Abstract:

Because of the advantages of surface nuclear magnetic resonance, such as direct, quantitative and unique inversion, it is widely used in hydrological environment investigation, early warning of disaster water source, and other fields. However, in practical applications, the nanovolt level SNMR signals is often submerged in the environmental noise and difficult to break down, which makes the result inaccurate. In this paper, a method for extracting SNMR signal envelopes based on local mean decomposition (LMD) is proposed. First, the real and imaginary envelopes of the noisy SNMR signal are decomposed sequentially from high frequency to low frequency. Then the noise interferences in the signal are removed, and the required signal components are extracted. Finally, the effective real and imaginary components are synthesized to obtain the target SNMR signal envelope. The results indicate that the fitting error of the initial amplitude of the SNMR signal envelope extracted by LMD  is within ±4.17%, and the error of the average transverse relaxation time is within±5.63%. The signal-to-noise ratio is improved by 30.3-37.2 dB.

Key words:  , surface nuclear magnetic resonance, envelope extraction, local mean decomposition, signal-to-noise ratio, error of fitting ,