Prediction of International Roughness Index of Flexible Pavements from Climate and Traffic Data Using Artificial Neural Network Modeling

Abstract

This study is done to predict the International Roughness Index (IRI) for flexible pavements using climate and traffic data by employing artificial neural network (ANN) modeling. The climate and traffic data are collected from the long-term pavement performance (LTPP) database. The ANN model is trained using 50% of climate, traffic, and IRI data and then rest 50% data is used to validate the model by comparing ANN predicted IRI and measured IRI for flexible pavement under a climatic zone. The trained model and the validated model are compared by calculating root mean square error (RMSE) of ANN predicted IRI and measured IRI. A flexible pavement located at the wet-freeze climatic zone, employing 7-7-1 ANN structure and using Pure Linear transfer function, the RMSE generated is 0.055. A better prediction ANN model is generated using 7-9-9-1 architecture using a non-linear transfer function and the RMSE further improved to 0.012.