Abstract
Pavement-management agencies progressively explore advanced forecasting tools. This study developed AI algorithms to predict the International roughness Index (IRI) of asphalt concrete pavements in Egyptian urban roads as a representative
example of arid regions. Three predictor variables were investigated in this study: pavement age, mean temperature, and precipitation. For this purpose, 921 data points were collected from the urban road network in the governorate of Beni Suif. A statistical analysis was performed to evaluate the correlation between IRI and the selected variables. Four different prediction models, namely, linear regression (LR), random forest (RF), XG boost (XGB), and artificial
neural network (ANN), were utilized. The predictions of the proposed models showed that the ANN yielded the best fit with a coefficient of determination ( R2) of 0.895 and 0.911, and root mean squared error (RMSE) of 0.139 and 0.131, in training and testing, respectively. The conducted statistical analysis demonstrated that age is the key variable of IRI. The other two parameters (mean temperature and precipitation) did not demonstrate appreciable statistical contributions. This highlights the necessity for utilizing pavement age for planning and maintenance of flexible pavement networks in modern pavement management systems in arid regions.
example of arid regions. Three predictor variables were investigated in this study: pavement age, mean temperature, and precipitation. For this purpose, 921 data points were collected from the urban road network in the governorate of Beni Suif. A statistical analysis was performed to evaluate the correlation between IRI and the selected variables. Four different prediction models, namely, linear regression (LR), random forest (RF), XG boost (XGB), and artificial
neural network (ANN), were utilized. The predictions of the proposed models showed that the ANN yielded the best fit with a coefficient of determination ( R2) of 0.895 and 0.911, and root mean squared error (RMSE) of 0.139 and 0.131, in training and testing, respectively. The conducted statistical analysis demonstrated that age is the key variable of IRI. The other two parameters (mean temperature and precipitation) did not demonstrate appreciable statistical contributions. This highlights the necessity for utilizing pavement age for planning and maintenance of flexible pavement networks in modern pavement management systems in arid regions.
| Original language | English |
|---|---|
| Journal | Geotechnical and Geological Engineering |
| Volume | 44 |
| Issue number | 34 |
| DOIs | |
| Publication status | Published - 9 Dec 2025 |
Free Keywords
- IRI
- Linear regression
- Prediction models
- Pavement management system
- Pavement deterioration models
- Machine learning
- Deep learning