<p>With the increasing service life of steel bridges, concerns regarding their structural safety have grown significantly. Early detection of fatigue cracks is crucial for ensuring the safe operation of such infrastructure. However, existing non-destructive testing methods often suffer from limitations, including low detection efficiency and insufficient precision. Electromagnetic ultrasonic-based defect detection has been proven to be an effective and reliable approach. In this study, a novel electromagnetic ultrasonic transducer (EMAT) is proposed for detecting fatigue cracks in steel bridges. A key innovation lies in the optimized coil design, specifically tailored for the detection of small cracks. Theoretical analysis and numerical simulations were conducted to systematically investigate the influence of coil geometry on wave propagation characteristics, focusing performance, and defect detection capability. Based on these insights, a prototype transducer was fabricated and validated through transmit-receive experiments. Results demonstrate that the proposed transducer efficiently excites focused surface waves, significantly enhancing the defect signal strength. Subsequently, artificial crack detection experiments were performed under varying crack lengths and depths to evaluate the detection performance of the transducer. The results indicate that the new design exhibits clear advantages in crack recognition accuracy, signal-to-noise ratio, and sensitivity to small defects. Finally, the transducer was applied to a steel bridge model and benchmarked against traditional transducers. The proposed EMAT not only demonstrated accurate and feasible crack detection but also showed superior performance in terms of signal strength and stability. This study provides an effective and reliable method for fatigue crack detection in steel bridges, offering both theoretical insight and practical value for improving bridge structural health monitoring.</p>