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.