基于多点关联性的尾矿坝位移监测序列异常值诊断

Diagnosis on abnormal values in displacement monitoring sequence of tailings dam based on multi-point correlation

针对尾矿坝位移监测序列中噪声和真实异常值的区分问题,提出1种基于多点关联性和改进孤立森林(IF)算法的异常数据诊断模型。通过IF算法对监测序列中的各样本点异常程度进行量化计算,引入云模型(CM)算法确定IF量化的异常得分与异常概念的相互映射关系以实现异常点的初步诊断,根据Apriori算法计算多测点序列间的关联性,找出强关联序列组合,结合序列关联性以及异常点诊断结果区分噪声与真实异常值。以某尾矿坝位移监测序列为例进行模型验证。研究结果表明:基于多点关联性的异常诊断模型能够有效区分尾矿坝位移监测序列中的噪声与真实异常值,提高监测系统的准确性。

Aiming at the problem in distinguishing the noise and real abnormal values in the displacement monitoring sequence of tailings dam,a diagnosis model of abnormal data based on the multi-point correlation and improve isolated forest(IF)was put forward.The abnormal degree of each sample point in the monitoring sequence was quantitatively calculated by using the IF algorithm,and the cloud model(CM)algorithm was introduced to determine the mutual mapping relationship between abnormal score and abnormal concept of IF quantification,so as to realize the preliminary diagnosis of abnormal points.The correlation between multi-point sequences was calculated according to the Apriori algorithm,then the strong correlation sequence combination was found out,and the noise and real abnormal values were distinguished combining with the sequence correlation and the diagnosis results of abnormal points.The model validation was conducted by taking the displacement monitoring sequence of a tailings dam as example.The results showed that the anomaly diagnosis model based on multi-point correlation could effectively distinguish the noise and real abnormal values in the displacement monitoring sequence of tailings dam,thus improve the accuracy of monitoring system.