Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression and Artificial Neural Networks

Abstract

The accurate prediction of forest road pavement performance is important for efficient management of surface transportation infrastructure and achieves significant savings through timely intervention and accurate planning. The aim of this paper was to introduce a methodology for developing accurate pavement deterioration models to be used primarily for the management of the forest road infrastructure. For this purpose, 19 explanatory and three corresponding response variables were measured in 185 segments of 50 km forest roads. Logistic regression (LR) and artificial neural networks (ANNs) were used to predict forest road pavement deterioration, Pothole, rutting and protrusion, as a function of pavement condition, environmental factors, traffic and road qualify. The results showed ANNs and LR models could classify from 82% to 89% of the current pavement condition correctly. According to the results, LR model and ANNs predicted rutting, pothole and protrusion with 83.5%, 83.00% and 81.75%, 88.65% and 85.20%, 80.00% accuracy. Equivalent single axle load (ESAL), date of repair, thickness of pavement and slope were identified as most significant explanatory variables. Receiver Operating Characteristic Curve (ROC) showed that the results obtained by ANNs and logistic regression are close to each other.