Expert-aware multi-stage AI-assisted image processing pipeline for accurate weld defect detection from radiographs: a use case for metal fabrication industries

Экспертно-ориентированный многоэтапный конвейер обработки изображений с поддержкой искусственного интеллекта для точного обнаружения дефектов сварных швов на радиографических снимках: пример применения в металлообрабатывающей промышленности
Avishek Mukherjee, Surjya Kanta Pal, Umesh Paswan, Nisith Ranjan Mandal, Tapas Roy
2025-10-26

YOLO v11multi-label classificationnon-destructive testingradiographic imagesweld defect detection
Radiography-based non-destructive testing is vital in shipbuilding, automotive, and aerospace industries, but manual inspection is labour-intensive and error-prone. While deep learning offers automation, it often lacks explain ability and expert-level insight. This paper presents a multi-stage, expert-aware AI pipeline for automated weld defect detection in radiographic images. The system targets five common defects: blowhole, porosity, wormhole, lack of fusion, and inclusion. Radiographic 16-bit DICOM images are enhanced and converted to 8-bit PNGs using histogram-based contrast adjustment guided by domain thresholds. Weld zones are detected using a YOLO v11 model (97.2% accuracy) and divided into square clips. Each clip undergoes binary classification via transfer learning (F1 score: 97%) to distinguish defective from non-defective regions. Defective clips are further analysed using a multi-label classifier to identify specific defect types. A secondary YOLO v11 model (93.1% accuracy) pinpoints defect locations, which are clustered and evaluated using ISO/NES standards. The pipeline integrates rules-based logic, simulated visual attention, and supervised learning to mirror expert inspection practices. It achieves 96% accuracy compared to human weld supervisors, offering a reliable, interpretable, and domain-informed solution for automated weld quality assessment.
1
An expert-aware, multi-stage AI pipeline automates weld defect detection from radiographs while incorporating rules-based logic and simulated visual attention.
2
Histogram-based enhancement converts 16-bit DICOM radiographs into 8-bit PNGs, followed by YOLO v11 weld-zone detection achieving 97.2% accuracy.
3
The complete pipeline achieves 96% accuracy compared with human weld supervisors and evaluates clustered defects using ISO/NES standards.
4
The system targets five defect types: blowhole, porosity, wormhole, lack of fusion, and inclusion.
5
Transfer-learning binary classification identifies defective clips with a 97% F1 score, while a second YOLO v11 model localizes defects with 93.1% accuracy.

weld defects in radiographic images of fabricated metal welds

automated detection, classification, localization, and standards-based evaluation of weld defects

Publication Details
Publication Date
2025-10-26
Journal
Publisher
ISSN
Access Type
Author Information
Authors
Avishek Mukherjee
Surjya Kanta Pal
Umesh Paswan
Nisith Ranjan Mandal
Tapas Roy
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%