Prediction of Concrete Mix Proportion using ANN Technique

Sourav Das, P Pal, & Singh
International Research Journal of Engineering and Technology   unpublished
Concrete mix design is carried out based on some empirical relationships and the experience of the engineer. To arrive at a satisfactory mix proportion a number of trial mixes are to be prepared and tested to check the various design parameters. Thus it is a time consuming task. An artificial neural network (ANN) can overcome these difficulties by predicting the mix proportions based on experimental mix design data. The use of ANN technique in the prediction of concrete mix proportions can be
more » ... roportions can be efficient and economical as it would reduce the need of preparing a large number of trial mixes. The learning processes in artificial neural networks use previous experimental mix design data to predict mix proportions specified by various input parameters. To train the ANN model a database of large number of mix proportions of M25 grade of concrete is prepared using PPC cement. To get the output as mix proportion of various ingredients, input parameters are Target Mean Strength, Workability in terms of slump, W/C Ratio, Specific Gravity of Cement, Sand & Coarse Aggregate and Grading Zone of Fine Aggregate. The trained network is validated with a set of five mix proportions which were not used in the training process. The average percentage error is observed as 0.193%. On comparison with linear regression analysis the ANN model is found to be more efficient.