Defect Assessment And Life Prediction Of Laser Additively Manufactured Ti-6Al-4V Alloy Using Support Vector Machine

Additive manufacturing(AM)technology has obvious advantages in forming and manufacturing complex structural parts due to its high design freedom and high material utilization,and has been widely used in aerospace,biomedical and other fields.It also shows great promise in the production of complex structures for highspeed trains.However,it has been found that the presence of internal defects in additive manufacturing metallic materials affects the fatigue performance of the materials,and their number,size,location distribution and morphology are complex,so it is difficult to analyze the influence of defects on the fatigue life of the materials by traditional methods.In recent years,the development of machine learning methods has provided a new way of thinking to evaluate the effect of defects on the fatigue life of additively manufactured metals.In this paper,the Selective Laser Melting(SLM)processed Ti-6Al-4V alloy is used as the research material,and its microstructure and basic mechanical properties are evaluated through various experiments.The size,location and morphological distribution of the internal defects of Ti-6Al-4V alloy were characterized by synchrotron X-ray microtomography,and the fatigue data of the specimens were obtained by fatigue tests.Based on the experimental results,a support vector regression(SVR)machine learning model was established from the principle of support vector machine(SVM),and the optimal parameters of the model were obtained by grid search and crossvalidation.The defect feature data and fatigue life data were first used to train the model,and then the prediction was performed.The coefficient of determination between the predicted and experimental life of the test set is 0.92,which indicates that the SVR model based on the defect features has a high accuracy in predicting the fatigue life of SLM formed Ti-6Al-4V alloy.The comparison with the results of other machine learning algorithms shows that the SVR model has a good generalization capability and is suitable for engineering applications.The SVR model was used to evaluate the effect of different features of the defects on the fatigue life.The results showed that all three features have an effect on the fatigue life,and the influence degree of defect features in order of decreasing significance is the defect’s morphology,dimension,and position.