Predictive modeling of punchouts in continuously reinforced concrete pavement: a machine learning approach

Al-Khateeb, Ghazi; Alnaqbi, Ali; Zeiada, Waleed · 2025 · Crossref

DOI: 10.1007/s43503-025-00057-7

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical need for accurate prediction of punchout deterioration in Continuously Reinforced Concrete Pavement (CRCP) to improve infrastructure management and maintenance planning. Punchouts, characterized by localized concrete spalling around reinforcing elements, significantly compromise pavement integrity and safety. Traditional assessment methods, such as visual inspections and empirical models, are limited by subjectivity, labor intensity, and an inability to capture complex, nonlinear relationships between pavement characteristics and distress. To overcome these limitations, the authors employ machine learning (ML) techniques to develop robust predictive models, aiming to identify key predictors and enhance the reliability of pavement condition assessments. The research utilizes a dataset extracted from the Long-Term Pavement Performance (LTPP) database, comprising 395 entries from 33 CRCP control sections across various climate zones. The dataset includes structural attributes (e.g., age, layer thickness), climate variables (e.g., temperature, freeze index), traffic parameters (e.g., Annual Average Daily Traffic, truck traffic), and performance metrics. The methodology involves descriptive statistical analysis, correlation analysis via heatmap matrices, and feature importance assessment using the Random Forest algorithm. Subsequently, multiple ML models—including linear regression, decision trees, support vector machines (SVM), ensemble methods, Gaussian process regression (GPR), artificial neural networks (ANN), and kernel-based approaches—were trained and evaluated. Data preprocessing included outlier removal using interquartile range analysis, with no imputation required due to complete records. Key findings reveal that pavement age, climate zone, and total thickness are the most significant predictors of punchout occurrence, as identified by the Random Forest feature importance analysis. Correlation analysis indicated a moderate positive relationship between pavement age and punchouts, while traffic loading, particularly truck traffic, showed a greater influence than total traffic volume. Climate factors exhibited weaker correlations, suggesting that structural and traffic loads are primary drivers of distress. In terms of model performance, ensemble methods, specifically boosted trees, and Gaussian process regression demonstrated superior predictive accuracy, characterized by low root mean square error (RMSE) and high R-squared values compared to other tested algorithms. These models effectively captured the nonlinear dependencies among input variables. The significance of this work lies in its contribution to advancing AI-driven pavement management strategies. By identifying key distress predictors and validating high-performing ML models, the study facilitates more informed decision-making regarding resource allocation and maintenance scheduling. The findings underscore the potential of ML to replace or augment traditional, less reliable assessment methods, thereby optimizing CRCP durability and reducing economic losses associated with road closures and repairs. The authors recommend future research focus on refining these models, exploring additional features, and validating results through real-world implementation trials to further enhance infrastructure resilience.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction