Exploring Machine Learning Techniques for Predicting Open Stope Stability in Underground Mining: Evaluating Accuracy and Applicability

• Author / Creator

  Szmigiel, Alicja

• Underground mining operations are inherently dangerous due to a variety of factors present in a mining environment. Firstly, the confined spaces and limited ventilation, the use of heavy machinery, explosives, and drilling equipment poses significant risks to the safety of workers. Moreover, underground mines are susceptible to geological hazards such as rockfalls, collapses, and seismic events.
  Collapsed and caving openings in underground mining are particularly hazardous due to the potential for catastrophic events. When openings collapse or cave in, they can trap workers underground, leading to injuries, fatalities, and the disruption of rescue operations. Furthermore, collapses can destabilize the surrounding rock mass, leading to further collapses.
  Observing and assessing the stability of underground openings is important for several reasons. Firstly, it ensures the safety of workers by identifying potential hazards before accidents occur. By monitoring the stability of openings, mining companies can implement preventative measures such as reinforcement and support systems. Additionally, assessing the stability of underground openings allows for informed decision-making regarding mining operations, ensuring the sustainability and efficiency of production while minimizing risks to personnel and equipment.
  Machine learning methods offer promising solutions to the stability assessment problem in underground mining. Through various techniques such as classification and feature importance analysis, machine learning algorithms can effectively predict and evaluate the stability of underground openings. Classification models can classify openings as stable, unstable, or caved based on input features such as geological characteristics and historical stability data. Feature importance analysis helps identify critical factors influencing stability, enabling targeted interventions.
  This research study presents a comprehensive investigation of various machine learning models applications, aimed to predict the stability of underground mining openings, particularly stopes. Open stopes are integral to underground mining operations, where they serve as excavated voids created during the extraction of mineral resources from underground deposits.
  Chapter 1 of this thesis presents the groundwork by providing a comprehensive overview of the research topic, outlining its primary objectives, the methodology employed, and the structure of the thesis.
  Chapter 2 of this study provides a comprehensive engineering background and overview of open stopes mining operations. The chapter begins with an explanation of the terminology associated with mining methods. Moreover, the chapter elaborates on the most popular methods used for rock mass classification.
  In Chapter 3, an extensive literature review of contemporary methods for assessing the stability of open stopes is presented. This review presents a diverse range of approaches, including empirical methods, statistical analyses, and applications of machine learning techniques, which have been proposed by various researchers to address the challenge of evaluating stope stability.
  Chapters 4, 5, and 6 present the results of various machine learning models, that were developed to predict the stability of open stopes. Chapter 4 utilizes a Potvin database, where stability number N’ and shape factor HR of each historical case were used, and each case had a stability assessment assigned. Random Forest (RF) and Logistic Regression models were employed, evaluated, and compared to achieve the most accurate predictions. In Chapter 5 an extensive analysis of the Potvin database was performed and used to develop the most effective Artificial Neural Network (ANN) model. In this study all the parameters that combine into the stability number N’ were employed and treated as a separate input features for the model. Various ANN model configurations were utilized and evaluated to find the most effective network configuration. The feature importance analysis was then performed to find the parameters that have the highest influence on the stability of an open stope. A final chapter 6, presents an analysis of a larger database obtained from literature, followed by a comparison of several machine learning models. The models’ results were then analyzed and the most important features for each model were determined.
  In essence, this thesis systematically integrates machine learning techniques to predict the stability of open stopes in underground mining. Through this approach, feature importance analysis was conducted to determine the parameters exerting the greatest influence on stope stability. By evaluating the dataset and identifying key influencing factors, the study enables better control over stope stability, it allows for a more precise understanding of the conditions that lead to failure, enhancing safety and operational efficiency in underground mining environment.

• Subjects / Keywords

  • Machine Learning
  • Stope Stability

• Graduation date

  Fall 2024

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/r3-drwj-sd72

• License

  This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.