Multi-Incremental Learning-Based Predictive Modeling for Unknown Distributed Parameter Systems Under Larger Working Region

Distributed parameter systems (DPS) are broadly present in numerous industrial manufacturing systems. Accurate modeling of DPS is critical for subsequent process monitoring and optimization. However, conventional modeling methods often ignore the larger working region in complex DPS. Besides, the inherent time-varying dynamic behavior of the system also brings challenges to spatiotemporal modeling. In this article, a new multi-incremental learning-based predictive modeling approach is proposed to solve the above concerns. First, the larger global working region is adaptively decomposed into multiple subspaces to extract local dynamics hierarchically. Then, a spatiotemporal forgetting-based incremental modeling method is further designed to cope with time-varying dynamics of local subspace. Finally, the global dynamic model is ensembled via multiple locally weighted incremental models to enhance modeling performance. Experiments on an industrial curing system demonstrated the effectiveness and superiority of the proposed modeling approach.