A tool cluster based strategy for the management of cutting tools in flexible manufacturing systems.
In this paper we address the tool management problem and present a tool cluster based strategy for managing tools and work flow. The main requirement of the strategy is to determine those parts with: a similar tooling requirement; requiring the same manufacturing work station type/group and: which can be manufactured in one tooling setup. Each tool setup is managed as a tool cluster. The cluster based strategy is proposed for managing and scheduling these tool clusters subject to batch quantity, tool life, tool mangazine size and part pallet capacity. This method also has potential application in sizing tool stores and selecting machine-resident tooling configurations. The focus is essentially on illustrating how a simple cluster identification algorithm can potentially be employed for work and tool scheduling. In a static mode, a Rank Order Clustering (ROC) algorithm, with enhancement for clustering tools as opposed to machines, has shown to be an effective method for achieving economy in tool provision. The clustering algorithm, arbitrarily selected for simplicity, is that suggested by King (1980, 1982). The algorithm coupled with new priority management rules provides the dynamics for managing a tool management system. This method is non-optimal.
The clustering algorithm (ROC and a tooling algorithm) is embedded in a Lotus 1-2-3 spreadsheet. This inexpensive and widely available spreadsheet software gives attractive visual content coupled with its availability on a number of hardware platforms. Readily available links to commercial data bases (like Ingres from Relational Technology or Oracle from the Oracle Corporation) and other software platforms such as commercial and generic simulators, expert systems, provide a cheap but effective solution to tool management problems.
The difficulties encountered in managing the cutting tool resource in flexible manufacturing systems points to the need for strategies to deal with specific problems such as tool economy, work and tool assignment and tool issue. The ultimate goal in planning and executiing a tool management strategy is reducing the distance travelled by the tool transporter, minimizing the machine idle time, maximizing the equipment utilization and eliminating tool redundancy and duplication, see Bell and de Souza (1987).
A system may be designed to allocate as many tools as possible to the machine-based tool store. This would generate a large tool inventory as many tools would be duplicated across several machines. The machine-based tool store would also need to be sufficiently large to accommodate all the required tools. Tool exchanges are minimized and flexibility of processing is increased at each machine because of this large local tool inventory. An alternative strategy is to keep the majority of tools in a cell tool store and transfer them to the respective machines just when required. A standard set of tooling would be available at each machine. Transferred tools are either held at the machine or returned on completion of a set of tasks. This effectively minimizes the tool inventory but requires many more tool exchanges. These two strategies are at the extreme ends of a continuum of possibilities. In reality many companies start at one extreme and migrate with learning and added mechanization to a leaner inventory. However, none of these strategies can claim to have a significant advantage over the other. The point is that each has value and benefit in a particular mechanical hardware and control scenario.
The tool provisioning solution adopted or selected for flexible manufacturing systems reflects the method of manufacture and the intrinsic value of the work. The higher the added value contributed to the work the less one would desire to repeatedly use tools, and thus, the solution would be inclined towards less sharing of tools and vice versa. The solutions adopted range from tool kitting for a batch, to the bulk issue of tools, to magazine capacity in a workpiece-based system, to the issue of tool clusters for a component range in a tool-based system.
CURRENT APPROACHES TO THE TOOL MANAGEMENT PROBLEM
Zavanella, Maccarini, and Bugini (1990), Bard (1988), Tang and Denardo (1988a, 1988b) and de Souza (188) among others, have addressed the general tool management problem, particularly that of deciding on and maintaining a static tool configuration and even the issue of tool kits, through the use of mathematical techniques. These techniques to a large extent involve complex mathematical formulations. Unfortunately, such approaches while theoretically valid are less useful on a practical level, see Newman (1987). The Bard (1988) and Tang and Denardo (1988a, 1988b) models search for optimum solutions to simplified problems such as a single manufacturing work station in work flow-dominated environments and often at the expense of computing power and simplicity. These solutions do not adequately provide the basis for the short-range scheduling of work and tool flows required to achieve economic performance. Some workers have also focussed attention on a group technology approach to tool management (see Burbidge (1989); and Ventura, Chen, and Wu (1990)). The latter approaches, although significant contributions to tool clusteing, mainly confine themselves to generating and maintaining static tooling configurations. This paper goes further by providing the dynamics of managing the change of tool clusters as well as providing for the case where resident tool clusters are desired. Thus, this work is more oriented towards the leaner tool inventory end of the continuum described above.
