Expert self-organizing neural network for the design of cellular manufacturing systems

Abstract

A pragmatic approach to the design of cellular manufacturing systems is often driven by multiple constraints and objectives. Use of neural networks has provided an opportunity to perform computationally intensive, multiconstraint tasks more efficiently. This paper presents a multilayered hybrid neural network to incorporate some of these constraints and objectives for practicality in simulating cellular manufacturing system design. The neural network, which is constraint-bound, is structured to include practical limitations such as duplicate machine availability and machine capacity during the cell design process. The expert system, which is interactive, takes its input from the neural network and uses alternate process plans to reassign any exceptional parts that may occur as a result of the constraint imposition during the initial cell design. Thus the hybrid neural net-expert system technique gives an added flexibility to the design approach by facilitating the incorporation of multiple constraints and objectives. The efficacy of the approach has been demonstrated with an example.