SOME PERFORMANCE ISSUES FOR A RECONFIGURABLE MANUFACTURING SYSTEM

Abstract

The modern manufacturing era is characterized by compelling motives to cope up with short opportunity for introduction of new products in conjunction with large fluctuations in product demand and mix. These drivers required a new manufacturing paradigm which must be well equipped to react to changes rapidly and in a cost effective manner. The inability of traditional manufacturing systems to deal with the present manufacturing arena has played a pivotal role for invention and implementation of new class of manufacturing system which possess high potential to respond quickly and economically to the ever changing products and process technologies as well as stochastic and uncertain markets. A Reconfigurable Manufacturing System (RMS) is considered to be a responsive system whose production capacity is adjustable to market fluctuations and whose functionality is adaptable to a variety of new products. The design of RMS evolves over a period of time in view of customer needs and market demands. In today’s manufacturing world, RMS is being recognized as a system for increasing productivity and profits despite of abrupt fluctuations in product volume and design and also the changes occurring in the process technologies in the global market. However, in spite of recognizing all the advantages associated with these modern systems, yet the practical implementation of RMS on shop floor has not seen the light of the day due to some concern issues pertaining mostly to costs, performance, design and operation. Efforts are still required to take up issues related to RMSs so that they can be economically and successfully implemented and subsequently one can achieve the long term advantages associated with these systems. Such efforts will not only demonstrate and justify the promises made by RMS but would also aid to gradually elevate the status of RMSs from its emerging conceptual stage to a mature stage with high acceptability for its implementation in the industry. In the present work some important performance issues pertaining to design and operation of RMS have been taken up which may facilitate and Abstract (v) motivate the modern industries to implement this state of the art technology to the shop floor level. The highlights of the research work carried out in this doctoral thesis are summarized below. In the first phase, a mathematical model based on probabilistic system states is proposed for production rate or throughput for a single part reconfigurable flow line with RMTs across stages of the production line. The developed model gives exact results for throughput of the system. Though, the proposed model is applicable not only to RMS and it can be extended to other traditional serial production systems. But such implementation is much easier and in RMS environment comprising of modular RMTs, the structure of which may be manipulated for varied capacity and functionality. For the development of mathematical model certain machine states are defined and the combination of these machine states decides the system states. These systems states are then classified as feasible and infeasible states depending upon the assumptions made. Steady state probabilistic working equations are formulated and are subsequently solved for throughput values. The developed model is then further extended to accommodate intermediate buffer spaces in between the stations. For that, the code used for representing the system states has been modified to hold the buffer capacities of some finite sizes in between the stations or stages. Also, in this case steady state probabilistic equations for feasible system states are formulated and subsequently solved to obtain the throughput values. In the next phase, an unbalancing strategy is suggested as a viable alternative for limited scalability of these systems. The unbalancing here simply means allocating higher work contents to the outer work stations or stages and a relative decreasing work contents towards the middle most work stations. In other words, allocating the mean processing or operation times at stations in such a manner that would gave a bowl shaped kind of time distribution across the flow line. The scalability philosophy is based on the fact that productivity alterations can be obtained by simply varying the shape distribution of processing time at stages without altering the total work content. Through, limited scalability is obtained using this strategy but the technique is unique in the sense as it Performance Issues for RMS (vi) does not require addition of duplicate resources for its implementation. It only requires an intelligent manipulation of RMTs for mean operation times to have bowl shaped mean processing time distribution across the line. The extent of scalability depends upon the optimum and critical degrees of imbalances. The proposed strategy is demonstrated by designing a three station serial single product reconfigurable flow line. Another vital area of concern is the part family formation, which is considered to be the central focus behind the design and operation of these systems. An effective and intelligent part family formation would definitely improve the performance of these systems. The earlier techniques and algorithms for part family formation were generally based on the classical concept of similarity indices. Considering part family formation as an important issue related to RMS, a methodology based on feature extraction and subsequent data classification for grouping of parts into families is proposed and demonstrated. The developed technique may prove to be advantageous over the earlier developed coding and classification techniques in terms of computational time and effort. An artificial neural network modeling is employed for feature extraction and subsequent grouping of the data related to part-machine relationships. The classification may take into account the data which might already exists on the shop floor for training the neural network. Further, if no such database exists, the neural network modeling may be clubbed with any of the already developed part family formation methodology like rank order clustering algorithm etc. Once a trained network is available, it can be utilized to group upcoming parts efficiently to the classified classes obtained by neural network. The technique is demonstrated with the help of a numerical illustration. The performance of RMS is greatly influenced by the selection of RMTs for carrying out the desired operations. Selection of reconfigurable machines with the performance indicators like high reliability and high operational capacity with lower costs would improve the overall performance of the RMS. But in majority of the situations these multiple performance characteristics associated with reconfigurable machines may conflict each other. Selecting machine configurations under these multiple performance measures is also crucial for RMS design and operation. For this objective a Genetic Abstract (vii) Algorithm (GA) based optimization technique is employed to optimally select the machine configurations for each stage of a serial single product reconfigurable flow line. The GA coding is based on real chromosome coding with mutations and crossovers. Optimal machine assignments based on fitness values are obtained after the desired number of generations are achieved. The proposed GA based methodology is explained using a numerical example. In the last phase, the RMS philosophy is extended to cater multiple part families. Most of the past studies done on optimum configuration selection problems are based on single part family RMSs. The selection policy for the part family is an action rule by which the manufacturer selects a family when dealing with multiple part families. This action rule is directly governed by the various types of cost associated with production and reconfiguration process. A brute force algorithm is utilized to determine the optimal configuration for RMS involving multiple part families based on an objective criterion formulated by incorporating the production and reconfigurations costs. Also, the sequence of executions of orders is important as this consideration may have a pronounced effect on the economic profits or benefits earned during the operation of RMSs. Thus, optimal part family sequences have been worked out for the configurations based on maximum benefit earned and the same has been demonstrated using a numerical example.