基于DBSCAN的绕坝渗流异常值自动识别方法研究

Research on automatic outlier detection for dam bypass seepage based on DBSCAN

绕坝渗流是大坝安全监测的主要工程地质问题之一,其安全监测结果对大坝稳定性评价十分重要。绕坝渗流监测异常值识别能从特定维度反映大坝稳定状况,为大坝管理提供科学依据。目前大多数大坝安全监测采用基于阈值的异常值识别模型,将绕坝渗流视为静态问题,忽略了渗流演变是一个动态系统的本质。针对该问题,提出利用DBSCAN方法构建渗流异常值识别模型,利用渗流数据分布紧密程度实现异常值自动识别。以腊寨大坝绕坝渗流为例,对该方法进行验证,结果表明该方法可准确进行异常值自动识别,在实际工程中有较高的应用价值。

Dam bypass seepage is one of the main engineering geological problems in dam safety monitoring,and its safety monitoring results are very important for the stability evaluation of dams.Outlier detection of dam bypass seepage monitoring can reflect the dam stability status from specific dimensions and provide scientific basis for dam management.At present,the outlier detection model used in most dam safety monitoring is based on threshold,which regards the dam bypass seepage as a static problem and ignores the essence of seepage evolution as a dynamic system.To solve this problem,the DBSCAN method is proposed to build the seepage outlier detection model,which realizes the automatic outlier detection by using the tightness of seepage data distribution.Taking the bypass seepage of Lazhai dam as an example,this method is verified,and the results show that this method can accurately identify outliers automatically,which has high application value in practical engineering.