基于数据驱动和物理模型的结构地震损伤识别方法研究

发布时间：2018-06-17 05:09

本文选题：数据驱动 + 物理模型　；参考：《哈尔滨工业大学》2013年博士论文

【摘要】：近年来世界频繁发生大地震并造成了巨大的经济损失、人员伤亡及社会影响，这是因为一方面人们对地震地面运动的特征认识不足，另一方面是人们还无法准确把握工程结构在真实地震作用下的地震损伤机理和破坏机制。近30年发展起来的结构健康监测系统通过布设传感器获取原型结构在真实服役环境及荷载作用下的实际响应，实时地诊断结构的损伤并及时地发出预警警报，是当前研究和揭示工程结构在真实地震作用下结构地震响应、损伤机理和破坏机制的最直接、最有效的手段。传统的基于振动信号的结构损伤识别方法一般认为结构损伤前后都是线性的，由于强地震地面运动的非平稳性使得工程结构在强地震作用下表征明显的非线性行为，如混凝土裂缝的张开与闭合，钢筋或钢材的屈服或屈曲等，这使得传统的结构损伤识别方法并不直接适合识别结构的地震损伤。因此迫切需要重新审视地震地面运动的非平稳特性，发展利用结构地震作用下非线性特征和非线性系统识别方法的结构地震损伤识别方法和损伤评估方法。为此，本文研究地震地面运动时频能量分布、以及基于数据驱动提取非线性特征的结构地震损伤识别方法和基于物理模型识别非线性系统的结构地震损伤识别方法。主要研究内容如下：研究基于匹配追踪分解算法和时频能量分布的地震地面运动生成方法。首先介绍匹配追踪分解算法的基本原理；然后构造具有相同峰值加速度和傅里叶幅度谱的简单人工地震地面运动，比较它们时频能量分布的差异以及时频能量分布对线性结构最大弹性位移的影响，进一步定义时频原子幅度、循环特征、滞后相位角和有效能量，分析地震地面运动有效能量与最大弹性位移的相关性；最后分析人工地震地面运动时频能量分布对弹塑性结构响应的影响规律，通过时频能量分布解释非线性结构地震响应的动态棘轮效应。研究基于层间位移时频特征分形维数的剪切型结构地震损伤识别方法。首先定义时频特征并采用分形维数定量刻画；然后结合模态叠加原理推导线性结构各自由度处时频特征分形维数的性质，结合子结构方法得到非线性结构各子结构时频特征分形维数的性质；最后以10层剪切型结构为数值仿真对象，模拟结构在单层损伤和多层损伤，，不同损伤程度，近场地震动和远场地震动、不同噪声水平下的识别效果，并讨论不同阻尼比和采用层间加速度对识别结果的影响。研究基于位移信号时频特征分形维数的框架型结构地震损伤方法。首先结合静态缩聚法、模态叠加原理推导线性框架型结构各主自由度处时频特征分形维数的性质；然后对于梁、柱端产生塑性铰的非线性框架型结构，研究拟力法推导塑性铰到非弹性位移的转换矩阵并得到各主自由度的时频特征分形维数的性质，对于局部安装非线性构件如阻尼器的非线性框架型结构，研究逆反路径法分析局部非线性对结构相对位移的影响并得到各主自由度的时频特征分形维数的性质。其次通过16层框架型结构数值算例验证结构单层和多层梁、柱出现塑性铰时的时频特征分形维数的识别方法；最后通过安装了摩擦阻尼器（模拟结构体系的局部非线性）的16层钢框架型结构振动台试验验证所提方法。研究基于扩展Kalman和无迹Kalman滤波相结合的结构地震损伤分散识别方法。首先介绍经典Kalman滤波，扩展Kalman滤波和无迹Kalman滤波的基本原理；然后研究分别用扩展Kalman滤波和无迹Kalman滤波识别线性和非线性子结构的扩展状态。其次以10自由度剪切型结构为数值仿真对象，模拟结构在单层损伤和多层损伤下的识别结果，并比较同基于扩展Kalman滤波和无迹Kalman滤波整体结构识别方法的识别结果。最后结合识别的力—位移曲线和修正的Park-Ang损伤指标对结构的地震损伤定量评估。
[Abstract]:In recent years, great earthquakes have occurred frequently in the world and caused great economic losses, casualties and social effects. This is due to the lack of understanding of the characteristics of earthquake ground motion. On the other hand, it is impossible for people to accurately grasp the mechanism and mechanism of earthquake damage under the effect of real earthquake. In recent 30 years The structural health monitoring system is designed to obtain the actual response of the prototype structure under the actual service environment and load, to diagnose the damage of the structure in real time and to send out the warning warning in time. It is the research and Revelation of the structural seismic response, the damage mechanism and the damage mechanism of the engineering structure under the real earthquake. The traditional method of structural damage identification based on vibration signals generally considers that the structural damage is linear before and after the damage. Due to the nonstationarity of the strong earthquake ground motion, the engineering structure is characterized by the obvious nonlinear behavior under the strong earthquake action, such as the opening and closing of the concrete cracks, the bending of steel or steel. The traditional structural damage identification method is not directly suitable for the seismic damage identification of the structure. Therefore, it is urgent to reexamine the non-stationary characteristics of the seismic ground motion, and develop the structural seismic damage identification method and damage assessment method using the nonlinear characteristics and nonlinear system identification methods under the structural seismic action. Therefore, this paper studies the time frequency energy distribution of seismic ground motion, structural seismic damage identification method based on data driven nonlinear characteristics and structural seismic damage identification method based on physical model identification nonlinear system. The main contents are as follows:
