期刊文献+

北京房山站神经网络高程时间序列分析

Analysis on height time series of Beijing Fangshan magmatic station of neural network
原文传递
导出
摘要 为了更好地分析高程时间序列的变化规律,使其能更好地应用到地壳运动监测.本文对房山站观测数据先采用PanTa方法剔除异常值,然后进行混沌动力学系统检验.经检验可知,北京房山站高程时间序列为非线性、非稳态时间序列,观测系统为部分确定性机制低阶混沌动力系统.因此对数据进行正则化RBF网络相空间重构和小波神经网络滤波,然后再对数据进行加窗谱估计、最小差距拟合和去除趋势FFT周期拟合计算分析.计算结果显示,北京房山站高程时间序列的平均谱功率为7694W,年周期为0.994~1.109(年),半年周期为0.438~0.617(年),年周期为0.247~0.378(年).分析计算结果表明,该时间序列斜趋势不明显,随机性不明显,其存在着年周期性、半年周期性和季周期性,其中年周期最为明显. These dates are widely applied to crustal movement monitoring,navigation and earth physics,and studying on other related scientific filed.Especially,applying to crustal movement application,for earthquakes often happen in some areas in China have caused heavy casualties and property losses.This paper first uses the PanTa method to eliminate outliers and based on chaotic dynamical system inspection.We apply to amplitude adjustment Fourier algorithm of constraint to generate alternative and Lyapunov index inspection time series nonlinear time series. Through the analysis,the elevation of time series Lyapunov index in the range of 0.0618~ 0.0618,and biggest Lyapunov index is positive,so the elevation time series for unsteady time series.According to the nonlinear model of the test,the elevation of time series Lyapunov index in the range of 151.3914~191.2036,so the time series was nonlinear.According to System uncertainty test,εεsamein the range of-0.1637~0.0759,so the time series observation system for part of the deterministic mechanism loworder chaotic dynamical system.Through the analysis,Beijing Fangshan magmatic station elevation for the nonlinear and unsteady time series,time series observation system for part of the deterministic mechanism low-order chaotic dynamical system. So date was phased space reconstruction by regularization RBF and wavelet neural network filter.According to windowing spectrum estimation, minimum distance decoding fitting and fitting of FFT cycle.Calculation results show that spectral power of time series of Beijing Fangshan was 7694 W,periodic year was range from 0.994to1.109(year),periodic half year is range from 0.438 to 0.617(year)and periodic season was range from 0.247 to 0.378(year),we can know this time series tendency is not obvious,randomness is not obvious,however exist in periodic year,periodic half year and periodic season,among them,year periodic is the most obvious.
出处 《地球物理学进展》 CSCD 北大核心 2014年第3期1084-1089,共6页 Progress in Geophysics
基金 国家自然科学基金(50674032)资助
关键词 高程时间序列 RBF神经网络 小波神经网络 加窗谱估计 最小差距法 elevation of time series RBF neural network wavelet neural network windowing spectrum estimation minimum distance decoding
  • 相关文献

参考文献15

二级参考文献185

共引文献314

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部