Artificial Intelligence-Driven Mine Dewatering to Improve Safety and Efficiency

Water accumulation in mines poses risks like flooding and hazardous in-rushes, compromising infrastructure and safety. Effective mine dewatering is essential for operational safety and efficiency, requiring continuous pumping to manage groundwater inflow. Traditional prediction methods, such as empirical equations and finite-element models, face limitations due to geological and hydrological complexities.

This paper describes an Artificial Intelligence (AI) based monitoring approach as a soft sensor for mine dewatering systems. Utilizing sensors to track water levels, flow rates, and pump power consumption, real-time data is fed into an AI/Machine Learning (ML) model to detect anomalies and predict potential issues. The system employs Principal Component Analysis (PCA) for anomaly detection, focusing on static window PCA due to its simplicity and effectiveness in stationary environments.

Alerts for abnormal behaviors, such as pump failures or excessive groundwater inflow, provide early warnings to maintenance personnel. An additional monitoring system enhances these alerts with specific parameter checks, including water flow, pump current anomalies, and groundwater inflow. These empirical rules improve the interpretability and actionability of alerts.

This AI-based soft alerting system has been implemented in a large mining operation. It has demonstrated its potential to enhance the efficiency, safety, and cost-effectiveness of mine dewatering processes. Continuous monitoring and feedback loops ensure the system adapts to operational needs, offering a robust solution for mine water management.