The DTW synchronized MPCA on-line monitoring and fault detection predicted with GCC

In general, Multiway Principal Component Analysis (MPCA) algorithm is hardly applied on batches process monitoring directly because of various time lengths of batches and mismatching of the pattern of characteristics within each time interval. After transformation of Dynamic Time Warping (DTW) algorithm to rearrange the similar segments of batches automatically, every batch becomes synchronous with the others. Furthermore, to improve the online monitoring and relevant fault diagnosis, a kind of Generalized Correlation Coefficients (GCC) are applied to search the similar trajectories from the history model library so as to predict the future part of the being tested batch. The outcomes of simulation of polyvinyl polymerization prove that the combination with GCC and DTW on the online MPCA monitoring helps to discover the abnormal of process earlier and improves the quality of monitoring.