结构状态监测中的模态参数辨识方法研究

【作者】 李炎华；

【导师】 樊可清；

【作者基本信息】 五邑大学 ， 模式识别与智能系统， 2009， 硕士

【摘要】 工程结构,特别是一些大型的土木工程结构,如大跨度桥梁、超高型建筑、大型水坝及海洋平台等,在长期使用过程中,由于环境侵蚀、材料老化及疲劳效应等因素共同作同,将不可避免的导致系统的损伤积累和抗力衰减,极端情况下可能会发生灾难性的突发事故。对这些结构的状态进行监测,从而及时对结构的健康状况做出评估,在灾难来临之前进行预测是十分必要的。基于模态参数辨识的结构状态监测(Structure Condition Monitoring)的基本原理是:由于结构的模态参数(固有频率、模态阻尼和模态振型)是结构物理特征(质量,阻尼和刚度)的函数,是结构动态特性的参数。结构损伤会导致结构特性的变化,从而影响结构动态特性的变化。因此通过对结构进行动态测试,观察结构的模态参数变化来进行结构状态监测是一种行之有效的方法,而且也是近年来土木工程领域研究的热点。相比传统的模态参数辨识方法,环境激励下的模态参数辨识是以车辆行人载荷、风载、大地脉动、海浪及其组合等作为结构的输入激励,仅从结构的振动响应数据中提取系统的模态参数,经济且省时,更适合工程实际应用。论文首先讨论振动信号预处理的几种常用方法,并对各种预处理方法的实现过程及实际工程应用做了详细的介绍;其次,针对三种常用的环境激励下的模态参数辨识方法:最小二乘复指数法、复模态指示函数法、随机子空间法等进行了深入的研究,同时结合仿真算例对三种算法进行了实验验证。然后论文提出一种新的基于独立分量分析的模态参数辨识方法,该方法主要思想是将系统各阶模态看作独立的虚拟源,即而将传统物理坐标下的模态分析问题转换为线性系统下的盲源分离问题,最后通过仿真算例和对实验室自制简支梁进行模态参数辨识,实验结果表明,该方法能很好的解决线性系统下的模态参数辨识问题。论文最后对本文所做的工作进行总结,并对今后的研究工作和发展方向进行了讨论。更多还原

【Abstract】 Engineering Structure, especially some large-scale civil engineering structures, such as the large-span bridges, ultra-high-type building, large dams and offshore platforms, because of the effect of the environmental erosion, material aging and fatigue effect during their long service period, it is inevitable that the damage of the system will accumulate and the resistance will decay, in extreme cases, sudden catastrophic accidents will happen. It is quite necessary to monitor these structures’s condition and assess their health status timely before the disaster coming. The basic principle of Structure Condition Monitoring based on the modal parameter identification is that modal parameters (natural frequencies, damping ratios and mode shapes) are the function of the structure’s physical characteristics (mass, damping and stiffness), and they are also the parameters of the structural dynamic characteristics. Change in structural characteristics due to damage subsequently affects the dynamic characteristics of structures. Therefore, it is an effective way to monitor the structure’s condition by observing the changes of its modal parameters which can be measured by the structural dynamic testing. Additionally, it has been widely used in civil engineering in recent years.Comparing with the traditional modal parameter identification methods, the modal parameter identification methods under environmental excitations always use the traffic, wind, wave, vibration of the earth, or the combinations of them as the excitation of structure, and only use the structure’s vibration response data to extract modal parameters. It is not only an economic and time-saving way but also quite fit the practical engineering application. Firstly, the thesis discusses the theory of some commonly used vibration signal pre-process methods, and presents the implementation and application of these methods in practical engineering application. Secondly, the research is focused on theory, algorithm and implementation of three popular modal parameter identification methods, the Least-Square Complex Exponential method, the Complex Mode Indicator Function method and the Stochastic Subspace Identification method, which are under the environmental excitations. At the same time, a simulating example is used to verify the algorithms respectively. Thirdly, a new modal parameter identification method by using independent component analysis is proposed. The main idea of this method is to interpret the normal coordinates of a dynamic system as virtual independent sources. Under this certain assumption, the traditional modal analysis problem is converted to the blind source separation problem. Finally, through the simulating example and modal parameter identification of laboratory-made simply supported beam, the experimental results show that this method can solve the modal parameter identification problem of the linear system properly.In conclusion, the thesis summarizes the main work of the research, and the future direction of research and development are discussed.更多还原