Abstract
A hybrid method with genetic algorithms and simulated annealing is proposed to invert surface-wave phase velocity to a 1D S-wave profile. The basic computational flow of the hybrid algorithm is principally based on the genetic algorithm by Yamanaka and Ishida (1995). However, a generation-dependent probability for choosing new models in the crossover operation is introduced in the hybrid method. The probability is defined in a similar manner as the simulated annealing using temperature that should be decreased with increasing generations. In addition to this operation, we used a real-number coding of parameters in the method. I examined the performance of the method in finding optimal S-wave velocities and thicknesses through numerical experiments using synthetic Rayleigh-wave phase velocity for a 4-layers model of deep sedimentary layers. It is concluded that the hybrid method can find an optimal model with less computational efforts than those the conventional genetic algorithm and simulated annealing. These features of the hybrid method can be also recognized in application to actual phase velocity data observed in microtremor explorations in the Kanto basin, Japan.
