Tensile behaviour prediction of laser powder bed fusion with engineered process parameters using a feedforward neural network

In the past few years, there are an increasing number of studies applying machine learning algorithms to laser powder bed fusion additive manufacturing process-property problems due to the complex physics behind. Most of the experiments are designed for human learning rather than for machine learning. Moreover, the authors identified an opportunity to utilise the massive number of studies regarding the physical phenomena in the process to predict the tensile properties. In this paper, a randomised experimental dataset is designed, and the resulting tensile properties demonstrate large variations. Despite the process parameters for laser powder bed fusion process, several engineered process parameters from literature have been fed to a deep feedforward neural network to predict the tensile behaviours. These engineered process parameters contribute to the high precision of predictions.