A knowledge-based system for VLSI process diagnosis
Tyson, Robert F.
MetadataShow full item record
In Very Large Scale Integrated (VLSI) circuit manufacturing environments, a great deal of effort is given to the detection and correction of yield degrading process problems. Process problems become evident at the final step in the VLSI process, mutliprobe testing. Process problems produce detectable patterns in the multiprobe test data. Using this failure pattern, in conjunction with the knowledge of the specific test failed, an inference can be made as to the process problem. In this effort, this heuristic approach is modeled in an autonomous, knowledge-based system for VLSI process diagnosis. This autonomous system perceives the patterns in the multiprobe test data through pattern recognition techniques applied to the multiprobe wafermaps. A forward chaining inference engine applies the patterns and the test failures to a knowledge base to render a diagnostic conclusion of the process problem. The knowledge base is built from human, heuristic experience in the domain of VLSI process problems. This approach to automated diagnosis will be of interest to those in the VLSI community, those studying the application of artificial intelligence to manufacturing environments, as well as to those interested in knowledge-based systems.