A novel framework using FEM and machine learning models with experimental verification for Inconel-718 rapid part qualification by laser powder bed fusion

Muhammad Arif Mahmood,Usman Tariq
DOI: https://doi.org/10.1007/s00170-023-12383-0
IF: 3.563
2023-10-04
The International Journal of Advanced Manufacturing Technology
Abstract:This study establishes a novel printability criterion for Inconel-718 parts by laser power bed fusion. For this purpose, the regions with D / t ≤ 1.15, L / W > 2.1, and W / D = 2.0 have been identified with lack-of-fusion, balling, and keyhole defects. Regimes within the processing maps related with defects have been regarded as a melt pool geometry function, derived using a FEM model with temperature-dependent thermophysical properties. The data was collected, via design of experiment technique, using validated simulation model. Following that, the acquired data was utilized to train and test a machine learning model based on a backpropagation artificial neural network (ANN). By linking melt pool dimensional ratios to defects, the validated ANN model was used to produce processing maps. The processing maps were validated using experimental analyses, which revealed a consistent correlation between experiments and simulations. The proposed processing maps can be utilized to quickly quantify the Inconel-718 parts generated by the LPBF.
engineering, manufacturing,automation & control systems
What problem does this paper attempt to address?