Application Research Of Fault Detection Based On Improved KPCA And SVDD

With the rapid development of automation system in industrial process,the traditional fault diagnosis method has been unable to solve the problem of on-line monitoring in complex industrial process.In the actual industrial production,there will inevitably be the aging of the system equipment,the transformation of the working conditions and components damage and other factors.If these small fault sources can not be processed in a timely manner,it will lead to paralysis of the industrial system affect the quality of products,when serious will lead to major accidents.Therefore,the industrial chemical process fault detection and recognition has become an important research area in the field of process control,and has been widely concerned by scholars at home and abroad.The data driven monitoring method can not only be limited by the mathematical model,but also can realize the on-line real-time monitoring of the complex industrial process by using the collected data in the industrial production process.In this paper,several kinds of fault detection models are introduced in detail.Based on the original detection algorithm model,the fault detection.This paper takes Tennessee Eastman industrial chemical process(TE)as a research platform,through the platform to complete the production process simulation and process data acquisition TE.In view of the characteristics of multi variable and high dimension in the production process of industrial chemical industry,the traditional principal component analysis(PCA)is used to establish the PCA fault detection model.Because the PCA method needs to assume that the sample data in the industrial chemical process obey the Gauss distribution and can only extract the linear principal component,the kernel principal component(KPCA)algorithm is used to extract the nonlinear feature principal component to construct the fault detection model.The basic idea of KPCA is the original space to a high dimensional feature space,then the principal component analysis in the feature space,so as to realize the data separability.The PCA model and KPCA model are applied to the TE simulation platform.The experimental results show that the KPCA model is better than the PCA model for fault detection.Aiming at the problems of nonlinear,non Gauss and many kinds of noises in the process of industrial production,the improved KPCA and SVDD fault detection model is proposed.Firstly,an improved KPCA method is proposed to reduce the influence of noise on the KPCA data extraction.In the KPCA feature space,the reconstruction error of the sample data is calculated,so that the sample data in the feature space can automatically identify and remove the abnormal value.The KPCA algorithm is applied to the nonlinear kernel mapping,and the kernel principal component is extracted.Then,in order to solve the problem of non Gauss characteristic in the process data,the support vector data description(SVDD)algorithm isintroduced in this paper.The SVDD algorithm does not need to assume that the industrial chemical process data obey the Gauss distribution,and has good classification results.SVDD method is used to establish the fault detection model of the kernel element in the improved KPCA.Finally,this paper puts forward the improved KPCA and SVDD method applied to the TE platform,the experimental results show that the method can achieve better fault detection effect.