在役桥梁动应变长期监测数据的极值估计  

Extreme-value Estimation of Dynamic Strain of Bridge in Service Based on Long-term Monitoring

作　　者：阳霞[1] 任伟新[1] 张静[1] 袁平平[2] 

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

出　　处：《铁道学报》2017年第11期97-102,共6页Journal of the China Railway Society

基　　金：中国博士后科学基金(2015M581982);安徽省自然科学基金(E080505)

摘　　要：车辆荷载效应是桥梁安全评估及可靠性研究的关键因素,基于长期应变监测数据的应变极值估计很有意义。以往估计车辆荷载效应模型的观测样本相对较少,常用的统计模型对于样本高分位点的估计也不够准确。采用太平湖大桥120d的动应变监测数据,选取车辆荷载作用下应变峰值为研究样本,确定所需最小采样时长。采用广义的Pareto分布作为样本超阈值概率模型,滤过的泊松过程作为样本超阈值随机过程的概率模型,结合极值统计理论,估计桥梁剩余服役期内车辆荷载引起的应变极值,给出桥梁不同承载能力情况下主梁的可靠指标及失效概率。结果表明:广义的Pareto分布模型可较好地拟合出桥梁动应变的右尾分布,能够比较准确地得出车辆荷载作用下应变极值的概率分布。Vehicle load effect is very important for the safety evaluation and the reliability analysis of in-service bridges,therefore the extreme strain estimation is very valuable based on the data acquired from structural health monitoring system.Generally,sampled data used for model study on the vehicle load effect is limited in number,and common statistical models are inaccurately in fitting high tail distribution.In this paper,based on 120 days of dynamic strain data of the Taipinghu Bridge,firstly,the required length of sampling time was determined,and strain peaks due to vehicle load were extracted as samples.The General Pareto distribution was chosen as the probabilistic model for the samples over threshold.And the filtered Poisson process was chosen as the stochastic process probabilistic model.Then extreme strain during the remaining service period was predicted.Lastly,the reliability index and failure probability of the girder were calculated under different cases.Results show that General Pareto distribution fit the right tail distribution of dynamic strain well,and the extreme value distribution of the strain induced by vehicle load can be estimated.

关 键 词：桥梁健康监测 应变极值 滤过的泊松过程 模拟退火算法 概率模型 广义PARETO分布 

分 类 号：U441[建筑科学—桥梁与隧道工程]