Soft sensing modeling based on support vector machine and Bayesian model selection

Soft sensors have been widely used in industrial process control to improve the quality of product and assure safety in production. The core of a soft sensor is to construct a soft sensing model. This paper introduces support vector machine (SVM), a new powerful machine learning method based on statistical learning theory (SLT), into soft sensor modeling and proposes a new soft sensing modeling method based on SVM. A model selection method within the Bayesian evidence framework is proposed to select an optimal model for a soft sensor based on SVM. In case study, soft sensors based on SVM are applied to the estimation of the freezing point of light diesel oil in distillation column. The estimated outputs of SVM soft sensors with the optimal model match the real values of the freezing point of light diesel oil and follow the varying trend of the freezing point of light diesel oil very well. Experiment results show that SVM provides a new and effective method for soft sensing modeling and has promising application in industrial process applications. (C) 2003 Elsevier Ltd. All rights reserved.