DETECTION AND DIAGNOSIS OF ABNORMAL BATCH-OPERATIONS BASED ON MULTIWAY PRINCIPAL COMPONENT ANALYSIS - WORLD BATCH FORUM, TORONTO, MAY 1996

The industrial application of a new monitoring scheme for batch and se mi-batch processes is presented. Multi-way Principal Component Analysi s is used to analyze the information from the on-line process measurem ents. The basic idea is to build a statistical model based on process measurements from past successful batches, which describes the normal operation of the process. Subsequently future batches are compared aga inst this model and characterized as normal or abnormal. The algorithm s and all the design equations are presented for setting up Statistica l Process Control charts which monitor the performance of a batch proc ess. Contribution plots for detected abnormal operations are developed to identify the measurement variables and time periods of abnormal op eration. Copyright (C) 1996 Elsevier Science Ltd.