Mathematical Modelling for Bond Strength of Recycled Coarse Aggregate Concrete Using Genetic Programming

Abstract

For the past several years, production of fresh concrete using recycled materials is being increasingly encouraged so as to reduce the environmental impact of concrete construction. In last two decades coarse recycled concrete aggregate (RCA), manufactured by processing of construction and demolition waste has received considerable attention as a potential substitute for natural coarse aggregate (NCA). However, structural application of RCA concrete has been slow primarily because of apprehensions that concrete containing RCA might be inferior to concrete made with NCA. The present work focuses on the effect of RCA on the bond strength between the concrete and the reinforcement. Further, a mathematical model is generated to related Bond Strength with Compressive Strength, bar dia (12mm, 16mm and 20mm), w/c ratio of the mixes. The experimental program consisted of studying the compressive strength and bond strength of mixes prepared at 5 different w/c ratios (0.42, 0.45, 0.48, 0.51 and 0.55) and using 4 replacement levels for coarse aggregates (0%, 30%, 60%, and 90%). The compressive strength was studied at 7 days and 28 days, while bond strength was calculated at 28 days. For studying bond strength, pull out tests were carried out as per IS:2770 (Part I), 1967 and IS: 432 (Part I), 1966 using 12mm, 16mm and 20mm dia of rebars. It was observed that both compressive strength and bond strength show a similar behaviour with respect to replacement ratio of RCA’s. with the increase in RCA content, initially bond strength and compressive strength decreased by small amount, then it increased at 90% replacement level of coarse aggregates. Using the experimental data, the mathematical model is generated using GP(Genetic Programming). The replacement ratios, water-cement ratios, diameter of the rebar and compressive strength have been used as input parameters for developing the mathematical expression. The bond strength of the mixes can be correlated well with the input parameters, with an average error of 13%.