车辆荷载作用下桥梁应变极值估计的阈值选取  

Threshold Selection for the Extreme Value Estimation of Bridge Strain Under Vehicle Load

作　　者：阳霞[1] 张静[1] 任伟新[1] 袁平平[2] 

机构地区：[1]合肥工业大学土木与水利工程学院,合肥230009 [2]中南大学土木工程学院,长沙410075

出　　处：《应用数学和力学》2017年第5期503-512,共10页Applied Mathematics and Mechanics

基　　金：中国博士后科学基金(2015M581982);安徽省自然科学基金(1708085ME101)~~

摘　　要：采用过阈法估计车辆荷载作用下桥梁的应变极值,合理的阈值选取十分关键.阈值选取过大,信息量少,阈值选取过小,广义Pareto分布模型参数估计偏差大.常用的阈值选取方法不能较好地适用于车辆荷载作用下的应变极值估计.基于太平湖大桥车辆荷载作用下1年的应变数据,对拟合结果较好的3种混合分布进行Monte-Carlo(蒙特-卡洛)抽样,对比同一样本基于不同阈值的广义Pareto分布模型的极值估计结果,提出了一种经验式的阈值选取方法.与常用阈值选取方法相比,根据文中方法所得阈值估计的周应变极值分布与实测结果更为接近,估计结果更好.The selection of a reasonable threshold is critical to estimate the extreme strain un- der vehicle load on bridges with the peak-over-threshold method. Little information can be used if the threshold is too high, while the bias of parameters of the general Pareto distribution will be large ff the threshold is too low. Common threshold selection methods are not suitable to be applied in estimation of the extreme strain under vehicle load. Based on 1-year strain data of the Taiping Lake Bridge, 3 types of mixed distributions for the strain peaks induced by vehicle load were chosen to generate a large number of samples with the Monte-Carlo method. The esti- mated extreme values of the samples based on the generalized Pareto distributions with differ- ent thresholds were compared and analyzed. Then, an empirical threshold selection method was proposed for the strain data induced by vehicle load. Finally, the Taiping Lake Bridge was cho- sen as the case verification. It is demonstrated that the estimated weekly extreme strain based on the threshold selected with the proposed method is more close to the measured results than those with the common methods.

关 键 词：桥梁健康监测 阈值选取方法 应变极值 广义PARETO分布 

分 类 号：U441[建筑科学—桥梁与隧道工程]