借助预警指标实时辨识大坝服役性态是大坝安全监控的重要手段。传统多依据大坝服役性态效应量测值序列的年极值为分析子样来估计预警指标。本文基于极值理论中的POT(Peaks over Threshold)模型,通过阈值的设定,以超限数据序列作为建模分析的对象,利用广义帕累托分布拟合超限数据子样,结合大坝的失事概率实现预警指标的估计。该方法不研究效应量测值序列的整体分布情况,只关注序列的超限值分布情况;充分考虑了所有较大测值出现的可能,更好地体现了数据样本的分布特征,因此得到的预警指标能更客观地反映工程实际。
Early-warning index should be built to identify dam service behavior and to monitor the safety of dam. The early-warning index is estimated based on the year extreme values in the serial observation of dam service behavior. The POT (Peaks over Threshold) model in Extreme value theory is introduced to implement the early-warning index estimation in this paper. According to the threshold, the analysis samples are obtained from the data exceeding the threshold, which satisfy the generalized Pareto distribution. The generalized Pareto distribution function of the data exceeding the threshold is determined. The early-warning index can be calculated by the combination of above distribution function and the dam failure probability. The analysis is focused on the distribution feature of the data exceeding the threshold, not the global distri- bution rule of the whole serial observation of dam service behavior. The obtained early-warning index can describe objectively the actual status of dam engineering.