Multi-layer Monitoring for Parallel Batch Processes with Input Trajectory Adjustment

This paper develops a multi-layer fault detection method for parallel batch process monitoring. Besides, an input trajectory adjustment strategy related to monitoring stage is implemented to improve the economic performance. Firstly, a global MPCA monitoring model is constructed with input-relevant variables for all parallel batches. Then, several individual BWPLS monitoring models are established to deal with the model uncertainty of local parallel batches. When no abnormal condition is detected by both monitoring layers, a new input trajectory with better economic performance for the current batch is calculated with input-relevant constraints defined by the global monitoring layer as well as a surrogate model. As a result, these layers are related to each other, which provide a reliable and effective monitoring and adjustment framework for parallel batches. A fed-batch reactor is introduced for performance evaluation and the result proves the effectiveness of the proposed method.