Loading...

Data-driven damage assessment of reinforced concrete shear walls using visual features of damage

Mansourdehghan, S ; Sharif University of Technology | 2022

96 Viewed
  1. Type of Document: Article
  2. DOI: 10.1016/j.jobe.2022.104509
  3. Publisher: Elsevier Ltd , 2022
  4. Abstract:
  5. This paper proposes a damage assessment framework based on the visual features of a damaged reinforced concrete shear wall, such as crack pattern distribution, crushing areal density, aspect ratio, and the presence of the boundary condition. The study contains two parts including: identifying the performance level of the damaged walls (i.e., Immediate Occupancy, Life Safety, and Collapse Prevention) and estimating the residual strength and drift ratio of the walls. The research database contains 236 images of 72 reinforced concrete shear walls tested in the laboratory under the quasi-static cyclic loadings at various drift ratios between 0 and 4%. To identify the performance level of a damaged wall, six supervised learning techniques, including Decision Tree, Random Forest, K_Nearest Neighbor, Gradient Boost, AdaBoost, and Naïve Bayes, are used for classification, and the most efficient method is introduced. Afterward, predictive equations are presented to estimate the residual strength and drift ratio of the walls. The proposed regression equations for drift and reserved capacity are finally used for the estimation of the backbone curve of the hysteretic cyclic loops. In other words, simultaneous employment of both predictive equations for drift ratio and reserved capacity allows for reconstructing the backbone of the cyclic curve just by looking at the progressive damage. The results of the performance level classification show that the Random Forest model is the most efficient method in comparison with other methods with 81.4% accuracy for the test dataset. Predictive equations are also capable of estimating the peak drift ratio and residual strength with an R-factor of 0.9 and 0.83, respectively. © 2022 Elsevier Ltd
  6. Keywords:
  7. Crack pattern ; Damage detection ; Health monitoring ; Machine learning ; Reinforced concrete shear wall ; Adaptive boosting ; Aspect ratio ; Classification (of information) ; Crack detection ; Decision trees ; Pattern recognition ; Reinforced concrete ; Statistical tests ; Supervised learning ; Crack patterns ; Damage assessments ; Drift ratio ; Performance:level ; Predictive equations ; Reinforced concrete shear walls ; Residual strength ; Symbolic regression ; Visual feature
  8. Source: Journal of Building Engineering ; Volume 53 , 2022 ; 23527102 (ISSN)
  9. URL: https://www.sciencedirect.com/science/article/abs/pii/S2352710222005228