Intelligent risk assessment and early warning for human–machine–environment coupling in coal preparation plants

Coal washing and preparation is a critical process for cleaner and more efficient coal utilization, while safety performance in coal preparation plants is often challenged by harsh operating conditions, tightly coupled equipment, and limited effectiveness of manual supervision. In this study, a coal preparation plant was used as an engineering scenario to investigate equipment safety risk grading and a deployable workflow for risk-based monitoring and on-site warning. Hazardous factors in key processes were identified through accident analysis and field investigation, and a multi-dimensional risk assessment model was established by integrating Fault Tree Analysis (FTA) and the Analytic Hierarchy Process (AHP). To enhance real-time supervision in high-risk areas, an optimized YOLOv10-based object detection model was developed, improving robustness under heavy dust, low illumination, and complex backgrounds. On this basis, a personnel safety monitoring system integrating monocular vision-based ranging and audible–visual alarms was implemented to enable real-time warning and traceable logging. Experimental and field results show that, under the reported deployment conditions, the proposed system achieved. These results suggest that risk grading can be combined with vision-based monitoring to support risk-informed warning and traceable logging in coal preparation plants.