AN ARTIFICIAL INTELLIGENCE APPROACH FOR THE ANALYSIS OF CASTING DEFECTS IN FOUNDRY

Abstract

The control of casting quality is a major problem for many foundries where a variety of defects can occur as a result of various causes. Knowledge based expert systems (Artificial intelligence approach) are computer programs which use a collection of facts and rules to suggest solutions to specific problems. Foundry practice is rich in thumb rules and knowledge base solutions, which can be implemented in such programs to aid the foundrymen in the diagnosis of casting defects. The characterisation of defects should be based on a technique which is very easy to learn and implement. It must be comprehensive. The system aims at bringing together the whole array of knowledge available with experts, foundry journals, text books etc. A methodology is developed to provide a systematic approach for the identification of the defect and causes that of leading to the elimination of rejected castings. The methodology consists of five steps defect identification, defect cause enumeration, defect cause determination, defect correction, and process monitoring. Knowledge representation is exclusively in the form of rules and facts. It uses a combination of static, heuristic and declarative knowledge to create its knowledge (v) base. This system which has been developed in Turbo Prolog, can be implemented on any IBM compatible PC. To make the system more comprehensive, the graphic display of the casting defects using AutaCAD software has also been achieved.