INTEGRATING EXPERT-SYSTEMS WITH A RELATIONAL DATABASE IN SEMICONDUCTOR MANUFACTURING

The designs and implementations of two expert systems for semiconductor manufacturing are described and compared. The first expert system is used for parametric test analysis in a CMOS wafer facility. It is designed as a user-interactive system, with all the expert's knowledge included in one knowledge base. This expert system tightly controls the flow of information between the knowledge base and the wafer data stored in a relational database. This expert system was implemented using the Knowledge Engineering Environment shell with interface programs to INGRES. The second expert system is used for parametric test analysis in an analog wafer facility. This system is partitioned into several independent C programs running on-line and continuously communicating their results through the relational database. The interactive system ig more flexible in representing expert knowledge and in explaining results, while the on-line system is faster, more open to new module integration, and better able to classify and diagnose large volumes of wafer data. Important factors that need to be considered for an effective expert system design in a computer integrated manufacturing environment include user interaction, algorithmic data processing, speed of response, knowledge representation, explanation capabilities and system access.