Porous media drying and two-phase flow studies using micromodels

Abstract

In this thesis, we report an investigation of porous media drying and steady-state two-phase flow behaviour at the pore scale using micromodels based on thin section images of real rocks. Fluid distributions (and the deposition of solid salt in the case of drying) were imaged in real-time using optical microscopy. Computer simulations of the two-phase flow was initially compared to micromodel experiments and then used to predict behaviour in geometries not available in the lab. We performed evaporation experiments on a 2.5D etched-silicon/glass micromodel based on a thin section image of a sucrosic dolomite carbonate rock at different wetting conditions. NaCl solutions from 0 wt% (deionized water) to 36 wt% (saturated brine) were evaporated by passing dry air through a channel in front of the micromodel matrix. For deionized water in a water-wet model, we observed the three classical periods of evaporation: the constant rate period (CRP) in which liquid remains connected to the matrix surface, the falling rate period (FRP) and the receding front period (RFP), in which the capillary connection is broken and water transport becomes dominated by vapour diffusion. The length of the deionized water CRP was much shorter for a uniformly oil-wet model, but mixed wettability made little difference to the drying process. For brine systems in water-wet and mixed-wet micromodels, the evaporation rate became linear with the square root of time after a short CRP. Although this appears similar to the RFP for water, salt continued to be deposited at the external surface of the matrix during this period indicating that a capillary connection was maintained. The reduction of evaporation rate appears to be due to the deposited salt acting as a partial barrier to hydraulic connectivity, perhaps allowing dry patches to grow on the evaporating surface. The mechanism causing the square root time behaviour is therefore unlike the case of deionized water where capillary disconnection from the fracture channel is followed by a diffusion controlled process. In completely oil-wet micromodels capillary disconnection prevented salt deposition in the fracture. The resulting permeability impairment was also measured, for the water-wet model, we observed two regions of a linear downward trend in the matrix and fracture permeability measurements. A similar trend was observed for the mixed-wet systems. However, for the oil-wet systems, fracture permeability only changes slightly even for 360g/L brine, a result of the absence of salt deposits in the fracture caused by the early rupture of the liquid wetting films needed to aid hydraulic connectivity. Overall, matrix permeability for all wetting conditions decreased with increasing brine concentration and was almost total for the 360g/L brine. Furthermore, drying with air was compared with drying with CO2 gas, with the latter having important applications in CO2 sequestration processes. We observed that using CO2 rather than air as carrier gas makes the brine phase somewhat more wetting especially in the deionized water case, with the result that hydraulic connectivity was maintained for longer in the CO2 case compared to dry-out with air. Steady-state two-phase flow experiments were also conducted to study the effect of viscosity ratio, flow rate and capillary number on flow regimes and displacement processes using a 2.5D etched-silicon/glass micromodel based on a thin section image of a Berea sandstone rock. Of particular interest here was a new type of pore-scale behaviour, termed dynamic connectivity, previously identified in steady-state two-phase flow experiments in real rocks at the transition to ganglia flow by X-ray tomography. Micromodels have the potential to resolve the dynamics of these displacement processes due to the high speed resolution of optical techniques. Depending on the mean-size, prevalence, and connectivity of the non-wetting phase, four flow regimes were identified: connected pathway flow (CPF), big ganglia flow (BGF), big-small ganglia flow (BSGF) and small ganglia flow (SGF). These flow regimes move from CPF to SGF as the capillary-viscous balance of the system is altered by increasing the total flow rate of the system. The boundaries of the flow regimes are indistinct, however the domain of the BGF increases (and/or SGF decreases) with a decrease in the viscosity ratio of the system. That is the BGF regime persisted to higher capillary number for the water/squalane system than the water/decane system because it is harder for big blobs to split into smaller blobs at low viscosity ratio. However, dynamic connectivity was not observed in these micromodel experiments even after replicating the experiments with the same fluid pair (Nitrogen/Deionized water) used in the real porous media experiment. Therefore, we speculate that the constant depth of the micromodel used in this study does not provide a suitable geometry for dynamic connectivity to develop. One potential reason for this is the compressed range of capillary pressures due to the single etch depth. Hence, a multi-depth non-repeat micromodel was designed based on a single confocal image of a Bentheimer sandstone. Prototypes of small sections of the multi-depth model were produced by 3D printing but it was not possible to fabricate a functioning model due to time constraints. Simulation was therefore used to explore the multiphase flow behaviour of the new geometry. Initially a Lattice Boltzmann code (developed in another project) was applied to simulate flow in a small region of the single depth geometry and compared to the experimental results as a validation step. The LB model was then used to predict flow behaviour in the multi-depth geometry, however only connected pathway and ganglia flow regimes were seen unambiguously. It is therefore likely that the lack of 3D connectivity rather than capillary pressure limitations prevent the appearance of dynamic connectivity.