PREDICTION OF STRENGTH OF FLY ASH CONCRETE

Abstract

Concrete, being widely used, is the most important building material in civil engineering. Concrete is a highly complex material, which makes modeling its behavior a very difficult task. Many attempts were taken earlier to develop suitable mathematical models for the prediction of compressive strength of different concretes, but not for flyash concrete. Those traditional methods have failed to map non-linear behavior of concrete ingredients. The present study has used artificial neural networks (ANN) to predict the compressive strength of fly ash concrete. The ANN model has been developed and validated in this research using the mix proportioning and experimental strength data of 6 different mixes. The artificial neural networks (ANN) model is constructed trained and tested (in MATLAB) using the previous researches data. A total of 149 different fly ash concrete mix design were collected from technical literature. For comparative study, 4 models were developed. Strength was modeled in ANN-1 model as a function of three input variables: w/b, cement, water. ANN-2 model was presented with 4 input parameters: w/b (water-binder ratio), cement, water and fly ash%. ANN-3 model consist of 6 input variables: w/b (water-binder ratio), cement, water, fly ash%, coarse and fine aggregates. In this study, an attempt was also made to develop a multiple regression model for predicting strength (in EXCEL) as it is being used largely by researches in prediction. Finally, these four models were compared using coefficient of determination and RMSE values, and resulted in the fact that ANNs models have performed better than MLR model in predicting compressive strength of flyash concrete. Also, ANN model presented with more description of system (with more input variables that affect strength) yield more accurate results showing better correlation with observed/ experimented/actual strength.