Dynamic nondestructive testing of pavements

The objectives of this work were to perform a comprehensive analytical study of the different factors affecting the dynamic response of pavement systems subjected to loads imposed by nondestructive testing (NDT) devices, and to use this information to solve the inverse problem of backcalculating the pavement layer properties. The surface response of an elastic half-space to harmonic or transient point and circular surface loads were studied. Closed form solutions found in the literature and solutions developed during this study were presented. The pavement system was modeled as a horizontally layered stratum resting on a rigid base or a half-space and each layer was considered isotropic and linearly viscoelastic. A continuous and a semi-discrete analytical model of this layered system subjected to NDT loads were presented. The response of some simple layered systems to harmonic and transient loads simulating the ones imposed by NDT devices were studied. A cut-off frequency for a layer resting over a rigid base was determined, below this frequency there is not propagation of surface waves and the dissipation of energy is nil or negligible. Analytical studies of the effect of various parameters (stiffnesses and thicknesses of the layers, depth to bedrock etc.) on the response of typical flexible and rigid pavements to loads imposed by the falling weight (FWD) and the spectral analysis of surface waves tests were deflectometer performed. The inversion of FWD data was studied. A least square algorithm was used to minimize the relative difference between measured and computed deflection basins. Three inversion procedures were proposed: (1) a static solution as done at present, (2) a pseudo-dynamic solution in which the values of the displacements that would be obtained if a static load were applied are first obtained by eliminating dynamic effects and a static inversion is then performed on these values, and (3) a dynamic solution in which full dynamic analysis are performed in each cycle of iteration. The accuracy and computational efficiency of these procedures were investigated.