Expert system for castability evaluation: Using a fixed-features based design approach

Abstract

In the metal-processing industry, casting is the most popular means for achieving a desired shape (net shape) in a single step. The purpose of this work is to predict possible casting defects at the design stage in order to minimize repeated trial and error in the casting process. A knowledge-based expert system called EXCAST has been developed to facilitate the design of axisymmetrical casting components and to evaluate the manufacturability of the design at the design stage using a fixed-features based design approach. Axisymmetrical casting parts are modeled into a simplified fixed-feature representation so that we can (i) simplify a complex geometrical shape for manufacturability evaluation, (ii) extract common rules for manufacturability evaluation, and (iii) produce a simple and dynamic database for use in other systems. The manufacturability evaluation is carried out for tolerancing, minimum thickness, draft angle, shrinkage, hot junction, parting line, rounding, boss, and solidification. The EXCAST system was tested for eight flywheels in production and found to predict major casting problems.