Neural Estimation of Bond Strength Degradation in Concrete Affected by Reinforcement Corrosion

Abstract

Bond between the reinforcement and its surrounding concrete is essential for their composite action. Due to corrosion of reinforcement and associated spalling/cracking of concrete, the bond strength reduces. Estimation of the reduced bond strength of the corroded reinforcing bar in concrete becomes important in assessment of the performance of the corrosion-affected concrete structure, particularly for residual life estimation. There are analytical as well as empirical models available in the literature for estimation of the reduced bond strength from the degree of corrosion of the reinforcing bar. In this paper, we develop ANN models for the aforementioned estimation for both pull-out tests and flexural tests. From the performance measures and the scatter plots, it is concluded that the ANN models perform comparably with the best empirical model (Bhargava 2008) though developed with 70–80% of the data.