Stochastic generation of multiscale 3D pore network models of building materials

Abstract

The current experimental determination of hygric properties of porous building materials are demanding in time and effort, merge ad- and desorption techniques, fuses static and dynamic methods, and finally do not yield complete nor robust results. Therefore, numerical pore-scale-based prediction of the hygric properties of building materials is on the rise as an alternative. For building materials, this is mostly based on pore network modelling (PNM), given that these are more efficient compared to lattice Boltzmann or particle hydrodynamic methods. Pore network modelling however requires data of the complete pore network for the building material. With the currently available characterization and visualization techniques, this cannot be readily obtained, as the pore sizes in building materials often span several spatial scales. The aim of this paper is to present a scale invariant stochastic generation. To realize this objective, distributions of direct parameters (pores’ sizes, shapes, positions, connections, ...) as well as indirect parameters (overall pore size distribution and open porosity value) are derived from the input data obtained by micro-CT and FIB-SEM and subsequently applied to generate a complete pore network of the porous building material. The quality of the generated PNMs is assessed by comparing them to the original PNMs.