Please use this identifier to cite or link to this item:
https://hdl.handle.net/20.500.13091/470
Title: | Determination of damage levels of RC columns with a smart system oriented method | Authors: | Doğan, Gamze Arslan, Musa Hakan Baykan, Ömer Kaan |
Keywords: | Reinforced Concrete Cantilever Column Earthquake Damage Image Processing Machine Learning Crack Identification Classification Buildings Strength Images |
Publisher: | SPRINGER | Abstract: | In this study, a method that is fast, economical and satisfying in terms of accuracy rate has been developed in order to determine the post-earthquake damage level of reinforced concrete column elements dependent on the damage image on the column surface. In order to represent the Turkish building stock, reinforced concrete columns were produced complying with the 2007 and 2018 Turkish Earthquake Code (TEC-2007 and TBEC-2018) and, in order to represent the existing building stock made before 2000, reinforced concrete columns which are non-complying with the code have been produced. A total of 12 reinforced concrete columns produced in 1/1 scale with square cross sections were tested under earthquake resembling reversible cycling lateral load and axial force. For each cycle, a data set was created by matching the surface images taken from the determined regions of the columns with the damage levels specified in TEC-2007 and TBEC-2018 depending on the load-displacement values measured on the column during the experiment. As a result of the experimental study, a total of 390 damage images were obtained for each load and displacement level. Image processing application was performed by using MATLAB on the damage images and the cracks on the column surface were separated. Parameters such as total cracks area, total cracks length, maximum crack length and maximum crack width have been obtained to represent the amount of damage on the column through the feature extraction process of the cracks in the images. The characteristics of the cracks were classified by support vector machines, decision trees, K-nearest neighborhood, Discriminant Analysis, Ensemble algorithms, which are machine learning classifiers, and the damage states for the columns were estimated. The estimation success from the classifiers ranges from 64 to 80%. In this study, it has been seen that the proposed and developed intelligent system will be open to development and will be a good alternative to existing conventional systems for the determination of column damage. | URI: | https://doi.org/10.1007/s10518-020-00826-y https://hdl.handle.net/20.500.13091/470 |
ISSN: | 1570-761X 1573-1456 |
Appears in Collections: | Mühendislik ve Doğa Bilimleri Fakültesi Koleksiyonu Scopus İndeksli Yayınlar Koleksiyonu / Scopus Indexed Publications Collections WoS İndeksli Yayınlar Koleksiyonu / WoS Indexed Publications Collections |
Files in This Item:
File | Size | Format | |
---|---|---|---|
s10518-020-00826-y.pdf Until 2030-01-01 | 3.57 MB | Adobe PDF | View/Open Request a copy |
CORE Recommender
SCOPUSTM
Citations
3
checked on Aug 3, 2024
WEB OF SCIENCETM
Citations
9
checked on Aug 3, 2024
Page view(s)
124
checked on Aug 5, 2024
Google ScholarTM
Check
Altmetric
Items in GCRIS Repository are protected by copyright, with all rights reserved, unless otherwise indicated.