Developing a Model to Predict Bleeding Areas in Asphalt Pavements Using Artificial Neural Network

Abstract

Bleeding can have a negative effect on the performance of a pavement, as the layer of asphalt on the surface can become slippery, reduce skid resistance, and lead to premature wear and tear. Therefore, predicting the bleeding area in asphalt pavements will help in maintaining better pavements, increasing the safety of the drivers, and delivering timely maintenance. The aim of the study is to develop a model using a highly sophisticated analysis tool called an artificial neural network (ANN) to predict the bleeding areas in asphalt pavements as output, with one hidden layer and several neurons and using several independent variables as inputs, such as lane annual average daily truck traffic, lane annual truck volume estimate, pavement thickness, average asphalt content, average annual temperature, annual total snowfall, and total annual precipitation. The Long-Term Pavement Performance database will be employed in this study to extract data for the states of Texas, Arizona, and New Mexico. The ideal architecture for forecasting the bleeding of asphalt pavements was concluded to be an ANN model with several neurons in one hidden layer and with 160 data sets, with R2 = 0.71. A standalone ANN equation was also extracted. Additionally, a sensitivity analysis was conducted to determine the sensitivity of the model to a particular change in an input parameter.