One dimensional nuclear magnetic resonance (1D NMR) logging technology is limited for fluid typing. Two dimensional nuclear magnetic resonance (2D NMR) logging can provide more parameters including transverse relaxation time (T-2) and diffusion coefficient (D) in multipore media, and the fluid typing is realized successfully according to the fluid properties of NMR. So, 2D NMR logging is more advantageous than 1D NMR logging. Point to the relaxation mechanism of 2D NMR at gradient field, the echo simulation and inversion of 2D NMR are discussed in detail. And a hybrid inversion algorithm is proposed, which is based on a damping least squares method (LSQR) and truncated singular value decomposition (TSVD). For verifying the hybrid inversion method, a series of spin echo trains in an ideal fluid model are firstly simulated with multiple echo spacing (TE) activation in gradient field. Then, these synthesized echo trains are inverted by hybrid inversion method. The inversion results including T-2 and D are well consistent with the ideal model presupposed, and the hybrid inversion method is more accurate and efficient than single inversion algorithm from the error analysis and runtime comparison, which indicates that the hybrid method is effective and suits to the inversion of 2D NMR logging. Furthermore,the LSQR-TSVD hybrid method is applied into the inversion tests of oil-water and gas-water models with different observation parameters such as magmatic field gradient and echo spacing group. From the inversion results analysis, the reasonable observation parameters are selected and optimized for fluid typing of (T-2, D) 2D NMR Logging.
