Exploiting Machine Learning Techniques to Predict the Stainless Steel Density Produced by Selective Laser Melting Additive Manufacturing

Porosity is one of the inherent defects that results from the Selective Laser Melting (SLM) additive manufacturing technique. The porosity related to fusion-solidification kinetics, results most often from non-optimally or poorly controlled manufacturing parameters. The density, a porosity indicator, affects the mechanical properties of the manufactured material (fatigue strength, cracks, deformations, etc.). Stainless steels are among the first materials developed by SLM because of their various industrial applications. Therefore, obtaining high-density steels remains very challenging and important to minimize the porosity impact on the mechanical properties. Nowadays, researchers are using machine-learning models to predict the rate of porosity or the maximum density of these materials as a function of the manufacturing parameters.
This work brings the advantage of machine-learning-based techniques such as artificial neural network, K Nearest Neighbor algorithm, and auto-encoder into additive manufacturing. The authors propose advanced techniques to develop their models, which predict the density of stainless steels given manufacturing parameters including laser power, scanning speed, hatch spacing and layer thickness. They validate their models using two indicators: the mean squared error and the coefficient of determination.