A comprehensive investigation on application of machine learning for optimization of process parameters of laser powder bed fusion-processed 316L stainless steel

Abstract

Metal 3D printing has gained a lot of attention among industries since it offers a practical solution to problems rising during manufacturing of parts and components with complex geometry. This is an additive technology that eliminated several fabrication steps and at the same time reduces material waste during manufacturing process. However, in all additive manufacturing technologies, the final properties of the parts are determined by the operational process parameters. In this study, several machine learning algorithms were examined to characterize the effects of the printing process parameters on relative density, hardness, yield strength, and tensile strength in manufactured parts. It was possible by using "Big Data" collected from a large number of previously published articles on application of Laser Powder Bed Fusion (LPBF) for 3D printing of 316L stainless steel samples. Among different process parameters, laser power, laser energy density, and scanning speed were proven to have the largest effects directly on physical and mechanical properties of LPBF processed parts. Six different classification models and five support vector machine regression-based models were tested to find the most accurate prediction algorithm. To validate the obtained results from the applied machine learning models, a set of 316L specimens were produced using LPBF technology using a random set of process parameters. The physical and mechanical properties of 3D printed samples were tested and compared to the ones those predicted from the optimum models from machine learning analysis. The results were in great agreement, which shows the high accuracy of the developed machine learning algorithms in this study.