Soft sensor modeling of the penicillin fermentation based on FCM clustering and LS-SVM

Abstract

In dealing with the problem that the important parameters of a penicillin fermentation process are hard to measure precisely, such as biomass concentration and production concentration, therefore, a soft sensor modeling for the penicillin fermentation based on fuzzy c-means clustering and least square support vector machine (LS-SVM) is proposed. First of all, features of sample data are extracted and the secondary variables are determined by principal component analysis (PCA). And then, in order to predict these important biological parameters, a fuzzy c-means clustering (FCM) algorithm is applied to group the training data into several clusters, and LS-SVM is used to construct models based on each cluster. The simulation example shows that the method could measure the important parameters which could not be measured online during the course of the penicillin fermentation with a high precision.