The method of seismic ground motion generation based on matching tracking decomposition algorithm and time frequency energy distribution is studied. First, the basic principle of matching tracking decomposition algorithm is introduced. Then, a simple artificial ground motion with the same peak acceleration and Fu Liye amplitude spectrum is constructed, and the difference of time frequency energy distribution and time frequency energy are compared. The influence of the cloth on the maximum elastic displacement of the linear structure, further defining the amplitude of the time frequency atom, the characteristic of the cycle, the lag phase angle and the effective energy, analysis the correlation between the effective energy of the earthquake ground motion and the maximum elastic displacement, and finally analyze the influence of the frequency energy distribution on the response of the elastoplastic structure when the time of the ground motion of the artificial earthquake is passed. The frequency energy distribution explains the dynamic ratcheting effect of nonlinear structures.
The seismic damage identification method of shear type structure based on the time frequency characteristic fractal dimension of interlayer displacement is studied. Firstly, the time frequency characteristics are defined and fractal dimension is used to describe the seismic damage. Then, the properties of the fractal dimension of the time frequency characteristics at each degree of freedom in linear structures are deduced with the principle of modal superposition, and the knot structure method is used to obtain the sub nodes of the nonlinear structures. The properties of the fractal dimension of the time-frequency characteristic are obtained. Finally, the 10 layer shear structure is used as the numerical simulation object to simulate the recognition results under the single layer damage and multi-layer damage, the different damage degree, the near ground motion and the far ground vibration, and the different noise levels, and discuss the influence of the different damping ratio and the interlayer acceleration on the recognition results.
The frame structure seismic damage method based on the fractal dimension of the time frequency characteristic of the displacement signal is studied. Firstly, the properties of the fractal dimension of the time frequency characteristic of the linear frame type structure are derived by the static condensation method and the modal superposition principle. Then the nonlinear frame structure of the plastic hinge at the end of the beam and the column is derived, and the derivation of the pseudo force method is studied. The property of the time frequency characteristic fractal dimension of each principal degree of freedom is obtained by the plastic hinge to the transition matrix of the inelastic displacement, and the inverse path method is used to analyze the influence of local nonlinearity on the relative displacement of the structure and obtain the fractal dimension of the time frequency characteristic of the main degrees of freedom for the nonlinear frame structure of the local installation of a nonlinear component, such as the damper. Secondly, a numerical example of 16 layers frame structure is used to verify the method for identifying the fractal dimension of the time frequency characteristic of the structural single and multi-layer beams and the columns when the plastic hinge appears. Finally, the proposed method is verified by the 16 layer steel frame type vibration table test of the friction damper (the local nonlinearity of the simulated structural system).
The structure of seismic damage identification based on the combination of extended Kalman and non trace Kalman filtering is studied. First, the basic principles of classical Kalman filtering, extended Kalman filtering and Untraced Kalman filtering are introduced. Then, extended Kalman filtering and Untraced Kalman filtering are used to identify the extended state of linear and nonlinear substructures. Secondly, the extended state of linear and nonlinear substructures is identified. The 10 degree of freedom shear type structure is used as the numerical simulation object to simulate the identification results of the structure under single layer damage and multi layer damage, and compare the identification results with the whole structure recognition method based on the extended Kalman filter and the Untraced Kalman filter. Finally, the structural earthquake is combined with the identified force displacement curve and the modified Park-Ang damage index. Quantitative assessment of damage.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TU311.3

【相似文献】