Research and Development of an Automatic Detection System for Weld Radiographic Film Defects Based on Deep Learning

View at publisher Summarise with VIRAT
Authors: Jiahang Dong, Jian Zhang, Jiawei Jiang, Chang Gao, Zhiguo Fu, Yixuan Du, Yifei Zhang, Xiaoyu Zhu
Cited by: 0
References: 5

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Radiographic testing (RT) is a critical method for weld defect evaluation. However, the industry still primarily relies on manual film interpretation for radiographic film defect assessment, which suffers from low efficiency, high subjectivity, fatigue-induced missed detections, and high professional requirements for inspectors. To address these issues, this paper develops an automatic weld radiographic film defect detection system via deep learning-based defect recognition technology. Firstly, ResNet18 network structure is adopted to construct a deep learning model. Test results indicate that the independently trained model achieves an overall recognition accuracy of 99.67% with no missed detections of non-defective welds, enabling efficient and accurate identification of common defects in weld radiographic films. Secondly, a visual interactive interface is designed and developed using Tkinter, featuring intuitiveness and ease of operation. Finally, PyInstaller is used to encapsulate the system into an executable file, supporting local deployment and lowering the usage threshold for on-site personnel. This system realizes full-process industrial application from image input to result output, contributing to the digital and intelligent upgrading of the weld defect inspection industry and empowering the high-quality development of high-end equipment manufacturing.

Fields of Study & Subjects

Industrial Vision Systems and Defect Detection Non-Destructive Testing Techniques Infrastructure Maintenance and Monitoring

Publication details

Journal / source
International Conference on Electrical Automation and Artificial Intelligence (ICEAAI)
Pages
1065–1070
Publication date
16 January 2026
Language
EN