Machine Learning Techniques for Group Technology Applications

Abstract

This paper presents a new approach of using machine learning techniques from artificial intelligence research to induce part families which can satisfy specific goals of group technology applications. A framework for god-directed part classification is proposed, and a case study is reported to demonstrate the framework. The approach treats forming part families as an inductive learning process whose purpose is to generate useful classification of a population of parts. Specially, conceptual clustering is used as an alternative to numerical taxonomy, and feature-based symbolic representation is employed in Lieu of numerical coding. The goal of GT applications can be incorporated by proper selection of relevant attributes for the clustering program, and/or by using background knowledge to bias the clustering process. The results from conceptual clustering are simple, conceptual interpretations of pan families which are easy for mastering the interaction among representation-classification-goal cycle for real-world GT applications. Furthermore, this research illustrates a meaningful application of advanced AI techniques to an important area of computer-integrated manufacturing.