Insulator fault detection using UAV inspection is essential for the safe and stable operation of smart grids. However, complex backgrounds and significant scale variations among fault targets pose considerable challenges to existing detection algorithms, often resulting in reduced accuracy. To address these specific issues, this paper proposes an enhanced fault detection algorithm based on YOLOv11n. A two-layer routing attention mechanism is introduced to precisely capture key regional features by integrating coarse-grained region screening with refined local modeling, which significantly improves the model’s ability to detect multi-scale targets in cluttered environments.