Soft Sensor Modeling of LS-SVM based on Bayesian Criteria for Marine Protease Fermentation Process

Abstract

Abstract. In order to solve the problems of real-time measurement of crucial biological parameters (such as substrate concentration, cell concentration, enzyme activity and so on) in the marine microorganism enzyme fermentation process, a soft sensor method of Least Squares Support Vector Machine(LS-SVM) based on Bayesian criteria is proposed. Take a typical marine enzyme-sea protease as the research object. Firstly, the auxiliary variables and the dominant variables of soft sensor model are determined based on the analysis of the mechanism of marine protease fermentation process. Secondly, considering the regularization parameter and kernel parameter are the difficulties in the LS-SVM modeling, Bayesian theory is used to optimal select parameters of LS-SVM, and then a soft sensor model is established by LS-SVM based on Bayesian Criteria. The simulation results show that the modeling has higher prediction precision and better generalization ability characteristics than LS-SVM.