基于小波神经网络的斜拉桥损伤识别方法研究

发布时间：2018-01-13 23:09

本文关键词：基于小波神经网络的斜拉桥损伤识别方法研究　出处：《长沙理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：我国桥梁数量日益增多,桥梁的健康监测变得越来越重要,而有效快速的识别桥梁结构可能发生损伤的位置和程度变得尤为重要。小波分析被誉为分析信号的数学显微镜,具有优良的时间和频率局部化特征。神经网络对信号处理方面具有良好的自组织、自学习、自适应能力,并且具有高度的非线性映射能力。因此,结合两者的优点,通过小波分析和神经网络有效地识别斜拉桥主梁的损伤位置和损伤程度,具有重要的理论意义和工程应用价值。论文以斜拉桥损伤识别为研究课题,取得了如下主要成果:本文通过对斜拉桥进行有限元分析,将提取的模态参数进行连续小波变换得到结构的小波系数图,由小波系数模极大值点确定损伤位置。以含损伤斜拉桥的前四阶固有频率作为神经网络输入参数构造神经网络模型,经过样本数据的训练与测试,从而实现对斜拉桥损伤程度的识别。给出了斜拉桥损伤识别的原理,建立了一种识别斜拉桥损伤位置和损伤程度的小波神经网络方法。通过对单塔斜拉桥含单损伤和多损伤分别进行研究,在运用有限元分析提取其转角和曲率模态参数的基础上,对模态参数进行连续小波变换,根据小波系数图中的模极大值点来识别损伤位置。利用有限元计算分析获得含损伤单塔斜拉桥的前四阶固有频率,对两种不同损伤工况分别构造了损伤程度为20%、30%、40%和18%、28%、38%的训练样本,通过BP神经网络构造的网络计算分析,样本数据训练和测试结果误差均在5%以内。本文研究了双塔斜拉桥含单损伤和多损伤的识别问题,通过建立含损伤的双塔斜拉桥有限元模型,对其动力特性进行有限元分析得到模态参数,然后进行连续小波变换有效地识别了损伤位置。再利用含损伤双塔斜拉桥的前四阶固有频率,对两种不同的损伤工况分别构造了损伤程度为20%、30%、40%和18%、28%、38%的训练样本,通过BP神经网络构造的网络计算得到的样本数据训练和测试结果均满足工程精度要求。本文所建立的小波神经网络算法,不仅能识别斜拉桥主梁的损伤位置,而且能有效识别斜拉桥主梁的损伤程度。该方法对斜拉桥损伤诊断的工程应用有一定的指导意义。
[Abstract]:The number of bridges in our country is increasing, bridge health monitoring is becoming more and more important position and degree of effective and rapid identification of bridge structure possible damage has become particularly important. Wavelet analysis is known as mathematical microscope analysis of time and frequency signal, has excellent localization features. The neural network for signal processing with self-organization good self-learning and adaptive ability, and has strong nonlinear mapping ability. Therefore, combining the advantages of both, effectively identify the damage location and degree of main girder of cable-stayed bridge by wavelet analysis and neural network, and has important theoretical significance and engineering application value. Based on the damage detection of cable-stayed bridge as the research topic. The main results are as follows: in this paper, based on finite element analysis of cable-stayed bridge, the modal parameters extracted by continuous wavelet transform to get the structure of the small Wave coefficient map, wavelet coefficients by modulus maxima to determine the location of the damage. The first four natural frequencies with damage of the cable-stayed bridge as input parameters to construct neural network model of neural network, after training and testing sample data, so as to realize the recognition of cable-stayed bridge damage. A principle of damage detection of cable-stayed bridge the establishment of the wavelet neural network method for identification of cable-stayed bridge, damage location and damage degree. Based on the single tower cable-stayed bridge with single and multiple damage were studied, based on the finite element analysis to extract the angle and curvature modal parameters, continuous wavelet transform is applied to the modal parameters based on Wavelet figure coefficient modulus maxima to identify the damage location. Using the finite element analysis of bearing damage first four natural frequencies of single tower cable-stayed bridge, on two different damage conditions are presented by injury. For 20%, 30%, 40% and 18%, 28%, 38% training samples, by constructing BP neural network analysis, training sample data and test results of error within 5%. This paper studied with single and multiple damage identification problem of Twin Towers cable-stayed bridge, through the establishment of damaged Twin Towers cable-stayed bridge finite element model, finite element analysis of the dynamic characteristics of modal parameters, then continuous wavelet transform to effectively identify the damage location. The first four order natural frequency reuse with damage of Twin Towers cable-stayed bridge, on two different damage conditions are constructed. The damage degree was 20%, 30%, 40% and 18%. 28%, 38% training samples, obtained by constructing BP neural network calculation of training sample data and test results meet the engineering accuracy requirements. The algorithm of wavelet neural network in this paper, not only can identify the damage position of main girder of cable-stayed bridge The method can effectively identify the damage degree of the main girder of the cable-stayed bridge. This method has certain guiding significance for the engineering application of the damage diagnosis of the cable-stayed bridge.

【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U446;U448.27

【相似文献】