Improvement of structural modeling of flexible pavements for mechanistic-empirical design.

This study was motivated by the need to improve structural modeling of flexible pavements for potential implementation into the Mechanistic-Empirical Pavement Design Guide (MEPDG). This involves improvement of computational efficiency of layered elastic analysis and development of an efficient tool for the viscoelastic modeling of layered pavement systems. In this study, several improvements in the numerical evaluation of layered elastic solutions were developed and implemented into a new layered elastic analysis program, MNLAYER. A comprehensive comparison of MNLAYER with two widely used layered elastic analysis programs, BISAR and JULEA, demonstrated that MNLAYER's accuracy matches or exceeds the accuracy of these programs. MNLAYER is more accurate than JULEA, especially at evaluation points located close to the surface or at large depths. For the same structural systems, the computational time of MNLAYER is about one-fifth of the computational time of JULEA and one-twentieth of the computational time of BISAR for benchmark tests. This makes MNLAYER an attractive candidate for incorporation into the MEPDG software. To enable a more realistic modeling of flexible pavement behavior, a procedure for analysis of the responses based on the solution for a boundary value problem for a multi-layered viscoelastic system was developed. To achieve this, Burmister's solution for layered elastic theory was generalized. The developed procedure involves reduction of the original viscoelastic BVP to a linear system of integral equations. The system is subsequently approximated by a series of linear equations, which permits an efficient and numerically stable solution. The algorithm described in this thesis was implemented into a computer program, MNLAYER-VE. The program was verified by a comparison with the analytical solution for a viscoelastic half-space and finite element solution for a two-layered system. All of the comparisons yielded very similar results with MNLAYER-VE. Extensive sensitivity studies confirmed that the program is self-consistent. For example, introduction of fictitious layers did not affect the solution. Based on the results of verification comparisons, and sensitivity analysis, it can be concluded that the code is robust.