This content is not included in your SAE MOBILUS subscription, or you are not logged in.
Machine Learning Based Parameter Calibration for Multi-Scale Material Modeling of Laser Powder Bed Fusion (L-PBF) AlSi10Mg
ISSN: 2641-9637, e-ISSN: 2641-9645
Published April 06, 2021 by SAE International in United States
Event: SAE WCX Digital Summit
Citation: Li, Y., Li, Z., Lai, W., Xu, H. et al., "Machine Learning Based Parameter Calibration for Multi-Scale Material Modeling of Laser Powder Bed Fusion (L-PBF) AlSi10Mg," SAE Int. J. Adv. & Curr. Prac. in Mobility 3(4):1526-1534, 2021, https://doi.org/10.4271/2021-01-0309.
Rapid development of Laser Powder Bed Fusion (L-PBF) technology enables almost unconstrained design freedom for metallic parts and components in automotive industry. However, the mechanical properties of L-PBF alloys, AlSi10Mg for example, have shown significant differences when compared with their counterparts via conventional manufacturing process, due to the unique microstructure induced by extremely high heating and cooling rate. Therefore, microstructure informed material modeling approach is critical to fully unveil the process-structure-property correlation for such materials and enable the consideration of the effect of manufacturing during part design. Multi-scale material modeling approach, in which crystal plasticity finite element (CPFE) models were employed at the microscale, has been previously developed for L-PBF AlSi10Mg. However, calibration of parameters for CPFE based on macroscale mechanical testing was found to be challenging due to the complexity and the high computational cost of the models. In the present study, the authors developed a machine learning based approach to tackle this challenge. With the training set data collected from preliminary runs of the multi-scale material models, surrogate models with different machine learning algorithms were constructed. It was found that the surrogate models using Gradient Boosting Machine (GBM) algorithm can well capture the response of the complex finite element based multi-scale material models. In the meantime, the study also showed that performing feature engineering can greatly enhance the efficiency and robustness of machine learning models. Specifically, the surrogate models are trained to predict the explicit physical descriptors instead of the summed error between stress-strain curves from testing and prediction. The obtained surrogate models were then utilized to find the optimal values of the parameters in microscale CPFE. Validation runs of multi-scale models using the optimal parameters found through surrogate modeling showed minimized differences between prediction and macroscale mechanical testing and thus proved the effectiveness of the developed approach.