Soft Sensor Technology On Avermectin Fermentation Process

Fermentation process is a strong nonlinear, time-variant and correlative process. To improve the production, more information about the process must be gained first. But with regard to a industrial process, cheap and reliable instrumentation suited to real-time monitoring is absent, only some of physical and chemical parameters can be measured on-line, and the others, such as bio-parameters which are more complex and important, are measured manually and the information is lagging. To overcome these problems, a way-out is to build a soft sensor model.This thesis mainly studies on the Avermectin fermentation process. With the mechanism and the soft sensor method, several schemes to solve the problem of the soft sensor are taken out. A new method based on KPCA (Kernel Principle Component Analysis ) and RBFNN (Radial Basis Function Neural Network) is proposed and applied in the soft sensor modeling of the biomass estimation.The main contributions of the thesis are as follows:1. An overview is made on the development and significance of fermentation process control, and the problems of estimation of important variables in the plants are given. To solve the problems, soft sensor must be used, which is briefly introduced, including its affecting factors.2. Study the mechanism of the fermentation process, analyze the corrletation of all process varialbes and discuss the relation of process modeling, control and optimization.3. ANN (Artifical Neural Network) is introduced, two kinds of wildly used Neural Network(BP Neural Network and RBF Neural Network) are studied, then their advantages and disadvantages in soft sensor modeling are discussed.4. With the comparison of all methods of soft sensor, a new method based on KPCA and RBF Neural Network is proposed, the method is applied to model the biomass estimation and the flow is given.5. With the process data, soft sensor models based on KPCA-RBFNN and KPCA-BPNN are established for biomass estimation in the process of Avermectin fermentation. Based on the analysis of the models' performance, we conclude that KPCA is better and promising.At the end of the thesis, an conclusion is made, and future research directions are proposed.