Automatic feature detection and interpretation in borehole data

Al-Sit, Waleed
Automatic feature detection and interpretation in borehole data. PhD thesis, University of Liverpool.

[img] Text
Al-SitWal_June2015_2014181 .pdf - Unspecified
Available under License Creative Commons Attribution.

Download (8MB)


Detailed characterisation of the structure of subsurface fractures is greatly facilitated by digital borehole logging instruments, however, the interpretation of which is typically time-consuming and labour-intensive. Despite recent advances towards autonomy and automation, the final interpretation remains heavily dependent on the skill, experience, alertness and consistency of a human operator. Existing computational tools fail to detect layers between rocks that do not exhibit distinct fracture boundaries, and often struggle characterising cross-cutting layers and partial fractures. This research proposes a novel approach to the characterisation of planar rock discontinuities from digital images of borehole logs by using visual texture segmentation and pattern recognition techniques with an iterative adaptation of the Hough transform. This approach has successfully detected non-distinct, partial, distorted and steep fractures and layers in a fully automated fashion and at a relatively low computational cost. Borehole geometry or breakouts (e.g.borehole wall elongation or compression) and imaging tool decentralisation problem affect fracture characterisation and the quality of extracted geological parameters. This research presents a novel approach to the characterisation of distorted fracture in deformed borehole geometry by using least square ellipse fitting and modified Hough transform. This approach approach has successfully detected distorted fractures in deformed borehole geometry using simulated data. To increase the fracture detection accuracy, this research uses multi-sensor data combination by combining extracted edges from different borehole data. This approach has successfully increased true positive detection rate. Performance of the developed algorithms and the results of their application have been promising in terms of speed, accuracy and consistency when compared to manual interpretation by an expert operator. It is highly anticipated that the findings of this research will increase significantly the reliance on automatic interpretation.

Item Type: Thesis (PhD)
Additional Information: Date: 2015-06 (completed)
Subjects: ?? TK ??
Depositing User: Symplectic Admin
Date Deposited: 30 Jul 2015 09:36
Last Modified: 17 Dec 2022 00:50
DOI: 10.17638/02014181