Shenvi, Mohit NitinSandu, CorinaUntaroiu, Costin D.Pierce, Eric2024-08-192024-08-192023-09https://hdl.handle.net/10919/120954Winter tires sold in North America are often tested using the ASTM F1805 testing process to determine if they can be labeled with the ‘mountain snowflake’ symbol which indicates better performance for snow usage. The standard dictates the requirements for testing and necessary track preparation methodologies. In addition, the requirements of the standard dictate the range of three major conditions for tests to be carried out, namely the CTI penetration measurement, the snow temperature, and the ambient temperature. However, these parameters cannot be directly used in the simulation stage of snow modeling for better evaluation of prototypes. It is well-known that snow properties depend on a wide variety of parameters, making the creation of an accurate and robust snow material model, and, consequently, applying a simulation-based approach for tire design, difficult. This work focuses on the analysis of a dataset of five winter seasons of a 14-in. Standard Reference Test Tire on snow used to benchmark the performance of a potential winter tire. The blinded data measured at Smithers Winter Test Center were used in the analysis to train regression models for predicting the traction coefficient and evaluating the extent to which the measured parameters affect the variation in the traction coefficient. This study utilized twenty-six different modeling approaches and implementation of principal component analysis. The findings of this study highlight the relative importance of the compression and shear characteristics of the snow on the traction of the tire. It was found that regression methods based on Gaussian processes were better at predicting the traction coefficient. The study also highlights the importance of utilizing a single physical tire as the reference tire for benchmarking according to the ASTM F1805.application/pdfencompacted snowwinter tire certificationsnow material propertiesregression learningprincipal component analysisArticle - Refereedhttps://doi.org/10.1016/j.coldregions.2023.103926213