Optimisation of hydraulic fracturing in mining through field data analysis and machine learning

Abstract This study presents a comprehensive analysis of field data obtained from a substantial hydraulic fracturing campaign in a hard rock mining operation, aiming to enhance prediction, understanding and optimise operational practices through parametric analysis and Machine Learning. Findings reveal that structural control predominates during the initiation stage, limiting the influence of operational parameters, while strong positive correlations emerge between operational variables and achieved pressures as fracturing progresses. Post-fracture shut-in pressure exhibits a moderate positive correlation across all categories, emphasising the importance of considering rock competency in project scheduling and resource allocation. However, seismic event prediction remains challenging, with operational influences showing slight correlations, suggesting the need for further research into understanding and predicting seismic risks. Recommendations for the industry include implementing real-time monitoring systems, incorporating geomechanical considerations into project planning, and investing in research initiatives focused on mining seismicity. Future research should focus on investigating the long-term effects of hydraulic fracturing, exploring advanced modelling techniques, and conducting interdisciplinary studies to optimise practices. This study provides valuable insights for enhancing operational efficiency and risk management in hydraulic fracturing operations.