Study On The Methods Of Fault Detection And Prediction In Non-linear Industrial Processes Based On Support Vector Machine

Due to the popularity of the globalization and informatization of large-scale industrial manufactures,there has grown up an urgent need the stability of the industrial production system,the efficiency of industrial production process and the requirements of the product quality.These mean a greater challenge around the corner.Therefore,in order to realize real-time monitoring and detection effectively and ensure production security and dependability,using support vector machine for the fault detection and prediction of nonlinear industrial process has important theoretical value and practical significance.In this paper,the basic theory of support vector machines(SVM)is analyzed.The modeling principle and process of the algorithm was deduced.Firstly.To deal with fault detection and prediction of nonlinear industrial processes,the kernel parameters of support vector machine is optimized using k-fold cross-validation.The SVM method is tested on the CSTH database and compared with an improved Partial Least Squares(IPLS)and Principal Component Analysis(PCA).The experimental results show that the SVM classifier in real complex industrial process has excellent ability to predict and ideal running time.For fault prediction problems of nonlinear industrial processes,the performance of S4 VM is validated where the classifier based on S4 VM outperforms those based on TWSVM in the processes of industrial fault conditions.S4 VM is not sensitive to the initial parameter value and gives a number of diverse large-margin separators,by optimizing the label assignment for unlabeled instances such that the worst-case performance.The results show that S4 VM is superior in the fault prediction of nonlinear industrial big data.