Two-step Principal Component Analysis for Dynamic Processes

A enhanced principal component analysis (PCA), termed as Two-step PCA (TS-PCA), is proposed to handle the dynamic characteristic of industry processes. Differently from the traditional dynamic PCA (DPCA) using the "time lag shift" structure, TS-PCA adopts a new structure to present the dynamic property in the process data. By using this new structure, TS-PCA can extract the time-uncorrelated components from the dynamic data and use it for process monitoring. In addition, it can update the expectation and standard variance of the process data at each step for data normalization.