A Conservative Darcy-Brinkman-Stokes Formulation for Digital Rock Flow Simulation with Embedded Second-Brinkman Effects

arXiv:2611.14348 · physics.flu-dyn · Submitted 2026-11-10 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "A Conservative Darcy-Brinkman-Stokes Formulation for Digital Rock Flow Simulation with Embedded Second-Brinkman Effects".

Title and authors: Kai: So this paper is about "A Conservative Darcy-Brinkman-Stokes Formulation for Digital Rock Flow Simulation with Embedded Second-Brinkman Effects." We talked about the core idea—how to do pore-scale flow without needing these massive boundary meshes, and how DBS1 misses that second Brinkman term.

Mira: And what this paper actually proposes is this new conservative formulation, DBS2, which embeds those second Brinkman effects directly into the standard finite volume discretization.

Lev: What I want to know is what they mean by "conservative formulation" here; does it just mean they’re using a specific way to write the equation that keeps mass balance correct?

Kai: It means they write the equation in its conservative form, and because of how they apply the standard central differencing scheme, it naturally retains the full DBS formulation without needing separate discretization for that second term.

Mira: So it’s about preserving computational efficiency while achieving accuracy, which is something I’ve always wanted to see more consistently in these continuum models.

Lev: Preserving efficiency is key because real hardware has limited memory and time; if you can achieve DNS-level accuracy without a separate costly discretization step, that's a huge practical advantage.

Kai: They tested this on three things: microchannel flow, a two dimensional heterogeneous micromodel, and that three dee Bentheimer sandstone image from micro-CT <ref:2611.14348#pg2>. The numbers they show are pretty telling.

Mira: It’s interesting to see how it performs across such different scales and complexities in the validation cases they chose.

Lev: I’m curious about the specific error metrics; what did DBS1 get on those tests, and how much better is DBS2?

The paper's summary: Kai: So to summarize, this paper shows that neglecting the second Brinkman term leads to errors in flow field characterization and permeability prediction. This is because that term arises from pore-scale velocity fluctuations.

Mira: And they’ve shown that you don't need a separate discretization for it; you can write the DBS equation in its conservative form and use a standard central differencing scheme to capture it naturally as DBS2.

Lev: So, the main takeaway is that this formulation corrects distributed viscous drag, which is what causes those errors in integral quantities like permeability, even when the overall velocity field looks similar.

Kai: It’s about correcting that distributed viscous drag because it’s essential for capturing those interfacial velocity predictions accurately.

Mira: They also mention that this conservative treatment of the viscous stress preserves the computational efficiency of DBS1 while achieving DNS-level accuracy in terms of the second Brinkman term inclusion.

Lev: If you can achieve DNS level accuracy without adding extra complexity, that’s a strong argument for using this in computationally expensive problems like those involving complex rock geometries.

Kai: They used an effective diffusion coefficient defined at cell faces, called nu eff f equals nu f over epsilon f, and they implemented it using a variable-coefficient Laplacian operator.

Mira: That variable-coefficient Laplacian operator naturally captures both the first and second Brinkman viscous terms through the divergence of that effective diffusion coefficient.

Lev: Capturing both terms this way sounds robust; I wonder if that formulation is stable when porosity epsilon approaches zero in the solid regions.

The paper's improvements: Kai: The improvements they highlight are pretty clear: DBS2 gets much better results than DBS1 across the board for permeability prediction.

Mira: In the horizontal microchannel test case, they show that the L2 error for DBS2 is effectively zero, while for DBS1 it’s about thirty-three point eight times ten to negative five.

Lev: Zero error on a benchmark like that is what we look for; it means the method is actually working where we expect it to be accurate.

Kai: For the two dimensional micromodel, DBS2 achieves a relative error of only one point five percent in permeability, compared to a fourteen point one percent error for DBS1.

Mira: That’s a significant reduction in error across that specific test case; it shows the correction is working well for complex geometry problems.

Lev: Fourteen point one percent is still quite high if you're trying to get reliable macroscopic properties from micro-CT data, but one point five percent is much closer to what we want.

Kai: And in the final three dimensional Bentheimer image case, DBS2 achieves a three point eight percent error in permeability, contrasting sharply with the thirty seven point nine percent error of DBS1.

Mira: That jump from thirty seven point nine to three point eight is quite dramatic; it really shows how much better this conservative formulation is at handling three dee complexity <ref:2611.14348#pg2>.

