SYSTEM IDENTIFICATION OF A STEAM DISTILLATION PILOTSCALE USING ARX AND NARX APPROACHES

Abstract

This paper presents steam temperature models for steam distillation pilot-scale (SDPS) by comparing Pseudo Random Binary Sequence (PRBS) versus Multi-Sine (M-Sine) perturbation signal Both perturbation signals were applied to nonlinear steam distillation system to study the capability of these input signals in exciting nonlinearity of system dynamics. In this work, both linear and nonlinear ARX model structures have been investigated. Five statistical approaches have been observed to evaluate the developed steam temperature models, namely, coefficient of determination, R 2 ; auto-correlation function, ACF; cross-correlation function, CCF; root mean square error, RMSE; and residual histogram. The results showed that the nonlinear ARX models are superior as compared to the linear models when M-Sine perturbation applied to the steam distillation system. While, PRBS perturbation exhibit insufficient to model nonlinear system dynamic