LAUSR

Abstract Low-field Nuclear Magnetic Resonance (NMR) is a non-invasive method widely used in the petroleum industry for the evaluation of reservoirs. Pore structure and fluid properties can be evaluated from transverse relaxation (T2) distributions, obtained by inverting the raw NMR signal measured at subsurface conditions or in the laboratory. This paper aims to cast some light into the best practices for the T2 data acquisition and inversion in shales, with a focus on the suitability of different inversion methods. For this purpose, the sensitivity to various signal acquisition parameters was evaluated from T2 experiments using a real shale core plug. Then, four of the most common inversion methods were tested on synthetic T2 decays, simulating components often associated with shales, and their performance was evaluated. These inversion algorithms were finally applied to real T2 data from laboratory NMR measurements in brine-saturated shale samples. Methods using a unique regularization parameter were found to produce solutions with a good balance between the level of misfit and bias, but could not resolve adjacent fast T2 components. In contrast, methods applying variable regularization – based on the noise level of the data – returned T2 distributions with better accuracy at short times, in exchange of larger bias in the overall solution. When it comes to reproducing individual T2 components characteristic of shales, the Butler-Reeds-Dawson (BRD) algorithm was found to have the best performance. In addition, our findings suggest that threshold T 2 cut-offs may be derived analytically, upon visual inspection of the T 2 distributions obtained by two different NMR inversion methods.