Fast Forward Modeling of Borehole Nuclear Magnetic Resonance Measurements in Vertical Wells
- Mohammad Albusairi (The University of Texas at Austin) | Carlos Torres-Verdín (The University of Texas at Austin)
- Document ID
- Society of Petrophysicists and Well-Log Analysts
- SPWLA 60th Annual Logging Symposium, 15-19 June, The Woodlands, Texas, USA
- Publication Date
- Document Type
- Conference Paper
- 2019. held jointly by the Society of Petrophysicists and Well Log Analysts (SPWLA) and the submitting authors
- 4 in the last 30 days
- 99 since 2007
- Show more detail
Nuclear Magnetic Resonance (NMR) logs are widely used to ascertain in-situ rock and fluid properties. The vertical resolution of the measurements is mainly controlled by the antenna length, logging speed, signal-to-noise ratio (SNR) and mud/formation electrical conductivity. To overcome SNR limitations, signal depth-stacking is used at the expense of vertical resolution. Hence, borehole NMR measurements suffer from non-negligible spatial averaging of formation properties. Forward modeling and inversion techniques can capture and quantify shoulder-bed, invasion and environmental effects on NMR logs across spatially complex rock formations. However, the computation time required to invoke the forward model restricts the use of inversion methods. Fast and accurate forward numerical methods can overcome these limitations and allow more reliable quantification of formation properties.
We develop a new fast and accurate forward algorithm using spatial sensitivity functions (SSF). The SSFs quantify the NMR porosity dependence on spatial formation property perturbations. The discrete adjoint method (DAE) is employed to efficiently calculate the SSF. Accordingly, a three-dimensional (3D) multiphysics forward model was developed to include NMR tool properties, magnetization evolution, and electromagnetic propagation to derive tool sensitivity maps. We apply the SSF-derived forward approximation to a series of synthetic cases in a vertical well, that include thin layers, variable fluid, and rock properties, and mud-filtrate invaded formations.
Results show that NMR spatial sensitivity depends on two formation properties (porosity and electrical conductivity), NMR pulsing-sequence and tool geometry. The NMR sensitivity is mainly influenced by mud/formation electrical conductivity and not the reference formation porosity. This behavior suggests that one SSF may approximate the NMR response across layered formation with low resistivity contrast. The proposed forward approximation can be computed in a few seconds of central processing unit (CPU) with maximum relative errors of 4%. The SSF-derived forward method constitutes a fast, reliable and efficient alternative for accurate modeling of NMR logs. Additionally, the forward method can be used with inversion algorithms to mitigate spatial smoothing effects across thinly-bedded and invaded sedimentary sequences, for instance.
|File Size||1 MB||Number of Pages||13|