AN EVALUATION OF ART1 NEURAL MODELS FOR GT PART FAMILY AND MACHINE CELL FORMING

This paper describes ART1 neural models for GT part family and machine cell forming. An ART1 neural model was first implemented in C and was tested with examples taken from the literature. The ART1 model was th en integrated with a feature-based design system for automatic GT codi ng and part family forming. It was finally incorporated into a three-s tage procedure for designing cellular manufacturing systems. Our evalu ation concludes that ART1, when compared with nonlearning algorithms, is best suited for GT applications due to its fast processing speed, f ault tolerance and learning abilities, ease of classifying new parts, etc.