Prediction of the hot asphalt mix properties using deep neural networks

Abstract

BackgroundMarshall design process is the most common method used for estimating the Optimum asphalt content (OAC) and this process is called the asphalt mix design. However, this method is time-consuming, labor-intensive, and its results are subjected to variations.ResultsThis paper employs artificial neural network (ANN) for the estimation of Marshall test parameters (OAC, Stability, Flow, Air voids, Voids in mineral aggregate) using the aggregate gradation as the input of the prediction process. Multiple ANNs are tested in order to optimize the NN hyperparameters and produce accurate predictions. Different activation functions, number of hidden layers, and number of neurons per hidden layer are tested and heatmaps are generated to compare the performance of every ANN. Results show that the optimum ANN hyperparameters change depending on the predicted parameter. Finally, the deep NN can predict the OAC, stability, flow, density, air voids, and voids in mineral aggregate with R values of 0.91, 0.8, 0.53, 0.65, 0.77, and 0.66.ConclusionThe linear activation function is the most efficient activation function and generates more accurate results than the logistic and the hyperbolic tangent functions. Additionally, it is shown that the deep neural network approach represents a major innovative tool for the prediction of the asphalt mix properties as results of this approach outperforms results of the shallow ANN that consists of a single hidden layer which is the only approach used in the literature. Thus, the use of the deep ANN can be useful during the phase of the design of the asphalt mix process because of its ability to predict variables with high accuracy. For example, the ANN with 3 hidden layers and 16 neurons per layer with the linear activation function can predict the OAC with high accuracy (R = 0.91), which can be helpful in the design process as the ANN can be employed for the prediction of the OAC of the asphalt mix.