• The International Concrete Abstracts Portal is an ACI led collaboration with leading technical organizations from within the international concrete industry and offers the most comprehensive collection of published concrete abstracts.

International Concrete Abstracts Portal


Title: Machine Learning Assessment of Fiber-Reinforced Polymer-Strengthened and Reinforced Concrete Members

Author(s): M. Z. Naser

Publication: Structural Journal

Volume: 117

Issue: 6

Appears on pages(s): 237-251

Keywords: bond-slip model; empirical analysis; fiber-reinforced polymer (FRP) reinforcement; fiber-reinforced polymer (FRP) strengthening; machine learning

DOI: 10.14359/51728073

Date: 11/1/2020

Fiber-reinforced polymers (FRPs) are often used as externally bonded systems or internal reinforcing elements to improve the performance and resilience of concrete structures. Oftentimes, the observed response of FRP-strengthened/reinforced concrete members, whether in the field or in experiments, does not match that predicted using codal provisions as such guidelines may not fully capture occurrence of complex phenomena such as debonding, or adhesive softening/cracking. To overcome such challenge, this study hypothesizes that a machine learning (ML) approach can be adopted to better comprehend the behavior of FRP-strengthened/ reinforced structures. This approach fuses artificial neural networks (ANNs) and genetic algorithms (GAs) to develop a new bond-slip model as well as to derive empirical expressions capable of accurately evaluating ultimate bending and shear capacity, as well as of identifying expected failure modes in FRP-strengthened/reinforced concrete structures. These expressions are developed and validated using observations obtained from over 600 experiments and hence are applicable to a wide variety of structural members and components. The proposed expressions are easy to apply (that is, in one step) and do not require tedious/iterative procedure nor advanced computing/simulation software.


Electronic Structural Journal


Please enter this 5 digit unlock code on the web page.