Laser Additive Manufacturing- Direct Energy Deposition of Ti-15Mo Biomedical Alloy: Artificial Neural Network Based Modeling of Track Dilution

Abstract

Direct energy deposition is one of the most popular laser additive manufacturing processes. However, the generation of geometry (or dilution and the resulting porosity) of the deposited tracks is not well understood as it is largely dependent on the laser process settings. In this work, the track dilution generated in single track deposition of Ti-15Mo alloy using the laser additive manufacturing – direct energy deposition (LAM-DED) process is accurately modeled and predicted, using the Matlab software based artificial neural network (ANN). Targeting a mean square error of zero, a 3–7-1 feed-forward, back-propagation, NN architecture with Levenberg-Marquardt training algorithm is developed. The percentage prediction error is calculated to validate the developed ANN model and is found to range between −0.981 and 0.584 establishing the goodness of the fitted model and the high prediction accuracy of track dilution. Further, regression analysis is also conducted for theoretical validation of the experimental track dilution. The correlation coefficient value obtained for 27 data is 0.982 which establishes high confidence in the fitted data. This study helps in accurate software based prediction of track dilution at different LAM-DED process parameter settings for possible fabrication of superior quality (low porosity) Ti-15Mo bio-implants.