Document Type : Original Research Paper

Authors

• Ali Aalianvari ¹
• Shirin Jahanmiri ²

¹ Department of Mining and Environmental Engineering, Faculty of Engineering, Tarbiat Modares University, Tehran, Iran

² Department of Mining Engineering, University of Kashan,Kashan,I.R Iran

Abstract

The increasing environmental risks associated with mining operations demand real-time, accurate risk assessment frameworks to prevent ecological damage and ensure operational safety. This study proposes an integrated real-time environmental risk assessment system utilizing multi-source sensor data and advanced machine learning techniques. Sensor arrays monitoring parameters such as turbidity, pH, conductivity, temperature, and geotechnical vibrations provide continuous high-frequency data streams, which are preprocessed and analyzed using feature engineering methods to handle noise and heterogeneity. A comparative evaluation of several supervised and unsupervised models—namely XGBoost, Random Forest, Long Short-Term Memory (LSTM), Autoencoder, and Isolation Forest—was conducted. The XGBoost model outperformed others with an accuracy of 95%, precision of 94%, recall of 94%, F1-score of 94%, and an AUC-ROC of 0.97, while maintaining a low inference time of 18.7 ms per instance, suitable for real-time deployment. Autoencoder models achieved the highest anomaly detection rate of 92%, indicating their effectiveness in identifying rare environmental hazards. Feature importance analysis highlighted turbidity, pH, and conductivity as the most influential predictors, corroborating environmental science insights. This framework demonstrates a robust, scalable, and interpretable solution for mining environmental risk management, enabling prompt hazard detection and facilitating proactive interventions

Keywords

• real-Time Environmental Risk Assessment
• Mining
• Sensor Data
• Machine Learning
• Anomaly Detection

Main Subjects

• Exploration