Advancements in multi‐phase unsaturated porous media fracture

Abstract

The focus in this research is on introducing an accurate and stable numerical modeling framework and to compare the numerical results with experimental data from the literature for desiccation‐induced fracturing of unsaturated porous materials. The macroscopic modeling approach is based on combined continuum porous media mechanics and a diffusive phase‐field method (PFM). In the case of unsaturated porous media, one has to deal with more than one pore fluid (e.g. water and air). In this, the mechanical behavior can be expressed via using the Bishop's effective stress principle, which considers the total stress, the capillary pressure (air pressure minus water pressure), and saturation degree. The onset of desiccation‐induced fracturing is driven by the capillary pressure, which leads to degradation not only in the effective stresses but also in the capillary pressure itself. In this contribution, we discuss via a numerical example the effect of considering the air pressure variation on the cracking process and numerical stability. Besides, we briefly discuss the potential inclusion of machine‐learning material laws, such as for retention curves and anisotropic permeability, using, e.g., deep recurrent neural networks (RNN) to improve the model accuracy.

A machine‐learning supported multi‐scale LBM‐TPM model of unsaturated, anisotropic, and deformable porous materials

“…This paves the way for the involvement of more efficient approaches, such as artificial intelligence techniques using, for example, deep neural networks (DNN), to bridge the gap between the different scales. Building DNN‐based constitutive models based on the lower‐scale simulations can be found in different works, such as within crystal plasticity, 8,9 elasto‐plasticity for composite materials, 10 multiphase porous media, 2,11,12 hyperelasticity with enforced constitutive conditions, as symmetry of the stress tensor, objectivity, material symmetry, polyconvexity, and thermodynamic consistency, 13 and other applications in ref. [14].…”