Experimental study and modeling of the effects of corrosion on concrete cover cracking time and bond behavior of deformed bars in reinforced concrete

Corrosion of steel reinforcement due to hostile environments is regarded as one vital structural health concerns in concrete structures. Specifically, the development of corrosion affects the resistance of concrete cover against cracking and the necessary bond strength of rebar in concrete contributing to the loss of resilience and possible structural failures. It is thus essential to understand the effects of corrosion on cracking time of concrete cover and bond strength so that remedial measures can be done on existing and deteriorating RC structures. Hence, this study investigated through laboratory experiments and Artificial Neural Network (ANN) modeling the effects of corrosion on time-to-cracking of protective cover and bond strength. Among all the variables considered in the modeling of cracking time of concrete cover, the compressive strength of concrete, rebar diameter, and concrete cover turned out to be the main influencing variables in the resistance of concrete to cracking. Results from the experimental test showed that at small amounts of corrosion less than 0.27%, the bond strength was observed to increase. At these levels, the amounts of corrosion products were sufficient enough to expand freely through the permeable structure of concrete and occupy the pore spaces. Beyond this level, however, the bond strength of concrete deteriorated significantly. The expansive and progressive internal radial stress due to corrosion resulted to the development of internal and surface cracks in concrete. These cracks contributed to the loss of confinement and mechanical interlock at the steel-concrete interface resulting to the reduction in the bond strength of the rebar. In the parametric investigation of the derived ANN model, the bond strength was also observed to decline continuously with the growth of corrosion derivatives as represented by the relative magnitudes of the ultrasonic pulse velocity (UPV). The prediction results of the model showed consistency and in good agreement with experimental results. All the proposed neural network bond models provided satisfactory prediction results and offered no significant difference relative to the actual experimental bond strengths. These models can be used as a non-destructive assessment tool that can provide baseline information for rehabilitation and repair to ultimately prevent premature damage of reinforced concrete structures