비파괴검사 품질향상을 위한 CNN 기반 용접결함 분류 및 검증에 관한 연구
A Study on Weld Defect Classification Using CNN to Enhance Non-destructive Testing
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
용접은 제조산업에서 가장 중요한 공정 과정이다. 하지만, 산업현장에서는 용접인력 부족 및 고비용의 문제와 전문가 수준의 기술을 습득까지의 오랜 경험이 요구되기 때문에, 제조공정에서 인공지능(AI)기술의 도입이 시급하다. 본 논문에서는 합성곱 신경망(Convolutional Neural Networks, CNN)기반 VGG 모델을 활용하여 제조공정에서 나타나는 용접결함에 대한 이미지 분류를 진행하고, 그 성능을 평가하고자 한다. 본 연구에서는 2가지 유형의 육안검사(Visual Test, VT)와 방사선 탐상검사(Radioisotope Test, RT)의 용접결함 이미지를 수집하였다. 용접결함은 4개의 용접결함(VT: 기공, 용입부족, 융합불량, 언더컷/RT:기공, 용입부족, 균열, 슬래그 혼입)과 1개의 정상으로 구성되어 있다. 용접결함 이미지는 원 데이터를 직접적으로 활용하였으며, 그 결과, VT, RT 이미지에 대하여 각각 98%, 92%의 성능을 확인하였다. 본 연구는 데이터 증강 등 데이터 정제 없이 진행한 점에 의의가 있으며, 향후 본 연구 결과를 기반으로 더 많은 자료수집을 통한 데이터의 확장과 더불어, 다양한 딥러닝 모델을 활용한 성능향상을 통하여 용접결함 검사 시스템 개발에 활용할 수 있을 것이다.
Abstract
Welding plays a crucial role in the manufacturing industry. Nevertheless, the manufacturing industry urgently requires integrated artificial intelligence (AI) technology to enhance productivity because of the scarcity of human resources, the high costs involved, and the long time needed to attain an expert level. This study aims to classify welding defects using the EGG model based on CNN (Constitutional Neural Network). This study collected two types of welding defect images: a visual test (VT) and a radioisotope test (RT). There are four categories of welding defects: VT defects, which include porosity, incomplete penetration, lack of fusion, and undercut, and RT defects, which consist of cracks, porosity, lack of fusion, and slag inclusion. We also collected normal data. We input the raw image, not data p reprocessing, and achieve classification performances of 98% in VT and 92% in RT, respectively. Our finding is significant because it was conducted without data purification, such as data augmentation. We are expected to expand the data and improve performance by utilizing a variety of deep-learning models. In addition, our study helps to create a welding defect inspection system that can accurately detect surface defects in welds in future works.
참고문헌 (15)
- BJ Jin, "Development of GTAW Non-destructive Big Data Generation System for Welding Defect Determination", Spring Conference of the Korean Welding and Joining Society, 72, pp.202~202 (2021)
- Tripicchio, P., "Welding defect detection: Coping with artifacts in the production line", The International Journal of Advanced Manufacturing Technology, 111, pp.1659~1669 (2020)
- Yun, J. P., "Automated defect inspection system for metal surfaces based on deep learning and data augmentation", Journal of Manufacturing Systems, 55, pp.317~324 (2020)
- Chang, Y., "A deep learning-based weld defect classification method using radiographic images with a cylindrical projection", IEEE Transactions on Instrumentation and Measurement, 70, pp.1~11 (2021)
- Singh, S. A., "Automated surface defect detection framework using machine vision and convolutional neural networks", Journal of Intelligent Manufacturing, 34(4), pp.1995~2011 (2023)
- WH Kang, "AI-Based Welding Defect Inspection for Vehicle Freams", 2023 fall Conference on Korea Institute of Information and Communication Engineering, 27(2), pp.23~27 (2023)
- HB Jo, "Development of Defect Classification Model for Smart Factory Data Using Machine Learning", Journal of The Korean Data Analysis Society(JKDAS), 25(5), pp.1637~1651 (2023)
- Golodov, V. A., "Approach to weld segmentation and defect classification in radiographic images of pipe welds", Ndt & E International, 127, pp.102597~ (2022)
- Harrouche, S., "A Comparative Study of Different CNN Models using Transfer Learning for Weld Defect Classification in Radiographic Testing", 2023 International Conference on Electrical, Communication and Computer Engineering (ICECCE), pp.1~7 (2023), IEEE
- Russakovsky, O., "Imagenet large scale visual recognition challenge", International journal of computer vision, 115, pp.211~252 (2015)
- Godoy, J., "ALAN: adaptive learning for multi-agent navigation", Autonomous Robots, 42, pp.1543~1562 (2018)
- Sokolova, M., "Beyond accuracy, F-score and ROC: a family of discriminant measures for performance evaluation", Australasian joint conference on artificial intelligence, pp.1015~1021 (2006), Springer