A comparative study of truck cycle time prediction methods in open‐pit mining

Abstract

PurposeThe purpose of this paper is to compare the predictive capability of three methods of truck cycle time estimation in open‐pit mining: computer simulation, artificial neural networks (NNs), and multiple regressions (MRs). The aim is to determine the best method. The most common method currently used is computer simulation.Design/methodology/approachTruck cycle times at a large open pit mine are estimated using computer simulation, artificial NNs, and MRs. The estimated cycle times by each method are in turn compared to the actual cycle times recorded by a computerized mine monitoring system at the same mine. The errors associated with each method relative to the actual cycle times are documented and form the basis for comparing the three methods.FindingsThe paper clearly indicates that computer simulation methods used in predicting truck cycle times in open‐pit mining underestimate and overestimate the results for short and long hauls, respectively. It appears that both NN and regression models are superior in their predictive abilities compared to computer simulations.Research limitations/implicationsThe cycle time prediction models developed apply to a specific mine site and one has to be careful not to directly apply these models to other operations.Practical implicationsThe paper describes the implementation of regression and NN modelling. An opportunity exists for mines to utilise the large volumes of data generated to predict truck haulage cycle times more accurately and hence, improve the quality of mine planning.Originality/valueThe paper addresses an area of need in the mining industry. Accurate prediction of cycle times is critical to mine planners as it impacts on production targets and hence, the budgets.