The above researchers also, in the main, focus on advance schedule generation over a certain time period. The disadvantage of such an approach is that the more detailed the schedule, the less robust it is likely to be in response to production disturbances. It is argues in this paper that an effective solution is to dynamically manage tool clusters using an algorithm derived from the literature on cluster analysis. Dynamic scheduling, though, requires knowledge of the system status at a given moment, but suffers in that there is a possible lack of sufficient foresight in making optimal assignments of tool clusters and palletized part.
THE INDUSTRIAL SIGNIFICANCE OF A TOOL CLUSTER BASED STRATEGY
Many companies have the tools, in their manufacturing facility, driven by work flow (work-oriented) as opposed to hving their work driven by the tooling configuration on the machines (tool-oriented), see "Software Strategies for Cell Management" (Anon. (1990)). Tooling solution are often selected without thorough knowledge of the options available, or a framework within which such systems can be assessed. The desie for flexibility in manufacturing has encouraged a range of hardware solutions, some very novel but with little practical usefulness. This desire for flexibility in work assignment has led some companies towards a tool kitting tool management decision, but many still operate with a static bulk tooling configuration established in advance and refreshed with new tools at convenient intervals.
A tool cluster based strategy offers added flexibility and reduced tool handling costs that takes the concept of tool eyond just kitting for a set of identical parts to one of kitting for a family of parts. In other words, the strategy seeks to minimize tool inventory, and balance the costs of greater computer cell control against the costs of installing more mechanical hardware. Generally, the more sophisticated software tool control systems require less tool transfer hardware, smaller tool magazines and a smaller central tool pool. It is recognized that where tools are a relatively small part of the total cost of the FMS, minimizing the cost of tooling, through clustering, may not be a critical objective and neither may it make a significant difference. In comparison to a kitting strategy, a tool cluster strategy reduces the number of tool exchanges. A tool cluster based strategy would further eliminate tool duplication across a part family and thus, reduce investment in tooling. Setup times would be reduced to loading far less tools in a lesser number of tool exchanges than would be possible in tool kitting in a workpiece-based strategy. The significant difference in work flow under a cluster based strategy is the added flexibility in routing and the greater local processing capability afforded by possession of a tool cluster rather than a kit.
It is helpful, at this stage, to regard a flexible manufacturing system as an activity area for work and tool flow involving a recirculating element of tooling and a proportion of worn (Figure 1). Assuming that all work stations are homogeneous, the use of a tool cluster based management strategy would allow each activity area to be configured or reconfigured in each manufacturing period into a number of "sub-cells." Each sub-cell being decided on has possession of one or more tool clusters. The possession of a tool cluster by a work station, in a sub-cell, activates manufacture of a family of parts within a tool cluster, in any sequence, and usually all in one setup.
Both tool kitting and the static tool configuration rely to a large extent on robust schedule generation, usually in advance of production. The techniques for generating this schedule are invariably based solely on work flow considerations with little attention given to tool economy. Thus, when tooling constraints complicate the scheduling it becomes necessary to re-engineer the schedule by combining or eliminating activities. This is normally necessary to avoid violation of the magazine capacity. Subsequently, this will lead to less efficient utilization of the tools, because the tendency is to defensively opt for larger tool magazines to minimize production disturbance. A secondary issue which becomes increasingly important in these circumstances is tool life or wear, see Bard (1988) and de Souza (1988). Tool life does not in itself influence a schedule but is essential when larger quantities of work require to be allocated to work stations to economize on work setups. It thus becomes apparent that magazine capacity plays a larger economic role than was originally conveived in the design of many FMS installations. One solution suggested by Tang and Denardo (1988a, 1988b) is to manipulate the sequence in which work is processed to economize on tool changeovers within the bounds of magazine capacity. All of these techniques rely on complex mathematical calculations to preschedule work to the machining stations. Tool kitting, as mentioned above, has been suggested by de Souza (1988) and Graver and McGinnis (1989) as a means whereby the decision may be partly shifted from advance schedule generation to a reactive role in manufacturing. The authors believe that a cluster based strategy advances beyond kitting to provide rapid on-line generation of tooling configurations where part mix and production ratios may be varied almost instantaneously on a spreadsheet and where rigidity in scheduling may be relaxed by allowing a tool cluster to attract any part in the cluster set. Thus, work batching and tool assignment are no longer as significant because a cluster enables a "random" batch (even a single part or part of a batch) to be machined just as long as this batch (or part) is a member of the tool cluster.
