Data-driven Thermo-mechanical Interfaces: Thermal Mismatch Induced Interface Debonding

Abstract

In this work, a model-free data-driven framework is proposed to simulate interface debonding induced by the thermal mismatch. The material phase space is expanded and a new thermo-mechanical distance norm is formulated, incorporating three canonical conjugate pairs: traction–separation, temperature–internal energy, and temperature gradient–heat flux. The thermo-mechanical coupling interactions of interface are naturally encoded by the given material database and the complex constitutive relations are not necessary herein. The proposed data-driven is enriched by an internal variable-based parameterization to construct a time-evolving database for enforcing monotonic increase of interface damage evolution. Several numerical examples are provided to evaluate the accuracy and efficiency of the proposed approach. First, the thermo-mechanical data-driven model is benchmarked against classical finite element results, with convergence behavior analyzed in detail. Second, parametric studies are conducted on key interfacial properties, including superficial heat capacity and interfacial thermal conductivity, demonstrating the capability of the developed data-driven method to capture essential features of heat transfer across interfaces. Finally, the proposed framework is applied to simulate interface debonding driven by thermal mismatch in different cases, successfully capturing the coupled evolution of heat conduction and interfacial degradation. The results confirm that the data-driven formulation provides a novel numerical tool to describe the interplay between thermal fields and damage processes at material of interfaces.