Two-Direction Residual Recursive Quality Prediction for Multi-Mode Multi-Phase Batch Processes

In batch processes, in order to meet different market demands, multiple operating conditions are required in the same production process, which forms a variety of process modes. In addition, the operating conditions in each batch process also have multi-phase characteristics. Therefore, different batch processing modes and different batch phase may have different behaviors and affect the final product quality. Based on the multi-phase residual recursive in a single mode, a two-direction residual recursive model is established for multi-phase and multi-mode batch processes in this paper. First, a phase mean model is established to represent the process variable correlations for the entire phase. According to the multi-phase characteristics, the residual prediction quality of each phase is obtained by recursion. The recursion is repeated in the within-batch direction. Then the same modeling strategy is applied for the inter-mode direction, and the two-direction residual recursive model for multi-phase and multi-mode batch processes is obtained. The application to the injection molding process has proved the effectiveness of the proposed method.