文献类型：期刊文献

英文题名：3D convolutional neural network based multi-modal data fusion for fatigue life prediction of additively manufactured Ti-6Al-4V alloy

作者：Zhang, Jianrui[1,2];Wu, Qimin[1];Wang, Haijie[1];Fan, Enxiang[3];Li, Bo[1,2]

机构：[1]East China Univ Sci & Technol, Sch Mech & Power Engn, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Addit Mfg & Intelligent Equipment Res Inst, Shanghai 200237, Peoples R China;[3]Shanghai Elect Grp Co Ltd, Cent Acad, Shanghai 200070, Peoples R China

年份：2026

卷号：209

外文期刊名：INTERNATIONAL JOURNAL OF FATIGUE

收录：;EI(收录号：20261220316580);WOS:【SCI-EXPANDED(收录号:WOS:001727242500001)】；

基金：This research work is sponsored by National Natural Science Foun-dation of China (No. 52205155 and No. 12411530109) , National Key R & D Program of China (No. 2024YFF0505204) , Postdoctoral Fellow-ship Program (Grade C) of China Postdoctoral Science Foundation (No. GZC20250420) , Shanghai Explorer Program (No. 24TS1411800) , Fundamental Research Funds for the Central Universities in China (No. JKG0123110) .

语种：英文

外文关键词：Fatigue life prediction; Multimodal fusion; Defect; 3D convolutional neural network; Additive manufacturing

摘要：The fatigue performance of additively manufactured (AM) components is influenced by multiple interdependent factors, posing substantial challenges to the accurate prediction of their fatigue life. In this work, a multimodal data fusion model based on 3D convolutional neural networks (3D-CNN) is developed to predict the fatigue life of laser powder bed fusion (LPBF) additively manufactured Ti-6Al-4V alloy. By concatenating 3D defect voxels of the region of interest (ROI) with one-dimensional (1D) modal data-including process parameters, mechanical properties, and loading conditions-this model integrates both text and image attributes, facilitating the crossmodal collaborative modeling and realizing end-to-end fatigue life prediction. To validate the predictive performance of the multimodal model, a defect feature- based deep neural network (DNN) is further developed as a comparative baseline. The results demonstrate that the multimodal model achieves higher accuracy in fatigue life prediction. This multimodal data fusion framework establishes a scalable modeling paradigm for fatigue life prediction driven by multi-source information.