基于物理参数时域辨识的结构损伤识别研究

项目名称： 基于物理参数时域辨识的结构损伤识别研究

项目编号： No.51208478

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 建筑环境与结构工程学科

项目作者： 王祥建

作者单位： 中国地震局工程力学研究所

项目金额： 25万元

中文摘要： 本项目将以典型结构（框架、悬臂和桁架结构）为研究对象，针对输入信息和输出信息的测试完备程度，分别采用基于经典最小二乘算法、复合反演算法、扩展卡尔曼滤波算法和广义复合反演算法的结构物理参数时域辨识方法，进行结构损伤识别研究。通过提出或引入的比例阻尼的伪线性化法、奇异值分解与修正Levenberg-Marquardt联合法、矩形窗法、渐消记忆技术等，完善基于物理参数时域辨识的结构损伤识别技术，研究噪声、参数初值选取、求解方法、矩形窗法等对结构损伤识别及地震动反演结果的影响。进行典型结构模型的振动台实验来验证结构损伤识别方法的有效性，并给出利用扩展卡尔曼滤波技术重构无效的测试信号。

中文关键词： 损伤识别；最小二乘算法；复合反演算法；广义复合反演算法；振动台试验

英文摘要： The typical structures, like frame, cantilever and truss structure, will be studied in this project. For different measured level of input and output information, identification methods in time domain of structural physical parameters based on classical Least-Square, hybrid inversion, extended Kalman filter and generalization hybrid inversion algorithms are used for structural damage detection. With the proposed or introduced methods, such as pseudo linearization of proportional damping, combined method of SVD and modified Levenberg-Marquardt, rectangular window, memory fading technology and so on, the structural damage detection technology based on physical parameter identification in time domain is improved. And some factors which are noise, initial values of parameters, solving methods and rectangular window method are studied for their influence on the results of structural damage detection and input inversion. Some shaking table tests of structural models will be done for verifying the effectiveness of the theoretical methods. And a new method using extended Kalman filter technology is proposed to estimate the measured failure signal.

英文关键词： damage detection；least squares algorithm；hybrid inversion algorithm；generalized hybrid inversion algorithm；shaking table test