Nonlinear identification of wavenet based NARX model for reactive distillation column.

Abstract

This work is aimed at establishing a reliable model for nonlinear behavior of a reactive distillation column. Nonlinear Auto-Regressive with eXogeneous input(ARX) modeling technique which consists of parallel combination of nonlinear and linear blocks described by nonlinearity estimation wavelet network was developed and then applied to reactive distillation process with of production of Methyl Tert-Butyl Ether (MTBE). The model was identified using the dynamic data simulated from CHEMCAD environment. Identification toolbox in MATLAB environment was used in estimation of the parameters and validation of the models. The obtained results gave a best fit of 71.26% and 87.13% for output purity and conversion respectively and the residual analysis results show that the model is a good model as the residual autocorrelation functions were within the confidence interval of 99%. The agreement between the simulated data and model estimation and validation shows that the model successfully predicts nonlinear behavior of a reactive distillation process.

Keywords: Nonlinear identification, NARX model, reactive distillation column, MATLAB, CHEMCAD.