基于 SIP-LOF 算法的地形变仪器监测数据异常识别方法

打开文本图片集

关键词：观测数据；典型事件；异常检测；局部异常因子

中图分类号：P315 文献标志码：A 文章编号：1000-0844（2026）01—0242—09

DOI：10.20000/j.1000-0844.20241021001

An anomaly identification method for monitoring data from crustal deformation instruments based on the SIP-LOF algorithm

FENG Xiaohan1，YANG Jiang1，2 （1.InstituteofSeismology，CEA，Wuhan430071，Hubei，China 2.Wuhan Institute of Seismic Scientific Instrument Co.，Ltd.， Xianning 4370oo，Hubei，China）

Abstract： This study proposes a data mining-based series important point-local outlier factor （SIP-LOF） algorithm to detect anomalous data from crustal deformation instruments. The primary objectives of this proposal are to improve the data availability rate of instruments and the efficiency of preliminary fault diagnosis by maintenance personnel. The initial observation sequence of deformation instruments is then separated into sub-sequences. The outlier distance and LOF of each point in the sequence are calculated，and the data point is identified as an outlier. This process enables the quantification of the anomaly degree of each data point. This approach facilitates the identification of anomalous events in precursor deformation observations， encompassing natural disturbances，equipment failures，and seismic precursors. The findings of the research demonstrate that， in comparison with conventional methods， the proposed method exhibits superior detection performance for precursor data from multiple stations，with a wider coverage of anomaly types. When the LOF threshold is set to 2.5， the average anomaly identification accuracy reaches its peak，which is of significant value for precursor data processing.

Keywords： observation data; typical events;anomaly detection; local outlier factor （LOF）

0 引言

近年来，我国地震监测工作已从单一观测发展到由测震台网、强震动观测台网和地球物理观测台网等组成的数字化、智能化、网络化综合观测体系。（剩余9462字）