Por Navid Kheirdast (Instituto Dom Luiz - IDL).
Abstract: Finite fault inversions is aimed to resolve the earthquake source process. the source process is explained with the source function which is a function of time and space. Finite fault inversion tries to find the source function using the observed displacements on the Earth surface. the current methods of finite-fault inversions involve high number of parameters to resolve, which comparing with a few number of observations from an earthquake on the earth surface, gives an under-determined problem. In order to decrease the number of parameters we used fuzzy systems to approximate the source function. Fuzzy systems are function approximators that detect the function behavior with a few number of patches. This patches works as a complete set of base functions that can approximate every functions with smooth base function. The combination of fuzzy systems with neural networks makes ANFIS network (Jang 1993) which is a standard tool to implement fuzzy systems to a set of data. In this work, we developed a finite-fault inversion method employing the approximation power of ANFIS to decrease the number of parameters of finite-fault inversion in a regularised fashion. The method has been tested on SIV1 benchmark problem (Mai el al, 2016) for finite fault inversions.
