A Novel 2-D Inversion Method for Low-Field NMR Working on Fluid Typing

Low-field (LF) nuclear magnetic resonance (NMR) technology has been widely used in reservoir identification, NMR logging and other geophysical exploration fields. Due to the advantage of providing abundant parameter information related to fluid (e.g. the longitudinal and transverse relaxation time, T ₁ and T ₂ ), it requires the advanced inversion approach for data interpretation from the measured echo data. As the present inversion schemes show weakness, especially in fluid typing and quantitative analysis, we developed a novel imaging scheme based on the classical Butler–Reeds–Dawson (BRD) algorithm frame. Employing the discrepancy principle, the new method aims to improve the selection for regularized parameter for inversion, and further increases the imaging accuracy. To verify the ability and reliability of this method with strong interference, we compare the fluid typing inversion results of simulated data for different noise level between the conventional and improved approach with variable data acquisition wait time and echo spacing. The results indicate that the estimation accuracy of T ₁ - T ₂ spectra for updated scheme gets about 8 percent improvement even with high-noise environments. To further evaluate the advantage, this algorithm is also tested with real oil-water fluid experimental data, which yields a distribution of fluid properties that matches the sample. In conclusion, the research in this paper shows significant for LF-NMR data interpretation, in particular for charactering pore-fluid and quantifying organic contamination in subsurface sediments in the geological survey.