Predicting the Mechanical Behavior of Cold Asphalt Mixtures Using Optimized Artificial Neural Networks

In this study, artificial neural network (ANN) models were developed to predict the indirect tensile strength (ITS) of cold asphalt mixes (CAMs) incorporated with different industrial waste materials. Input variables were determined as variables including emulsion content, curing time, sludge ash content, additive type, and aggregate grading, which were investigated the influence on ITS of CAMs from 163 experimental mix designs. The ANN model structure was formulated, and the multilayer feedforward perceptron was trained, while the number of neurons in hidden layer was set to adjust the structure of the network. The neuron number analyze process shown that 5:3:1 structure selected as the final pattern. The efficient of the trained model is reasonably high with a prediction correlation coefficient (R) of 0.963, and it gives a low Mean Absolute Percentage Error (MAPE) equal to 11.9%. The sensitivity analysis also showed that the curing time and water content were the two highest parameters impacting the ITS, and the emulsion content and initial moisture content were the two lowest parameters influencing the ITS. The test error is increased as the neuron of hidden layer was selected as 4, 6, and 7, however the minimum error is seated as the neuron number was 3. We recommend that ANN-based methodologies show great potential in being reliable and efficient tools that can be used to emulate the mechanical responses of the CAMs tested in this study, as well as aid in designing more sustainable pavement materials.