PREDICTION OF COMPRESSIVE STRENGTH OF SAW DUST ASH-CEMENT CONCRETE USING ARTIFICIAL NEURAL NETWORK METHOD

ABSTRACT

This research work focused on the Prediction of Compressive strength of Saw dust Ash-Cement concrete using Artificial Neural Network method. Neural Networks offer a number of advantages including requiring less formal statistical training, ability to implicitly detect complex non-linear relationship between dependent and independent variables, ability to detect all possible interactions between predictor’s variables, and availability of multiple training algorithms. A 5-20-1 network architecture was created. In all, a total of five hundred (500) data were selected and used in this work. Out of this number; three hundred and fifty (350) were used for training the network, seventy five (75) for validation and seventy (75) for testing the network. After training the network, the output and targets have an R - value of 0.97868 which greater than 0.9. This shows that the data used for training the network, have a good fit. The results obtained from the network are approximately the same as that obtained experimentally. The percentage error of the experimental results with respect to the network predicted results, ranges from 0.000 to 0.4145% which is very insignificant. The network was tested for adequacy at 0.05 significant levels using statistical student`s T-test, and it was found to be adequate. With the trained network, compressive strength of saw dust ash- cement concrete can be predicted from known mix ratios and vice-versa. Some problems often associated with the network which include prone to overfitting and overtraining, the ‘black box nature` of the network, greater computational burden, and empirical nature of the model development were taken care of by the validation checks.

Key words: Prediction, Neural network, Compressive strength, Mix ratio, Sawdust Ash, Concrete.