Finally, one can address the operational problems that a clustering strategy may pose. These usually center on tool life tracking and determination of current cell status. These issues are now readily resolved with the advent of embedded chip technology and developments in shop floors control.
A TOOL CLUSTER BASED TOOL MANAGEMENT STRATEGY
A tool clustering strategy is employed to achieve greater flexibility, reduce tooling costs, minimize tool setups and simplify the tool management. This strategy has evolved from research in an perceived views of industrial tool management practices, most notably in the Makino Max flexible manufacturing system described in The FMS Magazine (Anon. (1985)) and Makino (1985).
A too cluster based management system assesses the work schedule for a production period to find tool clusters and subsets of parts which can be manufactured with the same tooling setup. Operation of the system is centered on tool cluster scheduling. The tool scheduling system employs statistical cluster analysis to predetermine tool cluster configurations based on anticipated production volumes, work mix and tool wear. Tool cluster analysis considers each operation or part type as a work string which has an associated tool set required for manufacture of its sub-operations. Each work string is a palletized load which undergoes loading, processing and unloading. A collection of work strings forms a batch quantity. Cluster analysis attempts to sort these work strings such that sets of tools aggregate or cluster to generate tool clusters. A tool cluster provides each manufacturing work station with the flexibility to produce any work string included within its part set envelope in the same setup. This achieves savings in tool-changeover and tool-setup. It results in simple and efficient tool management. Tool clusters are changed to process another part set. As a part set may encompass a range of parts, complete (not partial) tool cluster change is essential if control is not to get too complex. The tool cluster content is only altered when changes are made to the part set or production ratios. An unchanged tool cluster on a work station, other than changing worn tools, requires less control effort and is afforded the title of "resident tool cluster." Tool handling requirements of a tool-based strategy are thus, at best, reduced to the marginal difference between tool cluster, but usually to servicing a resident tool cluster or to changing a whole tool cluster. The configuration of all tool stores is this managed on the basis of tool clusters and part sets and not as kits or single unrelated tools in a static magazine configuration (Figure 2).
MANAGING TOOL CLUSTER FLOW
The management of tool clusters is based on priorities (Figure 3). Each tool cluster is assigned a priority (the secondary priority) equal to the highest priority among all the work string priorities (the primary priorities) in the tool cluster envelope. The primary priority may be based on due dates or pallet priorities or a factor indicating progress towards a production target. In this discussion it is assumed that the higher the priority number the more urgent is the work string. The primary priority is used to schedule work strings on the respective tool clusters.
The secondary or tool cluster priority is used to schedule tool clusters to the manufacturing work stations. The cluster with the highest secondary priority number is scheduled to the first available and suitable manufacturing work station. Once a tool cluster is loaded, work strings of the current tool cluster can be sequenced to the manufacturing work station in either of two ways. Firstly, work strings may be assigned based on descending primary priority values; this permits all the highest priority work string to be machined first. Secondy, for greater flexibility available work strings may be assigned irrespective of their primary priorities.
In either mode of work string assignment, each cluster's secondary priority is altered with each completed work string. This is because each cluster's priority is derived from the highest unmachined work string primary priority. The assignment of work strings becomes more complex when tool clusters do not only have work strings without contiguous primary priorities but where the primary priority values are interspersed among several clusters. As clusters are assigned to work stations based on their secondary priorities, a possible consequence is that another tool cluster (say, cluster B) may, through a higher primary priority work string than that in the current cluster (say, cluster A), demand access to this manufacturing work station if no other work station is available. That is, cluster A's secondary priority is now lower than that of cluster B. The situation may be further complicated, given cluster B is on the work station, and after completion of say, just one work string, tool cluster A now possesses the next highest primary priority. This would thus signal another tool cluster change back to cluster A. This will result in clusters possibly oscillating between tool store and the work station. To resolve this conflict the concept of a termination criterion has been introduced.
