Pavement backcalculation based on nondestructive testing using a finite element model and optimization methods

Abstract

Inverse analysis is used in different applications to determine parameters from measured responses and an adopted model. When the model responses depend nonlinearly on the parameters, they can be determined by minimizing the difference between the experimental observations and the model responses (dependent on the parameters) using Nonlinear Least Squares (NLS) algorithms. An engineering application that uses this concept is the pavement backcalculation, in which the elastic moduli of pavement layers are estimated based on the results of nondestructive tests, such as the Falling Weight Deflectometer (FWD) test. It is widely used to assess pavement construction quality and monitor the structural condition throughout its lifespan. Several advances in the pavement backcalculation process have been made over the last few decades, both concerning the definition of the model to obtain the responses (forward module) and the optimization procedure for determining the parameters (backward module). Recognizing the limitations of existing methods, this work presents an efficient backcalculation approach based on minimizing the difference between the measured field deflections using the FWD and those obtained using a finite element model developed balancing accuracy and computational efficiency. The resulting NLS problem is solved using the Gauss-Newton and the Levenberg-Marquardt optimization methods, modified to incorporate bound constraints, thereby enhancing the robustness of the backcalculation process and ensuring physically sound results. Furthermore, the parameters of the algorithms of both methods were adjusted to application in pavement backcalculation. The efficiency, robustness, and accuracy of the proposed approach were assessed using numerical examples corresponding to highway and airport pavements. Comparisons between the two methods, with the literature, and with other programs were made. Moreover, the impact of the normalization factor on the backcalculation results was assessed and criteria for detecting defective deflection basins were defined. Finally, an application was carried out aiming at quality control of the rehabilitation of a highway through Light Weight Deflectometer (LWD) tests comparing surface modulus techniques and conventional multilayer backcalculation.