AIOT-Based Predictive Safety Framework for Underground Coal Mining: Integrating PINNs, Wearable Sensors, and Digital Twins in Indian Contexts
Abstract
Underground coal mining remains one of the most hazardous industrial activities worldwide, particularly in emerging economies where complex geological conditions, methane emissions, roof instability, dust exposure, and equipment-related accidents continue to threaten worker safety. India, the world's second-largest coal producer, operates numerous underground mines under challenging geotechnical and environmental conditions. Despite significant advancements in mechanization and monitoring technologies, accident investigations indicate that a substantial proportion of mining incidents remain attributable to delayed hazard detection, fragmented monitoring systems, and limited predictive capabilities. Conventional safety management approaches are largely reactive, relying on threshold-based alarms and post-event analysis rather than proactive risk prediction. This study proposes a Mining 5.0-oriented intelligent safety framework that integrates Artificial Intelligence of Things (AIoT), Physics-Informed Neural Networks (PINNs), wearable sensing technologies, and Digital Twin models for real-time hazard prediction and decision support in underground coal mines. The proposed framework combines data from methane sensors, temperature sensors, air velocity monitors, geotechnical instruments, equipment health monitoring systems, and wearable devices measuring worker location, physiological status, and environmental exposure. These heterogeneous data streams are fused within a Digital Twin environment that continuously replicates underground mine conditions. PINNs are employed to incorporate ventilation physics, methane transport dynamics, and geomechanical principles into machine-learning models, thereby improving prediction accuracy and interpretability under sparse or uncertain data conditions. The study identifies critical research gaps in existing mine safety systems, including inadequate integration of physical laws with AI models, limited utilization of worker-centric sensing technologies, and the absence of comprehensive Digital Twin platforms for proactive safety management. To address these gaps, a socio-technical framework is developed that enables continuous risk assessment, predictive analytics, and human-AI collaborative decision-making. The proposed approach is expected to enhance situational awareness, reduce accident probability, improve emergency preparedness, and support sustainable Mining 5.0 transformation in India.