A learning approach of wafer temperature control in a rapid thermal processing system

This paper presents a learning approach for wafer temperature control in a rapid thermal processing system (RTP). RTP is very important for semiconductor processing system and requires an accurate trajectory following. Numerous studies have addressed this problem and most research on this problem requires exact knowledge of the system dynamics. The various approaches do not guarantee the desired performance in practical applications when there exist some modeling errors between the model and the actual system. In this paper, iterative learning control scheme is applied to RTP without exact information on the dynamics. The learning gain of the iterative learning law is estimated by neural networks instead of a mathematical model. In addition, the control information obtained by the iterative learning controller (ILC) is accumulated in the feedforward neuro controller (FNC) for generalization to various reference profiles. Through numerical simulations, it is demonstrated that the proposed method can achieve an accurate output tracking even without an exact RTP model. The output errors decrease rapidly through iterations when using the learning gain estimated and the FNC yields a reduced initial error, and so requires small iterations.