University of British Columbia
Beware the undulation : failure mode prediction and rock mass classification in rock mechanics using deep neural networks and synthetic rock mass models
Abstract
dc:descriptionRock mass classification systems (RMCS) remain prominent within the rock engineering field. Although many of these systems have become ubiquitous, they are predicated upon a series of empirical case studies tailored to particular geographic locations and stress regimes. While each classification system may be valid for the project type and location for which it was created, modern applications extend the use of these systems beyond their initial raison d’être. Furthermore, the case studies which form the basis of these classification systems are scarcely representative of the conditions to which these systems are applied which poses a risk to safety, design stability, and reliability. Rock mass characterisation deals with assigning values to individual components that dictate rock mass behaviour. Rock mass classification attempts to characterise rock mass behaviour or strength based on observed characteristics. This research focuses on the effects of rock mass classification systems. Machine learning models may be trained on images from various locations, thus eliminating a key limitation of current classification systems: location relevance. This thesis examines the application of computer vision (artificial intelligence) techniques for joint mapping and investigates the impact of joint geometry on rock mass strength in synthetic rock mass (SRM) models. Current SRM models depict joint fractures as planar surfaces which risks oversimplifying the interactions observed in geometrically complex, undulated rock masses. Experiments conducted and explored as part of this thesis investigate the effects of large-scale undulation versus conventional flat joint representations on simulated rock mass strength. Multiple discrete fracture networks (DFNs) with flat, geometrically simplified, planar fractures were contrasted against geometrically complex surfaces. Results from this experiment reveal inconsistent patterns between explicit undulated surfaces and rock mass strength across various DFN realisations. These results challenge the prevalent assumption that the ramifications of geometric simplification may be mitigated by parametric adjustments. In fact, it re-enforces the notion that network topology and general fracture connectivity govern rock mass behaviour and strength.
Degree
thesis:*- Name thesis:degree_name
- Master of Applied Science - MASc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Mining Engineering
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ambah, Emmanuela
- Contributors dc:contributor
-
- Elmo, Davide
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/93175