Machine Learning Applications for Predicting Faulting in Jointed Reinforced Concrete Pavement

Abstract

Faulting predictive models are crucial for maintaining the structural integrity and safety of rigid pavements, ensuring a smooth and durable driving surface. Accurate predictions allow for timely maintenance, reducing long-term costs and extending pavement lifespan. The objective of this study is to advance faulting prediction methodologies for jointed reinforced concrete pavement (JRCP) to bolster pavement longevity and maintenance strategies. Using data from 22 distinct sections under the long-term pavement performance (LTPP) program, encompassing a wide array of climatic scenarios, the research leverages six cutting-edge machine learning algorithms: regression tree (RT), support vector machine (SVM), ensembles, Gaussian process regression (GPR), artificial neural network (ANN), and kernel methods. The methodology includes a detailed statistical analysis and an evaluation of feature significance to dissect the multifaceted interactions among key determinants of pavement performance. The results underscore the efficacy of machine learning in elevating faulting prediction precision. Among the algorithms tested, boosted trees demonstrated the highest accuracy, with a root mean square error (RMSE) of 0.68, a mean squared error (MSE) of 0.46, and an R-squared value of 0.78. The feature importance analysis highlighted that L4 Thickness, pavement age, L3 Type, and initial IRI were the most influential factors in predicting faulting, with importance scores of 0.2266, 0.1862, 0.1638, and 0.1594, respectively. This study demonstrates the significant potential of machine learning models in accurately predicting faulting in JRCP, paving the way for more efficient pavement maintenance and management strategies that can effectively address and mitigate pavement distress.