Research on multi-information sensing and graded warning technology for fires in electromechanical chambers

Abstract This study presents a comprehensive framework for intelligent fire prevention and control in underground coal mine electromechanical chambers. Addressing the limitations of traditional fire monitoring systems-such as narrow coverage, information silos, and delayed response-this research proposes a full-dimensional information perception and linked control system. The system integrates multi-parameter sensors, deep learning-based flame detection, and optimized sensor layout strategies to achieve real-time monitoring, early warning, and collaborative control. A multi-parameter sensor was developed for synchronous detection of temperature, smoke, and hazardous gases (CO, SO 2 , HCl), with enhanced portability and battery life. Flame detection was improved using an optimized YOLOv5 algorithm incorporating attention mechanisms and multi-scale detection layers, enabling accurate recognition of small-scale flames in complex underground environments. Sensor layout optimization was achieved through 3D spatial modeling and multi-objective intelligent algorithms, significantly improving monitoring coverage and reducing redundancy. Furthermore, a hybrid fire risk assessment model combining Fuzzy Neural Networks (FNN) and Grey Relational Analysis (GRA) was established, and a graded early warning mechanism based on the EPI model was developed. The results demonstrate that the proposed system enhances fire perception accuracy, response speed, and decision-making reliability, providing a robust technical foundation for intelligent fire safety management in underground mines.