Nonlocal multiaxial fatigue model based on artificial neural networks for predicting fretting fatigue life of dovetail joints

[Display omitted] •A nonlocal multiaxial fatigue model leveraging artificial neural networks (ANN) is proposed to forecast the fretting fatigue life of dovetail joints.•This model demonstrates exceptional accuracy in predicting the fretting fatigue life of dovetail samples within a 1.5× limit band.•This accuracy stems from combining the representation capabilities of ANN approaches with the integration of physics or domain knowledge. Fretting fatigue of dovetail joints is of paramount importance for ensuring equipment safety, where the swift and precise estimation of their fatigue life is crucial. In this study, we present a nonlocal multiaxial fatigue model based on artificial neural networks (ANN) to tackle these challenges. Initially, the damage parameters were calculated using critical plane approaches (CPA) and theory of critical distance (TCD) analysis, and the fretting fatigue stress was computed. Subsequently, these parameters were integrated as input features in the ANN model to predict the fretting fatigue life of dovetail joints. The predicted results demonstrate that this proposed model can accurately predict the fretting fatigue life of dovetail samples within a 1.5× limit band. Furthermore, a comparative analysis with other ANN models inspired by previous researchers also supports this viewpoint. This capability stems from its integration of ANN representation capabilities with physics and domain knowledge, such as CPA/TCD methods and fretting fatigue stress analysis. This approach not only establishes a theoretical foundation for predicting the fretting fatigue life of dovetail samples but also showcases promising practical applications.