【作者】 何龙军；

【导师】 练继建；

【作者基本信息】 天津大学 ， 水利水电工程， 2013， 博士

【摘要】 随着各类水工建筑物服役时间的增长以及安全运行要求的提高，结构健康监测和损伤诊断工作就变得日益重要。首先，为了应用有限的试验设备尽可能多地获取有效的测试信息，传感器优化布置研究就成为一个热点课题。但是，由于水工结构的局部或大部经常处在水下且测试需求多样，现有的传感器布置方法往往很难满足水工结构健康监测的需求。第二，模态参数识别是获取振动信号后的后续工作，也是进行损伤诊断的必要手段。环境激励下的水工结构受外界噪声影响较大，信号质量不易保证，而模态参数识别的精度将直接影响到水工结构损伤诊断的精确程度，因此寻求一种适合于环境激励下水工结构模态参数识别的方法十分关键。第三，尽早了解结构的损伤状况有助于提高结构的预期可靠性和安全性并减少维护费用。然而，由于环境激励输入的未知性和随机性以及传感器数目的限制等因素，使得环境激励下水工结构的损伤诊断研究同样遇到了很大困难。基于以上三个方面，本文从水工结构的整体动力特性出发，利用环境激励下水工结构产生的振动响应，通过新型传感器优化布置技术和信号处理技术保证结构模态参数的识别精度，进而提取对结构损伤敏感的特征量，实现对水工结构损伤的精确诊断。整套方法不影响结构的正常运行，且可以解决水工结构尤其是其水下部位损伤检测难的问题，有效填补了现有理论方法的不足。本文提出的水工结构损伤整体精细识别方法主要包括以下五个方面的创新性工作：（一）满足不同动力测试需求的水工结构传感器优化配置方法。为满足环境激励下水工结构的各类动力测试需求，对经典有效独立法的局限性进行了深入剖析，然后分别引入能量系数、距离系数和空间相关指数等概念对传统方法进行了全面改进，首次提出了能够满足强环境噪声背景下测试需求的三轴有效独立-驱动点残差法及不依赖有限元网格划分精度的距离系数-有效独立法和空间相关-有效独立法。新方法的有效性和适用性在厂房和大坝等水工结构的健康监测中得到了全面验证。（二）基于新型粒子群算法的水工结构测点最优化布置方法。为提高智能算法在求解水工结构传感器优化布置这一高维度组合优化问题时的全局寻优性能，首次提出了多种群整数编码粒子群算法和多种群克隆选择和粒子群混合算法等新型智能优化技术，通过对兼顾模态可测性和信息独立性的适应度函数进行优化，保证了智能优化技术在求解高维度传感器优化布置问题时的高效性和可靠性。（三）环境激励下基于NExT联合奇异熵定阶的模态参数群智能识别方法。该方法首先利用小波降噪等信号处理方法提升信号品质，然后应用NExT法计算结构的脉冲响应函数，再对该脉冲响应函数进行奇异熵定阶，最后借助一种新型变异粒子群算法的全局寻优得到结构的各阶模态频率和阻尼比等参数。该方法所需设置的参数少，识别精度高，对噪声不敏感，非常适合于环境激励下水工结构的模态参数识别。（四）基于模态频率的水工结构裂缝损伤整体识别方法。提出了一种利用实数编码克隆选择和粒子群混合算法优化模态频率指标的导墙结构损伤诊断方法。该方法仅需可测性强的低阶模态频率，非常适合于环境激励条件下的大型水工结构的无损动态损伤检测。将该诊断方法应用于某导墙的裂缝位置和程度识别中，证明该方法在算法全局寻优性能和识别准确性上均有较大优势，可尝试在各类水工结构的损伤诊断中推广应用。（五）融合传感器优化布置、模态识别和应变模态测试技术的水工结构损伤整体精细识别方法。提出了一种利用多种群实数编码粒子群算法优化应变模态指标的水工结构智能损伤诊断方法。该方法融合了新型传感器优化布置技术和信号处理技术，仅需有限测点的低阶应变模态，且无须对模态振型进行质量归一化，尤其适合于环境激励条件下大型水工结构的无损动态损伤检测。最后，针对不同噪声水平和不同测点数时的损伤组合进行研究，证明本文所提出的损伤诊断方法是十分有效的，可尝试在各类水工结构的损伤诊断中推广应用。更多还原

