基于长短期记忆神经网络的钢桥面板疲劳损伤预后

马开臣; 许杨铭; 江林松; 邢松; 王谐

doi:10.3724/j.gyjzG24060304

基于长短期记忆神经网络的钢桥面板疲劳损伤预后

doi: 10.3724/j.gyjzG24060304

广东省公路建设有限公司湾区特大桥养护技术中心, 广州 510120

详细信息

作者简介:
马开臣，博士，主要从事桥梁健康监测、智能巡检设备与技术、结构性能评估等方向的研究。电子信箱：kc-ma@outlook.com

计量
- 文章访问数: 44
- HTML全文浏览量: 12
- PDF下载量: 2
- 被引次数: 0
出版历程
- 收稿日期: 2024-06-03
- 网络出版日期: 2025-09-12

Prognosis of Fatigue Damage in Steel Bridge Decks Based on Long Short-Term Memory Neural Networks

Bay Area Great Bridge Maintenance Technology Center of Guangdong Highway Construction Co., Ltd., Guangzhou 510120, China

摘要

摘要: 正交异性钢桥面在大量的车流荷载下容易发生疲劳损伤，对于桥梁的安全健康运营造成严重威胁。利用海量的桥梁健康监测数据，实现正交异性钢桥面板的疲劳损伤度的预测，提出了一种基于长短期记忆神经网络的钢桥面板疲劳损伤预后方法，可以实现对钢桥面板的中长期疲劳损伤预后，提前了解桥梁的健康状况，为钢桥的疲劳评估提供有效参考。首先对长短期记忆网络和疲劳损伤序列的特点进行了研究分析，疲劳损伤序列本质上就是一种时间序列，十分适合使用长短期记忆网络进行建模预测，然后以某一大跨公路斜拉桥的监测数据作为研究对象，分析了顶板与纵肋焊接细节和横隔板与纵肋交叉构造细节两类疲劳构造细节的4个测点的应力特征和损伤序列特征，选择每小时损伤度序列为数据集进行训练，探究了模型隐藏层数、隐藏层神经元数目、预测输出步长、数据集规模、模型预测方式和模型更新方式对模型预测精度的影响，搭建了一个实现中长期的长短期记忆神经网络，能够实现对两个疲劳构造细节1个月后的疲劳损伤的精准预测。
- 钢桥 /
- 长短期记忆神经网络 /
- 疲劳损伤 /
- 疲劳损伤预后
Abstract: Orthotropic steel decks are prone to fatigue damage under repeated traffic loads， which poses a serious threat to the safe and healthy operation of bridges. By using large-scale bridge health monitoring data， the fatigue damage of orthotropic steel bridge decks can be predicted. In this paper， a fatigue damage prognosis method for steel bridge decks based on long short-term memory （LSTM） neural network is proposed. This method enables long-term fatigue damage prognosis of steel bridge decks， allowing early assessment of bridge health status and providing an effective reference for fatigue evaluation of steel bridges. First， the characteristics of LSTM neural networks and fatigue damage sequences were studied and analyzed. Fatigue damage sequences are essentially a type of time series， making them highly suitable for modeling and prediction using LSTM neural networks. The stress characteristics and damage sequence characteristics of four measuring points were analyzed for two types of fatigue details： roof and longitudinal rib welding details， and transverse partition and longitudinal rib cross structural details. The hourly damage sequence was selected as the dataset for training. The effects of the number of hidden layers of the model， the number of neurons in the hidden layer， the output step-size of the prediction， the size of the dataset， the model prediction method and the model update method on the prediction accuracy of the model were explored. A neural network with LSTM and long term memory was established to realize the accurate prediction of the fatigue damage after one month of the two fatigue details.
- steel bridge /
- long short-term memory neural networks /
- fatigue damage /
- prognosis of fatigue damage

HTML全文

参考文献(24)

