基于多通道奇异谱分析的结构时域损伤识别方法

发布时间：2019-06-19 12:01

【摘要】：工程结构的安全性与社会发展和民众生命财产安全直接相关,因此识别正在服役期的工程结构内存在的损伤是土木工程研究人员面对的重要课题之一。由于时域损伤识别方法具有响应信号易于测量、不存在中间过程误差、可以提供更多的损伤识别方程、可直接应用于非线性动力系统的辨识的优点,因此时域损伤识别方法具有广阔的前景。但是,传统时域损伤识别方法存在测量噪音严重影响识别精度的局限性,针对这一局限性,本文在考虑不同测量信号之间的相关性的基础上,开展了基于多通道奇异谱分析的结构时域损伤识别方法的研究,综合理论分析、数值模拟、试验验证等手段,进行了以下几个方面的深入研究:(1)研究了多通道奇异谱成分中噪音和损伤信息的分布规律,发现低阶多通道奇异谱成分中包含较少的噪音信息和较多的损伤信息,提出了定量分析多通道奇异谱成分中噪音信息和损伤信息含量的指标。本文将提取出这些含较少噪音、较多损伤信息的优质统计成分进行结构损伤识别,以减轻测量噪音对识别结果精度的影响。(2)在多通道奇异谱成分中噪音和损伤信息分布的研究基础上,进行了基于多通道奇异谱分析的结构时域损伤识别方法的理论分析。考虑不同位置测量信号的相关性,将传统灵敏度分析损伤识别方法与多通道奇异谱分析方法有机结合,并运用子空间投影方法得到优质投影空间上的基于多通道奇异谱分析的损伤识别方程。(3)详细探讨了影响基于多通道奇异谱分析的结构时域损伤识别方法精度的因素。首先讨论了三种不同投影空间和选取优质多通道奇异谱成分的标准,得出多通道奇异谱分析方法处理完好结构计算响应得到的空间为最优质的投影空间;最后研究了窗口长度对识别结果的影响,通过对比发现:窗口长度越长,优质成分的区间范围越大。(4)以二维平面桁架结构为数值模拟算例,应用本文提出的算法,在噪音水平高达20%和30%的情况下,进行了损伤识别数值模拟验证,并将结果与传统灵敏度分析方法识别结果进行了对比,验证了本文方法的有效性和稳定性。同时通过平面钢框架结构的损伤试验,验证了本文方法的实用性。
[Abstract]:The safety of engineering structure is directly related to social development and the safety of people's life and property, so it is one of the important topics faced by civil engineering researchers to identify the damage existing in the engineering structure which is in service. Because the time domain damage identification method is easy to measure the response signal, there is no intermediate process error, it can provide more damage identification equations, and can be directly applied to the identification of nonlinear dynamic systems, so the time domain damage identification method has a broad prospect. However, the traditional time-domain damage identification method has the limitation that the measurement noise seriously affects the identification accuracy. In view of this limitation, on the basis of considering the correlation between different measurement signals, this paper studies the time-domain damage identification method based on multi-channel singular spectrum analysis, comprehensive theoretical analysis, numerical simulation, experimental verification and so on. The following aspects are deeply studied: (1) the distribution of noise and damage information in multi-channel singular spectrum components is studied, and it is found that the low-order multi-channel singular spectrum components contain less noise information and more damage information, and the index of quantitative analysis of noise information and damage information content in multi-channel singular spectrum components is put forward. In this paper, these high quality statistical components with less noise and more damage information are extracted for structural damage identification in order to reduce the influence of measurement noise on the accuracy of identification results. (2) based on the study of noise and damage information distribution in multi-channel singular spectral components, the theoretical analysis of structural damage identification method based on multi-channel singular spectrum analysis is carried out. Considering the correlation of different position measurement signals, the traditional sensitivity analysis damage identification method and multi-channel singular spectrum analysis method are combined organically, and the damage identification equation based on multi-channel singular spectrum analysis in high quality projection space is obtained by using subspace projection method. (3) the factors that affect the accuracy of structural damage identification method based on multi-channel singular spectrum analysis are discussed in detail. Firstly, three different projection spaces and the criteria for selecting high quality multi-channel singular spectrum components are discussed, and it is concluded that the space obtained by the multi-channel singular spectrum analysis method to deal with the response of the intact structure is the best projection space. Finally, the influence of window length on the recognition results is studied. It is found that the longer the window length is, the larger the interval range of high quality components is. (4) taking the two-dimensional plane truss structure as a numerical simulation example, the numerical simulation of damage identification is carried out when the noise level is as high as 20% and 30%, and the results are compared with those of the traditional sensitivity analysis method. The effectiveness and stability of the proposed method are verified. At the same time, the practicability of this method is verified by the damage test of plane steel frame structure.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU317

【参考文献】