Lev: If you can drop that kind of error by almost a third in a complex three dee image, that’s a huge leap in predictive capability for things like geothermal energy recovery or carbon storage applications <ref:2611.14348#pg2>.

Conclusion: Kai: So to wrap up, the key finding is that accounting for the second Brinkman term in DBS2 significantly improves the method’s predictive accuracy by correcting distributed viscous drag.

Mira: This correction helps improve integral quantities like permeability even when the qualitative velocity field appears similar between DBS1 and DBS2.

Lev: It’s a practical pathway to bridge pore-scale physics with continuum-scale models because it allows you to accurately represent those higher-order viscous effects without needing separate numerical treatment.

Kai: They show that this conservative formulation is central to enabling accurate, computationally efficient representation of higher-order viscous effects in pore-scale microcontinuum simulations.

Mira: It supports a wide range of subsurface applications, like carbon and hydrogen storage and geothermal energy recovery, because the results are much more reliable.

Lev: I just want to say that the conservative formulation is essential for making this approach practical; it’s what allows us to get this accuracy without needing an extra layer of complex discretization on top.

Kai: Exactly. We've seen how DBS2 converges rapidly to the direct numerical simulation when you refine the mesh, which is a big indicator of its quality.

Mira: It seems like this approach provides a practical way to get high-fidelity results from micro-CT data without getting bogged down in the complexity of boundary-fitted meshes.

Lev: For me, it means we can start making more reliable predictions about subsurface flow patterns with less overhead for running those simulations on actual hardware.

Kai: So, this paper, "A Conservative Darcy-Brinkman-Stokes Formulation for Digital Rock Flow Simulation with Embedded Second-Brinkman Effects," gives us a much more accurate way to model viscous boundary effects in heterogeneous porous media.

Gospel E Stewart, Hannah P Menke, David Egya, Jacqueline Mifsud, Julien Maes

Institute for GeoEnergy Engineering, Heriot-Watt University

physics.flu-dyn

Submitted: 2026-11-10

Updated: 2026-11-10

Code: https://github.com/hannahmenke/DrawMicromodels

Importance score: 89/100

The gist: Accurate pore-scale direct numerical simulation (DNS) typically requires boundary-conforming meshes on fully resolved geometries…

Key concepts

Conservative DBS Formulation
This is a mathematical approach that treats the Darcy-Brinkman-Stokes equations in a way that preserves physical conservation laws. It allows the model to accurately capture complex viscous effects, including the second Brinkman term, without needing separate, computationally expensive discretizations for porosity gradients.
Second Brinkman Term (SBT)
This term represents a higher-order viscous effect in fluid flow within porous media. The DBS2 formulation treats this term implicitly by defining an effective diffusion coefficient at cell faces. This implicit treatment ensures numerical stability and correctly captures the distributed viscous drag that standard models miss.
Effective Diffusion Coefficient ($ u_{eff,f}$)
This coefficient is calculated using the ratio of fluid viscosity to porosity within a cell face. It is used in a variable-coefficient Laplacian operator to implicitly capture both the first and second Brinkman terms simultaneously. This method avoids storing explicit porosity gradients, maintaining efficiency even when solid regions are modeled.
DBS2 vs. DBS1
DBS1 systematically overpredicts permeability because it fails to account for the second Brinkman term. DBS2 corrects this by including it, leading to vastly improved accuracy. For example, in a 3D rock image, DBS2 error drops from 37.9% (DBS1) to 3.8% (DBS2), demonstrating superior predictive power.

Terminology

Summary

Accurate pore-scale direct numerical simulation (DNS) typically requires boundary-conforming meshes on fully resolved geometries, which is memory-intensive and often becomes impractical when modelling complex rock geometries. p. 0 The DBS2 formulation consistently improves agreement with direct numerical simulation (DNS), while the classical formulation (DBS1) systematically overpredicts permeability. p. 0

The core methodology involves a conservative Darcy-Brinkman-Stokes (DBS) formulation that naturally retains the full DBS formulation without separate discretisation of the second Brinkman term (SBT). p. 0 This is achieved by writing the DBS equation in its conservative form and applying a standard central differencing scheme in a finite-volume discretisation which naturally retains the full DBS formulation, referred to as DBS2. p. 0 The conservative treatment of the viscous stress preserves the computational efficiency of DBS1 while achieving DNS-level accuracy. p. 0

The paper validates this approach using three benchmark cases:

(i) microchannel flow.

