前驱波的小波能量与支持向量机分类
|
摘要
目的通过对震前地壳内长周期形变前驱波检测新方法与波形分析的研究,为未来地震预测寻找一个突破点。方法分析了采集到的超低频加速度前驱波信号,发现直接利用前驱波异常信号与地震发生的关系研究得到其关联度仅为51.59%;为提高分析效果,通过小波变换方法对前驱波异常信号进行分解,提取其各分量能量信息,并以此作为特征向量输入支持向量机进行信号分类。结果研究表明利用创新的前驱波检测技术和基于小波分解及支持向量机实现信号分类,异常前驱波与地震发生之间的关联度可达72.028%。结论所提出的方法在探索地震发生的前兆信息方面有一定价值,但在扩大采集地、更多的数据输入、核函数及其相关参数的优化选择等问题尚待进一步的研究。
Aim The earthquake prediction still remains a science problem worldwide,a new detection method and waveform analysis of long period crust deformation precursor wave before the event may become a breakthrough in future earthquake prediction.Methods In the analysis of the signal collected based on this technology,it is found that the accuracy of earthquake prediction using the relationship between the precursor abnormal signals and the earthquake is 51.59%;but in the classification based on the wavelet transform and support vector machine(SVM) method,the precursor wave abnormal signal is first decomposed using wavelet transform,and then the energy of each level is calculated and become the input of the SVM to be classified.Results The result shows that the accuracy of earthquake prediction can come to 72.028% by using the innovative precursor wave detection technology and the combined signal processing method of wavelet transform and SVM.Conclusion The proposed method has a certain advantage in predicting earthquake.Additional studies need to be carried out to further evaluate and improve the approach.
Aim The earthquake prediction still remains a science problem worldwide,a new detection method and waveform analysis of long period crust deformation precursor wave before the event may become a breakthrough in future earthquake prediction.Methods In the analysis of the signal collected based on this technology,it is found that the accuracy of earthquake prediction using the relationship between the precursor abnormal signals and the earthquake is 51.59%;but in the classification based on the wavelet transform and support vector machine(SVM) method,the precursor wave abnormal signal is first decomposed using wavelet transform,and then the energy of each level is calculated and become the input of the SVM to be classified.Results The result shows that the accuracy of earthquake prediction can come to 72.028% by using the innovative precursor wave detection technology and the combined signal processing method of wavelet transform and SVM.Conclusion The proposed method has a certain advantage in predicting earthquake.Additional studies need to be carried out to further evaluate and improve the approach.
引文
[1]梅世蓉,冯德益,张国民,等.中国地震预报概论[M].北京:地震出版社,1993.
[2]KANAMORI H,CIPAR J.Focal process of the greatChilean earthquake on May 22,1960[J].Physics of theEarth and Planetary Interiors,1974,9:128-136.
[3]KIZAWA T.Gravimeter recording before and after greatearthquake[J].Abroad Earthquake,1992,1:19-41.
[4]高金哲,吕政,张洪艳,等.地震前驱波观测与研究进展[J].华南地震,2005,25(1):53-58.
[5]陈德福,潮汐形变前驱波的时空特征[J].大地测量与地球动力学,2006,26(2):24-30.
[6]赵根模,杨港生,陈化然,寂静的前震与地震预测[J].地震,2001,21(1):69-77.
[7]中国地震台网中心.地震数据管理与服务系统[EB/OL].2012-03-11.http://www.csndmc.ac.cn/newweb/catalog-direct-link.htm.
[8]MALLET S G.A theory for multiresolution signal decom-position:the wavelet representation[J].IEEE Trans Pat-tern Anal Mach Intell,1989,11:674-693.
[9]VAPNIK V N.Statistical Learning Theory[M].NewYork:John Wiley,1998.
[10]朱天桥,黄力宇.单导癫痫脑电模糊特征提取的支持向量机发作预测[J].仪器仪表学报,2010,31(11):56-59.
[11]CHERKASSKY V,MA Y.Practical selection of SVM pa-rameters and noise estimation for SVM regression[J].Neural Networks,2004,17(1):113-126.
[2]KANAMORI H,CIPAR J.Focal process of the greatChilean earthquake on May 22,1960[J].Physics of theEarth and Planetary Interiors,1974,9:128-136.
[3]KIZAWA T.Gravimeter recording before and after greatearthquake[J].Abroad Earthquake,1992,1:19-41.
[4]高金哲,吕政,张洪艳,等.地震前驱波观测与研究进展[J].华南地震,2005,25(1):53-58.
[5]陈德福,潮汐形变前驱波的时空特征[J].大地测量与地球动力学,2006,26(2):24-30.
[6]赵根模,杨港生,陈化然,寂静的前震与地震预测[J].地震,2001,21(1):69-77.
[7]中国地震台网中心.地震数据管理与服务系统[EB/OL].2012-03-11.http://www.csndmc.ac.cn/newweb/catalog-direct-link.htm.
[8]MALLET S G.A theory for multiresolution signal decom-position:the wavelet representation[J].IEEE Trans Pat-tern Anal Mach Intell,1989,11:674-693.
[9]VAPNIK V N.Statistical Learning Theory[M].NewYork:John Wiley,1998.
[10]朱天桥,黄力宇.单导癫痫脑电模糊特征提取的支持向量机发作预测[J].仪器仪表学报,2010,31(11):56-59.
[11]CHERKASSKY V,MA Y.Practical selection of SVM pa-rameters and noise estimation for SVM regression[J].Neural Networks,2004,17(1):113-126.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心