Intelligent Fatigue Crack Identification Methods for Orthotropic Steel Bridge Deck Based on Deep Learning
This study presents a method to identify fatigue cracks in the orthotropic steel deck panels.The method is developed on the basis of deep learning and can improve the fatigue crack recognition efficiency.Specimens were prepared for indoor fatigue crack detection and image collection to create data sets,and the data sets were in turn processed by the image pre-processing technique.Combined with YOLOv5s(an image detection algorithm)and CBAM(Convolutional Block Attention Module)attention mechanism,a method to intelligently recognize ultrasonic phased array detection images is developed to complete the interpretation of the test results.It is shown that the image pre-processing method combining the CLAHE(Contrast Limited Adaptive Histogram Equalization)and adaptive median filtering suits well the pre-processing of ultrasonic phased arrays.Through introducing the CBAM attention mechanism,the image recognition accuracy is improved by 1.28％,reaching 98.74％.The presented YOLOv5s-CBAM image recognition algorithm can accurately frame and select the fatigue cracking zones,and the recognition confidence of the four types of common fatigue cracks reaches up to 0.91,and 0.85 for other types of fatigue cracks,proving that it can meet the intelligent detection requirements of fatigue cracks in indoor environments.
万方数据
维普