A termination criterion minimizes or eliminates the possibility of unnecessary tool cluster changeover before the tool cluster has been effectively utilized. The termination criterion operates by enforcing continued residence of the current tool cluster, by overriding the secondary priority, until the terminating condition is satisfied. Then, and only then, would the next candidate tool cluster be loaded. The progress of work through the cell is thus effectively managed. A termination criterion would need to be set for each tool cluster. The terminating condition could either be based on completion of a given number of work strings, or on expiration of a specified time (e.g., at the end of a shift or schedule). The choice of terminating condition depends on the manufacturing facility. Some manufacturing facilities may desire frequent tool cluster changes, like the MAX Cell in Makino (1985), whereas others may opt to hold a tool cluster in residence on a work station for a considerable time, weeks in the case of the Yamazaki system described by Kurimoto (1989). In the latter case, though, one could expect to service the cluster with fresh tools.
Thus, the choice of whether to adhere to primary priority order (even where contiguous priorities do not exist in a cluster) or follow a termination criterion is dependent upon the nature of production (that is, whether many parts in different clusters have the same priority or whether random presentation of work is desirable). In selecting the mode of priority management a significant consideration would, therefore, be the balance between tool cluster changeover time against work string cluster processing time, and tool cluster setup against desired work flow flexibility.
The use of the tool cluster based strategy thus creates the environment for controlled work and tool flow. Furthermore, by dynamically adjusting or assigning primary priorities not only can urgent tasks be pushed through the manufacturing system, but also those tasks required by upstream processes can be pulled. Increasing the termination time of each tool cluster will reduce the number of tool cluster changes required, thus resulting in higher work station utilizations. It should be noted that management of the system based on tool clusters requires the work station tool store capacity to be sufficient to accommodate the largest of the tool clusters, and a tool transfer system beneficially geared up for their transportation.
TOOL CLUSTER COMBINATION, WORK STATION ASSIGNMENT
AND TOOL ECONOMY
Three important factors in a tool cluster based management system are: cluster combination, work station assignment and tool economy. Clusters may be combined to form larger tool cluster sets. This is obviously constrained by tool magazine capacity. This cluster aggregation is not always straightforward. Cluster summing must aim to remove the duplicate tool types introduced when determining clusters. Furthermore, tool life factors for the tools in each cluster to be summed do not necessarily apply when tool life is calculated for the larger combined tool cluster. This is so because the summed residual tool life of each tool in each of the smaller individual clusters may now be cumulatively sufficient to enable further machining.
The assignment of clusters to manufacturing work stations is based on either tool replenishment or cluster changeover. In order to provide greater workpiece routing flexibility when a manufacturing work station breaks down, it may be necessary to add or duplicate a cluster set at another work station and thereby eliminate or minimize production delay. In this case the implications of cluster tooling combination have to be considered. This would include analysis of the constituent tools to maintain tool economy. Each cluster consists of the tool types and their duplicate or sister tools. It is the latter which has to be addressed when considering cluster combinations beyond the initial cluster formation stage. It may be preferable to develop production contingency plans if on-line clustering is not possible.
Tool economy through combination of clusters has two advantages. Firstly, it allows manufacturing to continue with the overlap of tools, if any between clusters thus minimizing the downtime occasioned by tool cluster sets changeover. Secondly, magazine capacity permitting, if all clusters can be accommodated as one merged entity within a single magazine, flexibility is enhanced through having the ability to manufacture a greater range of work strings. Tool changeover in this latter case is then reduced to tool replenishment. This greatly simplifies the tool management when the cell operates with a tool cluster in prolonged residence on the work station.
