Additive manufacturing energy consumption measurement and prediction in fabricating lattice structure based on recallable multimodal fusion network. (15th June 2022)

• Record Type: Journal Article
• Authors: Wang, Kang; Song, Youyi; Huang, Zhihao; Sun, Yibo; Xu, Jinghua; Zhang, Shuyou
• Is Part Of: Measurement. Volume 196(2022)
• Publication Date: 2022-06-15
• Subjects: Energy consumption measurement -- Additive manufacturing -- Lattice structure -- Multimodal fusion -- Laminated context recall; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111215
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: Energy shortage and excessive carbon dioxide emission caused by energy consumption in additive manufacturing (AM) have been increasingly severe and widely concerned. To address this issue, an additive manufacturing energy consumption (AMEC) measurement and prediction method for fabricating lattice structure based on Recallable Multimodal Fusion Network (RMFN) is proposed. The AMEC measurement model in 3D fabrication is first constructed according to the operating characteristics of 3D printer. As the backbone module of RMFN, the Multimodal Data Fusion Framework (MDFF) is then developed to predict AMEC by fusing the processing-, pixel- and geometric-level datasets, which are both generated during the design process of AM. In the light of the layer-wise fabrication principle in AM, a Laminated Context Recall Network (LCRN) is further designed to elegantly enforce the consistency of the contextual information among sliced layers, improving the regression accuracy of the AMEC prediction. Extensive numerical and physical experiments demonstrate that the proposed method performs better than state-of-the-art methods, motivating AM sustainability improvements and environmental performance. Highlights: Recallable Multimodal Fusion Network (RMFN) is proposed to predict energy consumption. Multiple modalities are merged by Multimodal Data Fusion Framework (MDFF). Laminated Context Recall Network (LCRN) is designed to distill geometry modality. Extensive measurement results signify the superiority of the proposed RMFN.