【作者】 蒋华；

【导师】 赵人达；

【作者基本信息】 西南交通大学 ， 桥梁与隧道工程， 2005， 博士

【摘要】 如何评价受损结构的残存能力和可靠性,以便对结构能否继续使用或是否需加固做出正确的决策,是当今土木工程研究的新课题。显然,解决这一新课题的前提是对受损结构的实际性态做出正确的判断。结构损伤诊断研究的主要目的,就是确定结构的实际性态,进而评定结构的可靠性。本论文受西南交通大学科技发展项目(2003A14)“基于静力测试数据的桥梁结构损伤识别研究”的资助,针对静力测试数据下的结构损伤检测及评估理论进行了系统研究,并根据高等学校博士学科点专项基金项目(98061307)“体外部分预应力高强混凝土连续梁结构行为研究”试验的静力测试数据进行了损伤诊断与评定,主要包括四个部分: 在论文的第一部分,笔者对损伤诊断的基准模型选择、损伤参数选择以及参数识别算法进行了研究。首先阐述了结构损伤识别基准参数模型选择的基本依据。考虑到桥梁结构的特殊性,笔者明确提出三维杆系有限元能够满足基准模型的选择条件,特别适于控制测试数据和损伤识别参数,应是桥梁结构损伤识别基准参数模型的较好选择。笔者从力学反问题入手,将损伤识别归结为系统参数识别问题,首次采用静力位移的相对残差矢量来构筑最小二乘估计准则函数,并采用约束极值算法来识别损伤参数。从结构损伤识别的角度讲,对于模型的每一个单元,只需一个参数就能够判别出该单元是否损伤及损伤程度如何,因此,笔者提出应选取刚度的折减系数作为待识别的唯一损伤参数。 在论文的第二部分,笔者对稀少测试数据条件下以及测试误差条件下的结构损伤识别理论进行了研究。对复杂结构而言,稀少的测试数据与大量待识别参数形成了矛盾,如何解决该矛盾是损伤识别中的一个关键问题。参数分组是目前条件下的一种较好解决方法。笔者在参考已有文献的基础上根据参数分组概念,提出了一种桥梁结构的损伤区域推断理论并对其进行了详细推导,且通过算例分析验证了其可靠性。考虑不确定性因素下的大型复杂结构的损伤检测,是当前国内外学术与工程界普遍关注和亟需解决的课题。笔者在概率统计理论的基础上,首先建立了结构的随机有限元参数模型以研究测试误差对损伤识别结果的影响;并基于假设检验理论,系统地研究了先验参数分布已知与未知条件下的结构损伤定位准则,并据此提出了结构损伤的定位指标;然后基于损伤参数的正态分布假设,提出结构损伤程度指标的概念,并建立了损伤程度指标更多还原

【Abstract】 In order to correctly judge whether the damaged structure could be continuously used or need to be reinforced, how to evaluate the remaining capacity and reliability of the damaged structure, is nowadays a new problem which must be studied in civil engineering. Obviously, in order to solve this problem, the premise is that the properties of the damaged structure must be actually evaluated. The main purpose of damage detection is to estimate the actual properties of the structure and assess its reliability, In this thesis, supported by a research project "Damage Detection of Bridge Structure from Static Measurements" sponsored by science and technology fund of Southwest Jiaotong University(2003A14), the author carries out systematic researches on the theories of structural damage detection and evaluation based on static test data. And according to the static test data from the research project "The Test Study on Externally Prestressed High Strength Concrete Continuous Beam " sponsored by the Doctoral Program of Higher Education(98061307), the damage detection of four test beams is conducted. The thesis includes four parts:In the first part, the author carries out the researches on the choice of the baseline model for damage detection, the damage parameters and the algorithm for parameter identification. Firstly the basic principles of the choice of the baseline model is discussed. Considering the specific characteristic of the bridge structures, the author believes that, the FEM model of the three dimensional bars, especially suitable for controlling the number of test data and the number of identified parameters, could be keeping with above principles. Therefore it is a good baseline model for the damage detection of the bridge structures. Starting with the Counter-question of the mechanics, the author counts damage detection to the problem of the system parameter identification and uses the relative residual error vector of the static displacement to construct the criterion function of the least squares method, and adopts constrained optimization method to identify the damage parameters. From the viewpoint of structural damage detection, for each element of the model, only one parameter is needed to distinguish whether the element damaged or not and assess the extent of the damage. Therefore the author selects the reduction factor of stiff-reduction factor of stiffness as the only damage parameter to be identified.In the second part, the author carries out the researches on the damage detection method under the rarely test data and the test error. To the large complicated structures, it is a contradiction between the scarcity of test data and the large quantity identified parameters of the structure. How to resolve it is a key problem within the damage detection. The parameter grouping algorithm is a good selection at present. The author brings up a damage area detection method of the bridge structures in line with the parameter grouping algorithm described in the reference literatures. This method is deduced in detail and its reliability is verified through the example analysis. The damage detection of large complicated structures taking the uncertainty into account is paid a widespread attention in the domestic and international academic and engineering field at present. Firstly, based on the probability and statistics theories, stochastic FEM parameter model of the structure is established to study the effect of the test error on the damage detection results. According to hypothesis test theories, whether the prior distribution of the parameter is known or unknown , the damage location criterion is studied systematically. The index of damage location is brought up based on this criterion. And thirdly , according to the assumption of normal distribution of the parameters, the index of damage degree is defined too. In order to calculate the index of damage degree based on the damage probability, the relationship between them is established. It makes the damage degree assessment have the probability meaning. Finally, The influence caused by model error to the damage detection is analyzed. The revision of the structural model error is counted to the process of damage detection under the specific condition. So the probability evaluation method is applied to revise the structural model error margin.hi the third part, the author carries out the researches on the theories of structural reliability and remaining life prediction. The traditional theories are based on the geometrical, physical properties of the component’s section derived from the local detection method. But the author, according to the results of the damage parameter identification directly, tries to establish the assessment theories of the structural reliability and remaining life prediction completely based on the viewpoint of the damage detection.hi the fourth part, the author applies the above theories to assess the damage of更多还原