Logging-while-drilling (LWD) nuclear magnetic resonance (NMR) tools play a crucial role in reservoir evaluation. However, in horizontal wells with thin layers smaller than the tool’s sensitive region, deciphering the contribution of each layer to the total signal poses a formidable challenge. The general thinking is that all the thin layers are contributing the same to the total acquired signal. However, the reality is more complex than just a box car response function. This complexity arises from the layer’s position relative to the cylindrical sensitive region, making it challenging to develop accurate forward models for LWD NMR tools in such scenarios. A theoretical response scheme was constructed to try to explain how the LWD NMR tools respond in such scenarios. However, conducting an experiment to validate the theory was a must to develop algorithms that can be used in forward modeling such complex response cases.
To address this challenge, we designed and executed an experiment within a controlled environment—a water tank—that replicates the intricate conditions found in horizontal wells with thin beds. To overcome the practical limitations of measuring specific thin-bed characteristics at various positions in a laboratory, we used the calibration setup of the Slim-LWD NMR tool for measurement purposes. The experiment involved a systematic reduction of the water level in the calibration tank in discrete increments, emulating the vertical movement of a thin bed within the horizontal cylindrical sensitive region of the tool. Data acquisition was conducted at predefined intervals as the water level progressively decreased until the tank was empty. The noise level during this experiment is the standard 6 p.u. per phase alternated pair (PAP), and the final stacking level used is five levels. As the tool is stationary in this tank, no motion effects (axial or lateral) are encountered during the experiment.
Through meticulous processing of the acquired water level data, we assessed the individual contribution of each level drop to the original signal. The culmination of this effort yielded a comprehensive lookup table containing the coefficients for horizontal layers as thin as 0.25 in. We believe that the outcome of this experiment can help log analysts have a better understanding of the LWD-NMR tool response in horizontal wells when more than one layer is measured at a single depth. As the experiment has validated the theoretical mathematical response we anticipated, a much thinner layer model can now be safely constructed with any layer thickness.
