Problem. Porosity and relaxation time distributions are the primary outputs of all downhole nuclear magnetic resonance (NMR) tools. Portions of the distributions provide valuable information, i.e., bound and free-fluid volumes, moveable fluid identification, and the estimation of oil viscosity within heavy oil or tar formations. Porosity measured by NMR is subject to environmental influences, such as temperature, local magnetic field, mud type, and, in particular, tool motion.

Motion causes a plethora of distortions to the data and affects all low-gradient nuclear magnetic resonance logging tools. Lateral motion distorts the echo train, at times manifesting as erratic noise or a complete loss of cohesion. Axial motion, on the other hand, can cause loss of magnetization or severe over-polarization. Tool motion also affects relaxation time distributions. Methods for axial motion artifact-free data exist, but real-time telemetry (mud pulse) bandwidth limitations have made practical implementation challenging.

The drilling environment also introduces challenges to real-time processing. To deliver porosity and distributions free from motion artifacts requires knowledge of the motion. Axial motion can be easily calculated at surface; however, processing the data from the time domain to a relaxation time domain—a process referred to as inversion—at the surface for acceptable accuracy requires more echoes than is feasible for mud-pulse transmission. To address both the requirement for surface data processing and the limits on data transmission, we introduce a split inversion in detail.

Methodology. The inversion is split into two parts. A first inversion is performed where the bulk of the information is processed downhole, the data is reconstructed, and then sent uphole. The reconstruction focuses on the primary points of interest for the measurement, such as amplitudes for a T1 sequence or sub-sample echoes for a T2 echo train. This smaller set of data, which retains the signal-to-noise ratio of the full dataset, effectively compresses the data and makes it easier for transmission. A second inversion, performed at the surface, accounts for the rate of penetration. This second inversion uses documented methods for motion-inclusive inversion or motion correcting the data prior to inversion. The motion-inclusive inversions demonstrated in this paper use a kernel with motion information to obtain porosity and distribution that are free from motion artifacts.

Results. This paper demonstrates the methodology for using the split inversion, surface rate-of-penetration (ROP) information, and computational modeling for motion-inclusive inversion to enable the delivery of reliable NMR data. The method is applied to an example synthetic log and lab-simulated log in which the porosity is known. Correct porosity and distributions are obtained for the lab data with the split inversion and compared to the real-time answer that would be delivered without the ability to rectify the over-polarization observed in the data. The results presented are applicable to the axial motion, which can be calculated at the surface using the time-depth log, unlike lateral motion. The split inversion method also has applications as a data-compression method for any NMR tool.

This content is only available via PDF.
You can access this article if you purchase or spend a download.