New Soft Sensor Method Based on SVM

Abstract

This paper proposes a soft sensor technique based on support vector machine(SVM) technique, firstly gives an introduction to LSSVM, then designs a training algorithm for LSSVM, finally uses it to identify Absorption Stabilization System (ASS) process variable. Case studies are performed and indicate that the proposed method provides satisfactory performance with excellent approximation and generalization property, soft sensor technique based on LSSVM achieves superior performance to the conventional method based on neural networks. approaches. The formulation of the SVM embodies the Structural Risk Minimization (SRM) principle, which has been shown to be superior to the traditional Empirical Risk Minimization (ERM) principle, employed in conventional neural networks. It is this difference that equips SVM with a greater ability to generalize, hence a better generalization ability is guaranteed. As an interesting variant of the standard support vector machines, least squares support vector machines (LSSVM) have been proposed by Suykens and Vandewalle(5,6) for solving pattern recognition and nonlinear function estimation problems. Standard SVM formulation is modified in the sense of ridge regression and taking equality instead of inequality constraints in the problem formulation. As a result one solves a linear system instead of a QP problem, so LSSVM is easy to training. This paper discusses the basic principle of the LSSVM at first, and then uses it as a soft sensor tool to identify Absorption Stabilization System (ASS) process variable. The method can achieve higher identification precision at reasonably small size of training sample set and can overcome disadvantages of the artificial neural networks (ANNs). The experiments of the identification have been presented and discussed. The results indicate that the SVM method exhibits good generalization performance.