基于局部异常系数的混凝土坝变形异常分析方法

针对回归分析方法在数值拟合方面存在的问题,根据变形监测历史数据,考虑大坝变形、水位、温度等数据的相关性,将大坝变形监测数据看作多变量时间序列,采用鲁棒性较强的Eros距离度量多变量时间序列间的相似性,并利用局部异常系数对变形数据进行挖掘和异常检测,提出了基于局部异常系数的大坝变形异常分析方法。实例应用结果表明,该方法与统计学方法所得结论一致,能较好地描述变形监测数据的异常,可用于混凝土坝变形异常评价。

Abnormal Deformation Analysis of Concrete Dam Based on Local Outlier Factor

Considering the correlation of deformation and water level as well as temperature, the dam deformation monitoring data can be treated as multivariate time series. The Eros distance, which is strong robustness, is used to measure similarity between multivariate time series. Further, deformation data is discovered and outlier detected by using local outlier factor. Finally, a new analysis method of dam abnormal deformation is established based on local outlier factor to solve the numerical fitting problem of regression analysis. The example results show that the conclusion obtained by the proposed method is consistent with statistical method. The proposed method can describe the abnormality of deformation monitoring data well, which can be used in the evaluation of concrete dam deformation.