摘要

With the recent rapid development of in situ real-time measurement for crystallization processes such as focused beam reflectance measurement (FBRM) and particle vision measurement (PVM), an increasing amount of process data has become available for developing multivariate statistical process control (MSPC) tools to monitor crystallization processes efficiently. To tackle the transitional phase changes because of process nonlinearity and time-varying characteristics in a semi-batch pH-shift reactive crystallization, an integrated monitoring method based on moving window multi-way principal component analysis (MPCA) combined with batch-wise unfolding of batch data arrays using crystallizer volume as an indicator variable is developed in this study. Simulation results reveal that, compared to the conventional MPCA and multi-way partial least squares (MPLS) methods, the proposed monitoring scheme not only can detect an abnormal batch efficiently, but it also reflects the contributions of the control actions to revert the process to an in-control state.

  • 出版日期2016-7