Accuracy analysis of modulus results considering the whole process of modulus back-calculation—based on GPR and FWD

Abstract

• The numerical hypothesis model was corrected by combining GPR and FWD. • The applicability of different regression models for deflection basins was clarified. • The sensitivity of modulus back-calculation results was analyzed. • The accuracy of the modulus back-calculation result has been significantly improved. In this study, based on the consideration of the whole process of modulus back-calculation, the measured deflection basins of asphalt pavements under dynamic loading were obtained by using the falling weight deflectometer (FWD), and the parameters such as thickness, initial modulus range, and Poisson's ratio of each structural layer in the numerical hypothesis model were corrected based on the detection results of the thickness and damage status of each structural layer by the ground-penetrating radar (GPR). At the same time, mathematical regression models such as the Exponential regression model, Parabolic regression model, and Sigmoid regression model were used to fit the theoretical deflection basin, and the accuracy of the relationship model between the deflection basin parameters and the pavement structural parameters constructed after fitting the theoretical deflection basin by different regression models were compared. It is observed that the change in the numerical hypothesis model and the regression model of the theoretical deflection basin can have a significant impact on the modulus back-calculation results. The research results show that when the numerical hypothesis model and the regression model are not corrected and optimized, the absolute value of the relative error between the back-calculation modulus and the dynamic modulus of the asphalt layer obtained in the laboratory test is 64 %. After correcting the numerical hypothesis model and fitting the deflection basin with the Sigmoid regression model, the relative error is within ±4 %. In summary, it is an important method to improve the accuracy of modulus back-calculation by correcting the numerical hypothesis model and optimizing the regression model of the deflection basin.