A multi-constraint neural network for the pragmatic design of cellular manufacturing systems

Multiple constraints and conflicting objectives govern the design of cellular manufacturing systems (CMS). The first step towards the design of a CMS is the development of an initial cell design which has evolved as a result of the consideration of a number of practical constraints. This paper presents a multi-layered neural network that can configure alternate cell designs by considering multiple constraints and objectives. These constraints and objectives are embedded within the network as transfer functions which help impose the practical constraints and guide the cell design process. Application of the approach to the configuration of a cellular manufacturing system shows how the neural network can generate alternate user-specified cell configurations which can be constrained by specifying, the availability of duplicate machines or a limit on machine capacities. The technique presented provides a quantitative basis for making important capital investment decisions while configuring cellular manufacturing systems.