Journal article
Vision-Based Quantification of Stiffness Degradation in Reinforced Concrete Shear Walls Using Graph-Based Surface Crack Features and Machine Learning
Journal of structural engineering (New York, N.Y.), v 152(5), 04026047
01 May 2026
Featured in Collection : Drexel's Newest Publications
Abstract
AbstractThe rapid and accurate assessment of structural damage in reinforced concrete shear walls (RCSWs) following seismic events remains a critical yet unresolved challenge in structural engineering. Current methods rely heavily on subjective visual interpretations, limiting their generalizability and practical deployment. This work streamlines and automates the approach of FEMA 306 and eliminates its subjectivity. In this regard, this study proposes a novel framework that quantitatively links observed surface crack patterns to structural stiffness degradation using a combination of mathematical graph representations and the walls’ design parameters. Crack images are converted into mathematical graphs, where nodes and edges represent the spatial distribution and connectivity of cracks, respectively. Afterward, graph-based features are extracted and combined with design properties such as horizontal and vertical reinforcement ratios to train supervised machine learning (ML) models. The primary innovation of this approach lies in its ability to predict the stiffness reduction factor (λk), a code-compatible damage index, without requiring displacement, drift, or force data. The framework is validated using a data set of 19 large-scale RCSWs tested under cyclic loading. The results demonstrate that the proposed method achieves reliable predictions, with the multilayer perceptron (MLP) model yielding the R2 score of 0.57 and a root mean squared error (RMSE) of 0.13 on the test set. This work offers a scalable and interpretable approach for postdamage evaluation of RCSWs using visually observable data.
Metrics
1 Record Views
Details
- Title
- Vision-Based Quantification of Stiffness Degradation in Reinforced Concrete Shear Walls Using Graph-Based Surface Crack Features and Machine Learning
- Creators
- Pedram Bazrafshan - Drexel UniversityRhythm Osan - Drexel UniversityArvin Ebrahimkhanlou - Drexel University
- Publication Details
- Journal of structural engineering (New York, N.Y.), v 152(5), 04026047
- Publisher
- American Society of Civil Engineers; RESTON
- Number of pages
- 14
- Grant note
- American Society of Civil Engineers (ASCE)Structural Engineering Institute (SEI) of ASCE for the O. H. Ammann Research Fellowship in Structural EngineeringNSF MRI: 2320600
The authors acknowledge the American Society of Civil Engineers (ASCE) and the Structural Engineering Institute (SEI) of ASCE for the O. H. Ammann Research Fellowship in Structural Engineering. Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the American Society of Civil Engineers; ASCE has not approved or endorsed its content. The authors also acknowledge the University Research Computing Facility (URCF) at Drexel University for providing HPC resources that have contributed to the research results reported within this paper. The authors are thankful for the access to the computational resources provided through the NSF MRI Award No. 2320600. The authors thank Dr. Hamed Momeni and Dr. Sina Basereh for data sharing.
- Resource Type
- Journal article
- Language
- English
- Academic Unit
- Civil, Architectural, and Environmental Engineering; Mechanical Engineering and Mechanics
- Web of Science ID
- WOS:001715496900001
- Scopus ID
- 2-s2.0-105031882917
- Other Identifier
- 991022164537504721