Leveraging Segment Anything Model (SAM) for Weld Defect Detection in Industrial Ultrasonic B-Scan Images
Использование модели Segment Anything Model (SAM) для обнаружения дефектов сварных соединений на промышленных ультразвуковых B-сканах
2025-01-06
SCID: 54.1/wzsbrx5m
Discuss with AI
LoRA fine-tuningSegment Anything Model (SAM)lack of fusion (LOF)ultrasonic B-scan imagesweld defect detection
Figures from the paper
Abstract (AI)
Automated ultrasonic testing (AUT) is a critical tool for infrastructure evaluation in industries such as oil and gas, and, while skilled operators manually analyze complex AUT data, artificial intelligence (AI)-based methods show promise for automating interpretation. However, improving the reliability and effectiveness of these methods remains a significant challenge. This study employs the Segment Anything Model (SAM), a vision foundation model, to design an AI-assisted tool for weld defect detection in real-world ultrasonic B-scan images. It utilizes a proprietary dataset of B-scan images generated from AUT data collected during automated girth weld inspections of oil and gas pipelines, detecting a specific defect type: lack of fusion (LOF). The implementation includes integrating knowledge from the B-scan image context into the natural image-based SAM 1 and SAM 2 through a fully automated, promptable process. As part of designing a practical AI-assistant tool, the experiments involve applying both vanilla and low-rank adaptation (LoRA) fine-tuning techniques to the image encoder and mask decoder of different variants of both models, while keeping the prompt encoder unchanged. The results demonstrate that the utilized method achieves improved performance compared to a previous study on the same dataset.
Key Findings
1
Experiments evaluated vanilla and LoRA fine-tuning of image encoders and mask decoders across SAM variants while keeping the prompt encoder unchanged.
2
SAM 1 and SAM 2 were adapted for automated lack-of-fusion detection in real-world ultrasonic B-scan images from automated pipeline girth-weld inspections.
3
The method transfers B-scan contextual knowledge into natural-image SAM models through a fully automated, promptable processing pipeline.
4
The resulting AI-assisted detection method outperformed a previous study conducted on the same proprietary ultrasonic B-scan dataset.
5
The study addresses practical AUT interpretation by developing an automated tool for detecting a specific weld-defect type in oil and gas pipelines.
Research Object
weld defects—specifically lack of fusion—in ultrasonic B-scan images from automated girth-weld inspections of oil and gas pipelines
Research Subject
AI-assisted detection and segmentation performance for lack-of-fusion defects, including the transfer and fine-tuning of SAM models for contextualized B-scan interpretation
Publication Details
Publication Date
2025-01-06
Journal
Publisher
ISSN
Cited by
16
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest