Predictions of Additive Manufacturing Process Parameters and Molten Pool Dimensions with a Physics-Informed Deep Learning Model

• DOI: 10.1016/j.eng.2022.09.015
• Publication Date: 2023-02-16T18:04:34Z
• Authors: Mingzhi Zhao; Huiliang Wei; Yiming Mao; Changdong Zhang; Tingting Liu; Wenhe Liao
• Keywords: Multilayer perceptron; Perceptron

Molten pool characteristics have a significant effect on printing quality in laser powder bed fusion (PBF), and quantitative predictions of parameters molten dimensions are critical to the intelligent control complex processes PBF. Thus far, bidirectional been challenging due highly nonlinear correlations involved. To address this issue, we integrate an experiment characteristics, mechanistic model, deep learning achieve both forward inverse key during The provides fundamental data, model significantly augments dataset, multilayer perceptron (MLP) predicts process based dataset built from model. results show that can be realized, with highest prediction accuracies approaching 99.9% mean over 90.0%. Moreover, accuracy MLP is closely related dataset—that is, learnability has crucial impact accuracy. 97.3% enhancement via while 68.3% when using only experimental dataset. largely depends as well. research demonstrate feasible for PBF, offer novel useful framework determination conditions outcomes additive manufacturing.