Symplectic Geometry Mode Decomposition-Based NMR Data Noise Attenuation

Nuclear magnetic resonance (NMR) logging is an effective method for evaluating petrophysical parameters and fluid identification, and it is significantly promising for unconventional reservoir exploration and development. However, the NMR signal detected is very weak in unconventional reservoirs due to low porosity and complex pore structure, which leads to a low data signal-to-noise ratio (SNR) and thus affects the accuracy of $T_{2}$ inversion. It is therefore necessary to undertake NMR data noise attenuation before inversion to improve data SNR, so as to enhance NMR inversion accuracy. A noise attenuation method for NMR data based on symplectic geometry mode decomposition (SGMD) was proposed in this article. The kurtosis was introduced in the SGMD process to determine the optimal decomposition group, and then the Pearson correlation coefficient was utilized to choose the optimal number of reconstruction components. The aim is to make the noise separated from the original signal adaptively with an efficient decomposition speed. The effects of the decomposition group and the reconstruction component were analyzed by numerical simulation. The validity and practicality of the proposed method were tested by comparing it with the wavelet transform method and the empirical mode decomposition (EMD) method through numerical simulations, core experiment data, and field NMR data.