Image Recognition of Weld Defect X-ray Detection Based on Deep Learning

With the development of modern manufacturing, welding technology plays a crucial role in joining materials, and its quality directly affects the performance and safety of products. Traditional manual inspection methods are limited due to their subjectivity and low efficiency. This study utilizes deep learning technology, especially Convolutional Neural Networks (CNN), to develop an automated welding X-ray image defect recognition system to improve the accuracy and automation level of detection. We employed the YOLOv5 model, a real-time object detector known for its high precision and efficiency. Through 150 rounds of training on two hundred welding X-ray images, we increased the mean Average Precision (MAP) from 0.50 to 0.97 and fine-tuned the model parameters. The experimental results show that the system can effectively identify various types of welding defects, providing an efficient solution for welding quality control. This study not only demonstrates the application potential of deep learning in the field of industrial inspection but also provides a new approach for the automated assessment of welding quality, which is of great significance for the intelligent transformation of future manufacturing industries.