Predictive Maintenance of Mining Haul Trucks Via Machine Learning

Abstract

Mining operations, on a global scale, experience challenges in ensuring high availability and utilization of their haul trucks for optimum production. Most of the shortfalls in availability are due to mechanical breakdowns of haul trucks. The main objective of this research work is to develop an early monitoring and warning system using predictive analytics for haul truck maintenance by applying an artificial intelligence-based approach. The predictive maintenance method relies on data for the prediction of the likelihood of equipment failure. This study employs Multinomial Logistic Regression and K-Nearest Neighbours on a dataset obtained from a haul truck. It is observed that multinomial logistic regression generally performs better than K-Nearest Neighbours in predicting haul truck failure. To improve this work in future, other artificial intelligence algorithms such as Artificial Neural Network can be used for prediction.