Machine Learning Analysis of Cave Mine Pillar Collapses
Loading...
Authors
Quevedo, Ricardo
Date
Type
thesis
Language
eng
Keyword
caving , cave mining , pillar stability , collapse , machine learning
Alternative Title
Abstract
The mining industry faces a series of modern challenges driven by the increased demand for minerals, depletion of superficial deposits, reduction in grades and increased depth of mining operations.
The preferred alternative to address this scenario are caving methods due to their high production capacity and low operational costs, which under the new deeper mining environments will also encounter critical stability challenges.
Stability evaluation in rock engineering relies on approaches like rock mass classification systems, empirical methods, and numerical modelling which are widely adopted in mining. Nevertheless, caving-related critical phenomena like rockbursts, inrush and massive collapse deformations have a complex multifactorial nature that goes beyond the scope of these traditional methods.
Machine Learning (ML) algorithms are capable of processing high volumes of data to produce models that capture patterns in their underlying structure. Due to their capacity for handling highly complex problems, they have been introduced in a wide range of scientific disciplines, although their adoption in rock mechanics domains is still in early stages.
These methods present opportunities to model, study and attain better knowledge about the mechanisms that lie behind the development of complex phenomena, and can account for factors outside the scope of traditional stability analysis methods.
This thesis illustrates this concept through two analyses, showing how different geological and design factors relate to pillar collapses in a cave mine in Northern Chile. An overview of the state-of-the-art applications of ML methods for rock mechanics is presented and discussed before the analyses are introduced.
In the first analysis, a tree-based ML model is developed highlighting the relationship between mine design parameters and geological conditions with pillar collapse likelihood at the production level of the cave mine. The second analysis generalizes such framework, developing a process for automated variable processing and using different ML models to obtain the most adequate effect attributions for the variables.
The results show that ML approaches are a viable alternative to model complex cave-related instabilities such as pillar collapses, allowing for a better understanding of the phenomena which can be used to generate mine design guidelines to mitigate their potential occurrences.
Description
Citation
Publisher
License
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution-NonCommercial-NoDerivs 3.0 United States
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution-NonCommercial-NoDerivs 3.0 United States