Thermomechanical fatigue behavior and lifetime prediction of nickel-based single crystal alloys under varying stress conditions

This study systematically investigates the thermomechanical fatigue (TMF) behavior of DD6 nickel-based single-crystal superalloys under varying stress conditions and obtains lifetime distribution data for two distinct phases. Fractographic and microstructural analyses reveal the failure mechanisms of the alloy at different stages. Furthermore, two machine learning-based lifetime prediction methods are proposed. The first method compares the predictive performance of multiple machine learning models, identifying the most effective model and conducting a detailed analysis of the most influential energy-related input features. The second method integrates a sequence learning model with a backpropagation neural network (BPNN), incorporating an attention mechanism to enhance prediction accuracy and generalization capability. The results demonstrate a strong correlation between experimental data and predictions, confirming the effectiveness of both approaches in TMF lifetime prediction. Notably, the sequence learning-based hybrid model outperforms in terms of accuracy and applicability, highlighting its potential for broad engineering applications.