Evolution-based genetic algorithms for analysis of non-destructive surface wave tests on pavements

Spectral analysis of surface waves (SASW) is an in situ non-destructive method for evaluation of near-surface stiffness profiles of pavement systems. This method is based on the dispersion phenomenon of Rayleigh surface waves in layered media. In the SASW method, the stiffness profile of a pavement is determined by fitting measured dispersion data with an adjustable theoretical model of the pavement system that depends on layer thicknesses and elastic moduli. The best-fit stiffness profile is usual ly found with search or optimization techniques. Unfortunately, the parameter space for this problem is highly complex and multi-modal and, consequently, traditional calculus-based search techniques often fail. In this study, a robust search technique employing genetic algorithms is applied. These algorithms are based on the principles of natural evolution, i.e. survival-of-the-fittest. A basic description of this search technique and its implementation for optimization in the SASW method are presented. Its effectiveness is also demonstrated on a difficult, three-layer pavement site. Crown Copyright 1998 Published by Elsevier Science Ltd.