THE COMPUTER ASSISTED TOOL CLUSTERING MODEL
The tool clustering model proposed here consists of a spreadsheet-based cluster model supported by a relational data base. The spreadsheet model specifies tool-work string groups (i.e. tool clusters, some or all of which may be independent), determines the minimum tool cluster configuration for a given order requirement, and suggests tool clusters and their assignment. The spreadsheet model provides input to the tool cluster simulation described below.
The tool cluster algorithm embedded in the spreadsheet is based on Rank Order Clustering (ROC), in King (1980), used for work station-work string group formation in production flow analysis. The ROC algorithm was selected primarily because it lent itself to easy embedding within a spreadsheet. Secondly, it was thought to be the simplest to employ for development of a useful and comprehensible shop floor tool. Using a spreadsheet was considered a widely acceptable platform which permitted ready manual involvement and rapid experimentation. ROC is a simple method of analysis that rapidly reveals tool cluster possibilities. A host of other techniques, reviewed in King and Nakornchai (1982) are available, and could have been used, ranging from similarity coefficient methods to analytical techniques. It is not the intention in this paper to compare the choices available but to illustrate how a tool cluster identification algorithm can be used to schedule work on tool clusters. A discussion of the ROC method may be found in King (1980).
The Rank Order Clustering algorithm has been adapted to focus on the commonality of tools as the criteria for tool cluster formation. The philosophy behind this application is essentially that in any manufacturing work schedule there will be quite a natural division of tools into tool clusters and parts into sets: what the model seeks to achieve is the identification of such sets or clusters. If a tool-part matrix is not capable of division into mutually exclusive groups, then one has to manually suppress those elements which cause the overlap until mutually exclusive tool clusters emerge. The suppressed tools are manually reinstated in the final matrix (see Figure 6). Hence, if additional tools of the same type are added for each case of overlap, then, it will always be possible to find a satisfactory solution. A relaxation procedure is suggested by King (1980) to determine the number of duplicated tool types required as well as their disposition in the block diagonal structure produced. This procedure may greatly increase the dimension of the matrix and is not employed in this work. King and Nakornchai (1982, 1986) provide ROC2 for larger problems. This was not considered necessary in this application as the spreadsheet already had the capability to handle relatively large problems.
Tool Cluster Formation
The clustering commences with the building up of a two dimensional array of work strings and tool types, the basic tools (Figure 4). The work string could represent either a whole parts manufacturing content or a single operation. This information is readily derived from the processing data of each part stored in the data base. Each operation may in itself be subdivided into tool activities or sub-operations. The sub-operation sequence of each part or operation is neither considered nor necessary in the formation of tool clusters. The algorithm can in its simplest form, be expressed as determining by a process of row and column exchanges of the array, a conversion from a haphazard pattern of entries into an arrangement whereby the entries are contained in mutually exclusive groups. The final solution yields tool clusters (Figured 5). It is then necessary to determine the number of tools to be contained within each tool cluster to service a given order requirement. For this purpose, a duplicate-tool specification algorithm was added to the ROC algorithm. This is described in the next section.
Tool Cluster Sizing
The model replaces each entry in a tool cluster with an accumulated tool usage time on a particular work string (Figure 7). The latter is the product of the order requirement specified for the manufacturing period and the processing time for the string. These times are summed for each tool type in each tool cluster to obtain a cumulative tool use time (Figure 8). The data base entries for initial tool life and percentage (which may be varied to account for certainty of the tool life value) gives rise to an available tool life. This life against the required cumulative tool use time will determine the least number of tools required for this tool type. This is in turn translated into a number of duplicate (sister) tools required within a particular tool cluster (Figure 8). The summation of the tool types and the sister tools will yield the minimum number of tools in each tool cluster.
Sizing a Tool Magazine
The minimum size of a tool magazine under a tool cluster based strategy is indicated by the size of the largest tool cluster (based on tool types; duplicated tools disregarded). This must be so since all tools for the work strings must be accommodated. The size of local tool storage is often a major factor in deciding work station type and its cost or the cost of maintaining local tool provision. It is advisable to allow some added capacity for duplicate tooling to accommodate for production volume and tool wear.
