Continuous diagnosis of low-pressure chemical vapor deposition reactors using evidence integration

Abstract

A diagnostic system is presented that uses the Dempster-Shafer (D-S) evidential reasoning theory to conduct real-time malfunction diagnosis on semiconductor processing equipment. This is accomplished by combing the continuous stream of information that originates from maintenance status records, from real-time sensor measurements, and from the differences between inline measurements and values predicted by equipment models. The effectiveness of this technique is demonstrated on an low-pressure chemical vapor deposition (LPCVD) reactor used for the deposition of undoped polysilicon