Développement d’une méthode numérique apte à simuler efficacement la propagation d’une rupture sismique de type « pulse »

Abstract

We have studied a numerical technique for simulating earthquakes, based on a boundary integral element method. In such a method, the stress at position on the fault is expressed as a space-time convolution over the whole slip history on the fault. In its basic formulation, the time convolution prohibits long simulations. The removal of the static stress contribution allows to truncate this convolution, though.We have applied a similar technique to extract the retarded static term, in which the static stress contribution is transferred at a finite velocity (that of the seismic wave in the medium), which is a much better approximation of the full elastodynamic solution. The resulting computational gain is typically of about a factor 2-3 for the tested cases. In addition, the features of the static and dynamic parts of our formulation are very promising for a specific application of this technique to earthquakes for which the duration of slip at each fault point is short (self-healing slip pulse).