Intelligent identification of welding defects in nuclear power pipelines based on deep learning

A welding defect intelligent recognition method based on the improved YOLOv5 algorithm is proposed for the needs of mobile welding inspection robots. The model adopts depth separable convolution for lightweight design to reduce the number of parameters while maintaining the feature extraction accuracy, and introduces the SE attention mechanism to enhance the feature expression ability. In order to further improve the efficiency and stability, the Hard Swish activation function is used instead to optimise the computational performance, and the C3 module in the Neck part is replaced by the DP-C3 module. The improved algorithm performs well in both efficiency and accuracy of weld defect detection, and provides an effective solution for mobile inspection tasks.