Compressive Strength Prediction of HVFA Control Concrete Using ANN and PSO-ANN Models

Utilization of high volumes of fly ash (HVFA) in concrete for cement replacement is a major step towards promoting sustainable developments with environmental benefits. Fly ash is abundantly available as secondary product from coal sourced power generation units. HVFA concrete has globally gained researchers' interest in development of cost effective mixes with minimum wastages. Traditionally used prediction methods have been replaced by modern computational methods which have proven its efficacy in solving highly non-linear real-life problems across various disciplines. Lately, design of HVFA concrete mixes with these approaches has become wide area of researches. In this study, 119 datasets of HVFA control concrete compressive strength (CS) collected from literature is used to train the artificial neural network (ANN) and particle swarm optimization based ANN (PSO-ANN) models; a dataset of 12 nos. each from two individual experimental studies along with their combination is used for testing both models. The models input parameters are cement, fly ash, water-binder ratio, superplasticizer, fine aggregate, coarse aggregate, specimen type and fly ash type for prediction of the HVFA concrete CS. A single hidden layer feed forward ANN using Levenberg-Marquardt algorithm is used, which is optimized by PSO. The models' efficiency is measured in terms of statistical parameters such as correlation coefficient, root mean square error and scatter index. From the results, it is observed that both ANN and PSO optimized ANN models have great potential in predicting the HVFA control concrete CS for a single experimental study.