Climate and traffic input based International Roughness Index (IRI) prediction model for rigid pavements using Artificial Neural networks (ANN)

Abstract

The International Roughness Index (IRI) is a widely used measure of the roughness of road surfaces and ride quality. The Federal Highway Administration (FHWA) has required States' Departments of Transportation (DOT) to include IRI values in their Pavement Management Systems (PMS) since 1990. However, IRI data collection can be challenging due to cost and resource constraints. This study presents an IRI prediction model for rigid pavements for three south Atlantic states of North Carolina, South Carlina, and Virginia. Utilizing climate and traffic data from the Long-Term Pavement Performance (LTPP) database, an Artificial Neural Networks (ANN) was developed to predict IRI. The R2 for the developed model is 0.84. Sensitivity analysis of the model showed that climate factors have more influence on IRI. In addition, a closed-form stand- alone equation is also extracted from the model, which Local transportation agencies can leverage to predict IRI using available climate and traffic data.