Utilizing Fuzzy Data and Rules for Discrete Control of Semiconductor Manufacturing Processes

Semiconductor wafer fabrication is a very complex manufacturing process. Many semiconductor processes are not very well understood. The control of semiconductor manufacturing processes is, thus, an active research area. Even though various algorithms have been developed for process control and optimization, the scope of application of these algorithms is not very well-characterized due to the complexity of the semiconductor manufacturing processes. However, information about the applicability of these algorithms can be gathered from various sources and can be roughly expressed in terms of rules of various types that relate the applicability of different algorithms to process conditions. We have applied fuzzy set theory to develop a multi-algorithm control system which can exploit this knowledge from multiple sources to make better control decisions. We have also developed generic decision making techniques capable of utilizing relatively contradictory information from heterogeneous sources. This system utilizes a database to store fuzzy rules that relate the suitability of use of different control algorithms to the process conditions. To design and implement this fuzzy rule-base we utilize FERM (Fuzzy Entity-Relationship Methodology), a design methodology for fuzzy relational databases, to develop a generic data model for fuzzy rules. Although the resulting system is targeted for control of a semiconductor manufacturing process, the proposed techniques are generic and can be applied to other manufacturing and control applications to model and integrate control rules from heterogeneous sources.