【作者】 黄晓敏；

【导师】 程赫明；

【作者基本信息】 昆明理工大学 ， 工程力学， 2013， 博士

【摘要】 由于桥梁结构的复杂化、大型化,使得桥梁结构健康监测的重要性和必要性越来越被人们所认识。桥梁结构健康的核心内容包括桥梁结构的模态参数识别、基于测试模态参数修正结构的有限元模型、基于测试模态参数识别桥梁结构损伤,本文针对上述问题进行了算法研究,具体内容包括：(1)针对桥梁结构的模态参数识别问题,本文以某一实际拱桥为例,实现了基于状态空间的时域模态参数识别算法在实际桥梁结构中的应用。主要研究了特征系统实现算法(ERA法)和随机子空间法(SSI法)在桥梁结构模态参数识别中的应用。研究结果表明上述两种方法具有较好的抗噪声能力,均适合于解决实际桥梁工程结构的模态参数识别问题。(2)针对时变温度对桥梁结构的动力特性影响问题,本文进行了试验研究。首先,对某一拱桥模型进行了长期动力测试,通过对测试加速度信息进行模态参数识别,得到了该结构在不同温度下的前6阶频率、阻尼比和振型；采用测试数据对该拱桥模型进行了有限元模型修正；其次,将结构的测试动力特性与有限元计算结果进行了比较分析,并讨论了该结构的动力特性随温度的变化规律,最后,建立了环境温度与结构频率的关系模型,并利用该模型对测试模型的结构性能进行了评价。(3)针对桥梁结构的有限元模型修正问题,本文研究了基于灵敏度分析的方差最小法在实际桥梁结构中的应用。在修正的过程中应用了实模态提取、自由度扩展、模型缩减、灵敏度计算、修正参数的选择等方面的一些方法。最终使理论的有限元模型同实际模型间的差异给予缩小。(4)针对桥梁结构的损伤诊断问题,本文对常见的钢筋混凝土简支梁结构进行了研究,利用静力加载试验模拟简支梁的损伤累积过程,通过突加荷载试验收集简支梁的自由振动信号,分析阻尼参数的变化机理,对已有的研究结果进行论证,并进一步提出一种基于阻尼特征的损伤识别指标,且验证了该指标的优越性,将其应用到实际桥梁的检测中,证明了其实用性(5)针对连续梁桥结构的损伤识别问题,本文在以往结构损伤识别的研究基础上,提出一种将损伤指标法和神经网络法相结合的连续梁桥结构损伤识别方法。首先,采用结构响应,通过时间序列分析方法得到AR(自回归)模型系数的残差标准差作为损伤指标,并将其用作神经网络法的输入；其次,采用人工神经网络实现桥梁结构的损伤识别；最后,采用一个连续梁桥的模拟算例对所提方法进行验证。更多还原

【Abstract】 The significance of structural healthy monitoring (SHM) of bridges has been paid more attention since the complxicity of bridges became the trend. The kernel of SHM of bridges goes to modal parameter identification of bridges, finite element model updating of bridges based on measured modal parameters, damge diagnosis of bridges. This study focused on couples of topics of the inverse dynamic analysis of bridges,(i) This study did some research on the application of modal parameter identification based on time-domain algorithm in practical bridges, and a practical arch bridge is taken as an example for verifying the effectiveness of ERA method and SSI method. The results showed that these two algorithms have strong ability to resist the influence of noise and it is promising for these two methods to solve the issues of identifying the modal parameters of practical bridges.(ii) To consider the effect of varying temperature on dynamic properties of bridge structures, a modal test for a arch bridge was carried out for about three months. Firstly, a long-term modal test is conducted, and the first six modes (frequencies, damping ratio and modal shapes) under different temperature are identified by modal parameter identification. Secondly, the comparison between the analytical dynamic properties and measured results are analyzed, and the changing regularity of this structure under varying temperature is summarized. The results show that the frequencies of this structure increase as the ambient temperature decrease, and that the damping ration and modal shapes are not sensitive with the ambient temperature. Finally, the relationship model between the environmental temperature and frequencies of this bridge model is obtained by regression analysis. The confidence interval of undamaged structure is obtained, and then this confidence interval is applied to assess the condition of this bridge.(iii) The minimum variance method based on sensitivity is employed to update the finite element model of a arch bridge. During the course of model updating some related aspects are also calculated including extracting the real mode shape, freedom expansion, model reduction, calculating sensitivities, choosing parameters for model updating. At last, the differences between measured and analytical modal parametes are reduced.(iv) With the development of sensor, signal acquiring and processing technologies, a larger number of vibration-based finite element model updating methods have been proposed for civil engineering structures. However, there are difficulties in applying these methods to real structures in operating environments. After a brief introduction to the vibration-based model updating methods developed over the last20years and their applications, this paper presents an analysis of the primary challenges in applying these model updating methods, and it concludes with a discussion of critical issues for future research in the area of vibration-based model updating methods.(v) A method combining the damge index and ANN (artificial neuro network) was proposed account on the damage detection of continous girder bridges. Firstly, the coefficients of AR model were obtained by time series analysis, which deemed as damage index. Secondly, the network considering many types damage was trained by using the damage index, and then this trained network could be applied to detect the damage of continous girder bridges. Finally, the effectiveness of proposed method was verified by using a numerical continous girder bridge.更多还原