Experimental study and neural network modelling of expansive sub grade stabilized with industrial waste by-products and geogrid

For highway construction projects, expansive sub grade improvement is one of the prime and major processes. The strength of the sub grade soil is indicated by its California bearing ratio (CBR) value which is quite expensive and time consuming. In order to overcome this situation, the present research aims in predicting the soaked CBR value for the stabilized soils by Multiple Regression Analysis (MRA) and Artificial Neural Network (ANN) modeling. Experiments were done to stabilize the expansive soils with the addition of varying percentages of industrial waste by-products (Coal ash, Bagasse ash and Groundnut shell ash) with geogrid layers. Ash type, Mix proportion, Atterberg limits, Maximum dry density, optimum moisture content and number of geogrid layers were taken as input variables and soaked CBR value as output variable for the regression based models. It is observed that ANN model is accurate than the MRA model in predicting the soaked CBR value of expansive soil stabilized with industrial waste materials, both the measured experimental values and predicted values are in good agreement. LevenbergMarquardt back propagation shows maximum R value of 0.94317 and minimum MSE value of 0.49. (c) 2021 Elsevier Ltd. All rights reserved. Selection and peer-review under responsibility of the scientific committee of the 2nd International Conference on Manufacturing Material Science and Engineering.