Soft Sensing for Propylene Purity Using Partial Least Squares and Support Vector Machine

Many important process variables in propylene distillation actual system is difficultly detected directly or not easy online survey, especially propylene purity which will vary with other process parameter. This paper presents an online soft sensing method by combining partial least squares and support vector machine. First multivariate data analysis was performed using partial least squares, then soft sensing regression model was constructed. Simulation shows that the proposed measuring scheme guarantees parameter estimation and predicting accuracy.