The study of 2D NMR logging inversion method based on the maximum entropy-Bayesian theory

Two-dimensional nuclear magnetic resonance logging has a clear advantage in the aspect of reservoir fluid identification and evaluation through measuring transverse relaxation time T
₂
and diffusion coefficient D of formation hydrogen nuclear simultaneously. Truncated singular value decomposition and standard Tikhonov inversion method are often applied to obtain the relaxation-diffusion spectra, but both methods must make nonnegative constraints, which not only increase the amount of calculation but also make the information loss easy. This paper proposes the maximum entropy-Bayesian inversion method of two-dimensional nuclear magnetic resonance, which adds a priori information in the process of inversion, and makes all solutions nonnegative at the same time, on the basic of relaxation-diffusion model of nuclear magnetic resonance. The measurement information is acquired to the greatest extent. Through simulation and experimental verification, the relative error of the transverse relaxation time T
₂
and diffusion coefficient D improve from 22.65% and 28.35% to 11.26% and 6.91% respectively. This method can be applied to the two-dimensional nuclear magnetic resonance logging.