Anomaly Detection of Seepage Monitoring Data of Asphalt Concrete Core Wall Dam based on WT-kNN
The quality of safety monitoring data is of significant importance for the analysis of the safety status of asphalt concrete core wall dam.Trend problems caused by time effects are the difficulties in detecting anomalies in seepage monitoring data.Modal decomposition methods can ef-fectively separate the trend component of time series and then identify abnormal signals.However,in the seepage monitoring data of earth-rock dams,modal aliasing of anomalies and real signals often exists.To solve these problems,the Wavelet Transform combined with local kNN weigh-ted regression(WT-kNN)anomaly detection method is introduced.In the proposed method,continuous wavelet transform is used to separate trend items,after which the detection results of wavelet transform are further scrutinized by the salocal kNN weighted regression,improving the accuracy of the model's anomaly detection.The results of engineering instance applications show that the recall rate of WT-kNN for monitoring se-quences with a gross error ratio of 2.5%～10%is more than 95%,and the misjudgment rate is less than 5%.The model and the WT-MAD method and the SSA-DBSCAN method comparative experiments have verified the effectiveness and superiority of WT-kNN.Sensitivity analysis re-sults show that the proposed model has low sensitivity to the proportion of the number of anomalies to the total data and the size of the anomaly fluctuation range,which can establish a basis for subsequent monitoring data analysis,processing,and early warning.
国家科技期刊平台
NSTL
万方数据
