Files
Abstract
The falling weight deflectometer (FWD) is a type of nondestructive testing equipment used to evaluate the stiffness of asphalt pavements. For the test procedure, the FWD briefly impacts the pavement, and displacement sensors measure the pavement's vibration at several locations. Afterward, a backcalculation method estimates the pavement's stiffness by analyzing the vibration measurements. Typical backcalculation methods comprise two components: a computer model that simulates the FWD test, and an optimization scheme to estimate the pavement's material properties. The optimizer intelligently varies the material properties in the model until the simulated deflections closely match the measured deflections. FWD testing has become the most widely used approach for nondestructive pavement evaluation, and numerous techniques have been developed to improve backcalculation accuracy. However, these developments increase the method's complexity, which brings forth new issues to be investigated.
The contents of this dissertation are organized into four key aspects. The first is the introduction of a new FWD backcalculation software, UGA-PAVE, which employs a state-of-the-art numerical pavement model, ViscoWave, and a novel gradient-based optimization algorithm, tandem trust-region. Second, verification and validation studies are conducted with the software which shed insight on ill-posed backcalculation and the effective frequency range---two prevalent issues hindering modern backcalculation efficacy. Third, a study comprising 720,000 simulated backcalculations is conducted to identify changes to the FWD testing procedure with the goal of increasing backcalculation accuracy. Fourth, improvements are presented for a widely used approach to approximate interconversion between two primary models of viscoelastic stiffness: dynamic modulus and relaxation modulus.
The results of these studies improve the efficacy of FWD backcalculation. The development, verification, and validation of UGA-PAVE highlights the importance of research on ill-posed optimization and the effective frequency range in FWD backcalculation, which is virtually nonexistent at present. The test procedure examination provides changes to the design of FWD equipment that can improve accuracy of current and future backcalculation approaches. Finally, the interconversion method is particularly useful for FWD backcalculation, because they are typically designed to output results in the form of one viscoelastic modulus but not the other.