桥梁汽车荷载响应的极值外推方法综述

Review on Extreme Extrapolation Methods for Bridge Traffic Load Response

摘要:

桥梁汽车荷载响应研究中，极值外推是短时车辆数据预测长周期荷载响应极值的重要方法，论文对汽车荷载研究领域的极值外推方法进行了系统综述，介绍了目前应用的几种主要外推方法，对影响外推极值的关键参数（基础数据和底分布）进行了分析，并重点比较了拟合经验外推方法和最大值外推方法。研究表明：汽车荷载响应极值外推的关键是描述底分布尾部数据走势；条件允许下，应至少选用21天实测数据作为极值外推基础，才能反应实际状况。经验极值外推方法具有一定可信度，但需解决可靠度统一问题；最大值外推方法在基础数据容量大的前提下，外推极值准确性和效率都很高；穿越次数外推方法中，不仅需寻找最优拟合起点、区间和优度检验方法，包含Rice公式在内的经验分布的选择，也至关重要。

Abstract:

Extreme extrapolation methods are critical to predict the extreme value of long-term load response on the base of traffic data acquired in a short time. In this paper, a review on the extreme extrapolation methods in the fields of traffic load research is presented. Several main extrapolation methods are introduced in current applications. Analysis on the key parameters, including basic data and bottom distribution, is performed with consideration to their influence on extrapolated extreme values. A detailed comparison between empirical extrapolation method and maximum extrapolation method is also provided. The research results suggest that the key of extreme extrapolation of traffic load response is the description of tail data tendency of the bottom distribution. Length of the measured data which would be adopted for extreme extrapolation should be at least 21 days to provide a reliable reflection on the practical situations. The empirical extrapolation method is credible in a way when the reliability standard problem is answered. The maximum extrapolation method is very accurate and efficient under the circumstance of basic data with large capacity. The extrapolation method based on level crossing method not only needs to search for the optimum fitting starting point, interval and goodness-of-fit test method, but also significantly depends on the selection of empirical distribution including Rick Formula.