Machine learning application for optimization of laser directed energy deposition process for aerospace component rapid prototyping in additive manufacturing

Abstract. The paper proposes a methodology for determining the optimal L-DED parameters based on the minimum number planned of L-DED trials. A dataset compiled from planned L-DED experiments was used to train a machine learning model. The algorithm demonstrated a robust ability to predict the output metrics with notable accuracy and proposed a theoretical framework that modeled the complex relationships between the input variables and the resulting critical welding properties for AM. The application of the developed model and its comparison with conventional methods thus offers a methodical approach to determining the optimum process parameters in advance. This is a step towards the development and production of additively manufactured components for future digital twin application in the aerospace industry.