Department

Civil and Environmental Engineering

Document Type

Article

Publication Date

Summer 8-3-2026

Embargo Period

8-10-2026

Abstract

Traffic noise is a critical public health concern affecting millions of highway users and adjacent residents worldwide. In response, many transportation agencies have adopted functional surface materials to reduce noise at the source on pavement, but assessing their effectiveness remains expensive and logistically challenging. Close Proximity (CPX) testing quantifies tire-pavement noise but requires specialized equipment costing $50,000-$126,000 and is limited to existing pavement, preventing proactive noise assessment during pavement design. This study develops machine learning models to predict CPX noise levels from readily available pavement characteristics, eliminating the need for costly tests during design and planning phases. To train and validate the models, CPX data were collected at three speeds (80, 100, 120 km/h) on 15 expressway routes surfaced with diverse functional materials and surface treatments. Both base learner (Random Forest, Gradient Boosting, and Extreme Gradient Boosting) and ensemble (Stacking and Voting Regressors) algorithms were implemented with key predictors, such as route, surface material, pavement age, vehicle speed, and ambient temperature. Feature importance analysis identified vehicle speed as the dominant predictor, followed by material type and age. Residual analysis confirmed unbiased predictions with no autocorrelation. The Stacking Regressor achieved superior performance with R² values of 0.98 during training and 0.95 during validation for equivalent CPX noise, corresponding to mean absolute errors of approximately 0.7–1.2 dB, well within ISO 11819–2 repeatability tolerances. Overall, the ensemble algorithms provide accurate, interpretable tools for CPX noise on pavement surfaces. The findings are expected to enable transportation agencies to make informed decisions regarding materials design, maintenance level and timing, and life-cycle noise management.

Journal Title

Case Studies in Construction Materials

Journal ISSN

2214-5095

Volume

25

Issue

2026

Digital Object Identifier (DOI)

https://doi.org/10.1016/j.cscm.2026.e06379

Download

Share

COinS