Title:
Data-Driven Prediction of The Bond Coefficient Between Fibre-Reinforced Polymer (FRP) Bars and Concrete
Author(s):
Nadia Nassif , M. Talha Junaid, Salah Altoubat, Mohamed Maalej, and Samer Barakat
Publication:
Symposium Paper
Volume:
360
Issue:
Appears on pages(s):
106-121
Keywords:
Fiber-Reinforced Polymer, Bond Coefficient, Concrete, FRP-Concrete bond, Data-driven, Artificial Intelligence
DOI:
10.14359/51740620
Date:
3/1/2024
Abstract:
Fiber-reinforced polymer (FRP) bars can serve as an appropriate substitute for steel rebar due to their lightweight, high strength, and good corrosion resistance. Nevertheless, the long-term success of FRP bars as promising reinforcement in concrete depends on understanding the bond between FRP bars and concrete. ACI 440.1R-15 recommends utilizing CSA S806-12 Annex S ‘‘Test Method for Determining the Bond-Dependent Coefficient of FRP Rods” for estimating the design value of the bond-dependent coefficient (kb). However, this testing method requires a four-point loaded 3.0-meter-long beam with continuous assessment of developed crack width. Due to the complexity of the test, studies were scarce in assessing the factors affecting the kb. Therefore, this study aimed to relate the experimental kb obtained from large-scale testing to a relatively simpler bond strength value, τu , obtained from smaller-scale FRP pull-out test. The relation was established utilizing data collection for both tests from experimental studies. Three machine learning techniques (Ensembled Trees Artificial Neural Network and Gaussian Process Machines) were then applied to mimic and understand the complex bond-behaviour at varying FRP and concrete properties. The results have shown promising relation (R2>0.8) between kb and τu for different surface textures and fibre types.