(ii) a 2D heterogeneous micromodel. p. 0

(iii) a 3D micro-CT image of Bentheimer sandstone. p. 0

Numerical implementation focuses on implicitly capturing the second-order Brinkman term through an effective diffusion coefficient defined at cell faces, denoted as νeff,f = νf/εf. p. 4 This is implemented using a variable-coefficient Laplacian operator, fvm::laplacian(fvc::interpolate(nu)/fvc::interpolate(eps), U). p. 4 This implicit treatment naturally captures both the first and second Brinkman viscous terms through the divergence of the effective diffusion coefficient, without requiring explicit storage of porosity gradients or specialised discretisation schemes. p. 4 This formulation remains well-defined even for ε ≈ 0, avoiding singular behaviour while consistently enforcing impermeability of the solid phase. p. 4

Model validation across different complexities demonstrates significant performance gains for DBS2 over DBS1. p. 5 In the horizontal microchannel test case, the L2 (RMS) error for DBS2 is effectively zero, whereas for DBS1 it is 33. p. 98 × 10−5. For the 2D micromodel, DBS2 achieves a relative error of only 1. p. 05% in permeability, compared to a 14. p. 61% error for DBS1. p. 9 In the final 3D Bentheimer image case, DBS2 achieves a 3. p. 08% error in permeability, contrasting sharply with the 37. p. 99% error of DBS1. p. 9

The findings indicate that accounting for the second Brinkman term in DBS2 significantly improves the method’s predictive accuracy by correcting distributed viscous drag, thereby improving integral quantities like permeability even when the qualitative velocity field appears similar. p. 9 This behaviour is consistent with the horizontal fracture benchmark where DBS2 converges rapidly to DNS with mesh refinement, whereas DBS1 remains offset. p. 7 The conservative formulation is central to enabling accurate, computationally efficient representation of higher-order viscous effects in pore-scale micro-continuum simulations. p. 4 This capability supports a wide range of subsurface applications, including carbon and hydrogen storage and geothermal energy recovery. p. 0

Future work could extend the current validation to multiphase and reactive-flow systems, where additional physical coupling may impose stronger demands on the conservative DBS formulation. p. 1 Although the DBS2 formulation demonstrates strong accuracy and efficiency, further work is needed to assess its behaviour across a wider range of pore geometries, particularly highly tortuous ones, where nonlinear pressure variations and complex local flow structures may not be fully captured at a given voxel resolution. Addressing these challenges would further improve predictive accuracy and broaden the applicability of the DBS approach in Digital Rock Physics. p. 1

**The overall conclusion is that the conservative DBS2 formulation accurately represents viscous boundary effects in heterogeneous porous media without requiring additional numerical treatment beyond the standard finite-volume discretisation. p. 4 This approach provides a practical pathway to bridge pore-scale physics with continuum-scale models, supporting a wide range of subsurface applications. p. 1

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Conservative DBS formulation implementation allows for capturing higher-order viscous effects without separate discretisation of porosity gradients, which preserves computational efficiency while achieving DNS-level accuracy. This means an improved AI system can perform accurate pore-scale simulations directly on voxelized micro-CT images without needing complex, computationally expensive boundary-fitted meshes for solid regions.

  2. The DBS2 formulation improves permeability prediction by correcting distributed viscous drag, as shown by the result DBS2 closely reproduces this value at 1.515×10−11 m2 in the 2D micromodel validation case. This capability enables AI systems to make more reliable predictions of macroscopic properties like permeability from micro-CT data compared to standard DBS models, reducing systematic overpredictions seen in DBS1.

  3. The conservative formulation is implemented using a variable-coefficient Laplacian operator that naturally captures both first and second Brinkman viscous terms through the divergence of the effective diffusion coefficient. This allows an improved AI system to handle sharp fluid-solid interfaces with better numerical robustness, avoiding singular behavior when porosity epsilon approaches zero in solid regions.

  4. The DBS2 method exhibits superior convergence rates compared to DBS1 for permeability prediction, as indicated by The DBS2 method converges at a rate nearly identical to that of the DNS method. This suggests an AI system utilizing this formulation will require fewer mesh refinements to achieve high accuracy when analyzing complex rock geometries like Bentheimer sandstone.

  5. The ability of the conservative DBS2 formulation to reproduce DNS velocity fields with significantly higher accuracy in the microchannel demonstrates that an improved AI system can accurately resolve detailed flow patterns, even at coarse mesh resolutions, which is crucial for understanding interfacial viscous dissipation.

Related papers