A novel damage mechanics and XGBoost based approach for HCF life prediction of cast magnesium alloy considering internal defect characteristics

•The characterization model of the effects of internal defects is presented.•The fatigue damage model considering the effects of internal defects is derived.•Experimental studies are investigated for the HCF behavior of casting alloys.•The XGBoost machine learning model is employed for the HCF life prediction. In this study, a novel method integrating machine learning with damage mechanics is proposed for predicting the high cycle fatigue life of ZM6 with internal defects. First, the characterization model of the effects of internal defects is presented, and the relationship between the stress concentration factor and the parameters of the ellipsoidal void is established. After that, the internal defect hazard factor is proposed to reflect the influence of ellipsoidal void on the fatigue behavior, and the fatigue damage model considering the effects of internal defects is derived. Second, the fatigue tests of ZM6 are carried out on smooth specimens with different stress ratios, and the 3D morphology of internal defects in standard test specimens is acquired by utilizing X-ray scans. A calibration method for material parameters based on the particle swarm optimization (PSO) algorithm is proposed. Based on the validated new damage mechanics model, a dataset is then generated to train the eXtreme Gradient Boosting (XGBoost) machine learning model. Lastly, the high cycle fatigue life of ZM6 is predicted with the given stresses, stress ratios, and hazardous defect coefficients, and a high prediction accuracy is achieved. The effects of the number of training samples and hyper-parameters on prediction performance are further investigated.