基于多尺度分解的微地震噪声压制与初至检测方法研究

Microseismic noise suppression and onset detection method based on ICEEMD

地面微地震数据信噪比很低,严重影响了初至拾取的精度及反演结果的可靠性。为此,对基于改进的完备总体经验模态分解(ICEEMD)的去噪方法与初至检测方法进行了研究,首先利用ICEEMD将非平稳信号分解为一系列相对平稳的固有模态函数,然后提出了一种自适应间隔阈值去除固有模态中噪声成分的方法,最后将去噪后的分量相加重构去噪后的信号。应用Hilbert变换计算每个分量的振幅,然后计算持续能量比,利用给定的阈值找到局部最大值,计算得到高能量的地震信号的到达时间。理论模型数据及实际微地震资料的处理结果表明,去噪后数据的信噪比得到了改进,相对于传统的空间域滤波与变换域阈值去噪,该去噪方法具有显著的优势及较好的应用价值,与Hilbert变换结合的初至检测方法可以有效地检测微地震信号初至。

Ground microseismic data have low SNR,which affects the accuracy of onset detection and the reliability of inversion results.This paper therefore proposes a seismic denoising method and an onset detection method based on improved ensemble empirical mode decomposition (ICEEMD).First,ICEEMD is used to decompose non-stationary signals into a series of relatively stationary individual components called intrinsic mode functions.Then,an adaptive interval threshold is used to remove the noise component in the intrinsic mode functions.Finally,the denoised components are added to reconstruct the denoised results.The onset detection method is combining the Hilbert transform on the basis of the ICEEMD denoising.The Hilbert transform is applied to the decomposed components to calculate the amplitude of each component,and the continuous energy ratio is then calculated.The arrival time of high energy seismic signals is calculated by finding the local maximum value through a given threshold.Test results on synthetic data and actual microseismic data showed that,in denoising,the proposed method is superior to traditional methods,such as spatial domain filtering and threshold denoising methods in transform domain.The onset detection method combined with the Hilbert transform can effectively detect the first arrival of seismic signals.