The detection of steel surface defects is crucial in industrial manufacturing to ensure product quality and operational efficiency. However, the diversity of defect types and the complexity of real-world environments pose significant challenges to the generalization capability of existing models, necessitating the development of improved detection algorithms. This study proposes an enhanced steel surface defect detection method based on an improved YOLOv11 model. Wavelet convolution is integrated into the backbone network to expand the receptive field and reduce computational complexity. Additionally, the original upsample structure in YOLOv11 is replaced with the EUCB upsample structure to further improve efficiency and accuracy. Experimental results on the NEU-DET and GC10-DET datasets show that the improved model achieves mAP50 scores of 74.5% and 61.1%, respectively, representing improvements of 1.7% and 2.6% over the baseline YOLOv11s model. These enhancements significantly improve the performance of steel surface defect detection, providing practical support for industrial applications.