Data-driven modeling of spalling in jointed reinforced concrete pavements using machine learning techniques

The serviceability and long-term performance of Jointed Reinforced Concrete Pavements (JRCP) are greatly impacted by transverse joint spalling, a localized and count-based distress. The intricate and nonlinear relationships between pavement structure, traffic loading, climate exposure, and initial pavement condition make it difficult to accurately estimate the occurrence of spalling. Using information taken from the Long-Term Pavement Performance (LTPP) database, this study suggests a thorough machine learning-based methodology to forecast spalling incidents in JRCP. A carefully screened dataset consisting of 22 JRCP sections and 184 observations, with no recorded maintenance or rehabilitation activities, was used to ensure that the observed distress reflects natural pavement deterioration. Multiple predictive models, including linear regression variants, decision trees, support vector machines, ensemble learning methods, Gaussian Process Regression, and kernel-based models, were developed and evaluated using a 25% holdout validation strategy. Model performance was assessed using RMSE, MSE, MAE, and the coefficient of determination (R²), along with computational efficiency metrics. The results show that nonlinear models significantly outperform linear methods. The Medium Gaussian SVM model achieved the highest predictive accuracy with RMSE = 2.30, MAE = 1.35, and R² = 0.66, while ensemble models also demonstrated strong performance with R² values approaching 0.60. These results confirm that advanced machine learning models can effectively capture the nonlinear interactions governing spalling development in JRCP pavements. Traffic loading, structural layer thicknesses, and environmental parameters were found to be the most significant elements influencing spalling development using Random Forest-based feature significance analysis. The acquired associations’ physical significance and consistency with known pavement degrading mechanisms were further validated by partial dependence analysis. The suggested approach supports proactive maintenance planning and sustainable management of jointed concrete pavements by offering a precise, comprehensible, and useful tool for anticipating spalling occurrences.