SELF-LEARNING FUZZY-LOGIC SYSTEM FOR IN-SITU, IN-PROCESS DIAGNOSTICS OF MASS-FLOW CONTROLLER (MFC)

An improvement in the yield of better quality wafers requires an accurate control of various process variables. The control should include timely diagnosis and appropriate in-situ, in-process adjustments for drifts in these variables. One such scheme, a self-learning fuzzy logic system is developed in this study for correcting drifts in the calibration of Mass Flow Controllers (MFC's) that control the flow of gases into a process chamber. It consists of two components; a diagnostic system and a self-learning system. The diagnostic system uses fuzzy logic to diagnose the problem and initiate suitable remedial action. The self-learning system automatically builds the knowledge base used for diagnosis. The knowledge base is initialized using clustering principle and is tuned for better performance using a set of heuristic rules. The system is capable of learning the behavior of different types of makes and models of MFC's under various flow rates. It has been tested on two different types of MFC's under different flow rates and encouraging results have been obtained.