Adaptive mesh reorientation for concrete fracture analysis using cohesive zone models with interface elements

Cohesive zone models with interface elements are widely used for simulating crack initiation and propagation, but their practical reliability is often limited by mesh bias: cracks can only propagate along predefined elements' boundaries, so predicted crack paths and global response may be strongly controlled by mesh orientation. In badly oriented automatically generated meshes, this can make physically admissible crack trajectories inaccessible and lead to inaccurate fracture patterns and artificial stiffness.This work introduces a novel adaptive mesh reorientation method to overcome this limitation while preserving the core strengths of interface-element formulations: discrete crack representation, natural crack initiation, branching, coalescence, and closure. No crack tracking and no predefined crack initiation locations are required. After each load step, mesh nodes are relocated so that intact interface elements progressively align with locally preferred fracture directions derived from the current stress state. The update is posed as a constrained discrete variational minimization problem and solved in a staggered manner after the mechanical equilibrium step.The approach is applicable in two and three dimensions. Its performance is demonstrated on established concrete fracture benchmarks, including heterogeneous mesoscale cases. The proposed method is robust and well suited for practical engineering analysis. It is readily integrated into standard finite element frameworks and directly applicable to complex geometries, material heterogeneity, and large-scale structural fracture simulations. Across seven benchmarks considered, the results consistently show improved crack-path prediction, reduced mesh sensitivity, and capture of crack initiation in the interior of the domain.