An efficient spatiotemporal region-decomposition algorithm for transient coupled heat transfer in multi-body enclosed cavities

At present, in engineering applications, numerical simulation algorithms for transient coupled heat transfer in multi-body enclosed systems involving solid heat generation and conduction, gaseous natural convection, and surface radiation still struggle to achieve a good balance among stability, convergence, and computational efficiency. To address these issues, this study proposes a multi-time-step loosely coupled algorithm based on spatiotemporal domain decomposition and heat flux regulation. The proposed method partitions the domain into conduction, radiation, and natural convection regions, and establishes a Main-SubCycle structure: the MainCycle solves transient heat conduction in solids, while the SubCycle is triggered when the wall temperature rise exceeds a threshold, initiating quasi-steady fluid convection and extracting the coupled heat flux. To improve the physical consistency and numerical stability of the heat flux boundaries, linear interpolation and time-averaging strategies are employed to ensure continuous heat flux updates. In addition, a PID controller based on the wall temperature response rate is implemented to dynamically adjust the temperature rise threshold, enabling adaptive control of the coupling frequency. Comparative validation against tightly coupled simulations demonstrates that the proposed method achieves a temperature prediction error below 0.41% and a 3.32-fold computational speedup, thereby achieving a dynamic balance between accuracy and efficiency. Experimental validation is further conducted using a three-layer equipment enclosure model placed in a closed cavity. The measured temperatures at the top, middle, and bottom points agree well with the numerical predictions, with relative errors of 2.07%, 1.91%, and 1.90% at 1200 s, respectively. This approach is well suited for efficient numerical modeling of complex thermal boundaries and multi-scale coupled processes, offering a general and scalable solution for long-duration, multi-source, and multi-mechanism thermal environment simulations.