Multi-level pseudo-random signal design and "model-on-demand" estimation applied to nonlinear identification of a RTP wafer reactor

Guidelines are presented for specifying the design parameters of multi-level pseudo-random sequences in a manner useful for "plant-friendly" nonlinear system identification. These multi-level signals are introduced into a rapid thermal processing wafer reactor simulation and compared against a well-designed pseudo-random binary sequence (PRBS). The resulting data serves as a database for a "model on demand" (MoD) predictor. MoD estimation is attractive because it requires less engineering effort to model a nonlinear plant, compared to global nonlinear models such as neural networks. The improved fit of multi-level signals over the PRBS signal, as well as the usefulness of the MoD estimator, is demonstrated on validation data.