Investigating deep energy method applications in thermoelasticity

This study presents a novel exploration into the deep energy method for addressing thermoelasticity problems. The deep energy method has recently been recognized as a robust numerical technique, demonstrating superior capability in managing intricate nonlinearities and delivering highly accurate results. Our research aims to extensively investigate the influence of network attributes, such as layers, neurons, and activation functions, on the accuracy of this method. Our technique's reliability and innovation are substantiated through the successful resolution of both 1D and 2D thermoelasticity problems. Additionally, our study delves into the use of coupled and sequential algorithms to determine the optimal combinations of activation functions, layers, and neurons, enabling the sequential method to efficiently achieve accurate results. Our technique displays remarkable consistency when compared with established analytical solutions and empirical evidence, thus highlighting the flexibility and efficacy of our approach. This research offers a potentially superior alternative to traditional numerical techniques in tackling complex multi-physics phenomena. The contributions of our work hold the potential to significantly influence the realms of computational physics and engineering, ushering in advancements in precision and efficacy when addressing diverse challenges.