Combining artificial intelligence with different plasticity induced crack closure criteria to determine opening and closing loads on a three-dimensional centre cracked specimen

Fracture due to fatigue crack growth remains a significant failure mode in both brittle and ductile materials. When dealing with crack tip plasticity induced phenomena, characterized by high strain and stress field gradients, only highly refined meshes around the crack tip can produce accurate results. Therefore, optimized mesh parameters must be used, in order to achieve high quality models with low computational costs. In this study, artificial intelligence models and a numerical three-dimensional model for a middle tension specimen were combined to enhance crack closure and opening loads assessment. The numerical accuracy was analysed based on the estimated stress and strain fields, plastic zone shape and size and crack closure and opening load values. Two artificial neural networks were trained using four different crack lengths, mesh sizes and simulated plastic wakes. The networks were capable of stress and strain field predictions and crack opening and closure load determination. It was verified that the crack stress criterion is strongly correlated with the principal strain field and the displacement field around the crack tip, providing a viable way to analyse plasticity induced crack closure.