In the context of defect detection in high-speed railway train wheels, particularly in ultrasonic-testing B-scan images characterized by their small size and complexity, the need for a robust solution is paramount. The proposed algorithm, UT-YOLO, was meticulously designed to address the specific challenges presented by these images. UT-YOLO enhances its learning capacity, accuracy in detecting small targets, and overall processing speed by adopting optimized convolutional layers, a special layer design, and an attention mechanism. This algorithm exhibits superior performance on high-speed railway wheel UT datasets, indicating its potential. Crucially, UT-YOLO meets real-time processing requirements, positioning it as a practical solution for the dynamic and high-speed environment of railway inspections. In experimental evaluations, UT-YOLO exhibited good performance in best recall, mAP@0.5 and mAP@0.5:0.95 increased by 37%, 36%, and 43%, respectively; and its speed also met the needs of real-time performance. Moreover, an ultrasonic defect detection data set based on real wheels was created, and this research has been applied in actual scenarios and has helped to greatly improve manual detection efficiency.
The detection of weld structures in service is significantly hindered by surface irregularities and blemishes, presenting a challenge to detection technology. To address this issue, the present study proposes a detection platform that integrates magnetic yoke excitation with induction scanning thermography technology for heating large-scale and complex structures. After the collection and preprocessing of thermography sequence data, a multi-feature fusion imaging method is devised based on multiple time-domain thermal features images input, and reevaluate the effectiveness of these individual features in distinguishing defects from backgrounds. Furthermore, an improved Phase Spatial Kurtosis (PSK) feature that utilizes dynamic detection signal mechanisms is proposed. The aforementioned features are combined through a trained artificial neural network (ANN) in order to obtain an estimation. The experimental results demonstrate that proposed method effectively detects and isolates the defect area in complex backgrounds at the pixel level, enhancing imaging accuracy to 99.6% while significantly suppressing interference from complex background information without generating defect false positives.
Ultrasonic testing is a highly efficient NDT technique, which is widely used and rapidly developing among the many NDT methods. With the rapid development of high-speed railways, existing testing speeds are often unable to meet demand. Wheel integrity is essential for safety. Existing ultrasonic testing techniques now rely more on the intuition and experience of skilled workers, which often results in a high rate of missed inspections. In this paper, images of wheels with different defect depths are firstly annotated and calibrated. Secondly, the structure of the YOLO v5 algorithm is improved for the characteristics of ultrasonic defects, making it more efficient for recognizing B-scan defect images of wheels. Finally, the data is validated and tested for each category. The experimental results show that the improved algorithm has significantly enhanced the performance of wheel B-scan defect recognition, reducing the false alarm rate while stabilizing the detection speed of one wheel within 10s.