Smithers recently contributed to a paper on the role of spin count in tire traction performance predictions using data on 14-inch standard reference tires (SRTT) from the Smithers Winter Test Center. The study, titled “Tractive performance analysis of the SRTT with respect to the Spin count and the ASTM F1805 parameters,” was published in the Journal of Terramechanics, and is available for free download through April 12, 2025.
Below is an interview with two of the paper’s co-authors: Chaitanya Sonalkar, a graduate research assistant at Virginia Tech, and Eric Pierce, Principal Engineer, Smithers Winter Test Center and Treadwell Research Park.
At Smithers, accurate data is one of our most enduring promises to our customers. This is an interesting challenge at the Smithers Winter Test Center, where we conduct winter traction testing. Snow is very sensitive to changes in the surrounding environment, including temperature, humidity, solar load, and countless others. The mercurial nature of snow makes it all but impossible to predict traction coefficients with 100% accuracy, but that doesn’t stop us from trying. We have invested considerable time and resources into deepening our understanding of snow.
In July 2023, Eric contributed to a paper written by Mohit Nitin Shenvi titled “Correlating tire traction performance on snow with measured parameters of ASTM F1805 using regression analysis.” The paper analyzed the use of a machine learning (ML) algorithm to predict the traction coefficient of a given tire, specifically a standard reference test tire (SRTT), based on three key parameters: CTI penetration depth, snow temperature, and ambient temperature. ML algorithms are valuable for predicting nonlinear data, and the study revealed some new insights.
After reviewing some of the data, the team decided to study a fourth variable: spin count. Specifically, the goal was to investigate the stability of the14-inch SRTT with respect to spin count, since many perceived the effect of this particular parameter on traction coefficient to be more or less static. Chaitanya began working with Mohit on the new study and eventually took over as the main author.
The study was conducted using five years’ worth of data collected at the Smithers Winter Test Center on 14-inch SRTTs. Mohit and Chaitanya implemented several machine learning algorithms to compare how well all four parameters—the three from the original study, plus spin count—correlated to traction coefficient. The algorithm ranked each parameter based on its contribution to the traction coefficient.
The study includes data from thirteen different tires, all 14-inch SRTTs, some with spin counts in the thousands, others in the low hundreds. Ideally, each tire would have the same spin count, but that simply isn’t practical under the circumstances.
The main conclusion we can draw is that spin count is the most contributing factor to the traction coefficient of a tire—once you pass a certain threshold. In other words, there is a point at which spin count becomes a larger predictor of traction coefficient for a 14-inch SRTT when compared to other variables. Over time, we hope to collect data that will allow us to do a comparable study on the 16-inch SRTT that was adopted in 2022.
There is, however, an exception. For tires tested at a higher ambient temperature, spin count was found to make a smaller contribution in predicting traction coefficient. One theory is that this could be due to the top layer of snow becoming softer at higher ambient temperatures, which reduces the tire's grip and results in lower traction. This phenomenon can be observed when analyzing the performance of tires E and N in the study data.
We also concluded that an optimizable Gaussian Process Regression (GPR) model was the most accurate method for training the algorithm and predicting the traction coefficient in this study.
Yes. Two in particular come to mind.
The biggest challenge at the test site itself is a common problem in the world of traction testing: getting weather to cooperate with the necessary environmental conditions. Some winters are too warm. Others are too cold. Even excess snowfall has its challenges, since it requires more surface grooming. As much as we may wish otherwise, we can’t control the weather, and if there’s not enough snow, then we can’t conduct testing and collect data. Data collection for projects like these is always at the mercy of the elements.
From a modeling and analytical perspective, we were faced with some surprisingly basic computer challenges. None of us on the project are computer science engineers, so we used a very powerful computer to train different methods of our algorithm and determine the best one. The computer needed a full week to process all those choices, and we couldn’t let it go idle or slip into sleep mode for even a minute, lest we delay the project. So we spent a lot of time babysitting the computer!
The reality is that, because snow is so capricious, we’re still a long way off from being able to accurately and consistently predict snow traction. But that only serves to motivate us! We’re never going to settle for “close enough” data. We’re always pushing the envelope and working toward a deeper understanding of snow, so we can share that knowledge with our clients and support their ambitious cold-weather performance goals.
For example, we worked with students from Lake Superior State University (LSSU) in 2024 to develop a better method for measuring snow compression and shear characteristics of compacted snow surfaces. The device they created allows us to quickly and easily measure these two parameters throughout the season, giving us greater insight into how tire behavior changes throughout the winter.
We recognize that certain variables have a greater impact under specific conditions. To enhance our understanding, we are planning future studies to collect and automate data to study these variables and their interactions, allowing us to identify more predictable factors over time.
In short, we’re always striving for better data, and this study is another important step in that direction.
The paper is free to download here through April 12, 2025.
