Asphalt Mixture Fatigue Life Prediction Model Based on Neural Network

摘要

An artificial neural network fatigue life prediction model with single hidden layer was built under Matlab environment. Three independent variables: asphalt content, air void, and strain level, and one outcome: fatigue life, were included in this model. Four point bending fatigue test was carried out under different conditions to obtain training data for neural network, and a total of 27 sets of samples were organized for training. The R~2 of the neural network fatigue life prediction model was 0.91. To make a comparison, the same data was used to fit the empirical fatigue life predicting model. Analysis exhibited that the neural network model prediction was better than the empirical fatigue life predicting. Meanwhile, the neural network model showed an ability to weigh the impact of variables.