A multi-kernel-based spatiotemporal modeling approach for energy transfer of complex thermal processes and its applications

• Published in: International journal of heat and mass transfer 2023-12, Vol.216, p.124597, Article 124597
• Main Authors: Xu, Bowen, Lu, Xinjiang, Bai, Yunxu, Xu, Du, Cui, Xiangbo
• Format: Article
• Language: English
• Subjects: Distributed parameter system; Forging process; Lithium-ion battery; Multi-kernel model; Spatiotemporal LS-SVM; Thermal processes
• ISSN: 0017-9310; 1879-2189
• DOI: 10.1016/j.ijheatmasstransfer.2023.124597

•A Gaussian mixture model strategy is developed to extract multi-features from the original spatiotemporal input/ output data.•Nonlinear model for each feature is constructed by LS-SVM method and the corresponding parameters are identified by cross-validation algorithm.•A novel spatial multi-kernel functions is developed to characterize the nonlinear spatial dynamics.•A multi-kernel spatiotemporal model is proposed for the nonlinear dynamics of DPSs in both space and time using data mapping technology. Most of actual thermal processes work in a large-scale operation region. In each region, the processes are along with complex energy transfer and featured by nonlinear distributed parameter system (DPS). Also, the dynamics at different regions is different, and the energy transfer between neighboring regions is complex. All these factors increase the modeling difficulty of thermal processes. In this paper, a multi-kernel spatiotemporal modeling approach is proposed to reconstruct dynamics of this kind of processes. First, a multi-kernel global spatial function construction method is developed to extract the global spatial features of thermal dynamics. For each local region, it constructs the corresponding spatial kernel function to represent the energy exchange on space. In order to reflect the global spatial feature, these spatial kernel functions are integrated by using Gaussian mixture strategy to form a global spatial function. Then, the temporal dynamics can be obtained and modeled by projecting the spatiotemporal data on this global spatial function. Integrating the global spatial function and temporal model, the spatiotemporal model is constructed for this process with large-scale operation region. The performance of this method is verified by theoretical analysis. Using experiments, its modeling ability for thermal processes are demonstrated in comparison with two commonly used modeling methods.