Print this pageFracture Characterization from Borehole Image: a Quantified Approach
Fracture analysis of borehole images is essential for characterizing reservoirs. Conventional analysis relies on a manual picking and classification of sinusoids. Such a picking is precise but time consuming, and almost purely qualitative because it is very interpreter-dependent. This presentation proposes a semi-automatic method in two steps, based on very new techniques in image analysis. The first step is the automatic extraction of each linear segment. The second step addresses the merging of these segments into sinusoids and their classification under the supervision of the interpreter.
On images, fractures often appear non-planar, bed-bounded, or not fully crossing the borehole. As a consequence, on images they appear as discontinuous traces named segments. So in a first step, the proposed method extracts the whole segments, each of them being characterized by a length, an orientation and an aperture. Statistical processing of these segments provides fracture density, in surface per volume of rock, and fracture porosity. Such statistics can be global or related to particular facies or mechanical layers. For a highly fractured reservoir, this new method allows portraying the overall fracturing without any consideration of orientation.
In the second step, individual fractures are identified and classified under the user’s supervision. For each set, the interpreter picks manually one or a few reference fractures and an automatic algorithm classifies segments with compatible orientation. In addition, the same algorithm merges segments that pertain to a same fracture and computes the corresponding orientation. From density and average aperture computed for each set or/and each facies, the user is able to deduce the influence of particular orientation in the future production.
This approach brings the user to a new way of thinking and characterizing fractures. Because the segment extraction is user independent, it provides quantitative results about the overall fracturing inside a particular well that can be compared to other wells no matter who the interpreter is. The classification into different fracture sets is not as good as manual picking, but it is much faster because it only requires the user to supply the main fracture orientations. The resulting loss of accuracy in the delineation of individual fractures is counterbalanced by the ability to quickly supply consistent, repeatable, and accurate fracture densities and statistics along the whole well.
AAPG Search and Discovery Article #90090©2009 AAPG Annual Convention and Exhibition, Denver, Colorado, June 7-10, 2009