采用Bootstrap抽样的靖远黄河大桥 模态参数识别不确定性量化

提出一种基于Bootstrap抽样的模态参数识别不确定性量化方法,从整体和局部的角度评价模态参数识别结果的可靠性.首先,基于动力测试的加速度时程数据,采用协方差驱动随机子空间(SSI-COV)法识别不同测试组的模态参数;引入Bootstrap抽样方法,对多组模态参数识别结果进行B次重复抽样,得到Bootstrap样本数据,并通过其概率统计特征值衡量整体不确定性.然后,对单个测试组中不同时间段的识别结果进行重复抽样,分析并量化单个测试组的模态参数识别的不确定性.最后,以靖远黄河大桥试验数据为例,对靖远黄河大桥竖向单个及多个测试组下的模态参数进行不确定性量化.结果表明:不同测试组识别的前3阶固有频率的均值分别为1.553 9,1.720 6,2.165 2,方差分别为0.076 1,0.042 9,0.096 5;单个测试组识别的前3阶固有频率的均值分别为1.526 5,1.788 0,2.306 0,方差分别为0.015 3,0.049 6,0.018 2;文中方法识别的固有频率值总体较为稳定.

Uncertainty Quantification for Modal Parameters Identification of Jingyuan Yellow River Bridge Using Bootstrap Sampling

An uncertainty quantification method for modal parameter identification based on Bootstrap sampling was proposed, and the reliability of modal parameter identification results is evaluated from the global and local perspectives. Firstly, the covariance-driven stochastic subspace identification(SSI-COV)method was adopted to identify the modal parameters of different test groups based on the acceleration time-history data of the dynamic test. Secondly, the Bootstrap sampling method was introduced to repeated B times sampling to obtain Bootstrap sample data in the result of multi-modal parameter identification. The overall uncertainty was measured by calculating probabilistic and statistical eigenvalues. Then, the results identified in different timeperiods in a single test group were sampled repeatedly to analyze and quantify the uncertainty of modal parameter identification in a single test group. Finally, taking the test data of Jingyuan Yellow River Bridge as an example, the modal parameters of vertical single and multiple test groups of Jingyuan Yellow River Bridge were quantified with uncertainty. The mean values of the first three natural frequencies identified by multiple test groups were 1.553 9, 1.720 6 and 2.165 2, the variances were 0.076 1, 0.042 9 and 0.096 5. The results show that mean values of first 3 natural frequencies identified by a single test group were 1.526 5, 1.788 0, and 2.306 0, the variances were 0.015 3, 0.049 6 and 0.018 2. The natural frequency identified by this method is generally stable.