A multiphysics deep energy method for fourth-order phase-field fracture with piezoresistive self-sensing

Abstract

Self-sensing conductive composites can reveal deformation and damage through measurable changes in electrical resistance, which makes them attractive for embedded diagnostics and learning-enabled structural health monitoring. This paper presents a physically consistent multiphysics Deep Energy Method (DEM) for brittle fracture in piezoresistive materials. The mechanical part is modeled by small-strain linear elasticity coupled to a fourth-order AT2-type phase-field fracture functional with tensile/compressive energy split and history-field irreversibility. To avoid artificial energetic mixing of mechanical and electrical quantities, the electrical problem is treated as a one-way coupled sensing subproblem: after solving the mechanics--fracture problem, the electric potential is obtained from a steady conduction problem whose conductivity depends on strain through a linearized piezoresistive law and on damage through a crack-induced conductivity degradation. The resulting formulation predicts crack evolution together with its resistance signature without assigning the electrical field an artificial crack-driving role. DEM is used to minimize the variational subproblems over admissible neural trial spaces with exact imposition of essential boundary conditions. A lean verification suite is used to validate the electrical building blocks and the fracture engine separately, followed by a numerical study of a tensile plate with stress concentrators and electrodes. In that study, the framework captures a nontrivial sensing regime in which appreciable damage growth leaves the global resistance nearly unchanged, followed by a sharp resistance increase once dominant conductive ligaments are disrupted and current paths reorganize strongly.