基于n个标准差法和箱线图法识别变形监测中异常值的应用探究

Application of Processing Abnormal Values in Deformation Monitoring Based on N-standard-deviation Method and Boxplot Method

以某工程多期沉降观测资料为检测对象,采用n个标准差法和箱线图法识别监测数据中的异常值,并对两种方法进行分析比较。结果表明:当精度要求较高或对变形量数据要求一般时,可适当改变两种方法中的限制参数n,以识别不同要求下的异常值;标准差法和箱线图法在识别异常值及极限异常值时效率高、作用明显;结合监测点的实际变化图,能更进一步验证两种方法对异常变化量准确识别的正确性;对于离散分布且监测精度要求较高的情况,箱线图法识别微异常值的效果更好。

With the multi-phase settlement observation data of Wanda Twin Towers in Kunming as the monitoring object,the n-standard-deviation and boxplot methods are used to identify abnormal data in the monitored data,and a comparative analysis is made of the two methods.The results show that when the requirement for accuracy is high or the requirement for deformation data is average,the limiting parameter n in the two methods can be changed appropriately to identify abnormal values under different requirements;both methods are efficient and effective in identifying abnormal and limit abnormal values;combining the actual change maps of monitoring points,the accuracy of the two methods in accurately identifying abnormal changes can be further verified;for discrete distribution and high requirements of monitoring accuracy,the box-and-line diagram method is more effective in identifying micro abnormal values.