用于地震预警系统的单台站防误触发算法研究

Interference prevention algorithm used for single station of earthquake early warning system

利用单台站的信息在较短的时间内识别地震事件和干扰事件是地震预警系统面临的关键问题。研究了基于波形特征参数的决策树方法来提高地震预警系统的单台抗干扰能力。为了研究电磁干扰对地震预警系统的影响,进行了7种电磁干扰试验。使用事件触发后3 s内的波形数据提取5个特征(事件结束否、Speed、Sparse、Pration、对称性),用决策树建立了特征值和事件类型之间的映射关系。使用了大量的强震记录和干扰记录(6359条强震记录、163条干扰记录)对训练后的决策树进行了验证。结果表明,所提方法识别率高,易转换成IF-THEN规则,能方便地写到计算机程序或专家系统中。该算法可以作为地震预警或报警系统的前端地震事件检测模块,以提高系统的可靠性。

Using information of single station to identify seismic events and interference in shorte time is a key issue in earthquake early warning system. Based on characteristic parameters of waveform,this paper studies the application of decision tree method for improving anti-interference ability of single station in earthquake early warning system. In order to study the influence of electromagnetic interference on earthquake early warning system,seven kinds of electromagnetic interference were tested in this paper. Five features(whether the event ended,Speed,Sparse,Pration,Symmetry) were extracted from waveform data in 3 seconds after event trigger,and relationship between feature values and the type of events was established by decision tree. The paper uses numerous strong-motion records and interference records(6359 seismic records and 163 interference records) to validate the trained decision tree,and the results show that this method achieves high recognition rate. This algorithm is easy to convert to IF-THEN rules and can easily be written in computer programs or expert system. This algorithm could be used as the seismic event detecting module for earthquake early warning or alarm system to improve their reliability.