Estimation of high cycle fatigue life of additively manufactured Ti6Al4V using data analytics

The α-β Ti-6Al-4V (Ti64) alloy possesses high specific strength and excellent corrosion resistance properties, established its utilization in aerospace and biomedical applications. The additively manufactured Ti64 alloy has shown its rapid growth in the recent years because of its design flexibility and manufacturing efficiency. As this Ti alloy is used in aerospace and bio-medical applications, it experiences variable loading conditions during actual service conditions and therefore, it is essential to estimate its fatigue life. The fatigue experimentations are highly expensive and time-consuming procedure to evaluate the life of the component, especially in the case of Ti alloys which is costly. Hence, destructive testing of this alloy is not desirable to estimate its fatigue life. To overcome these limitations, efficient methodology is required to predict the cyclic life of Ti alloy quickly and economically. Machine learning (ML) technique is one of the feasible solutions to fulfill such requirements and predict fatigue life using experimental data reported in the literature. Keeping this in view, Extreme Gradient Boosting (XGB) and Random Forest (RF) model have been used to predict the fatigue life of Ti alloys manufactured through laser powder bed fusion (LPBF). These XGB and RF have been trained using fatigue data of Ti alloy, which is influenced by process variables namely, Energy Density during manufacturing and Stress Amplitude during testing. The collected fatigue data is split in to train (80%) and test data (20%) and the XGB and RF are trained using the former and its predictive accuracy is estimated with the quantification of error. Subsequently, the trained XGB and RF are used to analyse the test data. The accuracy of XGB and RF for predictability of fatigue life of Ti alloy is assessed by using mean squared error and R2 scores.