基于模态参数小波神经网络的结构损伤识别方法研究
发布时间：2018-11-29 09:20
【摘要】：结构在使用的过程中,由于各种原因可能会出现不同程度的损伤,当这些损伤累积到一定的程度时,将会导致结构的刚度和承载力的下降,进而影响整个结构的使用性和耐久性,严重时还可能会引发灾难性的事故,造成巨大的经济损失和人员伤亡。因此,如何快速有效地识别出结构的损伤位置以及结构的损伤程度,已经成为当前工程结构损伤诊断研究领域的一项重要研究课题。小波分析作为一种时-频两域信号处理方法,能够在时域和频域较好的表征出信号的局部特性;神经网络算法拥有高度的非线性映射能力,对信号处理方面具有自组织、自学习、自适应能力。结合两者的优点,本文建立了基于小波分析和神经网络相结合的理论方法,通过小波分析得出的小波系数图判断出结构的损伤位置,并基于小波分析得出的小波系数模极大值,利用神经网络识别出结构的损伤程度,因此,通过小波分析和神经网络两种方法的结合,可以实现对结构损伤位置和损伤程度的有效识别。本文以含有损伤的简支梁为研究对象,建立了基于振型模态、转角模态、曲率模态的损伤识别方法,对简支梁含有一处损伤和多处损伤的裂缝位置进行有效的识别,并对比了这三种模态下的损伤识别效果；然后对梁的模态参数进行小波变换得出小波系数图,利用神经网络去模拟小波系数模极大值与损伤程度之间的非线性关系来识别结构的损伤程度。数值模拟分析表明,小波分析和神经网络的结合可以有效地识别出结构的损伤位置和损伤程度。本文以含有损伤的连续梁为研究对象,建立了含有一处损伤、二处损伤和多处损伤的有限元模型,对各损伤工况分别基于振型模态、转角模态、曲率模态下进行小波变换,通过小波系数图来识别出结构的损伤位置；然后利用神经网络去模拟小波系数模极大值与损伤程度之间的非线性关系,由神经网络的输出结果识别出结构的损伤程度。分析结果表明,将小波分析和神经网络相结合,准确的识别出了结构的损伤位置和损伤程度。因此本文方法对结构的损伤诊断具有重要的指导意义。
[Abstract]:During the use of the structure, there may be different degrees of damage due to various reasons. When the damage accumulates to a certain extent, it will lead to the decrease of the stiffness and bearing capacity of the structure. Furthermore, it will affect the durability and usability of the whole structure, and may lead to catastrophic accidents when serious, resulting in huge economic losses and casualties. Therefore, how to identify the damage location and damage degree of structures quickly and effectively has become an important research topic in the field of structural damage diagnosis. As a time-frequency two-domain signal processing method, wavelet analysis can better characterize the local characteristics of the signal in time domain and frequency domain. The neural network algorithm has high ability of nonlinear mapping, self-organizing, self-learning and adaptive in signal processing. Combining the advantages of the two methods, a theoretical method based on the combination of wavelet analysis and neural network is established in this paper. The damage location of the structure is judged by the wavelet coefficient graph obtained by wavelet analysis, and the modulus maximum of wavelet coefficient is obtained based on wavelet analysis. The damage degree of structure can be recognized by neural network, so the location and degree of damage can be effectively identified by combining wavelet analysis and neural network. In this paper, the damage identification method of simply supported beam with damage is established based on mode, rotation mode and curvature mode. The crack location of simply supported beam with one or more damage is effectively identified. The effects of damage identification in these three modes are compared. Then the wavelet transform of the modal parameters of the beam is carried out to obtain the wavelet coefficient graph and the nonlinear relationship between the modulus maximum of the wavelet coefficient and the degree of damage is simulated by using the neural network to identify the damage degree of the structure. Numerical simulation shows that the combination of wavelet analysis and neural network can effectively identify the damage location and damage degree of the structure. In this paper, a finite element model with one damage, two damage and multiple damage is established for continuous beam with damage. Wavelet transform is carried out for each damage condition based on mode, rotation mode, curvature mode, respectively. The damage location of the structure is identified by wavelet coefficient graph. Then the nonlinear relationship between the modulus maximum of wavelet coefficients and the degree of damage is simulated by the neural network, and the damage degree of the structure is identified by the output result of the neural network. The results show that the wavelet analysis and neural network are combined to identify the damage location and damage degree of the structure accurately. Therefore, this method has an important guiding significance for structural damage diagnosis.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU317
本文编号：2364607
[Abstract]:During the use of the structure, there may be different degrees of damage due to various reasons. When the damage accumulates to a certain extent, it will lead to the decrease of the stiffness and bearing capacity of the structure. Furthermore, it will affect the durability and usability of the whole structure, and may lead to catastrophic accidents when serious, resulting in huge economic losses and casualties. Therefore, how to identify the damage location and damage degree of structures quickly and effectively has become an important research topic in the field of structural damage diagnosis. As a time-frequency two-domain signal processing method, wavelet analysis can better characterize the local characteristics of the signal in time domain and frequency domain. The neural network algorithm has high ability of nonlinear mapping, self-organizing, self-learning and adaptive in signal processing. Combining the advantages of the two methods, a theoretical method based on the combination of wavelet analysis and neural network is established in this paper. The damage location of the structure is judged by the wavelet coefficient graph obtained by wavelet analysis, and the modulus maximum of wavelet coefficient is obtained based on wavelet analysis. The damage degree of structure can be recognized by neural network, so the location and degree of damage can be effectively identified by combining wavelet analysis and neural network. In this paper, the damage identification method of simply supported beam with damage is established based on mode, rotation mode and curvature mode. The crack location of simply supported beam with one or more damage is effectively identified. The effects of damage identification in these three modes are compared. Then the wavelet transform of the modal parameters of the beam is carried out to obtain the wavelet coefficient graph and the nonlinear relationship between the modulus maximum of the wavelet coefficient and the degree of damage is simulated by using the neural network to identify the damage degree of the structure. Numerical simulation shows that the combination of wavelet analysis and neural network can effectively identify the damage location and damage degree of the structure. In this paper, a finite element model with one damage, two damage and multiple damage is established for continuous beam with damage. Wavelet transform is carried out for each damage condition based on mode, rotation mode, curvature mode, respectively. The damage location of the structure is identified by wavelet coefficient graph. Then the nonlinear relationship between the modulus maximum of wavelet coefficients and the degree of damage is simulated by the neural network, and the damage degree of the structure is identified by the output result of the neural network. The results show that the wavelet analysis and neural network are combined to identify the damage location and damage degree of the structure accurately. Therefore, this method has an important guiding significance for structural damage diagnosis.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU317
【参考文献】
相关期刊论文 前1条
1 江雷;基于并行遗传算法的弹性TSP研究[J];微电子学与计算机;2005年08期
，本文编号：2364607
本文链接：https://www.wllwen.com/guanlilunwen/chengjian/2364607.html
