A hybrid neural network-feature-based manufacturability analysis of mould reinforced plastic parts

Abstract

The purpose of this research is to establish a method to perform manufacturability analysis of reinforced plastic components by using a hybrid system including automatic feature recognition and a feature-based assessment of manufacturability. Feature recognition plays a fundamental role and is usually the first step in downstream activities concerning the product development process, such as design for manufacturing, design for assembly and process planning. Critical features to successful reinforced plastic moulding are identified, and the relationship between geometric information of the model, expert geometric reasoning and knowledge of related manufacturing processes is clarified and predetermined in a useful and efficient manufacturability analysis system. A prototype system using solid modelling, object-oriented programming and a rule-based system, which is intended to consider the fuzziness of the expert reasoning about reinforced plastic component design, is under construction to test the proposed concepts. The major contribution from this work is a consistent and systematic methodology for analysing the geometry of models, allowing assessment of manufacturability. This methodology considers available manufacturing process capabilities, materials and tooling required. Some virtual parts have been used to test the system, showing promising results.