1374078 - Calculating Friction for the Entire Highway Network on an Annual Basis

Skid resistance is the most critical pavement surface property associated with road user safety, especially during wet weather conditions. There is a well-established inverse correlation between roadway friction levels and the number of wet weather crashes. In other words, as skid resistance decreases, the number of wet weather crashes increases. However, skid resistance remains a challenging property to measure at a network level due to the inefficiencies of the measuring equipment such as water usage, tire wear, and agency’s budgetary constraints.
These limitations have created a demand for an alternative method capable of predicting pavement friction accurately and continuously using readily available information, such as pavement texture. This research study collected a representative sample of 30 highway pavement sections in Texas that covered a wide range of pavement surfaces, textures, and friction levels. The data were subjected to rigorous quality control and were used to create machine learning models that predict surface type and friction.
The final model estimates friction in terms of skid number (SN) on a range of flexible pavements present in Central Texas with an accuracy in the order of 80% in terms of the coefficient of multiple determination. The innovative way in which machine learning algorithms are used in this study has the potential to allow highway agencies to accurately calculate friction for the entire highway network on an annual basis.