Application of Machine Learning to Assess the Stability of Underground Mining Drifts:A Comparative Study of Classification Algorithms

Authors

  • NEVAID DZIMUNYA Author
  • KUDAKWASHE DZWITI Author
  • ROBERT NYIRENDA Author

Keywords:

underground mining drift, numerical modelling, data mining, Naïve Bayes, k-NN, Random Forest, J48 algorithms

Abstract

The stability of underground mine development excavations such as drifts and crosscuts should be guaranteed throughout their lifespan. The stability assessment of these underground openings is a difficult task owing to factors such as the complexities in the geology, in situ stress regime and construction of excavation. This complexity is connected to uncertainties in rock mass characterisation as well as the variability of the rock parameters involved. However, the processes used in the design and analysis of the excavations yield large sums of geotechnical data which makes machine learning techniques suitable tools for analysis. In this article, popular classification algorithms (Naïve Bayes, k-NN, Random Forest and J48) were compared for applicability in underground mining drift stability assessment. The algorithms were trained and tested on a dataset developed from numerical modelling. The performance assessment of the four models indicates that the Random Forest and J48 classifiers can both be useful in predicting underground drift stability with acceptable accuracy.

Downloads

Published

2024-12-31