基于小波包分解和优化BP神经网络的桥梁结构损伤识别试验研究
DOI:
CSTR:
作者:
作者单位:

1.郑州大学;2.河南交通投资集团有限公司

作者简介:

通讯作者:

中图分类号:

U441+.4、TN911.71

基金项目:

国家自然科学基金项目(51408557);中国博士后科学基金(2013M541995).;河南省交通运输厅计划项目(2020J-2-6)


Experimental Study on Damage Identification of Bridge Structures Based on Wavelet Packet Decomposition and Optimized BP Neural network
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对单一损伤指标对桥梁损伤识别研究的局限性,根据小波包变换基本原理和神经网络原理,构造小波包能量相对变化率(Relative Energy Rate of Wavelet Packet Energy Spectrum, RES)作为损伤识别指标并建立优化反向传播(Backpropagation,BP)神经网络模型,提出基于小波包分析和优化BP神经网络的桥梁结构损伤位置和损伤程度识别方法,并对该方法进行试验验证,探讨了信号噪音和车速等因素对试验结果的影响。结果表明:以多个损伤工况作为遗传算法优化反向传播(Genetic Algorithm-Back Propagation;GA-BP)和遗传算法和模拟退火优化反向传播(Genetic Algorithm and Simulated Annealing-Back Propagation;GASA-BP)神经网络的训练集,两种神经网络模型在数值模拟工况和试验工况中均展现出了良好的识别能力,在数值模拟工况下,GASA-BP相较于GA-BP神经网络的最大平均误差提高了92.61%;在试验工况下,GASA-BP相较于GA-BP神经网络的最大平均误差提高了67.66%,由此可见GASA-BP神经网络具有更好的识别精度和较好的鲁棒性。该方法仅需少量的传感器即可对桥梁结构损伤位置实现精准定位。

    Abstract:

    Aiming at the limitations of a single damage indicator for bridge damage identification research, according to the basic principles of wavelet packet transform and neural network principles, the Relative Energy Rate of Wavelet Packet Energy Spectrum (RES) is constructed as a damage identification indicator and an optimised backpropagation (BP) neural network model is established. Backpropagation (BP) neural network model, put forward based on wavelet packet analysis and optimisation of BP neural network bridge structure damage location and damage degree identification method, and the method of experimental validation, to explore the signal noise and speed and other factors on the test results. The results show that: multiple damage conditions are used as the basis for Genetic Algorithm-Back Propagation (GA-BP) and Genetic Algorithm and Simulated Annealing-Back Propagation (GASA-BP). Propagation (GA-BP) and Genetic Algorithm and Simulated Annealing-Back Propagation (GASA-BP) neural networks, the two neural network models show good recognition ability in both numerical simulation and experimental conditions, and in numerical simulation, the maximum average error of GASA-BP is 92.61% higher than that of the GA-BP neural network, and in experimental conditions, the maximum average error of GASA-BP is 92.61% higher than that of the GA-BP neural network. In the experimental condition, the maximum average error of GASA-BP is 67.66% higher than that of GA-BP neural network, which shows that GASA-BP neural network has better recognition accuracy and better robustness. The method requires only a small number of sensors to accurately locate the damage location of the bridge structure.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-10-02
  • 最后修改日期:2024-11-08
  • 录用日期:2024-11-08
  • 在线发布日期:
  • 出版日期:
文章二维码
×
《国外电子测量技术》
财务封账不开票通知