Tool cluster 1 (in Figure 7) has 41 tool types (indicated by the first column). The required production volune specified in order quantities (indicated by the top row) are factored by the processing time required of each tool to determine the total number of tools required. It is thus, possible to size the magazine based on a tool type cluster size with added percentage for flexibility. The model in this instance provides an inexpensive and rapid decision tool when deciding on machine tool purchase, design and tool storage capacity.
Sizing a Cell Tool Store
The summation of all tool clusters, as suggested above, can indicate the total cell tool requirement for a selected order mix and production requirement. Secondly, manipulation of a user-input safe limit on tool usage and/or tool life value specification rapidly yields the sensitivity of the specified values not only to tool cluster size and hence magazine capacity but also on the number of captive tools in the cell and the cost of holding this inventory. Changes in process times are also able to be rapidly entered and their effects examined. The clustering process, however carried out, is only required to be activated if new parts are to be added or deleted or new schedules issued. For the rest of the time it lies dormant and allows other parameters such as production ratios to be examined for effects on cluster size.
If one suppresses the tool cluster boundaries imposed by clustering and considers the case of one large two-dimensional matrix (by merging the tool cluster formations shown in Figure 7), the model could be used dynamically to configure a common cell tool pool. This latter inventory could then be tested for size sensitivity by manipulation of the order quantity, tool life and percentage tool life availability parameters (Figure 8). A change in order quantity would ripple through the matrix effecting changes in processing demands on each tool thus generating a new tool requirement. A change in tool life for an individual or a group of tools could also be examined for its effect on size of inventory. However, a change is too life would exert no influence on inventory if the processing demands are satisfied by available tool life.
One could further impose a "percentage of tool life" to be made available to accommodate manufacturing learning on tool breakage. Thus instead of changing tool life one oculd specify a percentage life to be made available. Manipulation of this parameter for certain or even all tools could be examined for its effect on total tool inventory and cost.
Of course, all of these parameters could jointly be adjusted. The spreadsheet capability excels in this what-if mode. Finally, one can determine the quantity of duplicate tools that are required and be accommodated in a respective tool magazine if the tool cluster boundaries are restored.
Dynamically Assessing Cell Performance
Work has begun on embedding the tool cluster strategy and the priority management rules in a simulator. Prior to this decision no facility was available to act as a test bed for the evalaution of this strategy. The progress to simulation has provided the basis for current research in establishing station utilizations, throughput measures and cost comparisons between alternative strategies. The heuristic procedure for managing tool cluster flow has been explicitly embedded in a SIMAN simulation model, see Pegden et al. (1990). The simulation takes its prioritized tool cluster and work string input from the spreadsheet model and emulates the flow of tool clusters in tune with the control rules specified above. Each tool store (cell and work station) is modelled as a station. Tool clusters are modelled as queues of work strings ranked by their primary and secondary priorities. All the other elements one would expect in an automated flexible manufacturing installation, such as guided vehicles, buffers, etc., are included. The data-driven simulation permits the user to experiment with several scenarios and output detailed results on manufacturing performance and utilizations. Each simulation can be viewed and analyzed through the powerful animation capability of the CINEMA module. The results of this work are to be reported in a subsequent paper.
The following description is based on a flexible manufacturing installation already implemented. Because the cell is already commissioned, specific details may not be divulged. The cell is used in the above analysis to illustrate, from an academic viewpoint, the options available to managing such a system using a cluster based strategy. This real data is employed to illustrate how the total number of tools required at each machine can be guaranteed; when and how many tools are required to be exchanged and how added flexibility in part assignment can be achieved.
The cell comprises three identical machining centers, each of 120-tool capacity, serviced by a rail guided pallet transporter. The schematic diagram of the cell, with possible tool flow automation added, is shown in Figure 9. The cell can accommodate up to 22 pallets. Tool transfer, loading and unloading are performed manually, but automation is anticipated of these functions. The production strategy is such that a manufactured part (a complete set of work strings) is obtained after a visit to a single manufacturing work station. The cell produces eight part types (in a total quantity across eight parts of 40 per day). The average number of operations per part is 25.
