Classification of Drilling-Induced Fractures and Their Relationship to In-Situ Stress Directions

Bernt S. Aadnøy: Rogaland University, Stavanger, Norway J. Sebastian Bell: Geologic Survey of Canada

Abstract: Natural and drilling-induced fractures that can be identified on borehole walls are classified in terms of the failure initiation mechanisms that generate them and the principal stress orientations with respect to well trajectory. Fracture initiation mechanisms include tensile failure in vertical holes, tensile failure in inclined holes, tensile failure in elliptical boreholes, tensile failure resulting from plastic deformation, shear failure in vertical boreholes subject to low fluid pressures, shear failure in vertical boreholes subject to high fluid pressures, shear failure in inclined boreholes subject to low fluid pressures, shear failure in inclined boreholes subject to high fluid pressures, and failure in boreholes subject to horizontal stresses of equal magnitudes. Analyzing these situations assuming linear elasticity shows that the geometry of fracture traces on borehole walls depends on the in-situ stress tensor, the relative orientations of the stress tensor and the borehole, the fluid pressures in the borehole and in the surrounding rock, the rock's cohesive strength, and its friction angle for shear. Furthermore, hydraulically induced tensile fractures are governed by the least principal stress, but shear fractures are governed by the intermediate and the least principal stresses, contrary to previous conclusions. A classification based on the fracture initiation analysis is presented and is illustrated by examples of different types of borehole wall fractures recorded by image logging tools run in wells in western Canada.

Induction Tool with a Coplanar Coil System

P. R. de Carvalho: Faculdade de Ciências Agrárias do Pará-DCEE/FCAP, Belém, PA, Brazil O. P. Verma: Universidade Federal do Pará-DGf/UFPA, Belém, PA, Brazil

Abstract: Analog model studies were carried out comparing the electromagnetic responses of coaxial and coplanar two-coil induction tools in a borehole in order to determine their vertical resolution. For this purpose geoelectric scale models were constructed at a 1:20 scale. These models simulate well logging situations in stratified beds of varying thicknesses with or without fluid invasion. The sensitivity of the system to measure relative fields (secondary/primary) is of the order of 0.01%. The following conclusions were made: o In thin conducting beds placed in a relatively resistive formation, such as shale layers in a sandstone containing hydrocarbons, the coaxial system shows a better resolution than the coplanar both in the boundary demarcation and in estimating the thicknesses of thin beds. o Thin resistive beds in a conductive zone, such as sandstone layers containing hydrocarbons embedded in a shale formation, are better resolved by a coplanar coil system than by a coaxial array. o Shoulder effects are much more pronounced in the coaxial system than in the coplanar array. o The small horns in the coplanar profiles against the interfaces help in precisely demarcating these bed boundaries. However, these horns gradually disappear with the reduction in thickness of the beds.

Relaxation Time and Diffusion Measurements of Methane and n-Decane Mixtures

Sho-Wei Lo, George J. Hirasaki, Riki Kobayashi, and Waylon V. House: Rice University, Houston, Texas

(RESEARCH NOTE) Previous work has shown that pure alkanes, alkane mixtures, viscosity standards, and stock tank crude oils have NMR relaxation times that vary linearly with viscosity/temperature on a log-log scale (Brown, 1961; Zega et al., 1989; Tutunjian and Vinegar, 1992; Beznik, 1994; Kleinberg and Vinegar, 1996; Morriss et al., 1997). However, pure methane at some temperatures and pressures does not follow the trend because it relaxes by a spin-rotation mechanism in addition to dipole-dipole interactions, while other hydrocarbons relax only by dipole-dipole interaction (Oosting and Trappeniers, 1971; Jonas, 1973). The spin-rotation mechanism has a different dependence on viscosity. Live crude oil contains significant amounts of methane, and the linear correlation is not practical for live crude oil. Therefore, the study of methane-hydrocarbon mixtures, which are the main components of "live" crude oil, is of interest. In this work, we measured relaxation times of methane-n-decane mixtures at reservoir conditions and observed the relationship between relaxation times and viscosity. Diffusion coefficients of methane-n-decane mixtures were also measured by pulse field gradient measurements.

Use of Interpolation Neural Networks for Permeability Estimation from Well Logs

P. M. Wong: School of Petroleum Engineering, The University of New South Wales, Sydney NSW, Australia S.A.R. Shibli1: Sarawak Shell Berhad, Miri, Malaysia

Abstract: Reservoir permeability estimation from well logs is a difficult task for petrophysicists. Many studies have shown that backpropagation neural network (BPNN) is the most promising tool to date, because of its ability to learn and generalize. The use of gradient descent method in the backpropagation algorithm, however, often provides smoothed estimates. This article provides an inherently nonsmooth solution; namely, interpolation neural network (INET). It compares the performance of both neural networks by means of a case study. The results indicate that INET provides accurate and inherently nonsmoothed permeability profiles for the blind-test wells, compared to the BPNN method. It is also easy to implement in desktop computers.