A hybrid approach of support vector regression with genetic algorithm optimization for predicting spalling in continuously reinforced concrete pavement

Spalling in longitudinal joints of continuously reinforced concrete pavement (CRCP) can significantly impact pavement performance and longevity, making accurate predictive models essential for proactive maintenance and management. Effective spalling prediction models help in optimizing repair schedules, reducing lifecycle costs, and ensuring user safety. This study presents a hybrid approach for predicting spalling in longitudinal joints of CRCP using support vector regression (SVR) optimized with a genetic algorithm (GA). By leveraging the robustness of SVR in managing non-linear relationships and the optimization capabilities of GA, the proposed model aims to achieve higher prediction accuracy and provide deeper insights into the factors influencing spalling. The dataset, sourced from the long-term pavement performance (LTPP) database, includes 33 sections and 395 observations, encompassing a range of structural, climatic, and traffic-related variables. The GA-SVR model's performance was evaluated and compared to traditional models such as linear regression and decision tree. Results indicate that the GA-SVR model outperforms these traditional models, achieving a mean root mean square error (RMSE) of 17.2965 and a mean R-squared (R²) value of 0.90238 across all folds, indicating strong predictive accuracy. Key factors influencing spalling were identified through comprehensive data analysis, including age, layer thickness (L2, L3, L4), climatic conditions (temperature, precipitation, Freeze Index), and traffic loads (AADT, AADTT, KESAL). The study also utilized 3D interaction plots to visualize the relationships between Age and other variables, providing valuable insights into their combined effects on spalling. The residual plot and predictor importance analysis validated the model’s accuracy and robustness, highlighting its effectiveness in capturing complex patterns in the data. The findings underscore the potential of the GA-SVR model as a valuable tool for pavement engineers and policymakers, facilitating proactive maintenance and informed decision-making to enhance the longevity and performance of CRCP infrastructure.