基于时空神经网络的金属疲劳裂纹扩展预测

doi:10.16183/j.cnki.jsjtu.2024.090

上海交通大学学报 ›› 2026, Vol. 60 ›› Issue (3): 511-521.doi: 10.16183/j.cnki.jsjtu.2024.090

• 航空航天 • 上一篇

基于时空神经网络的金属疲劳裂纹扩展预测

梁佳铭, 余音(), 胡祎乐

上海交通大学航空航天学院;民机结构强度综合实验室, 上海 200240

收稿日期:2024-03-15 修回日期:2024-05-09 接受日期:2024-05-21 出版日期:2026-03-28 发布日期:2026-03-30
通讯作者: 余音,教授,博士生导师,电话(Tel.):021-34206640;E-mail:yuyin@sjtu.edu.cn
作者简介:梁佳铭(1999—),硕士生,从事疲劳裂纹扩展研究.
基金资助:
国家自然科学基金(U2241266)

Prediction of Fatigue Crack Growth in Metal Materials via Spatiotemporal Neural Network

LIANG Jiaming, YU Yin(), HU Yile

School of Aeronautics and Astronautics; Lab of Civil Aircraft Structures Testing, Shanghai Jiao Tong University, Shanghai 200240, China

Received:2024-03-15 Revised:2024-05-09 Accepted:2024-05-21 Online:2026-03-28 Published:2026-03-30

摘要/Abstract

摘要：

提出一种基于时空神经网络(STNN)的图像驱动模型,基于铝合金材料疲劳试验,开展了裂纹扩展预测研究.设计了边裂纹、0° 裂纹角度试验件,进行了15.0%极限载荷水平的疲劳试验.通过数字图像相关(DIC)设备拍摄试件变形图片,得到5 511帧位移场图像.对获得的试验图像数据进行了插值、数据增强和维度转换等,构建了STNN的数据集.分别利用卷积长短期记忆(Conv-LSTM)和SimVP两种STNN方法进行了疲劳裂纹扩展的预测,计算了其结构相似度指标(SSIM)和均方根误差(RMSE),比较了两种方法的预测准确性.结果表明,SimVP神经网络在测试阶段的预测效果更好,该方法可以准确预测疲劳裂纹扩展速率和路径,为损伤容限分析和结构监测周期的确定提供参考.

关键词: 时空神经网络, 图像驱动, 疲劳裂纹扩展, 疲劳试验, 数字图像相关

Abstract:

An image-driven model based on spatiotemporal neural network (STNN) is proposed for prediction of crack growth in aluminum alloy. Fatigue experiments with an initial edge crack angle of 0° and a 15.0% limit load level are designed, and images of specimen deformation are captured using digital image correlation (DIC) resulting in 5 511 frames of displacement field data used as datasets of STNN after interpolation, augmentation, and dimension-raising. Two neural netwroks, convolutional long short-term memory (Conv-LSTM) and SimVP, are employed to predict the fatigue crack growth, with their prediction accuracies further compared based on the structural similarity index measure (SSIM) and the root mean square error (RMSE). The results show that the SimVP neural network performs better in the test stage predicting fatigue crack growth rate and propagation path. This method provides a reference for damage tolerance analysis and determination of inspection intervals for structures.

Key words: spatiotemporal neural network (STNN), image driven, fatigue crack growth, fatigue experiment, digital image correlation (DIC)

中图分类号:

梁佳铭, 余音, 胡祎乐. 基于时空神经网络的金属疲劳裂纹扩展预测[J]. 上海交通大学学报, 2026, 60(3): 511-521.

LIANG Jiaming, YU Yin, HU Yile. Prediction of Fatigue Crack Growth in Metal Materials via Spatiotemporal Neural Network[J]. Journal of Shanghai Jiao Tong University, 2026, 60(3): 511-521.

