MPCA based phase identification method and its application to process monitoring

In order to characterize the intrinsic performance of multi-phase batch process further, a sub-phase partition method is proposed. According to the different numbers of principal components and variation direction of variable information, a two-step phase partition is realized for the phase partition of multi-phase process. After the two-step division, the entire time-slice matrices in the same sub-phase have the same number of principal components and similar variable variation direction. And the `fake´ phases, stable phases and transition phases are identified by combining the specific characteristics of batch processes. The proposed MPCA modeling methods and steps based on sub-phase partition are given and applied to online monitoring of penicillin fermentation process.