【Abstract】 The structural health monitoring and damage detection become increasinglyimportant as the scale of hydraulic structures increases and the working conditionsimprove. Firstly, research on optimal sensor placement (OSP) has become a veryimportant topic due to the need to obtain effective testing results from limited testresources. However, the existing methods for OSP are often difficult to meet theneeds of hydraulic structural health monitoring, because part of or most part of thehydraulic structures are always underwater and the test requirements are various.Secondly, modal parameter identification is the follow-up work of vibration signalmeasurement, and it is also the necessary means for damage detection. The noiseaffects the hydraulic structures under ambient excitation obviously, so the quality ofthe signal is difficult to guarantee. Because the accuracy of modal parameteridentification directly affects the accuracy of damage detection, it is very critical tofind an efficient method which is suitable for the modal parameter identification ofhydraulic structures under ambient excitation. Thirdly, early damage detection notonly improves safety and reliability of structures but also reduces maintenance cost.However, damage detection is difficult to implement in hydraulic structures underambient excitation because of the uncertainty of ambient excitation and the limitationof sensors.Based on the above three reasons, this study considers the global dynamiccharacteristics of hydraulic structures, uses the vibration responses of hydraulicstructures under ambient excitation, emploies the OSP technique and signalprocessing technique to ensure the accuracy of modal parameter identification,extracts the features which are sensitive to structural damage, and achieves theaccurate diagnosis for hydraulic structural damage. Without affecting the normaloperation of hydraulic structures, the entire methodology can solve the difficulties indamage detection of hydraulic structures, especially the underwater parts, and can fillup the lackness of existing theoretical methods. There are five innovative points in theglobal and fine identification theory for hydraulic structural damage: (1) The OSP methods of hydraulic structures for various dynamic testrequirements. In order to meet the various test requirements for hydraulic structuresunder ambient excitation, this study dissects the limitations of typical effectiveindependence (EfI) method, introduces the energy coefficient, distance coefficient andspatial correlation index to improve the EfI method comprehensively, and pioneersthe triaxial effective independence driving-point residue (EfI3-DPR3) methodsatisfying the test requirements under strong noise environment and the distancecoefficient-effective independence method and spatial correlation-effectiveindependence method which do not rely on the finite element mesh generation. Theeffectiveness and applicability are comprehensively verified in the health monitoringof the hydraulic structures, such as the hydropower house and dam.(2) The OSP methods for hydraulic structures based on novel particleswarm optimization (PSO) algorithms. To further improve the global optimizationeffectiveness of intelligent algorithm dealing with OSP problems of high dimensions,this study first proposes an integer-encoding multi-swarm particle swarm optimization(IMPSO) algorithm and a hybrid intelligence algorithm of clonal selection algorithm(CSA) and discrete particle swarm optimization (DPSO), optimizes the fitnessfunctions considering the modal observability and information independence, andthen assures the efficiency and reliability of intelligent algorithms when dealing withOSP problems of high dimensions.(3) The swarm intelligence modal parameter identification method based onnatural excitation technique (NExT) and order determination of singularentropy under ambient excitation. The method empolies the wavelet de-noisingtechnique to improve the quality of signals, uses NExT method to calculate theimpulse response function, ensures the order of singular entropy for the impulseresponse function, and finally obtains the structural modal frequency and dampingratio based on a novel mutation particle swarm optimization algorithm. The method isvery suitable for the modal parameter identification of hydraulic structures underambient excitation due to fewer parameters, high identification accuracy andinsensitivity to the noise.(4) The global damage identification method for hydraulic structural cracksbased on modal frequency. A new damage detection method, which employs a realencoding hybrid algorithm of clonal selection and particle swarm optimization tooptimize the modal frequency index, is proposed for guide wall structures. The proposed method only requires low modal frequency, thus making the methodsuitable for nondestructive dynamic damage detection of large hydraulic structuresunder ambient excitation. Taking a guide wall structure as an example, results showthat this method has advantages in the global searching performance and identificationaccuracy. The proposed method is effective and can be applied in many types of largehydraulic structures.(5) The global and fine identification theory for hydraulic structural damageby the OSP technique, modal identification method and strain modal testtechnique. This study proposes a new damage detection method that employs the realencoding multi-swarm particle swarm optimization algorithm and fitness functionsevolved from strain modes to find the optimal match between measured and simulatedmodal parameters and to determine the actual condition of structures. The proposedmethod merges the OSP technique with signal processing technique, requires lowfrequency modes and incomplete modes and does not require mass normalization ofparameters, thus making the method suitable for nondestructive dynamic damagedetection of large hydraulic structures under ambient excitation. The efficiency of theproposed method was analyzed by using different noise levels and sensor numbers.Results show that the proposed method is effective and can be applied in many typesof hydraulic structures.更多还原