&&ReWrAp:HEADERFOOTER:0:ReWrAp&&

Author

Mehdi Parvini

Date of Award

10-1997

Degree Type

Thesis

Degree Name

Doctor of Philosophy (PhD)

Department

Civil Engineering

Supervisor

Professor D.F.E. Stolle

Abstract

In order to assess the structural charateristics of a pavement-subgrade system, non-destructive, in-situ tests together with backcalculation procedures are widely used. Traditionally, the analytical models adopted for this proccess are deterministic, however, in reality, the quantities involved in the problem may be random variables. Neglecting the variable nature of the system parameters, e.g., highway material properties, may affect the reliability of the pavement response prediction. On the other hand, inverse solutions to pavement problems are often ill-conditioned and sensitive to the input parameters. Past experience has shown that the estimated values of a blackcalculated parameter by different agencies may vary by several orders of magnitude, representing a high level of uncertainty in the estimated paramter. Unless the uncertainty is quantified, practitioners are forced to resort to higher safety factors, which is neither economical nor always conservative. The present study investigates, rigorously, the behavior of a pavement-subgrade system from a stochastic point of view, and addresses the sensitivity of response variation to variations in layer properties. The results of a forward analysis are utilized to establish a relation between input and output statistical moments in order to interpret the pavement deflection data stochastically. The proposed framework in this research allows one to quantify the uncertainty level in backcalculated system parameters. It also provides a tool to infer the accuracy of the pavement performance prediction based on mechanistic models. For the purpose of introducing the stochastic approach, the perturbation technique is applied to an idealized, two-layered, pavement-subgrade system for the case of: (a) a static solution based on Odemark definition of equivalent layer thickness; and (b) a frequency domain solution to a single degree of freedom (SDOF) system using an impedance function. The methodology is then extended to a stochastic finite element framework in order to analyze boundary-valued problems of more complex geometry and distribution of material properties. The perturbation method is a mean-based, second-moment analysis for the second-order accurate expected value, and first-order accurate cross-covariance function. For the dynamic analysis, viscoelastic response of the pavement is obtained by using the periodic-load analysis approach and Fourier synthesis. Based on the results of the simulations, it is demonstrated that, the sensitivity of surface deflections is significantly higher to the subgrade properties than those of the surface and base layers, both in a static and a dynamic analysis. Consequently, it is concluded that, the low dominant frequency of the falling weight deflectometer (FWD) load limits the capability of this test in characterizing surface layer properties. Using the concept of coefficient matrix, it is illustrated that, the low sensitivity of deflections to surface layer properties can be interpreted as a high level of uncertainty in the estimated pavement moduli in a backcalculation exercise. It is indicated that uncertainties in backcalculated parameters often result in an unacceptable pavement performance prediction. Moreover, the physical behavior of the layers are identified by finding the contribution of each layer to the total deflection of the system using the notation of contribution ratio.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS