基于RBF神经网络的网架结构损伤识别试验研究

针对网架结构损伤识别中模态密集、自由度高等困难,利用RBF网络良好的容错性和鲁棒性,依据损伤前后的网架结构模态参数发生变化理论,提出了基于模态参数和RBF神经网络的网架结构损伤识别方法。以一个6 m×7.5 m的正放四角锥网架结构为研究对象,首先依据连续倒塌理论计算各杆件的重要性系数,确定模拟损伤杆件位置;然后以损伤前后结构的标准化频率平方变化率及标准化位移振型的组合参数作为RBF神经网络的损伤指标,利用有限元分析得到学习样本,试验分析结果作为测试样本。采用二阶段损伤识别方法,首先在所有杆件中排查出可疑受损杆件位置,最后再精确识别损伤位置和程度。结果表明,该方法能够很好地识别网架结构的损伤位置和程度。     

基于模态应变能的梁结构损伤识别

单元模态应变能法诊断结构损伤时理论上需要完整振型,而实测振型往往是不完备的。本文以梁结构为例,基于理论自由度和实测自由度相匹配的原则,对不完备振型信息进行扩充处理,然后再应用模态应变能的方法进行损伤识别。结果表明,当仅测得不完备的振型信息时,单元模态应变能法能很好应用于结构损伤识别,为实际工程的损伤检测提供了依据。     

基于模态扩阶法的结构损伤识别方法应用研究

动力参数法诊断结构损伤时往往需要完整的自由度信息,而在实际工程中很难测量到结构的全部自由度信息.以一钢筋混凝土简支粱为例,通过模态扩阶法将实测的不完备信息扩充到完备的实测信息,使理论自由度与实测自由度相匹配,在此基础上用振型转角差法和模态应变能法对该简支梁结构进行损伤识别.结果表明利用振型转角差法和模态应变能法均较好地...     

基于振型扩充的模态应变能法在结构损伤识别中的应用

单元模态应变能法诊断结构损伤时需要完整振型,而实测振型往往是不完备的.以平面框架和板为例,基于理论自由度和实测自由度相匹配的原则,对不完备的实测振型信息扩充后进行损伤识别.结果表明,当仅测得不完备的振型向量时,单元模态应变能法能很好地对损伤结构进行破损位置和程度的识别.该方法简单易行,可为实际工程应用提供参考.     

基于模态应变能与神经网络的钢网架损伤检测方法

神经网络通过对样本的学习,获得结构模态参数与损伤之间的映射关系。目前基于神经网络的损伤检测已经越来越广泛地使用在非破坏性损伤诊断当中。但对于大型结构而言,它的训练样本数量过大,将消耗大量的计算。所以如何降低神经网络的计算量使其可用于大型结构的损伤诊断是一个亟待解决的问题。为了解决这个问题,提出了空间钢网架损伤的两步诊断法:第一步,利用模态应变能对结构损伤的敏感性,判断出结构损伤的可能位置;第二步,利用神经网络从可能发生损伤的杆件中定位出实际损伤的位置,并进行损伤程度的判断。利用一个空间网架作为数值算例,进行可行性验证。结果表明此方法可以准确判断出结构的损伤位置以及损伤大小,是一种行之有效的方法。     

基于振型相关性的结构模态参数频域自动识别

环境激励下,频域分解法(FDD)具有良好的结构模态频率和振型识别能力,但识别过程中需要基于频率准则人为判断奇异值曲线的峰值,无法准确识别出现近频交叠或重频情况的结构模态参数,而且不能自动进行模态参数的识别。对此,提出了基于振型相关性的MAC-FDD模态参数识别方法,该方法能自动搜索目标模态所对应的频率范围,并利用反映实测振型与理论振型良好相关性的最大模态置信准则(MAC)值判断和识别结构真实振型。给出了该方法的具体实现过程,对平面桁架和空间网架算例进行了模态参数识别。结果表明,该方法能够准确识别出各阶模态频率和振型,可以有效避免模态遗漏,且具有很强的抗噪能力。该方法弥补了常规FDD的不足,适用于对频率密集或具有重频现象的空间网格结构的模态参数识别,且其过程便于编程实现。     

空间网格结构抗震薄弱区识别的静力推覆方法

空间网格结构在强震下出现薄弱区的原因是该区域杆件的地震内力与决定其截面配置的非抗震设计内力差异较大,且该内力差异主要来自于少数模态的贡献。为此,利用非抗震设计和多遇地震验算的最不利内力,提出了一种疑似薄弱区杆件的简便判别方法。基于多遇地震下由振型分解反应谱法计算得到的结构响应,可以确定疑似薄弱区杆件地震内力的主要贡献模态。考虑这些主要贡献模态并参考振型质量参与系数,构造了能够近似反映最不利单向地震响应的综合模态,并基于结构应变能相等的原则确定了罕遇地震水平的等效静力推覆荷载。给出了一种能够计入三向地震动贡献的静力推覆方法,并对一个三心圆柱面双层网壳算例进行了推覆分析。通过与动力弹塑性时程分析结果对比发现,只要在建立推覆荷载时组合模态包括了疑似薄弱区杆件地震内力的主要贡献模态,并且所有组合模态的振型质量参与系数之和大于90%,则该静力推覆方法可以有效识别到该结构在罕遇地震下可能形成的薄弱区。     

