Estimating vehicle braking distance over wet and rutted pavement surface through back-propagation neural network

Abstract

Due to repeated load cycles and climate impacts, road pavement deteriorates. One of the primary causes of pavement degradation is the formation of rutting under the wheel path on the road surface. Rutting's impact on vehicle performance, especially on rainy days (since the rain would fill the rutting depression), leads to longer braking distances compared to dry conditions. This study developed a MATLAB-based model for calculating vehicle braking distance on wet asphalt pavement affected by rutting, using dynamic skid resistances generated through Back-Propagation Neural Network (BPNN) analysis. This study addresses the worst-case scenario in which rutting is filled with water, then calculates the required vehicle braking distance under various Water Film Thickness (WFT) conditions. The developed model can perform these evaluations for different operational conditions across various input ranges, such as precipitation intensity, number of lanes, lane width, cross slope, average texture depth, rutting depths, and accumulated WFT. As an outcome, the vehicle braking distance can be estimated as a function of Rutting Depth (RD) within a known vehicle speed interval. After validating the proposed model against existing approaches from the literature, several sensitivity analyses are conducted to assess the impact of influencing parameters on the results. Moreover, the study examines the relationship between AASHTO braking distance requirements and the RD threshold levels adopted by several highway agencies. Furthermore, this model is also applicable to real-world case studies, enabling the calculation of vehicle braking distances with varying RDs in the presence of various WFTs on the pavement surface.

Keywords

• Machine learning
• Back-propagation neural network
• Vehicle braking distance
• Rutting
• Dynamic skid resistance
• Water film thickness