Data supporting the publication: A Novel Tire-Pavement Related Parameter for Improved Rolling Resistance Predictions

This dataset supports the study “A Novel Tire-Pavement Related Parameter for Improved Rolling Resistance Predictions” and contains field measurements, processed surface texture data, and derived parameters used to investigate rolling resistance (RR) of asphalt pavements. The dataset combines physics-based and data-driven approaches by introducing a novel tire–pavement interaction parameter (δ), representing the average tire penetration depth within the contact patch.Data were collected from 22 pavement sections on the Dutch road network (primarily PA16/ZOAB-type asphalt). Rolling resistance coefficients (RRC) were measured using a dedicated trailer system equipped with a Standard Reference Test Tire (SRTT) at 80 km/h under controlled tire pressure and temperature conditions. Surface texture data were obtained using a Laser Crack Measurement System (LCMS) with approximately 1 mm spatial resolution, from which standard indicators such as Mean Profile Depth (MPD), Root Mean Square (RMS), Estimated Texture Depth (ETD), and skewness were derived.A novel workflow was developed to quantify the δ parameter. This includes: (1) a custom-built portable device using a tire and polymer-based texture pads to capture tire penetration imprints under static loading; (2) a computer vision pipeline based on Structure-from-Motion (SfM) applied to high-resolution smartphone video recordings (4K, 60 fps) to reconstruct 3D pavement surfaces and texture pad geometries; and (3) numerical simulation using validated finite element (FE) tire–pavement interaction models. The FE models incorporate hyperelastic material behavior (Neo-Hookean model) and realistic contact conditions to estimate δ under full-scale rolling resistance test conditions.Machine learning models (Multiple Linear Regression, Random Forest, and Artificial Neural Networks) were developed using features including MPD, skewness, and the δ parameter. Data preprocessing steps include cleaning, mesh refinement, outlier removal, scaling, feature normalization (based on training data only), and grouped cross-validation to prevent data leakage. Hyperparameter tuning and model validation were performed using randomized search and k-fold cross-validation. Statistical analyses (e.g., Anderson-Darling and Wilcoxon signed-rank tests) were conducted to assess data distribution and model improvements.The dataset enables reproducibility of the study’s findings by providing the necessary variables for reconstructing the δ parameter, texture indicators, and rolling resistance relationships. However, the raw data originate from collaborative projects involving TU Delft, TNO, and Rijkswaterstaat, and include infrastructure-related measurements. Data collection was conducted on road infrastructure without involvement of human or animal subjects. Therefore, no ethical approval was required; however, data usage complies with institutional agreements, data governance policies, and applicable legal requirements related to infrastructure data ownership and confidentiality.