自由度匹配技术在框架结构损伤识别中的应用研究

实测自由度与理论自由度不匹配是结构损伤识别技术应用于实际工程中的一大难题,模型缩聚和模态扩阶技术是解决这一问题的主要方法。通过单层框架结构有限元模型的数值模拟,利用振型转角差损伤识别方法,较好地实现了测试信息不完备情况下的框架结构损伤识别。结果表明,利用Guyan减缩法与模态扩阶法进行自由度匹配后再进行损伤识别,其结果非常准确,可以应用于框架结构的损伤识别。     

基于单元应变模态差的网架结构损伤诊断研究

对具有一定程度损伤的网架结构来说,常规的结构有限元分析或实验模态分析得到的位移模态和频率难以有效地反映结构的损伤状况。为了提高诊断效率与诊断结果的可靠性,根据空间杆系结构的受力特点,利用两节点空间铰接杆单元有限元法,由节点位移模态推导出单元应变模态,提出采用结构损伤前后的单元应变模态差作为网架结构损伤定位的识别指标,并以损伤单元应变模态的差值大小确定损伤程度。通过对一个典型网架结构的数值模拟研究表明:该方法能够在低阶模态条件下,有效识别网架结构不同位置和程度的局部损伤;且在一定噪声水平下具有较强的鲁棒性,适用于实际观测条件下的网架结构损伤定位。     

应变模态指标在平板网架结构损伤识别分析中的应用

平板网架结构在实际工程中应用广泛,针对此类结构的损伤识别研究显得尤为重要.由于避免了重新建模引入的系统误差,无模型的损伤识别理论正日益受到重视,并已被应用于梁式构件的健康监测.文中引入了一种应变模态指标,将无模型损伤识别理论应用于平板网架结构中.以折减杆件截面面积的方式来模拟杆件损伤,数值分析了5种具有代表性的损伤工况,在数值计算结果中引入了白噪声,并取部分杆件作为监测杆件.损伤识别分析结果显示,损伤杆件的应变模态指标为结构所有杆件中最高的.通过比对各杆件的应变模态指标,便能对发生在结构中上弦杆、下弦杆以及腹杆等不同位置的损伤进行识别.同时,损伤杆件周围杆件的应变模态指标也会比无损杆件的指标较大,在损伤点周围会出现一个损伤指标显著的区域.另外,当结构中两根或多根杆件同时损伤时,应变模态指标也能识别出全部损伤杆件.     

Structural damage assessment with incomplete and noisy modal data using model reduction technique and LAPO algorithm

Abstract

The present article proposes an efficient method for damage assessment of structures with incomplete and noisy modal data using model reduction method and Lightning Attachment Procedure Optimisation (LAPO) algorithm. The damage identification problem is formulated as an optimisation problem, in which the extent of damaged elements and a combination of both natural frequencies and modal flexibility changes are considered as the continuous design variables and a hybrid objective function, respectively. The LAPO algorithm, a recently developed algorithm, is for the first time extended to solve the formulated optimisation problem. In addition, due to incomplete modal data, an iterated improved reduction system (IIRS) technique is adopted to condense structural physical properties. The efficiency and robustness of the proposed method are investigated using three types of structures and the results are compared with those obtained by other well-known optimisation algorithms. It is shown that regardless of the effect of incomplete and noisy modal data, the proposed method not only successfully determines the location and extent of single and multiple damage cases in the structures, but also has lower computational cost than the other optimisers.     

基于最小秩方法的结构损伤识别

针对结构损伤识别中的最小秩方法存在的问题,经过研究发现,对测试模态进行关于质量矩阵的正交归一化可保证反演后刚度矩阵的对称性;提出了一种迭代修正算法,可保持反演结果的稀疏性;基于模态力余量,定义了一种损伤指标来预先判定结构损伤单元的位置,并可据此选取合适的测试模态阶数进行反演计算。数值试验结果表明,改进后的方法在考虑测试模态误差的情况下可对结构的损伤进行精确的定位和标定。     

网架结构模型损伤试验研究

本文通过模拟网架结构构件的损伤及支座沉降,采用实际应用较为成熟的结构静力响应测试及低阶模态测试方法对一平面足尺钢网架模型进行测试,探讨网架结构在构件损伤时结构特性及响应变化特点规律,试验成果可为大跨结构的健康监测及损伤识别提供依据.     

