Abstract Three-dimensional (3D) nuclear magnetic resonance (NMR) provides more reliable fluid typing and petrophysical evaluation in tight sandstone than traditional one-dimensional (1D) NMR. However, neglecting the effects of mud invasion saturation and rock wettability under invasion conditions can lead to misinterpretation. To address this, a numerical simulation approach is introduced to examine their effect on 3D NMR responses and the accuracy of fluid identification across three reservoir scenarios: water invaded by oil-based mud (OBM), water invaded by oil, and oil invaded by OBM, each evaluated at varying saturation levels. The numerical simulation algorithm was developed based on tight sandstone pore geometry, constituent fluid distributions, and NMR signal source parameters. Numerical solutions to simulate the longitudinal relaxation time (T1) and transverse relaxation time (T2) distribution are obtained by integrating differential and integral relationships between relaxation time and pore-scale properties. The formulation incorporates statistically distributed bulk relaxation and diffusion parameters modeled as log-normal Gaussian functions. Analyzing the 3D NMR response demonstrates that, in conditions of water-wet and oil-wet, a decrease in wetting fluid saturation increases surface relaxation, resulting in shorter T2 and T1. Meanwhile, the non-wetting fluid relaxation time remains unchanged, with only an amplitude change observed. In contrast, fluid saturation significantly influences the fluid phase’s relaxation time and signal amplitude under mixed-wettability conditions. Moreover, the fluid type recognition could become complex. The T1 to T2 geometrical mean (T1lm / T2lm) ratio was analyzed as a potential parameter for classifying different reservoirs and, consequently, to help identify fluid types. Although the analysis revealed a strong correlation between T1lm / T2lm and oil/OBM saturation under varying wettability conditions, the similarity of T1lm / T2lm across the simulated reservoir scenarios limits its effectiveness for precise classification.
Ishag et al. (Thu,) studied this question.