A surrogate model for the prediction of permeabilities and flow through porous media

In this contribution we present an approach to generate a data driven surrogate model for the prediction of permeabilities and flow through two dimensional random micro-heterogeneous materials. The laminar is well described by Darcy's law. order achieve efficient computational tool generation database (up 103 realizations), needed training neural networks, apply stochastic based on Brownian motion. stationary state resulting solves Darcy equation can be iteratively solved Monte Carlo applied particle simulation. Improved numerical efficiency yield usage related transition matrix.