信息熵理论在混凝土结构损伤动力识别中的应用研究

发布时间：2019-06-10 15:36

【摘要】：本文在全面回顾结构损伤识别方法的基础上,提出使用信息熵理论对基于振动的混凝土结构进行损伤识别,部分解决目前损伤识别领域存在的若干关键技术问题,为混凝土结构的健康监测提供配套判据。根据信息熵理论,以自功率谱熵为例提出适合分析混凝土结构损伤的熵指标函数构建方法。通过两根不同的钢筋混凝土简支梁试验,验证自功率谱熵对基于振动的混凝土结构的损伤识别效果。对比研究自功率谱熵和功率谱熵在损伤识别方面的异同点,通过实验研究论证本文提出方法的优势。对原始的振动信号添加不同强度的噪声,考察自功率熵的抗噪能力。改进近似熵理论的程序编制方法,提升其计算效率。通过自功率谱密度函数对衰减信号进行分类,针对不同的信号模式提出不同的近似熵计算方法进行结构的损伤识别。使用相关函数将不具规则性的衰减信号转化成具有规则性的互相关时间序列,采用数值模拟结合试验的方法论证对衰减信号引入近似熵理论的可行性。试验研究表明,利用近似熵理论对混凝土结构进行损伤识别时,具有很强的敏感性和抗噪能力。引入改进前和改进后的线性化传递熵理论,利用梁上的加速度信号进行动力损伤识别研究。利用Abaqus软件分析只在主平面内弯曲振动的混凝土简支梁,通过折减指定单元弹性模量的方法模拟梁的弹性损伤。基于梁上多点的加速度信号,利用核密度估计方法验证了直接使用线性化传递熵理论的合理性。通过定义合适的损伤指标,表明改进的线性化传递熵在对混凝土结构进行损伤识别时具有很强的敏感性和鲁棒性。基于Matlab软件对承受高斯激励的Duffing软弹簧振子进行数值研究,计算传递熵的解析解,并以此验证线性化传递熵以及核密度估计传递熵的合理性。通过调整非线性因子模拟系统受不同程度的非线性损伤,利用传递熵结合替代数据方法对该系统进行非线性损伤识别,识别结果表明该方法可有效的捕获系统的非线性特征。基于Abaqus软件,分析一根带裂缝的混凝土简支梁在高斯激励下的振动特性,以裂缝的长度模拟结构的非线性损伤程度,结合替代数据方法,对连续系统进行非线性损伤识别。通过3根不同的钢筋混凝土简支梁试验,验证传递熵结合改进的替代数据方法对实际结构的非线性损伤识别效果。通过定义合适的损伤指标,量化该方法的损伤识别效果,使用完整梁和缺陷梁验证该方法对不同损伤模式的识别能力。从采样频率、抗噪能力以及激励方式等角度,验证传递熵结合替代数据方法在损伤识别时的鲁棒性。全文结合理论分析、数值模拟和试验研究,验证了几个不同的信息熵指标在混凝土结构损伤动力识别方面具有的可靠性和优势,旨在为混凝土结构的损伤识别领域提供一种新的、较为可行的研究思路。
[Abstract]:Based on a comprehensive review of structural damage identification methods, this paper proposes to use information entropy theory to identify concrete structures based on vibration, which partly solves some key technical problems in the field of damage identification. It provides a supporting criterion for the health monitoring of concrete structures. According to the theory of information entropy, taking the self-power spectral entropy as an example, a method of constructing entropy index function suitable for analyzing the damage of concrete structures is proposed. Through two different tests of simply supported reinforced concrete beams, the effect of self-power spectral entropy on damage identification of concrete structures based on vibration is verified. The similarities and differences between self-power spectral entropy and power spectral entropy in damage identification are compared and studied, and the advantages of the proposed method are demonstrated by experimental research. The noise resistance of self-power entropy is investigated by adding different intensities of noise to the original vibration signal. The programming method of approximate entropy theory is improved to improve its calculation efficiency. The attenuated signals are classified by self-power spectral density function, and different approximate entropy calculation methods are proposed for structural damage identification for different signal modes. The non-regular attenuation signal is transformed into a regular cross-correlation time series by using the correlation function. The feasibility of introducing approximate entropy theory to the attenuated signal is demonstrated by using the method of numerical simulation and experiment. The experimental results show that the approximate entropy theory has strong sensitivity and anti-noise ability in damage identification of concrete structures. The linear transfer entropy theory before and after improvement is introduced, and the dynamic damage identification is studied by using the acceleration signal on the beam. The simply supported concrete beams with bending vibration only in the main plane are analyzed by Abaqus software, and the elastic damage of the beams is simulated by reducing the elastic modulus of the specified element. Based on the acceleration signal of multiple points on the beam, the rationality of using the linear transfer entropy theory directly is verified by the kernel density estimation method. By defining the appropriate damage index, it is shown that the improved linear transfer entropy has strong sensitivity and robustness in damage identification of concrete structures. Based on Matlab software, the numerical study of Duffing soft spring oscillator excited by Gao Si is carried out, and the analytical solution of transfer entropy is calculated, and the rationality of linear transfer entropy and kernel density estimation transfer entropy is verified. By adjusting the nonlinear factor to simulate the system subjected to different degrees of nonlinear damage, the transfer entropy combined with alternative data method is used to identify the nonlinear damage of the system. The identification results show that the method can effectively capture the nonlinear characteristics of the system. Based on Abaqus software, the vibration characteristics of a simply supported concrete beam with cracks under Gao Si excitation are analyzed. The nonlinear damage degree of the structure is simulated by the length of cracks, and the nonlinear damage identification of continuous system is carried out combined with the alternative data method. Through the tests of three different reinforced concrete simply supported beams, the effect of transfer entropy combined with improved alternative data method on nonlinear damage identification of practical structures is verified. By defining the appropriate damage index, the damage identification effect of the method is quantified, and the complete beam and defect beam are used to verify the recognition ability of the method to different damage patterns. From the point of view of sampling frequency, anti-noise ability and excitation mode, the robustness of transfer entropy combined with alternative data method in damage identification is verified. Combined with theoretical analysis, numerical simulation and experimental research, this paper verifies the reliability and advantages of several different information entropy indexes in damage dynamic identification of concrete structures. The purpose of this paper is to provide a new and feasible research idea for the field of damage identification of concrete structures.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TU317;TU37

【参考文献】