Physical Hierarchical Neural Network for Thermomechanical Fatigue Life Prediction of Compacted Graphite Cast Iron

ABSTRACT

A Physical Hierarchical Neural Network (PHNN) model was developed in this study for material‐level thermomechanical fatigue (TMF) life prediction. First, TMF experiments were conducted on compacted graphite cast iron (CGI) to investigate the evolution of its mechanical properties during the testing process. Subsequently, existing low‐cycle fatigue (LCF) tests were incorporated, and the input data were unified to establish a low‐cycle‐thermomechanical fatigue (LCF‐TMF) experimental database. Finally, the predictive capability of the PHNN model for material‐level TMF life prediction was evaluated and compared with that of the Sehitoglu model. The results indicate that the PHNN model exhibits higher prediction accuracy, stronger nonlinear fitting capability, and better extrapolation performance than the Sehitoglu model. These results demonstrate that the PHNN framework has been successfully applied to TMF life prediction of material, further enhancing its generality.