Truck haulage data from openpit mine sites are usually massive and multidimensional with multipeak Gaussian distributions. This study used artificial neural networks (ANNs) coupled with Gaussian mixture modeling (GMM) to analyze these data. ANNs are well-known machine learning (ML) methods for handling massive and multidimensional data, while GMM is an efficient clustering technique for processing the data under multipeak Gaussian distributions. For the first time, this study adopted three ANNs — back-propagation neural network (BPNN), extreme learning machine (ELM) and Bayesian regularized neural network (BRNN) — in combination with GMM to process complex truck haulage data and build the weighted ensemble models (WE-BPNN, WE-ELM and WE-BRNN) to forecast mining truck productivity. The results showed that the proposed weighted ensemble models performed better than the benchmark models in predicting truck productivity, and haul distance was the most crucial input variable affecting truck productivity. This study provides engineers with a new means of solving real-site complex data and provides accurate prediction models of mining truck productivity, thus improving mine planning and decision-making.
Full-text paper:
Mining, Metallurgy & Exploration (2023) 40:583–598, https://doi.org/10.1007/s42461-023-00747-9
