Dynamic inversion of the rupture parameters on fault system with complex geometry: A GPU parallel genetic algorithm based on BIEM

In this study, we attempted to perform an earthquake source dynamic inversion to obtain dynamic parameters on fault system with complex geometry. The forward modeling of the spontaneous rupture process is carried out using a boundary integral equation method (BIEM) based on unstructured meshing, and the inversion method is implemented by a genetic algorithm based on a parallel acceleration of the GPU. The source model in this study is a branched fault, which is described by two physical parameters, the initial stress T⁰ and the critical slip-weakening distance D_c. We investigated the effect of the inherited parameters on the accuracy and convergence of the inversion simulation. Numerical results showed that if a set of parameters are assigned properly, the inversion of rupture parameters is accurate and converges fast. It is easy to converge to a local optimal solution during the inversion process if inappropriate inherited parameters are selected. Compared with T⁰, D_c has better convergence and accuracy in the inversion process.