图/表 13

图1

图2

图3

表1

图4

图5

表2

图6

图7

图8

图9

表3

图10

参考文献 16

[1]	ZERBST U, MADIA M, KLINGER C, et al. Defects as a root cause of fatigue failure of metallic components. I: Basic aspects[J]. Engineering Failure Analysis, 2019, 97: 777-792. doi: 10.1016/j.engfailanal.2019.01.055 URL
[2]	CHAI M, HOU X, ZHANG Z, et al. Identification and prediction of fatigue crack growth under different stress ratios using acoustic emission data[J]. International Journal of Fatigue, 2022, 160: 106860. doi: 10.1016/j.ijfatigue.2022.106860 URL
[3]	ALSHOAIBI A M, FAGEEHI Y A. 2D finite element simulation of mixed mode fatigue crack propagation for CTS specimen[J]. Journal of Materials Research and Technology, 2020, 9(4): 7850-7861. doi: 10.1016/j.jmrt.2020.04.083 URL
[4]	何龙龙, 刘志芳, 顾俊杰, 等. 基于XFEM的疲劳裂纹扩展路径和寿命预测[J]. 西北工业大学学报, 2019, 37(4): 737-743.
	HE Longlong, LIU Zhifang, GU Junjie, et al. Fatigue crack propagation path and life prediction based on XFEM[J]. Journal of Northwestern Polytechnical University, 2019, 37(4): 737-743. doi: 10.1051/jnwpu/20193740737 URL
[5]	CHEN J, LIU Y. Fatigue modeling using neural networks: A comprehensive review[J]. Fatigue and Fracture of Engineering Materials and Structures, 2022, 45(4): 945-979. doi: 10.1111/ffe.v45.4 URL
[6]	张松林, 马栋梁, 王德禹. 基于长短期记忆神经网络的板裂纹损伤检测方法[J]. 上海交通大学学报, 2021, 55(5): 527-535. doi: 10.16183/j.cnki.jsjtu.2020.095
	ZHANG Songlin, MA Dongliang, WANG Deyu. Method for plate crack damage detection based on long short-term memory neural network[J]. Journal of Shanghai Jiao Tong University, 2021, 55(5): 527-535.
[7]	HIMMICHE S, MORTAZAVI S N S, INCE A. Comparative study of neural network-based models for fatigue crack growth predictions of short cracks[J]. Journal of Peridynamics and Nonlocal Modeling, 2022, 4(4): 501-526. doi: 10.1007/s42102-021-00062-1
[8]	WANG B, XIE L, SONG J, et al. Curved fatigue crack growth prediction under variable amplitude loading by artificial neural network[J]. International Journal of Fatigue, 2021, 142: 105886. doi: 10.1016/j.ijfatigue.2020.105886 URL
[9]	DO D T T, LEE J, NGUYEN XUAN H. Fast eva-luation of crack growth path using time series forecasting[J]. Engineering Fracture Mechanics, 2019, 218: 106567. doi: 10.1016/j.engfracmech.2019.106567 URL
[10]	付强, 王华伟. 基于多层LSTM的复杂系统剩余寿命智能预测[J]. 兵器装备工程学报, 2022, 43(1): 161-169.
	FU Qiang, WANG Huawei. Intelligent prediction for remaining useful life of complex system based on multi-layer LSTM[J]. Journal of Ordnance Equipment Engineering, 2022, 43(1): 161-169.
[11]	STROHMANN T, STAROSTIN PENNER D, BREITBARTH E, et al. Automatic detection of fatigue crack paths using digital image correlation and convolutional neural networks[J]. Fatigue and Fracture of Engineering Materials and Structures, 2021, 44(5): 1336-1348. doi: 10.1111/ffe.v44.5 URL
[12]	MELCHING D, STROHMANN T, REQUENA G, et al. Explainable machine learning for precise fatigue crack tip detection[J]. Scientific Reports, 2022, 12(1): 9513. doi: 10.1038/s41598-022-13275-1 pmid: 35680941
[13]	SHI X, CHEN Z, WANG H, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]// Proceedings of the 28th International Conference on Neural Information Processing Systems. USA: MIT Press, 2015: 802-810.
[14]	TAN C, GAO Z, LI S, et al. SimVP: Towards simple yet powerful spatiotemporal predictive learning[EB/OL]. (2023-04-26)[2024-03-15]. https://doi.org/10.48550/arXiv.2211.12509.
[15]	张明义, 袁帅, 钟敏, 等. 金属材料和结构的疲劳寿命预测概率模型及应用研究进展[J]. 材料导报, 2018, 32(5): 808-814.
	ZHANG Mingyi, YUAN Shuai, ZHONG Min, et al. A review on development and application of probabilistic fatigue life prediction models for metal materials and components[J]. Materials Reports, 2018, 32(5): 808-814.
[16]	ZHANG A, LIPTON Z C, LI M, et al. Dive into deep learning[M]. Cambridge, UK: Cambridge University Press, 2023.

