Automatic Detection of Anomalous ULF Electromagnetic Changes and Noise Reduction in the Seismo-Electromagnetic Measurements

为实现电磁法地震短期预报，发展了地震电磁测量中异常变化的自动检测和降噪方法。主要研究结果如下：(1)对于时间序列数据异常变化的自动检测，提出了线性自适应滤波概念和非线性神经网络滤波概念的信号预报器，利用过去的数据估计时间序列中的下一个数据。将所设计的预报器应用于高层大气物理中的瞬变信号--Trimpi效应的自动检测。结果表明，在信噪比较好的情况下，两种方法都较好，但在信噪比较差的情况下，神经网络优于神经网络。(2)研究了地震电磁学的信号识别方法。它们是主成分分析(PCA)、分形分析和基于小波变换的站间传递函数(ISTF)方法。将主成分分析应用于…数据。更多的人在伊豆半岛的军队系统服役，这与2000年的伊豆群岛震群有关。发现与太阳活动有关的一个因子在第一主成分中占主导地位。在第二主成分中，人为噪声的影响比较严重，第三主成分与地震相关的电磁现象有关。超低频磁场变化的分形分析结果表明，在大地震前一个月或几天，该阶段似乎是临界的(SOC)。提出了一种基于小波变换的ISTF方法，用于去除由日地相互作用产生的全球外源场。结果表明，在磁场和电场(t&lt；1000秒)下，几乎都能成功地消除外场变化。残差包括人工噪声和地震相关信号等区域信号。人工信号与地震相关信号的分离是未来的问题。较少

To realize the short-term earthquake prediction with using electromagnetic approach, developments of methodologies for automatic detection of anomalous changes and noise reduction in the seismo-electromagnetics measurements. The main results are as follows.(1)As for the automatic detection of anomalous changes in the time series data, signal predictors with a linear adaptive filter concept and a nonlinear neural network filter concept has been developed, estimates the next datum in time series using the past data. The designed predictor is applied to the automatic detection of a transient signal in the upper atmospheric physics, Trimpi effect. The results show that both are good at good SNR but NNF is superior when SNR is bad.(2)The signal discrimination methods for seismo-electromagnetics have been developed and examined. They are principal component analysis (PCA), fractal analysis, and interstation transfer function (ISTF) method using wavelet transform. PCA is applied to the data o … More bserved at the army system in Izu peninsula, which are associated with the 2000 Izu islands earthquake swarm. It is found that a factor associated with solar activity is dominant in the first principal component. In the second principal component, the influence of artificial noise is rather severe, and the third principal component is concerned with the earthquake-related electromagnetic phenomena. Results of the fractal analysis of ULF magnetic variations indicate that the stage seems to be critical (SOC) at one month or a few days before the large earthquake.The ISTF method with wavelet transform has been developed for remove global external source fields originated from solar-terrestrial interaction. The results show the external field variations are almost removed successfully in both magnetic and electric fields (t<1,000sec). The residual includes the regional signals such as artificial noises and earthquake-related signals. The signal separation between artificial and earthquake-related is future problem. Less