Using a static bulk tooling configuration, for a pre-scheduled "work-to" list, it was found that a maximum of 58 tools were required on any one work station over 24 hours. With the available tool magazine capacity of 120 tools, no tool changeover was evident as the magazine capacity of 120 tools, no tool changeover was evident as the magazine capacity was sufficient to accommodate all of the tools at the start of the production run (enough for at least three days). But a substantial time was involved in loading the magazine with tools. Thus, the time before machining could commence was considerable. The total tool requirement across the three machining stations grossed 145 tools for the static configurations. Employing the tool cluster based strategy for the eight-part types generated gross tool requirement of 50, 39, 49 (totalling 138) for the stations (see Figure 7). This amounts to a saving of just seven tools. Both strategies, however, immediately demonstrated the over-capacity in the tool magazines, verifying the belief that companies veer towards over-specification of magazine capacity.
Where a clustering strategy excels is in the elimination of robust advance schedule generation, required for determination of the bulk tooling configuration, and in on-line flexibility in choice of work station. In advance schedule generation, parameters like schedule period, tooling requirement, and work and tooling assignment have to remain relatively rigid. In the cluster based strategy, all these parameters may be rapidly changed and dynamically re-clustered to provide new schedules either in reaction to a disturbance or in response to a rolling schedule, different part mix, changed production ratios or different tool life specifications, as described in section on "The Computer Assisted Tool Clustering Model."
The issue of routing flexibility is paramount in deciding on a strategy to implement. The static configuration strategy is inflexible in part assignment and sequencing. This is because the strategy is invariably based on pre-scheduling of work flow. The total cluster based strategy is much more flexible and well suited to changing environments as it is relatively insensitive to the work string sequence.
The company required that a part should be able to be assigned to either of any two out of the three manufacturing work stations. The part assignments arbitrarily decided on by the company were that parts 1 and 2 should be able to visit either work station 2 or 3; parts 3 and 4 to visit either work station 1 or 2; and parts 5 to 8 to visit either work station 1 or 3. This made it necessary to consider cluster assignments and a magazine capacity for holding sufficient tools.
The cluster based strategy, see Figure 7, could satisfy the company part assignment requirement by permitting cluster 1 to be allocated to work stations 2 and/or 3; and clusters 2 and 3 to work stations 1 and/or 3. This meant that if one only considered sole cluster residency on a work station and that tool change is manually supported, then work station 1 would require a tool magazine capacity of 35 (the size of cluster 3 with only basic tools); work station 2 would require a tool magazine capacity of 41 (the size of cluster 1); and work station 3 a magazine capacity of 41 (the size of the largest of the three basic-tool clusters). If the company desired no cluster changeover, it becomes clear that only work station 3 requires a tool magazine capacity of 104 (the size of all the basic-tool clusters). However, this is an upper bound since some basic tools might have been duplicated in the clustering process.
It was evident that due to a lack of proper tool management and control the company had opted for the 120-tool magazine to cover all the scheduling possibilities. The 104-tool requirement of the three merged clusters provides this evidence. The cluster based strategy suggests that through efficient tool organization and management, the company can still retain flexibility in assignment (while also achieve a reduction in tool inventory through a saving of 243 tool magazine slots (i.e. [120 tool slots x 3 work stations] - [35 + 41 + 41 required tool slots] = 243).
This paper has discussed the requirements of a tool management system and presented a case for a tool cluster based strategy for managing work and tool flow for short range schedules. A comparatively simple method of grouping tools into clusters was described and it was shown how these tool clusters can be priority managed as resident or flexible work station tool configurations. The cluster based strategy provides a method that occurs in direct response to the prevailing manufacturing conditions using priority rules to guide the decision.
A tool cluster based strategy, firstly, rapid determines the copies of tools required based on a changing product mix, production ratios and tool life. Secondly, it provides a strategy for operation of a tool management system based on fewer tool setups and static or dynamic configurations of cell and work station tool stores. Finally, it provides a means whereby better economic performance may be achieved by negating the impact of tool magazine capacity, providing for flexibility in routing of work strings, reducing tool commitment per schedule by greater sharing, eliminating the need for robust advance scheduling and by establishing relatively smaller tool magazine capacities. To ocnclude, the paper illustrates the application of a common clustering technique to an area not previously explored in detail and provides a direction for further research effort and wider exploration of the issues in tool management. Research is currently directed at simulation models to determine how to set the cluster termination criterion, how often new schedules should be issued and re-clustering occur, and to investigate how the clustering strategy performs in relation to other tool management strategies.
