Time-series models for predicting dump truck travel time in an underground mine: A case study in Peru

Predicting dump truck travel time in underground mining is critical because of its direct influence on productivity, operational scheduling, and haulage costs, particularly in environments with high variability and geometric constraints. This study aimed to develop and compare time-series models for predicting dump truck travel time in an underground mine located in Pataz, Peru, using consolidated daily records from a fleet of 10 dump trucks. The methodology included data preprocessing, exploratory and temporal diagnostic analysis, and the development and validation of ETS, SARIMA, ARIMAX, SARIMAX, and TBATS models. A chronological split of 70% for training (n=255) and 30% for testing (n=110) was applied, together with time-series cross-validation and the metrics MAE, RMSE, MAPE, sMAPE, MASE, and Theil's U. The results showed that the series exhibited trend, serial dependence, and low-to-moderate seasonality. Among the evaluated models, ARIMAX achieved the best predictive performance in the test set, with MAE=1.620, RMSE=2.176, MAPE=1.557%, and sMAPE=1.554%. The 30-day forecast projected a stable average travel time of 121.91 min/trip. These findings indicate that incorporating exogenous operational variables substantially improves forecasting performance in underground haulage systems.