Connectionist models for part-family classifications

Abstract This paper presents a neutral network approach to clustering and classification of parts into families, as applied to Group Technology principles. Kohonen's self-organizing feature maps have been used for clustering parts into families and the part-family associations thereby obtained are fed as training inputs to a simple feedforward back-propagation network. This network has been seen to model the part-family relationships well and is also capable of accurate family prediction for test parts that were not used for training. The methodology of clustering followed by classification and generalization is not problem specific, and can be applied to other problems in the fields of dynamical system modeling, recognition, prediction and control.