Non-Proportional mixed mode plastic zones via finite elements and artificial neural networks

Abstract

The plastic zone developed around a fatigue crack tip can affect both fatigue and fracture material behaviour. Predicting the plastic zone shape and size under cyclic conditions can, therefore, enhance fatigue crack propagation analysis. While there are theoretical solutions for the elastic stress and strain fields under pure or mixed mode conditions around the crack tip, solutions for plastic fields must be determined using experimental or numerical approaches. The Compact Tension Shear (CTS) specimen has been extensively used to analyse plastic zones under proportional conditions, but when non-proportional conditions are applied the number of necessary analyses for a reasonable understanding of the plastic zone shape and size around the crack tip can increase exponentially. To address this problem, a combined approach was used to reduce the number of required plastic zone simulations. First, a hand selected number of loading configurations were simulated using the Finite Element Method (FEM), predicting the plastic zone shape and size. Then, an Artificial Neural Network (ANN) was trained to predict the plastic zone under different conditions. Using only 18 configurations for 3 different loading conditions, the trained ANN was able to accurately predict the plastic zone shape and size for both tensile and shear propagation modes. The network can now be used to predict the plastic zone influence on fatigue and fracture behaviour, without the need for further numerical analysis. The paper results also show that crack propagation direction can be correlated and predicted using the applied loads and the resulting plastic zone.