Research On Fault Diagnosis Of Beer Brewing Process Based On Sparse Support Vector Machine

The chemical production process often has characteristics of nonlinear,timevarying and strong coupling,It is difficult to establish an accurate analytical model in process monitoring,At the same time,the automated control system can record a large amount of data in the production process,which promotes the development of the theory that data-driven process monitoring.In this paper,based on the typical batch chemical process,the fault detection method based on statistical analysis theory is studied,The main work focuses on the improvement of the least squares support vector machine fault detection algorithm.The clustering algorithm is used to preprocess the data,In order to solve the sparsity problem in LS-SVM,which makes the LS-SVM method have the same characteristics as the standard svm.A sparse least squares SVM fault diagnosis model based on kernel distance clustering is proposed,aiming at Parameter selection of penalty factor C and kernel function in LS-SVM model,utilizing particle swarm algorithm to optimiz the parameter and enhancing the generalization ability of the model.Based on the simulation study of beer fermentation process,the proposed sparse model can improve the classification accuracy and training time,which can also be more efficient to separate the fault and has good anti-interference ability.