DEVELOPING PREDICTION MODELS FOR SLOPE VARIANCE FROM THE INTERNATIONAL ROUGHNESS INDEX

Abstract

Road roughness is considered a primary indicator of pavement condition and serviceability, and the performance of paved roads is linked to road roughness. The focus of this study is to develop a relationship between two important roughness indicators, namely the international roughness index (IRI) and slope variance (SV), based on actual road roughness data to achieve a suitable correlation between these two indices using artificial neural networks (ANNa) and gene expression programming (GEP) techniques. Different study areas were selected to develop the prediction model. The first study area is the Desert Highway in Jordan, while the three remaining study areas are located in the US. A total of 533 data sets were used in this study to develop a model to predict the IRI from the SV. The GeneXproTools 5 software package was used to build the GEP model, while MATLAB 2019 was employed to develop the ANN model. The results showed that the GEP and ANN models outperformed all other previous models. The GEP-Based model showed a better performance and more precise results than the ANN model according to the coefficient of determination (R2).