DEVELOPMENT OF JUDGEMENT SUPPORT SYSTEM OF SURFACE PREPARATION FOR STEEL STRUCTURES USING DEEP LEARNING
-
- OHYA Makoto
- 松江工業高等専門学校 環境・建設工学科
-
- SUWA Taiki
- 金沢大学 理工学域地球社会基盤学類
-
- GOBARA Tatsuya
- 松江工業高等専門学校 生産・建設システム工学専攻
-
- TAKEBE Masamichi
- 松江工業高等専門学校 環境・建設工学科
-
- HIROSE Nozomu
- 松江工業高等専門学校 環境・建設工学科
Bibliographic Information
- Other Title
-
- 深層学習を用いた鋼構造物の素地調整時の除錆度判定システム
Abstract
<p>The quality and durability of repair coatings for steel bridges are significantly affected by the surface preparation prior to repair coating. Surface preparation by blasting stipulated by ISO Sa2 1/2 or higher is required as surface treatment before repair coating. The judgement of the preparation grade is enforced by visual inspection. However, it is difficult for inexperienced field supervisors to not only accurately judge of the preparation grade, but also make quantitative decisions. In this paper, we try to develop a system that supports the judgment of the preparation grade on steel surface before repair coating using deep learning. This study targets rust grade D of weathering steels. As training data for deep learning, we use blasted images pre-graded by experts. An unknown image is applied to the established system, and it is evaluated whether it can be appropriately classified into four classes from ISO Sa1 to Sa3. However, this study did not provide enough teacher images to ensure accuracy. Therefore, two approaches were tried to solve this problem. The system was applied, firstly, an extension technique of number images and, secondly, the use of a trained model. It was confirmed that the developed system using deep learning enables highly accurate classification of the preparation grade of rust removal after blasting. </p>
Journal
-
- Intelligence, Informatics and Infrastructure
-
Intelligence, Informatics and Infrastructure 2 (J2), 771-776, 2021
Japan Society of Civil Engineers
- Tweet
Keywords
Details 詳細情報について
-
- CRID
- 1390853038534382336
-
- NII Article ID
- 130008118381
-
- ISSN
- 24359262
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- CiNii Articles
-
- Abstract License Flag
- Disallowed