The Research On Soft Measurement Of Cell Concentration Based On Support Vector Machine

The cell concentration directly reflects the growth of bacteria in the course of fermentation.It is the main attribute of the course of fermentation and plays a decisive role on the product quality.Limited to the level of biosensor,the cell concentration is difficult to be detected in real time online.According to the key variables to characterize the growth of bacteria in the course of fermentation,an accurate soft measurement model about cell concentration can be established,it is can be used to monitor the course of fermentation in real time and optimize the fermentation technology.Support vector machine(SVM)is skilled in small sample learning based on statistical learning theory.This paper studies on the course of fermentation of pichia yeast and measures the cell concentration based on SVM.The course of fermentation is complex with long production cycle.It is difficult to collect the sample data.This paper analyzes the course of fermentation of pichia yeast and finds the key factors affecting the cell concentration.Two methods of SVM and BP neural network are used to measure the cell concentration.Through MATLAB simulation,the prediction performances of the two soft measurement methods are compared in the case of small samples.The improved genetic algorithm which is based on the genetic algorithm and the hormonal regulation mechanism is used to optimize the SVM parameters.Through the simulation and analysis,the differences of the rapidity of convergence and the optimization effect are compared.The course of fermentation is complex and changeable,and the characteristics of various stages in the process of microbial growth are significant.The soft sensing is an important method to solve this problem.In this paper,the fuzzy mean clustering algorithm is used to divide the course of fermentation.The SVM soft sensor is used to predict the cell concentration of each stage.Compared with the global prediction model,both of the prediction performance can be judged.Based on the control system of the existing fermentation process,the real-time prediction of the cell concentration in the course of fermentation is realized by soft measurement technology.