Point-Cloud Neural Network Using Transfer Learning-Based Multi-Fidelity Method for Thermal Field Prediction in Additive Manufacturing
Proceedings of the ASME 2021 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference. Volume 3A: 47th Design Automation Conference (DAC)., V03AT03A038
*Corresponding author
Cite
@proceedings{10.1115/DETC2021-67963,
author = {Huang, Xufeng and Hu, Zhen and Xie, Tingli and Wang, Zhuo and Chen, Lei and Zhou, Qi},
title = "{Point-Cloud Neural Network Using Transfer Learning-Based Multi-Fidelity Method for Thermal Field Prediction in Additive Manufacturing}",
volume = {Volume 3A: 47th Design Automation Conference (DAC)},
series = {International Design Engineering Technical Conferences and Computers and Information in Engineering Conference},
year = {2021},
month = {08},
abstract = "{Melt pool modeling is critical for model-based uncertainty quantification (UQ) and quality control in metallic Additive Manufacturing (AM). Finite element (FE) simulation for thermal modeling in metal AM, however, is tedious and time-consuming. This paper presents a multi-fidelity point-cloud neural network method (MF-PointNN) for surrogate modeling of melt pool based on FE simulation data. It merges the feature representations of low-fidelity (LF) analytical model and high-fidelity (HF) FE simulation data through the theory of transfer learning (TL). A basic PointNN is firstly trained using LF data to construct correlation between the inputs and thermal field of analytical models. Then, the basic PointNN is updated and fine-tuned using the small size of HF data to build the MF-PointNN. The trained MF-PointNN allows for efficient mapping from input variables and spatial positions to thermal histories, and thereby efficiently predict the three-dimensional melt pool. Results of melt pool modeling of electron beam additive manufacturing (EBAM) of Ti-6Al-4V under uncertainty demonstrate the efficacy of the proposed approach.}",
doi = {10.1115/DETC2021-67963},
url = {https://doi.org/10.1115/DETC2021-67963},
note = {V03AT03A038},
eprint = {https://asmedigitalcollection.asme.org/IDETC-CIE/proceedings-pdf/IDETC-CIE2021/85383/V03AT03A038/6800892/v03at03a038-detc2021-67963.pdf},
}
Abstract
Melt pool modeling is critical for model-based uncertainty quantification (UQ) and quality control in metallic Additive Manufacturing (AM). Finite element (FE) simulation for thermal modeling in metal AM, however, is tedious and time-consuming. This paper presents a multi-fidelity point-cloud neural network method (MF-PointNN) for surrogate modeling of melt pool based on FE simulation data. It merges the feature representations of low-fidelity (LF) analytical model and high-fidelity (HF) FE simulation data through the theory of transfer learning (TL). A basic PointNN is firstly trained using LF data to construct correlation between the inputs and thermal field of analytical models. Then, the basic PointNN is updated and fine-tuned using the small size of HF data to build the MF-PointNN. The trained MF-PointNN allows for efficient mapping from input variables and spatial positions to thermal histories, and thereby efficiently predict the three-dimensional melt pool. Results of melt pool modeling of electron beam additive manufacturing (EBAM) of Ti-6Al-4V under uncertainty demonstrate the efficacy of the proposed approach.