Modeling Of Ferment Process Based On Support Vector Machine

Because many popular and uneasily synthesized biological products with complex structure can be made by genetically engineered bacteria, genetic engineering is becoming more and more important in modern society. Escherichia coli have the characters of simple structure, short growing cycle and clear growing conditions. These characters make it widely used as an expression vector, so the fermentation of E. coli is becoming one of the hot points.Some critically questions in the fermentation restricts the adjusting of parameters which affects the results of fermentation and further optimizing and controlling in the E. coli fermentation process, therefore, it is significant to model the E. coli fermentation process. In the fermentation process, the most popular method is neural network modeling approach, which is based on the principle of ERM, but it is easy to fall into over learning and local minimum. However, the way of SVM is based on the principle of SRM, and it can effectively avoid the above problems by introducing kernel function and high-dimensional space, so the article uses the method of SVM to establish the forecasting model of fermentation process. Then, comparing with the SVM method and BP neural network prediction model, the simulation results show that SVM is more suitable for high density fermentation of E. coli model.By using the SVM method, we need to select some parameters before establishing the model. The existing methods of selecting parameters are based on experts experience and the cut and'try method, which haven't a high precision. Aiming at this problem, the thesis analyzes the principle and the advantage of ant colony algorithm, proposes a SVM method with high precision. Simulation results show that the established prediction model in this way can search the SVM parameters with high precision in a short time. The SVM model with modified parameter has the good ability of fitting and generalization. It has high prediction accuracy and provides a more accurate prediction model for the control of high-density process of E. coli.The SVM has different parameters toward different data. Based on the thinking of using ant colony algorithm and given data to search parameters. The established model can be easily applied in other substances in the fermentation. It provides a more generic approach for modeling of microbial fermentation.