Introducing a New Solution for Network Level Pavement Structural Evaluation Using Traffic Speed Deflectometer Device (TSDD): A Case Study of Mississippi Department of Transportation

Abstract

This paper introduces a new approach to evaluate the structural performance of road networks using data collected by a traffic speed deflectometer device (TSDD). The advantage of this platform over traditional backcalculation methods is that thickness information for the pavement structure is not required. This is important for network-level evaluations as the information about layer thicknesses is rarely available for agencies. The platform proposed in this study determines the effective structural number (SN_eff) with inputs readily available from testing. In this study, a database containing 20,000 randomly generated deflection basins was simulated, and the SN_eff for each simulation was calculated. The simulated data were then used as input to train a deep learning model. The trained model on the synthetic database showed an R² of 0.97 on both the test and training sets showing its robustness. The model was then validated with 12.7 mi on a major highway in Mississippi. The determined SN_eff values were then compared with those of the American Association of State Highway and Transportation Officials (AASHTO) 1993 model as well as those obtained from backcalculation. The results indicate that the proposed model can accurately measure SN _eff , with the notable advantage of not requiring thickness as an input. This finding is a big step forward in processing TSDD results at large scale. Following the initial analysis, the model was used to process additional TSDD data collected by MDOT. The time to process 1.5 million MDOT records was less than 5.2 s, which shows the tremendous promise of the proposed platform as a useful tool for real time structural evaluation of the roads.