Accurate identification of crack parameters in cantilever beams is of great significance for structural health monitoring. Aiming at the problems of disjointed modeling and detection, difficult micro-crack extraction, and large inversion search space, this paper proposes a crack parameter identification method by deeply integrating the energy method and lightweight YOLO visual detection. The method establishes the relationship between crack parameters and structural responses through mechanical modeling, uses a lightweight YOLO network to achieve high-precision multi-scale crack detection, and constructs a closed-loop fusion mechanism with visual prior constraint, energy method inversion, and bidirectional verification. Experimental results show that the YOLO detection module constructed in this paper achieves an mAP of 0.931 with a micro-crack miss rate below 7.9%. The fusion method obtains a position error within ±3 pixels ( ≈ ±2.34 mm) and a relative depth error within 5.2% with respect to the ultrasonic-calibrated crack-depth reference. By complementing vision and mechanics, the method overcomes the bottlenecks of single approaches, effectively improves identification accuracy, robustness and efficiency, and provides reliable technical support for health monitoring and damage assessment of cantilever beam structures.

cantilever beam, crack parameter identification, energy method, YOLOv8n, structural health monitoring