296First Fit 11,801 4831 7782 7534 3837 3240 7880Next Fit 11,551 4791 7762 6864 3882 3306 8166Best Fit 12,646 7866 9667 10,254 7987 4526 10,131Dynamic ToolReplacement PolicyWorst Fit 11,501 4466 7432 6829 3451 2546 7111Table 2. Number of times the machine was down.DATA SETSReplacement PolicySelectionPolicy I II III IV V VI VIIFirst Fit 8 3 4 4 2 2 4Next Fit 6 2 5 2 2 3 6Best Fit 11 13 12 13 11 7 12 Series ToolReplacement Policy Worst Fit 6 1 3 2 0 0 3First Fit 9 2 8 5 1 2 7Next Fit 7 2 8 6 1 2 7Best Fit 8 10 6 10 13 5 12Parallel ToolReplacement PolicyWorst Fit 7 1 6 5 0 0 5First Fit 7 2 4 4 1 2 5Next Fit 6 2 4 2 1 2 7Best Fit 9 10 9 11 12 5 11Dynamic ToolReplacement PolicyWorst Fit 6 1 3 2 0 0 3Figure 1. Tool cost on applying first fit tool selection policy with different tool replacement policies.Figure 2. Tool cost on applying dynamic tool repl acement policy with four tool selection policies.D. G. RAO ET AL. 411Figure 3. Total cost on appl ying three tool replacement policies with four tool selection policies on dataset 1.simulation model, these results cannot be directly gener-alized to any other data set, but the simulation can becarried out for other data sets also to select the best com-bination of selection and replacement policies.8. References[1] P. Tomek, “Tooling Strategies Related to FMS Manage-ment,” FMS Magazine, Vol. 5, 1986, pp. 102-107.[2] W. Eversheim, M. Lenhart and B. Katzy, “Information Mod-elling for Technology Oriented Tool Selection,” Annals ofthe CIRP , Vol. 43, No. 1, 1994, pp. 429-432.doi:10.1016/S0007-8506(07)62246-X[3] S. Pal, “An Expert System Approach for Scheduling andTool management in an FMS Environment,” UnpublishedM.Tech. Dissertation, I.I.T. Kanpur, Kanpur, 1991.。