Inversion on S-Wave Velocity Structure of Shallow Soil Layer Site Using Parallel Genetic Algorithm

Abstract

Microtremors was developed to inverse S-wave velocity structure of sites because it costs little and easy to monitor, can be performed at any place even in a densely populated city with non-destructive measurements, the genetic algorithm is widely used in inversion and there are many disadvantages in using genetic algorithms to solve practical problems, so the authors did a lot of efforts to overcome these disadvantages. In order to solve these disadvantages, a coarse-grained parallel genetic algorithm(PGA) based on personal computer cluster was proposed to inverse S-wave velocity structure of shallow soil layer of actual engineering sites, the simulated annealing algorithm and parallel technique message passing interface(MPI) were adopted to implement the coarse-grained parallel compute. The subpopulations were collaboratively optimized through individual migration strategy and the dynamic populations were adopted to balance the computing load. The shallow S-wave velocity structures of two examples and the actual engineering sites were inversed through a 4-node PC cluster test system, the results showed that the algorithm has a good parallel efficiency and can be used in engineering site.