Knowledge transfer using Bayesian learning for predicting the process-property relationship of Inconel alloys obtained by laser powder bed fusion
Cuiyuan Lu, Xiaodong Jia, Jay Lee, Jing Shi
Abstract
In this study, we investigate the transferability of the process-property relationship between two Inconel alloys for laser powder bed fusion (LPBF). By developing a Bayesian learning approach, the process-property model of Inconel 625 learned from Inconel 718 demonstrates high accuracy with R of 0.95, which verifies the feasibility of this innovative concept. It is further found that the accuracy of the knowledge transfer model of Inconel 625 is increased if the data on Inconel 625 is more abundant. In this regard, the mean RMSE and MAE on relative density are decreased by about 0.45% and 0.35% when the Inconel 625 dataset size is increased from 15 to 60. In addition, both accuracy and robustness of the Inconel 625 model are increased with transferred knowledge from Inconel 718 regardless of the dataset size of Inconel 625, in which the mean RMSE and MAE are decreased by up to 0.7% and 0.5%, respectively.