Prediction of Suitable Percentage of Waste Materials for Improving Geotechnical Properties of Black Cotton Soil Using AI Approaches

Abstract

Black cotton (BC) soil consists of montmorillonite minerals. Montmorillonite is one of the reasons to show the swell and shrink behaviour of black cotton soil. The unequal settlement develops under the structures due to this behaviour of black cotton soil. Thousands of researchers and investigators conducted an experimental study to improve soil black cotton soil properties using different materials. These researchers and investigators reported that the geotechnical properties of soil can be enhanced using waste materials. Still, the determination of the geotechnical properties of soil by the experimental procedure is a cumbersome and time-consuming task. A suitable quantity of waste materials is predicted to improve the geotechnical properties of BC soil. The regression analysis, random forest (RF), support vector regression (SVR), decision tree (DT), Gaussian process regression (GPR), and artificial neural networks (ANNs) AI approaches are used to predict the suitability of waste materials in the present research work. The artificial neural network models are developed by one to five hidden layers with ten neurons. The hidden layers are selected in the range of one to five. The performance of MLR, SVR, GPR, RF, DT, LMNN_1H10, LMNN_2H10, LMNN_3H10, LMNN_4H10, LMNN_5H10 is 0.4447, 0.3593, 0.8788, 0.7296, 0.6341, 0.5025, 0.7770, 0.6320, 0.5389, and 0.3856, respectively. The GPR model is identified as an optimum performance AI model and used to predict the suitability of waste materials to improve the geotechnical properties of soil.