A statistical approach for pavement layer moduli backcalculation as a function of traffic speed deflections

The backcalculation of pavement layer moduli is salient for assessing the performance of existing pavement structures and for making appropriate maintenance/rehabilitation decisions. The current backcalculation practice is traditionally dependent on falling weight deflectometer (FWD) measurements. However, due to the stationary nature (i.e., stop and go) of the FWD test, there are some safety, logistic, and technical challenges associated with it. In this paper, a statistical approach is presented for backcalculating pavement layer moduli as a function of traffic speed deflections collected by the Rapid Pavement Tester (RAPTOR) utilizing SPSS, MATLAB and curve expert professional software. The study utilized a pre-developed finite element traffic speed deflection database to generate pavement deflections under the RAPTOR loading at different lateral offsets.Kindly check anfd c Then, nonlinear regression pavement deflection models were developed as a function of several parameters including RAPTOR travel speed, asphalt mid-depth temperature, asphalt thickness, base thickness, |E*| of the asphalt corresponding to the asphalt mid-depth temperature and RAPTOR loading frequency, base modulus, and subgrade modulus. A total of 1250 synthetic data points were used to develop the regression models, which were capable of explaining more than 80 of this data. Such feasible prediction models provided an attractive alternative for making a better primary prediction about pavement deflection in a quite short time with a very low error rate. This was followed by developing an algorithm to minimize the error of the target function which estimates the difference between predicted and measured deflection bowls by changing pavement layer moduli in an iterative procedure. The developed methods were validated against laboratory and field moduli values obtained for two pavement sections located on the National Center for Asphalt Technology (NCAT) test track. The results revealed an acceptable prediction accuracy for the developed models yielding an average root mean square error (RMSE) of 14.41 for the asphalt layer, 15.12 for the base layer, and 9.50 for the subgrade layer, respectively. © 2022, Springer Nature Switzerland AG.