题目：Applications of statistical methods to batch processes
主讲内容：For ensuring operation safety and consistent product quality, multivariate statistical methods, such as multiway PCA/PLS and phase-based PCA/PLS, have been widely adopted in chemical batch process monitoring and quality prediction. However, these methods still have certain limitations, which may not work well in many specific cases. In this presentation, the applications of statistical methods to batch processes will be discussed through three examples: 1. final product quality prediction, 2. batch-to-batch steady state identification and startup monitoring, and 3. online monitoring of resin transfer molding. In the first example, it will be shown that the conventional modeling methods may not fully utilize available process information, which affects both prediction accuracy and model interpretation. In the later two examples, process data do not obey the assumptions of the conventional statistical monitoring methods. This presentation will provide solutions to these problems and show how shallow process knowledge can help statistical modeling of batch processes.