Page 34 - ESAM-1-4

Engineering Science in
Additive Manufacturing Machine learning for biomedical metal AM

Figure 7. Workflow utilized to predict mechanical properties, including data collection, feature extraction, machine learning model training, predictive
tasks, and model analysis. 86
Abbreviations: AI: Artificial intelligence; E modulus: Elastic modulus; GB: Gradient boosting; GPM: Gaussian process model; LR: Linear regression;
NN: Neural network; RF: Random forest; SHAP: Shapley additive explanations; SVM: Support vector machine; UTS: Ultimate tensile strength; YS: Yield
strength.

A B

Figure 8. Swarm plot (A) and mean SHAP plot (B) for the XGBoost model predicting yield strength. 86
Abbreviations: SHAP: Shapley additive explanations; XGBoost: Extreme gradient boosting.

2.5. Fatigue life forward prediction enhance prediction accuracy. For instance, Zhan et al.
89
Fatigue is also a predominant cause of metal failure, developed a hybrid framework combining continuum
particularly for biomedical metallic materials serving in damage mechanics with a random forest model. Using
environments under cyclic loading. 87,88 Consequently, the stress concentration factor, maximum stress, and
accurate prediction of fatigue life is crucial for ensuring the stress ratio as inputs, the model effectively predicted the
long-term safety and reliability of implants. In recent years, fatigue life of additively manufactured titanium alloy,
hybrid approaches integrating physical models with data- demonstrating significantly higher accuracy than pure
driven methods have emerged as a promising pathway to physical models. This method proves particularly suitable

Volume 1 Issue 4 (2025) 12 doi: 10.36922/ESAM025440031