Loading...

Backbone Curve Prediction for Reinforced Concrete Shear Walls Using Data-Driven Methods

Ebrahimi, Pouya | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57508 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Mohtasham Dolatshahi, Kiarash
  7. Abstract:
  8. The aim of this research is to predict the backbone curve of reinforced concrete shear walls by identifying different failure modes. Failure mode identification is performed by measuring the contribution percentage of flexure, shear, and combined flexure-shear failure modes in the force-displacement response of the wall under cyclic loading. The approach used in this research is a probabilistic and data-driven method, which leads to the calculation of model uncertainty in detecting the wall's failure mode and, consequently, the prediction uncertainty of the backbone curve. The database used in this research consists of 253 reinforced concrete shear walls, obtained from a review of existing experimental studies, including wall damage images at the end of the tests, cyclic and backbone curves, geometric properties, and mechanical characteristics. Initially, all the walls in the database are classified into three groups—shear, flexure, and combined flexure-shear—based on the reported failure modes, damage images, and cyclic curves, following existing guidelines. Then, supervised and unsupervised learning algorithms are used to measure the contribution of flexure and shear failure modes in the overall seismic behavior of the wall and predict them. The unsupervised model employs the k-means clustering algorithm, which achieves over 90% accuracy in matching the wall failure labels. Based on the results of this model, a hybridity index is introduced to indicate the contribution of each failure mode in the final failure of the reinforced concrete shear wall during seismic testing. The supervised model, on the other hand, uses regression algorithms to predict the hybridity indices based on the geometric and mechanical properties of the walls. Among various regression models, the Extra Trees model shows the best performance, with determination coefficients of 0.94 and 0.86 for the training and test datasets, respectively. In the next step, using the hybridity indices, the backbone curve of reinforced concrete shear walls is predicted by estimating 37 points of the curve. Finally, the research investigates the impact of uncertainty in the hybridity indices (uncertainty in the wall's failure mode) and material uncertainty on the backbone curve prediction, and the results are presented. In the conclusion, two case studies are provided to evaluate the model's performance in predicting failure mode contributions and the backbone curve
  9. Keywords:
  10. Reinforced Concrete Shear Wall ; Backbone Curve ; Failure Modes ; Composite Index ; Data Driven Method

 Digital Object List

 Bookmark

No TOC