Adaptive Kernel Principal Component Analysis (KPCA) for Monitoring Small Disturbances of Nonlinear Processes

Abstract

The Tennessee Eastman (TE) process, created by Eastman Chemical Company, is a complex nonlinear process. Many previous studies focus on the detectability of monitoring a multivariate process by using TE process as an example. Principal component analysis (PCA) is a widely used dimension-reduction tool for monitoring multivariate linear process. Recently, the kernel principal component analysis (KPCA) has emerged as an effective method to tackling the problem of nonlinear data. Nevertheless, the conventional KPCA used the sum of squares of latest observations as the monitoring statistics and hence failed to detect small disturbance of the process. To enhance the detectability of the KPCA-based monitoring method, an adaptive KPCAbased monitoring statistic is proposed in this paper. The basic idea of the proposed method is first adopting the multivariate exponentially moving average to predict the process mean shifts and then combining the estimated mean shifts with the extracted components by KPCA to construct the adaptive monitoring statistic. The efficiency of the proposed monitoring scheme is implemented in a simulated nonlinear system and in the TE process. The experimental results indicate that the proposed method outperforms the traditional PCA and KPCA monitoring schemes.