A domain-decomposed PINN method for identifying piecewise moving heat-flux boundary conditions

This study proposes a domain-decomposed physics-informed neural network method for identifying piecewise moving heat-flux boundary conditions in thermite-induced transient heat transfer. The axial domain is divided into three subdomains, each with an independent heat-flux profile and propagation speed. The local heat-flux profiles are parameterized using radial basis functions and identified from outer-wall temperature measurements. Interface continuity conditions are imposed to ensure physical consistency between adjacent subdomains. Numerical results show that the proposed method can recover the main features of the piecewise heat-flux profiles and identify the local propagation speeds with relative errors within approximately 5 %. The method provides a flexible framework for describing locally non-uniform moving thermal inputs.