MOdeling Of Steam Distillation System Using Hammerstein-Wiener model

Abstract

This paper presents a new method to model a steam temperature in distillation system by using system identification. Three nonlinear models have been compared, i.e. a Hammerstein model, a Wiener model and a Hammerstein-Wiener model. In this work, we propose the utilizing of the piecewise-linear and sigmoid network Hammerstein-Wiener model for single-input single output processes. All the models have been optimized with respect to initial state, search criterion and number of iterations. The testing of the trained model will be based on percentage of best fit (R2), Final Prediction Error (FPE) and loss function (V). Among three model tested, the most accurate model is the Hammerstein-Wiener model with piecewise linear and sigmoid network estimators. This model produce highest percentage of best fit, the lowest FPE and loss function.