Artificial Neural Networks, A Tool for Predicting Compressive strength of Recycled Aggregate Concrete

Abstract

Now a days designing of mix proportions is more costly and time taking process. The final mix proportions are arrived from several trail mixes. Due to this there is a lot of wastage of material and also it requires at least 7 days to finalizing the mix proportions. Present study deals with the prediction of compressive strength of Recycled Coarse Aggregate Concrete (RCA) utilizing Artificial Neural Networks (ANN). Finding the compressive strength using conventional methods involves usage of various mixes, materials, and most importantly time. Based on this problem, there is a need to develop a software based strength design rather than going for experimental based design. The compressive strength as modeled is a function of five inputs and one output variable. The input variables are Coarse Aggregate, Recycled Coarse Aggregate (RCA), Fine Aggregate, Cement, Water, and output variable is Compressive Strength. The required data is collected from the previous research works carried on Recycled Coarse Aggregate Concrete by several researchers from detailed literature review, about 60 mix proportions are selected and considered to predict the compressive strength of RCA by using VISUAL GENE DEVELOPER software. The selected values are trained, validated and predicted for analytical based mix proportions. To validate the predicted strength values, regression values are checked. The analytical predicted values are compared with the experimental values. From results it is concluded that the artificial neural network based results shown high prediction accuracy and the outcomes exhibited that the utilization of ANN in evaluating compressive strength of concrete is valuable in fostering the blend extents for various evaluations of cements.