Application Of An Adaptive Moving Window KPCA Fault Campus Diagnosis Method In Process Monitoring

With the continuous development of global industry,the control of industrial production process becomes more and more important.The mechanization and intellectualization of industrial process have lead to the quantification and multi-dimensional of production process data.How to screen and simplify these data so as to monitor production process more effectively has become a difficult problem.Multivariate statistical methods based on Principal Component Analysis(PCA)are widely used in the troubleshooting of industrial processes to improve the quality and productivity of processes.But the PCA-based monitoring model is time-invariant after it is established,and most real industrial processes are time-varying.The Kernel Principal Component Analysis(KPCA)method can use the nonlinear kernel function and the integral operator to calculate the principal components in the high-dimensional feature space,and solve the eigenvalue problem in the case of linear state.Using the adaptive moving window algorithm to reconstruct the KPCA model can effectively update the eigenvalues and better solve the time-varying problem.Based on the PCA and KPCA algorithms,a variable moving window algorithm is used at the same time.After comparing the advantages and disadvantages,the variable moving window KPCA method is studied in combination with the Tennessee Eastman(TE)process.Monitoring performance,main research contents and conclusions are as follows:(1)This paper introduces the basic idea of moving window algorithm,and studies an adaptive moving window algorithm for updating the traditional monitoring model.The simulation results show that by adding variable forgetting factor to the moving window algorithm,the optimal size of mobile window can be selected continuously ? It avoids the problem that the eigenvalues and eigenvectors caused by the window being too large or too small are not representative,so that the tracking change is less efficient or less accurate.(2)A new variable moving window KPCA method is studied.In order to keep the kernel matrix with enough new samples,we can add the latest samples and discard the oldest samples through the moving window,and then select the optimal size of the moving window,which greatly improves the efficiency and accuracy of process monitoring.The simulation results show that the method is effective.(3)The variable moving window KPCA method is applied to TE chemical process monitoring.The results show that the method can effectively improve the extraction efficiency of specific values in TE chemical process,thus improving the overall chemical process monitoring rate and reducing the false alarm rate.It proves that the method has high accuracy and good scalability,which can effectively improve the efficiency of industrial process monitoring.Finally,through summarizing the full text,the paper puts forward the problems that need to be further discussed and studied in the process monitoring based on the adaptive moving window KPCA method.