Malambule, Thulani Mduduzi2025-08-132024Malambule, Thulani Mduduzi . (2024). Application of Data Analysis and Machine Learning to Develop a Maintenance Strategy for Load-Haul-Dump (LHD) Machines at Booysendal Mine [Master`s dissertation, University of the Witwatersrand, Johannesburg]. WIReDSpace. https://hdl.handle.net/10539/45879https://hdl.handle.net/10539/45879A research report submitted in fulfillment of the requirements for the Master of Science in Engineering, In the Faculty of Engineering and the Built Environment, School of Mining engineering, University of the Witwatersrand, Johannesburg, 2024This report focuses on applying data analysis and machine learning to develop a maintenance strategy for load-haul-dump (LHD) equipment at Northam Platinum’s Booysendal underground mine. This operation predominantly relies on trackless mobile machinery, with a significant emphasis on LHD machines. The mine maintains daily records of mechanical equipment breakdowns. However, Booysendal's reliance on a reactive “run-to-failure” maintenance system has led to operations outside a predefined maintenance plan. The objective of this research was to apply data analytics to understand trends in the dataset and extract meaningful insights regarding LHD breakdowns. It aimed to develop a predictive maintenance model for LHD machines using appropriate machine learning models. Lastly, to design a data-driven maintenance strategy based on insights from data analysis and machine learning (ML). The hydraulic system and transmission components were found to contribute 80% towards downtime. The K-Nearest Neighbours (KNN) regressor was chosen as the best regression model, achieving the lowest Root Mean Square Error (RMSE) of 2.02, while the Support Vector Machines (SVM) classifier was selected as the best classification model with the highest accuracy of 60%. Recommendations on how the predictive models could be improved were highlighted. Finally, a hybrid maintenance strategy is proposed for proactive optimisation. The strategy entails integrating predictive analytics, real time condition monitoring and threshold-based alerts to enable proactive maintenance actions. The proactive actions include ensuring availability of critical spares and conscientizing LHD operators on failures related to bad operating practises.en© 2024 University of the Witwatersrand, Johannesburg. All rights reserved. The copyright in this work vests in the University of the Witwatersrand, Johannesburg. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of University of the Witwatersrand, Johannesburg.UCTDdata analysis and machine learningLoad-haul-dump (LHD)LHD machinesDissertationUniversity of the Witwatersrand, JohannesburgSDG-9: Industry, innovation and infrastructure