【作者】 张亮；

【导师】 宋玉普；

【作者基本信息】 大连理工大学 ， 结构工程， 2008， 硕士

【摘要】 网架结构是近些年来发展迅速的一种空间结构。这种结构适合于大空间、超大空间建筑，投资巨大，往往是人群密集或配置重要设施的场所。如果发生事故将可能引起重大的经济损失和生命危险。近年来发生了许多大跨度空间结构倒塌的实例。因此，如何检测和监测网架的运行状况有很大的实际意义。结构的检测有静力方法和动力方法，动力方法可以得到结构的频率、振型等参数，便于从整体上把握结构的运行性能，本文主要研究网架检测的动力方法。本文以一足尺网架模型的试验研究为基础，通过动态测试试验得出网架的频率、振型等动态参数，并根据这些参数分析网架结构损伤检测的模态方法。以实验网架为分析原型，结合Ansys有限元软件，比较了不同模态指标在网架损伤识别中的有效性，通过添加噪声比较各个指标的抗噪能力。通过比较得出模态应变能方法和最小二乘支持向量机方法结合比较适合网架损伤识别。具体做法是模态应变能初步定位疑似损伤杆件，再结合最小二乘支持向量机对损伤准确定位和定量。通过数值模拟，假设杆件单损伤并添加一定噪声，识别结果证明文中采用的方法是可行的。最后，考虑到实际检测中，网架结构自由度数目巨大，不可能得到所有自由度上的振动信息，会遇到测量模态不完备的情况，文中分析了传感器的优化布置，以及得到不完备测量数据后如何对网架结构的运行情况进行评估。文中通过单损伤工况，通过比较不同模态扩阶方法，得出采用有效独立法选择自由度，使用动态扩阶得到完备模态，再使用模态应变能和最小二乘支持向量机相结合的方法进行损伤识别能够得到满意的结论。更多还原

【Abstract】 Steel grid structure, which can fulfil the needs of long-span spatial sructures, has gained growing popularity at home and abroad. As this structure has a large span and can serve many people , an accident will threaten people life and cause huge economic losses and property loss. There are many collapse and accidents of long-span spatial structures in latest years, so it is important to identify and monitor space grid structure. In the field of identification, there are many mature methods which are divided into two parts: Static and Dynamic Methods. It is convenient to get the whole state of structure by dynamic method which is operated by getting the mode parameters such as frequency and mode shape and is more useful for space truss.In this paper, based on the experiment of a truss, modal methods are used and analysed with the mode parameters such as frequency and mode shape obtained by dynamical tests.The validity and robustness of different methods are compared by adding white noise to the datum of numerical simulation with Ansys. Results show that the damage of truss can be effectively identified by the method of modal strain energy plus LS-SVM (least-square support vector machine). Modal strain energy is used to identify the probable damage elements and the exact damage location and extent can be attained by using LS-SVM. This method is verified by numerical simulation of a space truss with some nosie.Finally, optimal sensor placement and how to identify space truss with incomplete mode data are analyzed because it is impossible to get the complete mode data of space truss which has too many elements. Results attained by numerical simulation of signal damage element verify that the method combined by effective independence method, dynamic expansion and LS-SVM is efficient for space truss.更多还原