Skip to content

DRP-10 Estaillades Verification Overview

This report documents a controlled DRP-10 benchmark where the current voids image-to-network workflow is compared against the published Estaillades reference values reported by Muljadi et al. (2016).

The reproducible artifact for this report is notebook notebooks/31_mwe_drp10_estaillades_raw_porosity_perm.ipynb.

Goal

The benchmark answers the following question:

Given the same DRP-10 Estaillades binary image, how closely does the current voids extraction-plus-PNM workflow reproduce the published-reference porosity and absolute permeability?

In this documentation split, DRP-10 is classified as verification because the paper reference values were obtained from a numerical OpenFOAM workflow, not from a laboratory core-flood experiment.

Studied Notebook

  • 31_mwe_drp10_estaillades_raw_porosity_perm

Sources

Current Notebook Setup

The DRP-10 notebook currently uses:

  • full-sample analysis (500 x 500 x 500 voxels), no ROI cropping
  • RAW decoding with C-style voxel ordering (order='C')
  • void convention raw == 0
  • optional pre-trimming to axis-percolating paths (trim_nonpercolating_paths = True)
  • native maximal-ball extraction as the primary voids workflow
  • comparison extraction with PoreSpy snow2 default settings
  • comparison extraction with PoreSpy snow2 plus compatibility geometry repairs
  • conductance model valvatne_blunt
  • pressure-dependent water viscosity from thermo, 298.15 K
  • reference outlet pressure 5.0 MPa
  • imposed pressure gradient 10 kPa/m
  • directional permeability in x, y, and z for each backend
  • scalar summaries using arithmetic-mean and RMS Kabs across solved axes

Figures

DRP-10 Estaillades slices

Representative orthogonal slices of the segmented Estaillades volume used in the benchmark.

DRP-10 directional permeability comparison

Directional permeability from the current voids run, backend-comparison workflows, backend mean values, and the paper flow-axis reference (Kx, Table 2).

DRP-10 extracted-network statistics

Extracted-network diagnostics for pore/throat counts, size distribution, and topological quality indicators.

Results

The full CSV outputs are available here:

Porosity and Permeability Against Paper Reference

Quantity voids estimate Paper reference Relative error
Full-image porosity [%] 10.8180 10.8 +0.17%
Network absolute porosity [%] 10.8180 10.8 +0.17%
Primary-backend permeability Kx [mD] 369.8386 172.0 +115.02%

Directional Permeability By Backend (Current Run)

Backend Kx [mD] Ky [mD] Kz [mD] Mean Kabs [mD]
Native maximal-ball 369.84 60.40 189.90 206.71
PoreSpy snow2 defaults 187.81 21.12 98.15 102.36
Compatibility-repaired snow2 121.92 17.91 92.44 77.43

Network Statistics (Current Run)

Metric Value
Np 2477
Nt 4392
Mean coordination 3.55
Max coordination 24
Mean pore diameter-equivalent [um] 33.77
Mean throat diameter-equivalent [um] 20.53
Connected components 1
Giant component fraction 1.00
Dead-end pore fraction 0.199

Interpretation

The DRP-10 case shows strong agreement in full-image porosity, while the permeability estimate remains strongly extraction-dependent. The primary native maximal-ball workflow over-predicts the paper flow-direction permeability in this snapshot (+115.02% in Kx). The PoreSpy snow2 default comparison is much closer to the paper Kx value, while the compatibility-repaired snow2 variant is lower.

All three backend configurations preserve the same qualitative anisotropy (Kx > Kz > Ky), but not the same absolute scale. That spread is expected to affect which scalar permeability is comparable to a paper-reported directional value.

For this reason, the current DRP-10 benchmark is best interpreted as workflow-sensitivity evidence under explicit extraction and closure assumptions, not as universal physical calibration.

Limits Of This Verification

Important limits and assumptions:

  • single DRP-10 sample (Estaillades v2) under one current workflow setup
  • no ROI/subvolume sensitivity analysis in this notebook
  • permeability comparison uses paper directionality assumptions from Table 2
  • extraction and constitutive choices can shift Kabs independently of linear solver correctness
  • the native maximal-ball extraction is still under active verification against backend-comparison and intermediate-quantity checks
  • agreement with one OpenFOAM-referenced paper does not imply agreement with all carbonate datasets or all simulator setups

Reproducible Artifacts

  • Notebook: notebooks/31_mwe_drp10_estaillades_raw_porosity_perm.ipynb
  • Outputs:
  • examples/data/drp-10/Estaillades_v2_estimated_properties.csv
  • examples/data/drp-10/Estaillades_v2_kabs_directional.csv
  • examples/data/drp-10/Estaillades_v2_kabs_directional_by_backend.csv
  • examples/data/drp-10/Estaillades_v2_kabs_summary_by_backend.csv
  • examples/data/drp-10/Estaillades_v2_network_stats.csv
  • examples/data/drp-10/Estaillades_v2_slices.png
  • examples/data/drp-10/Estaillades_v2_kabs_comparison.png
  • examples/data/drp-10/Estaillades_v2_network_stats.png
  • examples/data/drp-10/Estaillades_v2_network_plotly.html