To address the challenges of diverse defect types, difficulty in detecting tiny defects, and the trade-off between detection speed and accuracy in industrial PCB defect inspection, an improved PED-YOLOv11n algorithm is proposed. A lightweight backbone network based on Partial Convolution (PConv) is constructed to reduce computational complexity. An Efficient Multi-scale Attention (EMA) mechanism is introduced to enhance key feature extraction, while a Dynamic Adaptive Detection Head is designed to improve adaptability to defects with varying shapes. In addition, the Shape-NWD loss function is employed to optimize bounding box regression for small defects. Experimental results on the Peking University PCB defect dataset show that the proposed algorithm achieves a mAP50 of 92.2%, representing a 1.1% improvement over the baseline model, while reducing the number of parameters and computational cost by 18.6% and 25.4%, respectively, and improving inference speed by 15%. The proposed method achieves a favorable balance between high accuracy and efficiency in PCB defect detection, providing a feasible technical solution for intelligent quality inspection in industrial automation.