Knowledge-based manufacturing quality management: A qualitative reasoning approach

Manufacturing fault diagnosis is the problem of determining the manufacturing fault(s) responsible for any critical dimensions or performance tests of the final assembly that fall outside their tolerance limits, as measured by on-line probes inserted at several chosen points of the assembly. An effective means for manufacturing fault diagnosis is crucial for controlling the quality of products rolling out of the manufacturing system. The current practice of this fault diagnosis process is to employ a computer-based information system to monitor the in-line testing results and have a diagnostic expert interpret the data when problematic measurements or performances are observed, so that any aberrations of the manufacturing system can be corrected. However, such an approach usually creates information overload and production-line disruptions, making the diagnostic task burdensome and prone to judgmental errors. The objective of this research is to automate the diagnostic process by an artificial intelligence (AI) approach. The approach is characterized as qualitative reasoning; it makes diagnostic decisions by "explaining" the undesirable test measurements and building causal links based on the qualitative model of the product assembly. We illustrate the fault-diagnosis approach by studying the quality management decision process of a torque-converter manufacturing system. Empirical manufacturing data is used to illustrate the procedure for validating the model.