基于逐步回归分析和Bootstrap重抽样的铁路钢桁桥不确定参数识别  

Identification of Uncertain Parameters of Railway Steel Truss Bridge Based on Stepwise Regression and Bootstrap Resampling

作　　者：骆勇鹏[1] 黄方林[1] 伍彦斌[1] 鲁四平[1] 苏泽平 

机构地区：[1]中南大学土木工程学院,湖南长沙410075

出　　处：《公路交通科技》2016年第11期104-110,共7页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(51378504);贵州省交通厅科技项目(2013-122-001);中南大学中央高校基本科研业务费专项资金项目(2015zzts058)

摘　　要：为获得不确定性参数的概率统计特征,提出一种基于逐步回归分析和Bootstrap重抽样的不确定性参数识别方法。该方法首先根据合理试验设计确定试验点并计算试验点所对应的响应,基于逐步回归分析方法构造表达设计参数和响应关系的最优响应面模型。进而在小样本的情况下,利用Bootstrap重抽样技术对实测响应进行抽样得到大样本数据,结合响应面模型,通过优化反演来求得各个Bootstrap样本所对应的参数值,经概率统计分析得到参数的均值和标准差。采用一组试验钢板和铁路钢桁桥算例来验证所提方法的可行性和可靠性。结果表明,所提方法可准确地识别出不确定性参数的概率统计特征值;与随机模型修正相比,所提方法在保证精度的前提下,具有更高的计算效率,可用于复杂工程结构的不确定性参数量化分析中。In order to obtain the probabilistic statistical feature of uncertain parameters, a new uncertain parameters identification method based on stepwise regression analysis and Bootstrap resampling is proposed. First, according to the reasonable experiment design, the test points is determined and the corresponding responses of test point are calculated in the method, the optimal response surface model which expresses the relationship of design parameters and responses is constructed based on the stepwise regression analysis. Secondly, in the case of small samples, the Bootstrap resampling technology is used for the resampling of the measured responses to get the large-scale samples. Combining with the response surface model, the optimization inversion procedure is constructed to obtain the corresponding parameter values of each Bootstrap sample, and the means and standard deviations of the parameters are obtained by probabilistic statistical analysis. Finally, the feasibility and reliability of the proposed method are verified by the calculation examples of a set of test steel plates and of an existing railway steel truss bridge. The result shows that （1） the proposed method can accurately identify the probabilistic statistical feature values of uncertain parameters;（2） compared with the stochastic model updating, the proposed method has a higher computational efficiency in the case of high precision and can be used in the quantitative analysis of uncertain parameters of complex engineering structure.

关 键 词：桥梁工程 不确定性参数识别 Bootstrap重抽样 铁路钢桁桥 模型修正 

分 类 号：U446.1[建筑科学—桥梁与隧道工程]