Estimating Ore Production in Open-pit Mines Using Various Machine Learning Algorithms Based on a Truck-Haulage System and Support of Internet of Things

This study aimed to develop and assess the feasibility of different machine learning algorithms for predicting ore production in open-pit mines based on a truck-haulage system with the support of the Internet of Things (IoT). Six machine learning algorithms, namely the random forest (RF), support vector machine (SVM), multi-layer perceptron neural networks (MLP neural nets), classification and regression tree, k-nearest neighbors, and M5Tree model, were developed and investigated to estimate ore production at a limestone open-pit mine in South Korea. Systems of IoT were used to collect a massive database of 16,217 observations (big data). To process this big data, a downscaling method was applied to reduce the size of the original observations to improve the computational cost of the machine learning models. Subsequently, three validation datasets were selected from the original observations and used to evaluate (after downscaling the observations) the performance and accuracy of the machine learning models in practical engineering through various performance metrics. The results revealed that the models used can be potentially used for predicting ore production in open-pit mines. The SVM, MLP neural nets, and RF models demonstrated high accuracy, with the SVM model exhibiting the most superior performance and the highest accuracy. An assessment of the validation datasets also confirmed the feasibility and stability of the models for predicting ore production. Furthermore, a sensitivity analysis indicated that the relative operation start time, relative operation end time, and interval between the operation times were the most important input variables for improving ore production in practical engineering.