Optimization of a Knowledge Based Design System Using AHP & Pattern Recognition

Abstract

Present work addresses the problem of identifying the existing parts that, are similar, in one or many characteristics, to a new part at the design stage .The environment is UG/NX (3.0) where manufacturing partners share product related data to come up with new, customized, high quality products within a minimal time period or we can say with minimal lead times. The proposed method is based on the principles of Group Technology, Pattern Recognition and on the definition of Similarity of critical design attributes. A two step procedure is proposed: (First) A search procedure, which acquires and processes the search intent to retrieve the similar parts. (Second) A sorting procedure, which ranks these parts on the basis of their similarity using Similarity Index to the previously teach patterns. The first one is based on Binary Conversion and template matching using Analytical Hierarchy Process while the other one uses the ranking criteria based on Comparative index. The approach employs a systematic procedure to identify the closest similarity once the weights has been modified and generated for a particular file. A software system using C and Object Oriented Technology has been developed for Pattern Recognition using AHP concept. A Neural Net code based on Feed Forward Network for Pattern recognition is also developed to compare the capability of the AHP in the field of Group Technology. Same inputs were given to Neural Network code and the outputs were compared with those from AHP. Finally the nearness of outputs is compared for the weighted samples.