Self-learning fuzzy modeling of semiconductor processing equipment

摘要

A qualitative equipment model for a low pressure chemical vapor deposition (LPCVD) process is presented. The model is based on fuzzy representation of input-output relationships and utilizes self-tuning membership functions. To demonstrate this concept a fuzzy inference system has been built for polysilicon grain size prediction based on deposition and annealing temperatures. After the system is trained with experimental data, it automatically tunes its membership functions to accommodate additional experimental data.