Research Of Fermention Process Modeling Based On Support Vector Machine

Support vector machine (SVM) theory is one of the most popular research topics studied by domestic and overseas scholars nowadays. The SVM is a small sample statistics and has the advantages of good generalization and global optimization and very suitable for the modeling of nonlinear process. Now SVM is developing promisingly either in theory or applications.Fermention process has the characteristic of strong nonlinear, time varying and uncertainty, and its inherent mechanism is very complex. Due to the advantages of SVM, using SVM to modeling for fermention process has important actual value. The article using SVM to model for a typical fermention process: penicillin ferment process.Aiming at the disadvantage of SVM,least square support vector machine,fuzzy least square support vector machine, providing a sparse solution based on fuzzy least square vector machine. Using the fuzzy membership to describe the importance of the samplesn and set a value to cancel the samples which have the fuzzy membership is lower than the value. Then using a recursive algorithm to simplify the model and improve the operation speed. This solution only preserves the samples which involve important information, and in the prediction, not all samples are attending the calculation. It means the solution gets the sparseness. The simulation results shows that this solution has nice modeling effect and preserving the modeling accuracy, comparing with other support vector machine methods, its speed is greatly improved and gets the sparseness.