Abstract

Mining companies continue to set aggressive targets for substantial decarbonization by the 2030s and complete decarbonization by mid-century.  Achieving these targets, especially those in the coming decade will be complicated by a lack of commercially viable zero emission load-haul mining equipment.  Therefore, any successful path to achieving these interim goals must reflect the continued operation of fleets partially comprised of diesel-powered trucks. 

Fortunately, much can be done to improve the efficiency of load-haul mining activity with present day equipment; and while capturing gains in diesel efficiency will not be sufficient to achieve even interim decarbonization goals, they provide several crucial benefits.  The first is an immediate incremental reduction in carbon emissions and energy costs.  The second is the accelerated viability of next-generation zero emissions equipment, including battery electric and hydrogen powered mining assets. 

The author will present a framework for applying machine learning techniques to identify, capture and audit efficiency and decarbonization gains presently in use in major mining markets.  This will be followed by case studies examining the most successful applications of this framework.   

At its core, this approach exploits the predictive and explanatory power of machine learning models trained on extensively characterized haul cycles.  A haul cycle, in this context, includes all activity performed by a mining truck between 2 consecutive dump events.  Haul cycle characterization includes every measurable element known to influence energy expenditure. This goes far beyond time, tonnage, and distance to include elements such as vertical travel, idle time, operator id, equipment id, mining area etc.  With a robust collection of such haul cycles, machine learning models are constructed and interrogated to expose the relationships between haul cycle properties and energy efficiency outcomes.  This is the starting point for surgically directed mining interventions designed to reduce energy use and save cost across the haulage operation. 

The majority of interventions fall neatly within four categories; equipment interventions occur when machine learning models point to specific pieces of equipment that consume more energy than expected for the work performed; operator interventions occur in response to a model’s identification of operators who are similarly identified; road interventions are triggered by findings attributing energy expense to road design or construction; and finally technological advancements are evaluated for their effectiveness without need to interfere with equipment allocation or dispatch.  

In this presentation the author will provide real world example of mines that have put the framework to use to implement efficiency triggered maintenance programs, operator performance management strategies, road correction programs and various technology evaluation studies.  The combined results of these concerted efforts have resulted in significant energy and cost savings that in many cases correspond directly to benefits in zero emission vehicles including range extension, reduced battery size, reduced charging infrastructure, that are likely more significant that those captured today in a diesel-powered mining environment.