Fine-Scale Modeling and Monitoring of Wide-Range Nonstationary Batch Processes With Dynamic Analytics

Abstract

Batch manufacturing has attracted extensive attention as a crucial way in process industry. It has typical wide-range nonstationary characteristics subject to time variant conditions, not only within each batch but also for different batches. The presence of closed-loop control complicates the task of batch process monitoring which shows that closed-loop control can significantly alter the process characteristics with both static deviations and varying temporal correlations. It is of a considerable challenge for batch process analysis and monitoring under closed-loop control which may work quite unlike that under open-loop control. In this article, a fine-scale process modeling and monitoring method are developed with dynamics analytics for wide-range nonstationary batch manufacturing with duration uncertainty. Both static and dynamic monitoring statistics are designed with clear physical interpretation which can work together to achieve fine-scale batch process status identification and monitoring. On one hand, it can identify frequent normal switching of operation status at different time by simultaneously checking both static and dynamic deviations regulated by closed-loop control. On the other hand, it can detect small disturbances in early stage where the impact may be overshadowed by the regulation of closed-loop controllers, making faults difficult to detect in time by conventional methods. Experiments have shown the validity of the proposed fine-scale monitoring method for wide-range batch process.