Digital twin and deep learning-based metal weld defect detection system

The quality of metal welding is critical to product performance and safety in the industrial manufacturing process, and traditional welding defect detection methods often rely on manual inspection or traditional machine vision methods, which have the problems of low efficiency and insufficient accuracy. In order to improve the accuracy and efficiency of welding defect detection, this paper proposes a welding defect detection system based on digital twin technology and improved Vision Transformer model. Firstly, the virtual and physical two-way mapping model of the welding process is constructed through digital twin technology, and the welding data is obtained and fed back in real time to provide high-quality defect sample datasets. Secondly, combined with deep learning technology, an improved Vision Transformer model was designed to extract the high-dimensional features of welding defects, and the detection accuracy and robustness of the model to welding defects were improved by optimizing the attention mechanism and introducing a multi-scale feature fusion module. Experimental results show that the proposed method achieves excellent performance on multiple metal welding defect datasets, and significantly improves the accuracy and recall of defect detection compared with the traditional Convolutional Neural Network (CNN) method, which provides an efficient and reliable solution for automated quality control in industrial welding process.