Design and planning for cellular manufacturing: Application of neural networks and advanced search techniques.

A cellular manufacturing system (CMS) is a manufacturing structure organized based on the group technology (GT) concept. The main advantages of the CMSs include the low material handling costs, short setup times and reduced work in process. This study addresses the machine/part grouping and group scheduling (i.e., part/part family scheduling) problems, the two key issues in the CMS design and planning. The machine/part grouping problems can be classified into binary and comprehensive grouping problems depending on whether or not the processing times and the machine capacities are considered. The binary grouping problem arises if the part demands are unknown when the CMS is being developed. If the part demand can be forecast accurately, both the processing times and machine capacities have to be included in the analysis. This gives rise to comprehensive grouping. Both the binary and comprehensive grouping have been proved to be NP-complete problems which cannot be solved in polynomial time. Considering the large number of parts and machines involved in the industrial design problem, efficient solution methods are highly desirable. In this study, a novel neural network structure, Ortho-Synapse Hopfield Network (OSHN), has been designed to solve the binary grouping problem. Due to its significantly reduced number of synapses and unique structure, the OSHN is very computationally efficient and training-free. An objective-guided search approach has been developed to lead the OSHN search process to tune the network parameters and escape the local optima. To solve the comprehensive grouping problem, two approaches are proposed. The first one is a simulated annealing (SA) method based on a generalized grouping efficiency index. The SA method is used jointly with the OSHN algorithm to improve the computational efficiency. The second method is a modified OSHN algorithm. The objective of the modified OSHN is to maximize the generalized grouping efficiency subject to machine capacities. Our computational results compare favorably with solutions obtained in the literature. The group scheduling problem has also been proven to be NP-hard. Furthermore, due to the limited time available for scheduling decisions, computational efficiency is more critical. To this end, a combined tabu search/simulated annealing (tabu-SA) approach is developed to solve the group scheduling problem. The main advantage of this approach is that the simulated annealing search can be accompanied by a short term memory to avoid cycling and thus improve solution quality and computational efficiency. This has been tested and demonstrated in our computational experience