Waveform inversion of cross-well data with cooperative coevolutionary differential evolution algorithm

Waveform inversion with global optimization methods don't need to compute the gradient and can work well without good guess of initial model. Cooperative coevolutionary differential evolution (CCDE) had exhibited superior performance compared to some other methods in the application of 1D large size waveform inversion problem. In CCDE, the large size problem is decomposed into some subcomponents, and a mutation operator is designed to guide the mutation direction of each subcomponent with the corresponding local fitness value. In this paper, we extended the CCDE for the waveform inversion of cross-well data. The new decomposition strategy, local fitness function and crossover operator are proposed for 2D model. We applied our algorithm to a synthetic cross-well data. Numerical results showed that we can greatly raise the efficiency of DE by introducing the decomposition strategy and local fitness function concept.