Abstract: To address issues such as slow detection speed and low automation levels in existing monitoring technologies, this paper proposes a mine air door monitoring system based on PLC and deep learning. The proposed EAW-YOLO model, based on YOLOv11n, introduces several key improvements: (1) the integration of the EMA attention mechanism into the C3k2 module to form a C3k2-EMA module, enhancing feature extraction capability; (2) the incorporation of ADown convolution to reduce channel dimensions while preserving critical information; and (3) the adoption of the WIoU loss function, which dynamically adjusts the weight of anchor boxes to accelerate regression convergence. Experimental results demonstrate that the system can accurately identify various targets and precisely control mine air doors. The EAW-YOLO model achieves a Precision (P) of 91.8%, mAP@0.5 of 98.1%, and an FPS of 86.7, with only 2.1M parameters, meeting the requirements for real-time performance, high accuracy, and ease of deployment in underground mining environments.