Wind load identification using wind tunnel test data by inverse analysis

The method for modal wind load identification from across-wind load responses using Kalman filter is presented and verified using the wind tunnel test data. The Kalman filter is utilized for the inverse identification from limited measured responses and the closed-form of Kalman filter gain in modal space is derived for different types of measured response solving the Riccati equation. The wind induced responses used for the verification are measured responses from an aeroelastic wind tunnel test of a rectangular shaped concrete chimney. The displacement responses of the top part of the model are measured and used for the wind load identification, but the acceleration responses obtained by numerical differentiation of displacement are also used in order to evaluate the effect of response type on the identification result. It is found from the identification results that the proposed method identifies the modal across-wind load from measured responses with quite accuracy and the acceleration response yields more accurate wind load identification than displacement response.     

基于有限元模型和不完备模态测试信息的结构损伤识别方法研究

本文在有限元模型的基础上,建立了一种基于不完备模态测试信息的损伤识别方法。采用Levenberg-Marquardt非线性最小二乘算法对建立的杆系结构损伤识别方程进行求解,并通过一拱桥的数值计算结果表明,该方法可以同时进行结构的损伤定位和定量研究,在测试数据不完备及一定数据噪声水平条件下,本文的方法仍有较好的损伤识别能力。     

Axial Strain Accelerations Approach for Damage Localization in Statically Determinate Truss Structures

This work proposes an efficient and reliable method for damage localization in truss structures. The damage is localized on the basis of measured acceleration signals of the structure followed by simple statistical signal processing. It has three main advantages over many existing methods. First, it can be directly applied to real engineering structures without the need of identifying modal parameters or solving any global optimization problem. Second, the proposed method has higher sensitivity to damage than some other frequently used methods and allows to localize damage as small as a few percentages. Third, it is a model‐free method, which does not require precise finite element model development or updating. Validation of the method has been conducted on numerical examples and laboratory‐scale trusses. Two types of frequently used trusses have been selected for this study, namely, Howe and Bailey trusses. The presented experimental validation of the method shows its efficiency and robustness for damage localization in truss structures.     

Two-stage damage diagnosis approach for steel braced space frame structures

In order to diagnose the location and extent of damage in steel braced space frame structures, a two-stage damage diagnosis approach is proposed. This approach is comprised of the damage locating vectors method and eigensensitivity analysis. By deriving formulas used to calculate characterizing stresses in space frame members, and by defining characterizing stresses in connections, the damage locating vectors method is extended to locate damage in space frame members and connections. In addition, the simplified calculation of modal mass-normalization constants for damaged structures is improved. The first- and second-order sensitivities of the modal parameter discrepancies with respect to the structural model parameters and the stiffness matrix of beam elements with one damaged end are utilized. To verify the effectiveness of the proposed approach, numerical simulation analysis and experimental testing of a steel braced space frame model are performed. Ten and seven damage patterns are simulated in the numerical example and experimental testing, respectively. Modal parameters of the undamaged and damaged structures are extracted from the acceleration data using the natural excitation technique (NExT) and the eigensystem realization algorithm (ERA). The extended damage locating vectors method is utilized to determine potentially damaged elements. Based on the identified modal information, the extent of damage of the potentially damaged elements is estimated using the second-order eigensensitivity analysis. It is demonstrated that the two-stage damage diagnosis approach is effective when the damage of the members or connections in steel braced space frame structures reaches a certain level.     

Structural Model Updating and Health Monitoring with Incomplete Modal Data Using Gibbs Sampler

Abstract:   A new Bayesian model updating approach is presented for linear structural models. It is based on the Gibbs sampler, a stochastic simulation method that decomposes the uncertain model parameters into three groups, so that the direct sampling from any one group is possible when conditional on the other groups and the incomplete modal data. This means that even if the number of uncertain parameters is large, the effective dimension for the Gibbs sampler is always three and so high-dimensional parameter spaces that are fatal to most sampling techniques are handled by the method, making it more practical for health monitoring of real structures. The approach also inherits the advantages of Bayesian techniques: it not only updates the optimal estimate of the structural parameters but also updates the associated uncertainties. The approach is illustrated by applying it to two examples of structural health monitoring problems, in which the goal is to detect and quantify any damage using incomplete modal data obtained from small-amplitude vibrations measured before and after a severe loading event, such as an earthquake or explosion.     

空间网格结构节点螺栓松动损伤定位有限元分析

以单层柱面网壳试验模型为背景进行损伤定位研究,采用ANSYS软件对其螺栓松动损伤进行了有限元分析,利用神经网络的模式分类功能进行了螺栓松动的分步损伤定位,即首先以整体结构为研究对象,采用面向子结构的损伤定位,确定结构是否存在损伤及损伤的大致位置;然后采用面向节点的损伤定位法,确定子结构中螺栓松动损伤的节点位置。研究结果表明：损伤识别的成功率与杆件灵敏度密切相关,为了保证识别结果准确性,其灵敏度达到2%以上才能正确识别,当杆件的灵敏度过低时将无法通过动力测试识别;采用分步损伤定位方法可以显著减少损伤精确定位时训练样本的数量,适合于不完备的模态数据,利用低阶模态数据即可准确识别损伤的子结构与节点位置。