Abstract: The induced voltage of surface nuclear magnetic resonance (NMR) based on conduction model is a nonlinear function of water content distribution. A model constrained linear iteration inversion method was introduced to dissolve the nonlinear problem. And during the inversion, two constrained models, flattest and smoothest model, were incorporated into the sensitivity matrix calculated by theoretical formula. The inversion simulation was taken for the homogeneous half space, layered conductive model data and field data. And the results showed that the model constrained iteration inversion methods can get more veliable water contents and distribution from the indued volotage of surface NMR, and these results can be inverted by model constrained inversion from an initial homogeneous water content model. For noise free surface NMR signals, both flattest and smoothest model constrained inversions can infer correct water content distributions. However, to noisy data, by smoothest model constrained inversion, the inferred water content became lower and distributed in a larger range than real model, which can be inverted by flattest model constrained inversion correctly.

Key words: surface nuclear magnetic resonance, model constrain, inversion, electrically conductive medium, groundwater exploration