Anon. "Software Strategies for Cell Management." Machinery and Production Engineering, September 7, 1990, 32-35.
Anon. "MAX Combines Flexibility and Productivity." The FMS Magazine, January 1985, vol. 3, no. 1, pp. 26-28.
Bard, J.F. "A Heuristic for Minimizing the Number of Tool Switches on a Flexible Machine." IIE Transactions, vol. 20, no. 4, December 1988, 382-391.
Bell, R., and R.B.R. de Souza. "The Management of Tool Flows in Highly Automated Flexible Machining Installations." In Proceedings of the Second International Conference on Computer Aided Production Engineering, Edinburgh, April 1987, 233-238.
Burbidge, J.L. Production Flow Analysis for Planning Group Technology. Oxford Series on Advanced Manufacturing, J. R. Crookall and M. C. Shaw (eds.) Oxford: Clarendon Press, 1989.
de Souza, R.B.R. The Management of Tool Flow in Highly Automated Batch Manufacturing Systems. Ph.D. Thesis, Loughborough University of Technology, Loughborough, England, 1988.
Graver, T.W., and L.F. McGinnis. "A Tool Provisioning Problem in an FMS "International Journal of FMS vol. 1, 1989, 239-254.
King, J.R. "Machine-Component Grouping in Production Flow Analysis: An Approach Using a Rank Order Clustering Algorithm." International Journal of Production Research, vol. 18, no. 2, 1980, 213-232.
King, J.R. and V. Nakornchai. "An Interactive Data-Clustering Algorithm." Flexible Manufacturing Systems: Methods and Studies, A. Kusiak (ed.) Essex, UK: Elsevier Science Publishers, 1986, 285-300.
King, J.R., and V. Nakornchai. "Machine-Component Group Formation in Group Technology: Review and Extension." International Journal of Production Research, vol. 20, no. 2, 1982, 117-133.
Kurimoto, A. Yamazaki Machinery, UK Ltd. Personal communication on "Total Quality Management in an Advanced Manufacturing Environment," 1989.
Makino Milling Machine Co., Japan. Personal communication, 1985.
Newman, P.A. "Scheduling in CIM Systems." Artificial Intelligence: Computer Integrated Manufacturer, A. Kusiak (ed.) IFS Publications, 1987.
Pegden, C.D., R.E. Shannon, and R.P. Sadowski. Introduction to Simulation Using SIMAN. Pennsylvania, Systems Modelling Corporation. New York: McGraw Hill, 1990.
Tang, C.S., and E.V.Denardo. "Models Arising from a Flexible Manufacturing Machine, Part I: Minimization of the Number of Tool Switches." Operations Research, vol. 36, no. 5, September-October 1988a, 767-776.
Tang, C.S., and E.V. Denardo. "Models Arising from a Flexible Manufacturing Machine, Part II: Minimization of the Number of Switching Instants." Operations Research, vol. 36, no. 5, September-October 1988b, 778-784.
Ventura, J.A., FF. Chen, and C-H Wu. "Grouping Parts and Tools in Flexible Manufacturing Systems Production Planning." International Journal of Production Research, vol. 28, no. 6, 1990, 1039-1056.
Zavenella, L., G.C. Maccarini, and A. Bugini. "FMS Tool Supply in a Stochastic Environment: Strategies and Related Reliabilities." International Journal of Machine Tools Design and Manufacture, vol. 30, no. 3, 1990, 389-402.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Special Issue on Group Technology and Cellular Manufacturing|
|Author:||de Souza, R.B.R.; Bell, R.|
|Publication:||Journal of Operations Management|
|Date:||Jan 1, 1991|
|Previous Article:||A comparative evaluation of nine well-known algorithms for solving the cell formation problem in group technology.|
|Next Article:||Countering the negative impact of intercell flow in cellular manufacturing.|