Rutting is a major cause of pavement deterioration resulting in deformation in the bituminous layers under heavy traffic and high temperatures thereby reducing service life of the pavement. This study evaluates the rutting susceptibility of the bituminous mixtures using two wheel-tracking procedures: a rubber wheel test following BS EN 12697-22 and a steel wheel Hamburg Wheel Tracking Test (HWTT) following AASHTO T-324. A review of existing literature indicates that while steel-wheel testing performance specifications are well-established, the thresholds for rubber-wheel applications remain relatively limited and less consistently defined. To address this, both laboratory-compacted and field-extracted samples of Stone Mastic Asphalt (SMA-2) with 19 mm NMAS and Dense Bituminous Macadam (DBM-I and DBM-II) with 37.5 mm and 26.5 mm NMAS, respectively, incorporating VG-40 and PMB-40 binders, were subjected to steel-wheel and rubber-wheel tracking tests. The results showed that the rubber-wheel configuration produced 23–32
Road agencies are increasingly pressured to improve safety and asset maintenance outcomes under constrained budgets, making predictive maintenance (PdM) and data-driven decision support highly appealing for road infrastructure management. This study evaluates expert perceptions of PdM adoption readiness and its perceived impact on road safety management, considering awareness, expected benefits, and key technical, organizational, and regulatory barriers. A cross-sectional survey of 100 experts was conducted using an 11-item Likert-scale instrument. Data were analyzed through reliability analysis, exploratory factor analysis, and descriptive and correlational statistics to identify latent dimensions of readiness, perceived benefits, and perceived barriers. The PdM scale showed high internal consistency (Cronbach’s α = 0.954) and a two-factor structure representing perceived impact and acceptance, and technical-organizational readiness, jointly explaining approximately 90
High-fill embankments often suffer excessive, uneven settlement that compromises construction quality and operational safety. Machine learning offers data-driven predictions, but most existing models are single predictors with limited generalization ability and interpretability. This study proposes an interpretable settlement prediction method for high-fill embankments based on stacking ensemble technique and local interpretable model-agnostic explanations (LIME). Firstly, hyperparameters of five representative machine learning models are optimized using a genetic algorithm. Then, to build the ensemble model, base learners are selected using the Spearman correlation algorithm, with the best-performing model identified as the meta-learner. Next, the prediction results of the ensemble model are evaluated and analyzed, and LIME is employed to interpret the decision-making mechanism of the ensemble model. Finally, the applicability and robustness of the proposed method are validated through an application on a different engineering project and a noise interference analysis. The results show that the ensemble model achieves R2 = 0.9940 and a mean absolute percentage error (MAPE) of 0.26
This study explores the application of machine learning algorithms, specifically Random Forest (RF) and Support Vector Machine (SVM), to predict the International Roughness Index (IRI) for the Tehran-Karaj Freeway in Iran. The research utilized a comprehensive set of features, including geometric, pavement, weather, and traffic-related variables. After preprocessing and selecting nine key features, hyperparameters for both models were optimized through random search to improve prediction accuracy. Comparative analysis indicated that the RF model outperformed the SVM, with a Mean Squared Error (MSE) of 0.1480 versus 0.2155 during validation. Following retraining with optimized parameters on 85
The degradation of skid resistance on wet asphalt pavements represents a major cause of driving risk, particularly during rainfall. Conventional assessment methods generally lack real-time responsiveness and fail to account for the dynamic vehicle–road interaction. To overcome these limitations, this study proposes a driving-safety-oriented framework for evaluating skid resistance risk. The framework introduces the Dynamic Skid Resistance Safety Margin (DSRS), which quantifies the safety margin between the actual pavement friction and the minimum friction required to maintain vehicle stability, braking control, and lateral safety. A complementary Dynamic Skid Resistance Index (DSRI) enables continuous and quantitative grading of skid-related risk. Furthermore, an expected skid resistance performance model is formulated by incorporating vehicle–road coupling effects, thereby enhancing the interpretability of the framework. Field experiments conducted on a 2-km asphalt pavement section under varying rainfall intensities and vehicle speeds confirm that the proposed framework effectively captures friction variations and associated safety risks. Overall, the proposed framework provides a practical foundation for dynamic risk evaluation and early-warning systems, supporting adaptive decision-making in intelligent transportation systems under wet-weather conditions.
