Fuel Consumption Analysis in Diesel Haul Trucks Using Machine Learning

Abstract

Fuel consumption in high-tonnage mining trucks accounts for approximately 32% of the total energy consumed in open-pit mine, and its contribution to operating costs can reach up to 22%. This is influenced by multiple factors, including the characteristics and mechanical condition of the equipment, haul road design and conditions, weather conditions, mine planning, driver behavior, among others. Studying this phenomenon and understanding its key variables enables decision-making aimed at minimizing fuel consumption while maintaining the desired productivity of the equipment. Based on this premise, this thesis analyzes the most influential factors in fuel consumption for open-pit mining trucks using data collected from sensors installed in currently operating equipment. In this research, the analysis is conducted through the use of both supervised and unsupervised machine learning models applied to real-world data. These models provide insights into attributes such as travel speed, road inclination, payload, accelerator pedal usage, and weather conditions, as well as truck operation parameters like temperatures and pressures in different engine components. This thesis first explores the key factors influencing fuel consumption, analyzing the impact of design and operational decisions. Secondly, a methodology is presented to assess the health status of mining trucks and quantify its effect on fuel consumption. The proposed methodologies support decision-making in various areas related to mine operation, maintenance, and design, contributing to the reduction of fuel consumption in mining trucks. The results obtained indicate that, per haul cycle, a 1% reduction in ramp grade can lead to fuel savings of up to 1.6%. Reductions in the maximum ramp speed may result in increases in fuel consumption of up to 7.4%, and trucks overloaded by 20 tons consume approximately 2.7% more fuel. Additionally, deterioration in a truck's mechanical health can lead to an additional 3.4% increase in fuel consumption. Although the developed Machine Learning models suggest specific improvements based on the datasets, the proposed methodologies can be applied to any diesel haul truck fleet for mine-specific recommendations.