A knowledge based manufacturing advisor for CAD

This thesis presents a knowledge based manufacturing advisor for Computer Aided Design. The aim of the project has been to develop techniques that can help designers to evaluate the manufacturability of moulded parts during the design process. One of the major achievements in the research has been the development of a novel feature recognition approach to allow moulding features to be recognised from a CAD model. Existing feature recognition techniques are not appropriate for moulded parts and a new feature recognition approach has been developed that uses a mid-surface abstraction from the CAD solid model as the basis for feature recognition. The feature recognition methodology presented in this research is a graph based technique that uses an attributed mid-surface adjacency graph as the basis for feature recognition. Feature recognition algorithms have been developed for a range of common moulding features. Two evaluation techniques are presented that measure the quality of the feature recognition results and provide a confidence measure for the manufacturing advice that is generated. The first evaluation technique uses the Hausdorff distance to measure the similarity of the mid-surface model to the CAD solid model that was used to generate it, and the second uses a face mapping technique to evaluate the completeness of the feature recognition results. A demonstrator for the knowledge based manufacturing advisor has been developed which incorporates the feature recognition software and an expert system for moulding advice. The expert system provides an interactive software environment in which the user can input details about their part design, and receive tailored manufacturing advice for a range of moulding processes. The demonstrator has been tested on a range of realistic moulded parts.