Prediction of Marshall Parameters of Modified Bituminous Mixtures Using Artificial Intelligence Techniques

ABSTRACTThis study presents the application of artificial neural networks (ANN) and least square support vector machine (LS-SVM) for prediction of Marshall parameters obtained from Marshall tests for waste polyethylene (PE) modified bituminous mixtures. Waste polyethylene in the form of fibres processed from utilized milk packets has been used to modify the bituminous mixes in order to improve their engineering properties. Marshall tests were carried out on mix specimens with variations in polyethylene and bitumen contents. It has been observed that the addition of waste polyethylene results in the improvement of Marshall characteristics such as stability, flow value and air voids, used to evaluate a bituminous mix. The proposed neural network (NN) model uses the quantities of ingredients used for preparation of Marshall specimens such as polyethylene, bitumen and aggregate in order to predict the Marshall stability, flow value and air voids obtained from the tests. Out of two techniques used, the NN based model is found to be compact, reliable and predictable when compared with LS-SVM model. A sensitivity analysis has been performed to identify the importance of the parameters considered.