Machine Learning Applications for Predicting Longitudinal Cracking in Continuously Reinforced Concrete Pavement
Ali Alnaqbi, Ghazi G. Al-Khateeb, Waleed Zeiada
Abstract
This study addresses the critical issue of longitudinal cracking in Continuously Reinforced Concrete Pavement (CRCP), a paramount concern in pavement engineering due to its significant impact on infrastructure performance and longevity. methodology. The research uses a dual-phase methodology to leverage data from the Long-Term Pavement Performance (LTPP) database. Conventional regression models provide insights into influential factors, but recognizing their limitations, the study extends to machine learning models. Gaussian Process Regression with a Squared Exponential kernel emerges as a standout performer, emphasizing its efficacy (RMSE: 11.84, R-squared: 0.78). Ensemble Tree models, especially Boosted Trees, also exhibit competitive results. Feature importance analysis highlights critical variables like temperature and AADTT. The study's findings underscore the superiority of certain machine learning models over traditional regression methods in predicting longitudinal cracking, offering practical implications for optimizing maintenance strategies and enhancing CRCP infrastructure longevity.