Monitoring batch processes using multi-model discriminant partial least squares

Multiway principal component analysis (MPCA) has been successfully used to monitor batch processes. But, there exist two problems when MPCA is used to monitor batch processes. One is that the process data must submit normal distribution. Another is that MPCA need complete batch data for online monitoring, so the values from the current until the end of its operation must been estimated or filled in. To overcome those problems, a novel monitoring method, multi-model discriminant partial least squares (multi-model DPLS), is presented. The proposed method is proved to be effective by the application of an industrial streptomycin fermentation process.