Influence of thickness and interface condition variability on backcalculated layer moduli from surface deflection

Evaluation of pavement structural performance by Non Destructive Test (NDT) has been growing since the introduction of the Benkelman Beam at the WASHO Road Test in the early 1950's. The deflection basin data have to be analyzed d to obtain the structural properties of pavement layers and subgrade, that are typically used for pavement evaluation and rehabilitation design. One of the more common methods for analyzing deflection data is the back-calculation, in which theoretical deflections are compared with measured deflections and the assumed moduli are then adjusted in an iterative procedure until theoretical and measured deflection basins match acceptably well. The iterative process assume the pavement model as an input, and the accuracy and the precision of the backcalculation process can be reduced by errors due to the inaccurate input evaluation. Particularly pavement layer thickness and interface condition between layers are affected by a construction and measurement related variability. As matter of fact layer thickness information are measured accurately from core, in a limited number of test points because of their destructive nature, and are rough estimated from Ground Penetrating Radar equipment along the entire profile section. Layer interface properties depends on many factors (such as temperature, coat binder, normal pressure etc.) and are very difficult to be estimated. In this study a sensitive analysis to evaluate the combined impact of the layer thickness and interface condition variability was performed. A back-calculation program to accurately evaluate the layer moduli, based on a linear elastic multilayer model, was developed and a simulation procedure, based on Monte Carlo method, was applied to represent the thickness variability. The study showed that layer thickness and interface condition variability have a strong impact on the back-calculated results, therefore it was proposed to perform the design of maintenance treatment according to suitable percentile value of their probability distribution. This former could be evaluated by the simulation procedure developed, as a case study reported shows.