An FOP based recursive PCA for adaptive process monitoring
Principal Component Analysis (PCA) has been found wide applications. However, when standard PCA is applied to slow or parameter-varying process, it will lead to many false alarms. In this paper, to deal with the problem, we propose one recursive PCA algorithm, which efficiently updates the covariance matrix and will decrease the computation cost. The algorithm is based on First-Order Perturbation (FOP) theory, which is a rank-one update of the eigenvalues and its corresponding eigenvectors of a observation covariance matrix. We also propose two new statistics, one of which is similar to the Hawkin’s statistic but without the numerical drawback motivated by an analysis of the existing test statistic. In comparison with the SPE index, the threshold of the new statistic is computationally simpler. The effectiveness of the proposed RPCA algorithm and tow new statistics has been evaluated with an application of monitoring a continuous stirred tank heater simulation system.
Process monitoring and fault diagnosis Recursive Principal Component Analysis First-Order Perturbation
Hu Zhikun Gui Weihua Yang Chunhua Song Yan-po Peng Xiao-qi
School of Physics & Electronic, Central South University, Changsha, 410083 School of Information Science and Engineering, Central South University, Changsha, 410083 School of Energy Science and Engineering, Central South University, Changsha, 410083
国内会议
厦门
英文
1-6
2012-08-01（万方平台首次上网日期，不代表论文的发表时间）