试验件编号	N	数据集类型
E-0-15.0-1	2 099	训练集
E-0-15.0-2	1 795	验证集
E-0-15.0-3	1 617	测试集

模型	训练				验证				测试
模型	e_RMSE,u/ mm	e_RMSE,v/ mm	e_SSIM,u	e_SSIM,v	e_RMSE,u/ mm	e_RMSE,v/ mm	e_SSIM,u	e_SSIM,v	e_RMSE,u/ mm	e_RMSE,v/ mm	e_SSIM,u	e_SSIM,v
Conv-LSTM	0.110	0.060	0.834	0.942	0.236	0.180	0.667	0.875	0.143	0.076	0.798	0.918
SimVP	0.043	0.021	0.934	0.989	0.065	0.032	0.909	0.979	0.057	0.014	0.933	0.992

裂纹扩展阶段	周期数	裂纹长度真实值/mm	Conv-LSTM模型预测的裂纹长度/mm	Conv-LSTM模型预测误差/%	SimVP模型预测的裂纹长度/mm	SimVP模型预测误差/%
稳定扩展阶段	63 480	14.2	14.1	1.2	14.4	1.0
	63 540	14.4	14.2	1.4	14.4	0.0
	63 600	14.8	14.5	2.3	14.7	0.9
稳态撕裂的裂纹扩展阶段	87 180	45.6	44.9	1.5	45.5	0.1
	87 240	46.2	45.4	1.6	45.6	1.2
	87 300	47.1	46.0	2.2	46.8	0.5

基于时空神经网络的金属疲劳裂纹扩展预测

Prediction of Fatigue Crack Growth in Metal Materials via Spatiotemporal Neural Network

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 16

相关文章 8

编辑推荐

Metrics

本文评价

[1]	. 通过变化子区调整数字图像相关测量精度及其在飞艇蒙皮中的应用[J]. J Shanghai Jiaotong Univ Sci, 2025, 30(2): 239-251.
[2]	王进潇1, 程斌1, 2, 3, 向升1, 李思达1, 闫兴非4. 玻璃纤维复合材料胶栓混合节点在拉伸荷载下的静力与疲劳力学性能[J]. J Shanghai Jiaotong Univ Sci, 2024, 29(5): 817-830.
[3]	王猛, 梁光强. 输氢海底管道强度设计准则研究 [J]. 海洋工程装备与技术, 2023, 10(3): 41-47.
[4]	唐嘉琪，张文光，蔡丰隆，尹雪乐. 基于柔顺机构的减振电极的植入损伤对比分析[J]. 上海交通大学学报, 2018, 52(2): 188-193.
[5]	周凯荣, 陈建, 陈玉喜, 陈华斌, 陈善本. 基于原位测量的钨极惰性气体保护焊焊接过程高温应变表征与计算[J]. 上海交通大学学报, 2016, 50(10): 1588-1591.
[6]	胡梦, 陈凯, 张乐福. 基于直流电压降法的三点弯曲试样疲劳裂纹扩展速率测量方法[J]. 上海交通大学学报, 2015, 49(12): 1778-1784.
[7]	陈凯1，杜东海1，陆辉1，张乐福1，石秀强2，徐雪莲2. 690合金传热管疲劳裂纹扩展研究[J]. 上海交通大学学报（自然版）, 2014, 48(11): 1639-1643.
[8]	侯善芹,许金泉. 基于疲劳特征长度概念的一种新疲劳理论 [J]. 上海交通大学学报（自然版）, 2010, 44(10): 1444-1449.