Prediction of pavement friction using Markov chain at the project and network levels

Pavement friction studies continue to focus on reducing wet-weather accidents as lack of friction or skid resistance has been linked to a portion of those vehicle accidents on wet pavements. Probabilistic models for predicting pavement friction or skid resistance as a function of surface age, at both the project and the network levels, are presented. The models deploy the discrete-time Markov model to predict the deterioration of both asphalt (AC) and concrete (PC) pavements. A pavement management system can utilise such predictive models for monitoring and rehabilitation of segments with marginal pavement friction. Two case studies are presented to demonstrate the potential uses of the proposed models. A non-homogeneous Markov chain was used to predict the future friction number (FN) for 12 sample projects using an analysis period of up to 10-year with the estimation error being less than (0.001). On the other hand, a homogeneous chain was employed to estimate the future state probabilities for both AC and PC sample networks using a 5-year analysis period, resulting in high correlations between observed and estimated state probabilities. Data used in the studies were obtained from the Long-Term Pavement Performance Program (LTPP) database. The data was treated such that AC friction records were corrected for testing temperature variations and friction data was sorted by type of test tyre used.