文摘
Statistical quality prediction methods for multiphase batch processes have gained much attention in recent years. While most methods are focused on the data information inside each phase, the relationships among different phases have rarely been explored and used for quality prediction, although it may have significant impacts on prediction of the final quality. In this paper, a two-level partial least squares (PLS) model is proposed, in which the relationships among different single phases are modeled and incorporated for quality prediction. In the first level of this method, a representative intraphase-PLS model is built for each single phase, while in the second level, a series of interphase-PLS models are constructed to capture the relationships among different phases. With the incorporation of the additional interphase information, the multiphase quality prediction performance can be improved, which is evaluated through an industrial case study.