[1]	HU N，DAI G L，YAN B，et al. Recent development of design and construction of medium and long span high-speed railway bridges in China［J］. Engineering Structures，2014，74:233-241.
[2]	张清华，卜一之，李乔. 正交异性钢桥面板疲劳问题的研究进展［J］. 中国公路学报，2017，30（3）:14-30，39.
[3]	XIAO Z G，YAMADA K，INOUE J，et al. Fatigue cracks in longitudinal ribs of steel orthotropic deck［J］. International Journal of Fatigue，2006，28（4）:409-416.
[4]	JI B，LIU R，CHEN C，et al. Evaluation on root-deck fatigue of orthotropic steel bridge deck［J］. Journal of Constructional Steel Research，2013，90:174-183.
[5]	KAINUMA S，YANG M，JEONG Y S，et al. Experiment on fatigue behavior of rib-to-deck weld root in orthotropic steel decks［J］. Journal of Constructional Steel Research，2016，119:113-122.
[6]	张清华，崔闯，卜一之，等. 钢结构桥梁疲劳 2019 年度研究进展［J］. 土木与环境工程学报（中英文），2020，42（5 ）:147-158.
[7]	孟凡超. 钢桥面板抗疲劳关键技术［M］. 北京:人民交通出版社，2018.
[8]	李爱群，缪长青，李兆霞，等. 润扬长江大桥结构健康监测系统研究［J］. 东南大学学报（自然科学版），2003（5）:544-548.
[9]	PRŽULJ N，MALOD-DOGNIN N. Network analytics in the age of big data［J］. Science，2016，353（6295）:123-124.
[10]	YAN W，DENG L，ZHANG F，et al. Probabilistic machine learning approach to bridge fatigue failure analysis due to vehicular overloading［J］. Engineering Structures，2019，193:91-99.
[11]	DENG Y，ZHANG M，FENG D M，et al. Predicting fatigue damage of highway suspension bridge hangers using weigh-in-motion data and machine learning［J］. Structure and Infrastructure Engineering，2021，17（2）:233-248.
[12]	HENG J，ZHENG K，FENG X，et al. Machine Learning-Assisted probabilistic fatigue evaluation of Rib-to-Deck joints in orthotropic steel decks［J］. Engineering Structures，2022，265，114496.
[13]	SUN Z，SANTOS J，CAETANO E. Data-driven prediction and interpretation of fatigue damage in a road-rail suspension bridge considering multiple loads［J］. Structural Control and Health Monitoring，2022，29（9），2997.
[14]	宋刚，张云峰，包芳勋，等. 基于粒子群优化 LSTM 的股票预测模型［J］. 北京航空航天大学学报，2019，45（12）:2533-2542.
[15]	杨青，王晨蔚. 基于深度学习 LSTM 神经网络的全球股票指数预测研究［J］. 统计研究，2019，36（3）:65-77.
[16]	CHEN S，GE L. Exploring the attention mechanism in LSTM-based Hong Kong stock price movement prediction［J］. Quantitative Finance，2019，19（9）:1507-1515.
[17]	欧阳红兵，黄亢，闫洪举. 基于 LSTM 神经网络的金融时间序列预测［J］. 中国管理科学，2020，28（4）:27-35.
[18]	张宇帆，艾芊，林琳，等. 基于深度长短时记忆网络的区域级超短期负荷预测方法［J］. 电网技术，2019，43（6）:1884-1892.
[19]	ALHUSSEIN M，AURANGZEB K，HAIDER S I. Hybrid CNN-LSTM model for short-term individual household load forecasting［J］. IEEE Access，2020，8:180544-180557.
[20]	朱凌建，荀子涵，王裕鑫，等. 基于 CNN-Bi LSTM 的短期电力负荷预测［J］. 电网技术，2021，45（11）:4532-4539.
[21]	GUL M J，URFA G M，PAUL A，et al. Mid-term electricity load prediction using CNN and Bi-LSTM［J］. The Journal of Supercomputing，2021，77（10）:10942-10958.
[22]	MEMARZADEH G，KEYNIA F. Short-term electricity load and price forecasting by a new optimal LSTM-NN based prediction algorithm［J］. Electric Power Systems Research，2021，192，106995.
[23]	王祥雪，许伦辉. 基于深度学习的短时交通流预测研究［J］. 交通运输系统工程与信息，2018，18（1）:81-88.
[24]	TIAN Y，ZHANG K，LI J，et al. LSTM-based traffic flow prediction with missing data［J］. Neurocomputing，2018，318:297-305.