Application of Soft Computing for Estimation of Pavement Condition Indicators and Predictive Modeling

Abstract

The pavement management system is recognized as an assertive discipline that works on pavement indices to predict the pavement performance condition. This study used soft computing methods such as genetic algorithms and artificial intelligence to propose a modern generation of pavement indices for road networks in Jordan. The datasets used in this study were collected from multiple roads in Jordan, and 128 data points were used in this study. The input variables are the pavement condition index (PCI) and the international roughness index (IRI) in the artificial neural network (ANN) and gene expression programming (GEP) models. The output variable is the pavement serviceability rate (PSR). The results show an efficient performance benefit of using these techniques. In addition, the ANN and GEP models were able to predict the output variable with a reasonable accuracy, where the ANN model has an R 2 value of 0.95, 0.87, and 0.98 for the PCI, IRI, and PSR, respectively. The ( R 2 ) values of the GEP model are 0.94, 0.89, and 0.99 for PCI, IRI, and PSR, respectively.