Speaker
Description
Immiscible two-phase flow in porous media exhibits different flow regimes depending on driving parameters like the capillary number and viscosity ratio. In the steady state, these regimes correspond to characteristic pore-scale flow patterns, such as ganglion flow and drop-traffic flow. By considering pairwise fluid-fluid correlations in the pores and maximizing the entropy, we derive a configurational probability distribution for steady-state two-phase flow that characterizes these pore-scale patterns. The energy function in the probability distribution resembles that of an Ising spin system. Using Boltzmann machine learning applied to configurational data from dynamic pore-network simulations, we estimate the coupling constants in the energy function. We find the couplings are disordered with both positive and negative values similar to those in a spin-glass system, and their distribution depends on the applied pressure drop. Such distributions introduce frustration in a spin-glass system. We investigate the implications of this frustration in the two-phase flow system by measuring magnetization, spin-glass order parameter and susceptibilities from pore-scale configurations. These quantities allow us to characterize the flow regimes and reveal a spin-glass like transition. While our analysis uses steady-state configurations from a dynamic pore-network model, the method is equally applicable to data from other computational approaches or experiments.
| Country | Norway |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
