The New PCA for Dynamic and Non-Gaussian Processes

In order to address the non-Gaussian and dynamic features in the process, this paper combined the two- step principal component analysis (TS-PCA) and the preliminary-summation-based PCA (PS-PCA). By using summation and two-step PCA strategy, PS-TS-PCA can eliminate the influence of dynamic relationship among the process data and address the non-Gaussian feature at the same time. And hence PS-TS-PCA achieved good performance in both dynamic and non-Gaussian process monitoring. In addition, this paper also proposed new strategy to eliminate the correlation relationship between two data sample, and proved that least squares algorithm can estimate the dynamic structure in non-Gaussian process.