Fast scanning pattern selection in laser directed energy deposition via simulation data-driven based deep regression method

doi:10.1016/j.jmapro.2024.11.072

IMR OpenIR

	Fast scanning pattern selection in laser directed energy deposition via simulation data-driven based deep regression method
	Ma, Liang 1,2; Kong, Xiangwei 1,4,5; Cheng, Liu 1; Liang, Jingjing 2; Li, Jinguo 2; Jin, Zhibo 1; Jiao, Zhidong 3; Zhang, Xue 2; Sun, Cong 1
通讯作者	Kong, Xiangwei(shawnkongneu@163.com) ; Liang, Jingjing(jjliang@imr.ac.cn) ; Li, Jinguo(jgli@imr.ac.cn)
	2025-01-17
发表期刊	JOURNAL OF MANUFACTURING PROCESSES
ISSN	1526-6125
卷号	133 页码:367-379
摘要	Thermal accumulation in the laser directed energy deposition (LDED) process is a crucial issue that can significantly affect the microstructure and texture of the deposited layer. In addition to the process parameters (including beam type, power, and scanning speed), the scanning strategy is also a significant factor that affects the temperature field. Most scanning strategies used in the past are heuristic methods due to the computation limit of the finite element model. In this paper, an optimal framework for the fast selection of scanning patterns for the LDED process is proposed. Firstly, a temperature field dataset is built based on the finite element simulation results of random scanning patterns. Secondly, a deep regression model based on skip-connected 3D convolutional autoencoder (SC-3DCAE) is developed and trained by the dataset for temperature field prediction. Afterward, optimal scanning patterns are quickly selected from randomly generated candidates using the validated SC-3DCAE model based on the thermal variance criterion. Finally, finite element simulations and a 50layer thin-wall deposition process are performed with the recommended optimal scanning pattern. The trained SC-3DCAE model shows high accuracy and fast prediction of the temperature field. The optimized scanning pattern achieves higher cooling rates and temperature gradients compared to the continuous deposition pattern. Smaller dendrite spacing and grain size are found in the samples with the optimized scanning pattern.
关键词	Laser directed energy deposition 3D regression networks Deep learning Texture Scanning pattern selection Thermal field prediction
资助者	National Science and Technology Major Project ; National Key Research and Development Program of China ; State Ministry of Science and Technology Innovation Fund of China
DOI	10.1016/j.jmapro.2024.11.072
收录类别	SCI
语种	英语
资助项目	National Science and Technology Major Project[2019-VII-0019-0161] ; National Key Research and Development Program of China[SQ2019YFB1704500] ; State Ministry of Science and Technology Innovation Fund of China[2018IM030200]
WOS研究方向	Engineering
WOS类目	Engineering, Manufacturing
WOS记录号	WOS:001370497500001
出版者	ELSEVIER SCI LTD
引用统计
文献类型	期刊论文
条目标识符	http://ir.imr.ac.cn/handle/321006/181313
专题	中国科学院金属研究所
通讯作者	Kong, Xiangwei; Liang, Jingjing; Li, Jinguo
作者单位	1.Northeastern Univ, Sch Mech Engn & Automat, Shenyang 110819, Peoples R China 2.Chinese Acad Sci, Inst Met Res, Shenyang 110016, Peoples R China 3.CRRC Qingdao Sifang Co Ltd, Qingdao 266000, Peoples R China 4.Northeastern Univ, Key Lab Vibrat & Control Aeroprop Syst, Minist Educ, Shenyang 110819, Peoples R China 5.Northeastern Univ, Liaoning Prov Key Lab Multidisciplinary Design Opt, Shenyang 110819, Peoples R China
推荐引用方式 GB/T 7714	Ma, Liang,Kong, Xiangwei,Cheng, Liu,et al. Fast scanning pattern selection in laser directed energy deposition via simulation data-driven based deep regression method[J]. JOURNAL OF MANUFACTURING PROCESSES,2025,133:367-379.
APA	Ma, Liang.,Kong, Xiangwei.,Cheng, Liu.,Liang, Jingjing.,Li, Jinguo.,...&Sun, Cong.(2025).Fast scanning pattern selection in laser directed energy deposition via simulation data-driven based deep regression method.JOURNAL OF MANUFACTURING PROCESSES,133,367-379.
MLA	Ma, Liang,et al."Fast scanning pattern selection in laser directed energy deposition via simulation data-driven based deep regression method".JOURNAL OF MANUFACTURING PROCESSES 133(2025):367-379.