Author ORCID Identifier

https://orcid.org/0009-0000-5374-5960

Date Available

5-6-2025

Year of Publication

2025

Document Type

Master's Thesis

Degree Name

Master of Science in Mining Engineering (MSMIE)

College

Engineering

Department/School/Program

Mining Engineering

Faculty

Steven Schafrik

Faculty

Joseph Sottile

Faculty

Zach Agioutantis

Abstract

This thesis addresses significant safety challenges presented by powered haulage fatalities in underground mining by developing and evaluating a machine learning-driven Collision Avoidance Information System (CAIS). The research utilized a ZED 2i camera to capture both RGB and depth data for enhanced spatial awareness in visually limited underground environments. A specialized dataset of underground mining equipment was captured from limestone and zinc mines, annotated, and used to train a segmentation network. The CAIS was field-tested in diverse underground settings, achieving an accuracy of 82% in a limestone mine similar to the training data, but a lower accuracy of 45% in a zinc mine with different equipment, highlighting the need for environment-specific training. The research demonstrates the potential of Machine Learning systems to provide real-time hazard detection and improve operator awareness, contributing an adaptive approach to collision avoidance in underground mining. The findings underscore the importance of expanding training datasets and refining models to enhance the reliability of CAIS across various underground mining environments, ultimately working towards the goal of zero mining fatalities.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2025.53

Funding Information

2023-2025 Center for Disease Control (CDC)/ National Institute for Occupational Safety and Health (NIOSH): Under the Central Appalachian Regional Educational Collaborative (CARERC)

Contract number 75D30123C17307: “in-mine underground collision avoidance information system”

Download

Share

COinS