Dynamic process monitoring based on probabilistic principle component regression

Probabilistic principal component analysis (PPCA) has been proposed to monitor the industrial process in the probabilistic manner. However, the traditional PPCA method is invalid when the observed data is strongly auto-correlated. For monitoring the probabilistic dynamic process, a probabilistic principal component regression (PPCR) model is proposed to extract the dynamic information of the data, based on which new monitoring statistics are constructed. The Expectation-Maximization (EM) algorithm is utilized for model training. Then, the PPCR based state space model is constructed to monitor the dynamic process. For performance evaluation, a case study on the Tennessee Eastman (TE) benchmark process is provided.