Prediction of unconfined compressive strength of treated expansive clay using back-propagation artificial neural networks

Abstract

The multilayer perceptrons (MLPs) artificial neural networks (ANNs) that are trained with feed forward back-propagation algorithm was used in this study for the simulation of unconfined compressive strength (UCS) of cement kiln dust-treated expansive clay. Artificial neural networks (ANNs) are yet to be efficiently extended to soil stabilization aspect of geotechnical engineering. As such, this study aimed at applying the ANNs as a soft computing approach to predict the UCS values of Nigerian expansive clay. For each of the three ANN model development, eight inputs and one output data set were used. The mean squared error (MSE) and R-value were used as yardstick and criteria for acceptability of performance. In the neural network development, NN 8-11-1 that gave the lowest MSE value and the highest R-value were used for all the three outputs in the hidden layer of the networks architecture which performed satisfactorily. For the normalized data set used in training, testing and validating the neural network, the performance of the simulated network was satisfactory having R values of 0.9812, 0.9783 and 0.9942 for the 7, 14 and 28 days cured UCS respectively. These values met the minimum criteria of 0.8 conventionally recommended for strong correlation condition. All the obtained simulation results are satisfactory and a strong correlation was observed between the experimental UCS values as obtained by laboratory test procedures and the predicted values using ANN.