Use of Artificial Neural Networks (ANN) to Predict Pavement Management Data Attributes

Mississippi Department of Transportation (MDOT) has a pavement management system (PMS), which began in the late 1980s, for which the University of Mississippi (UM) did the original studies and models. The research team has collected condition and distress data approximately every two years since 1991, with imaging and other technology changing throughout. The team has decision trees, as well as prediction models developed in the form of Markov transition probability matrices; however, the team would like to take another look at the models. The models apply to attributes such as various types of cracking, roughness (IRI), rutting, and faulting. Artificial neural networks (ANN) offer a possibility for updating and simplifying these models. This study will investigate the possibility of using ANN to update the PMS prediction models.