基于神经网络的钢梁损伤识别研究

发布时间：2018-01-04 22:25

本文关键词：基于神经网络的钢梁损伤识别研究　出处：《昆明理工大学》2014年硕士论文　论文类型：学位论文

更多相关文章： 损伤识别 ANSYS 神经网络 曲率模态

【摘要】：我国国力的不断增强、经济的持续快速发展以及现代化科技水平的提高,随之而来的是我国的道路交通运输业也在同步快速的发展。桥梁作为我国公路建设的一个不可或缺重要分支,从古至今一直肩负着运输的重任,同时也在不断的发展。特别是近年来,各种不同类型、不同功能的桥梁在各地如雨后春笋般不断涌现。然而由于年代久远、超负荷运转以及自然环境的恶化等因素造成许多桥梁产生了不同程度的损伤,这些表面或隐藏在内部的损伤造成的重大的安全隐患。诸多桥梁的安全事故的发生不但造成了巨大的经济损失,也严重的危及人们的人身安全。对桥梁结构的的监测和对结构可能出现的损伤进行诊断并且及时对产生的损伤进行补救成为了桥梁维护的重点工作。在整理总结工程界国内外学者对于桥梁结构的损伤识别的诸多工作的基础上,综合了结构振动的模态分析法和神经网络理论对损伤结构进行识别分析。首先利用大型通用仿真分析软件ANSYS对检测目标进行建模,并进行模态分析,把通过模态分析采集结构的各种模态参数导入到MATLAB软件的神经网络工具构造神经网络并对网络进行训练,最后利用已经训练好的神经网络对试验方法得到的结构损伤数据进行测试。实验的结果表明,综合了模态分析理论和神经网络原理的方法能够有效的检测出结构的损伤位置以及结构的损伤程度。本文在阐述了桥梁结构损伤检测的背景及意义的基础上,具体分别一一列出了近些年来国内外学者在结构损伤识别研究工作上的重要进展和理论的创新情况。介绍了结构振动的基本理论和模态分析的基本原理,讨论了几种模态参数在损伤识别中的利用并比较几种参数的精确程度。基于有限元理论的分析软件ANSYS的理论基础、分析方法,同时介绍了神经网络的原理并分析了几种神经网络的应用于损伤检测各自的利弊。对被检测结构的无损以及不同损伤情况下的工况利用软件进行建模并进行模态分析,把模态分析得到的数据进行作为单一参数或者互相组合来构成神经网络的输入参数,并对神经网络进行训练。训练好的神经网络可以用来对结构进行损伤位置检测和损伤程度判定。总结本文的主要研究工作,以结构的指定点的曲率模态值和频率结合作为综合参数,利用BP神经网络对结构单损伤以及双损伤都能准确的定位,且对于损伤程度的判定精确度较高。
[Abstract]:China's national strength, the sustained and rapid economic development and modernization of the science and technology level, followed by China's road transportation industry also synchronized development rapidly. As an important branch of China's highway construction of the bridge, straight shoulders from ancient times to the present transport task, but also in the unceasing development especially in recent years, a variety of different types, different functions such as bridge in the country continue to emerge. However, like bamboo shoots after a spring rain due to the passage of time, overload and the deterioration of the natural environment and other factors caused many bridges have varying degrees of damage, the surface or hidden in the internal damage caused by the occurrence of major security risks. Many accidents of bridge not only caused huge economic losses, but also seriously endanger people's safety. The monitoring of bridge structure and the structure It is the key work of bridge maintenance to diagnose the possible damage and to remediate the damage in time.
In the summary of domestic and foreign scholars on the basis of the engineering of bridge structural damage identification on the comprehensive structural vibration modal analysis method and neural network theory for structural damage identification analysis. Firstly, using the general simulation software ANSYS for the detection of target modeling and modal analysis, the analysis of all kinds of import modal parameter acquisition structure to the MATLAB software through the modal neural network tool to construct neural network and the network training, and finally tested by structural damage data has trained neural network to test method. The experimental results show that the comprehensive method of modal analysis theory and neural network principle can detect the structural damage the location and extent of damage of the structure effectively.
Based on the description of the background and significance of the structural damage detection of bridges on the specific list in recent years domestic and foreign scholars in the innovation of structural damage important progress and the research work on the theory of recognition. This paper introduces the basic principle of basic theory and modal analysis, discusses the use of several modal parameters in damage identification and accuracy comparison of several parameters. Analysis method of the theoretical foundation, the theory of finite element analysis software based on ANSYS, and introduces the principle of neural network and analyzes the application of several neural network in damage detection of the respective advantages and disadvantages. The modeling and modal analysis using the software structure and different nondestructive damage under the conditions tested, the modal analysis of the data as a single parameter or combined with each other to form the input of the neural network The neural network is trained. The trained neural network can be used to detect the damage position and determine the damage degree of the structure.
The main research work in this paper is summarized. Taking the combination of curvature modal value and frequency of the specified point as the comprehensive parameter, the BP neural network is used to locate the single damage and double damage of the structure accurately, and the accuracy of the damage degree is high.

【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U445.71

【参考文献】