基于经验模态分解的核磁共振去噪方法研究

核磁共振（NMR）在孔隙结构评估和流体识别方面具有独特的优势，但NMR信号很容易受到噪声影响。根据NMR噪声的时域和频域特征，提出了基于一种经验模态分解（EMD）的NMR去噪方法。首先，利用EMD将信号由高频到低频分解为一系列的本征模态函数，以此分解噪声和噪声NMR信号，然后，使用曲线趋势法和改进的过零点率曲线确定信号噪声分离准则，将有用信号叠加到剩余项以获得去噪信号。通过岩芯数据和测井数据对比发现，基于EMD的去噪方法可以提高信噪比的同时保留孔隙结构信息，其去噪效果优于小波阈值和EMD小波阈值法，计算得到的孔隙度接近实际孔隙度。

Research on NMR Denoising Method Based on Empirical Mode Decomposition

NMR has unique advantages in pore structure evaluation and fluid identification, but NMR signals are easy to be influenced by noise. This paper presents a kind of denoising method according to time and frequency domain characteristics of NMR noise based on EMD. Firstly we decompose noise and noisy NMR signals using EMD. Then we determine guidelines of signal-noise separation using curve trend method and improved zero-crossing rate curve. Finally we add useful signals to residual term to obtain pure signals. It is confirmed by core data and logging data experiments that the denosing results are better than wavelet threshold and EMD-wavelet threshold method and porosity calculated is closer to real porosity and the inversion results are consistent with actual pore structure. The denosing method based on EMD can improve signal to noise ratio and reserve pore structure information.