Purpose and evidence boundary

NeqSim implements many calculations used to study petroleum reservoir fluids. Their existence does not establish accuracy for every fluid or operating range. This matrix connects the topic scope of Pedersen, Christensen and Shaikh, Phase Behavior of Petroleum Reservoir Fluids, third edition (2024), to current source and representative tests. The publisher’s contents establish the book scope. Full chapter numerical examples were not available for this audit; no claim is made that every example in the book has been reproduced.

The initial audit used master 98d185ded48c5c0c12b2faec374544b6b6b043b1 on 24 October 2026. The swelling correction and independent numerical check below are additions to that baseline. Follow the complete work in roadmap #4564.

Use three distinct evidence levels:

A numerical fixture captured from earlier NeqSim output is a regression guard. It is useful but does not independently establish experimental accuracy. The representative suites in the matrix are evidence pointers, not a claim that all of those suites or all physical models have been requalified in this change.

Capability and evidence matrix

Book topic Current implementation and representative evidence Status and next qualification
Fluid types (1) Compositional SystemInterface fluids, named components, TBP and plus fractions Implemented inputs; select public black-oil, volatile-oil, condensate and gas cases with complete model definitions
Sampling and analysis (2) Fluid readers, composition inputs and recombination workflows; WhitsonPVTReaderTest Software routes exist; certify composition basis, sample quality, recombination ratios and component balances per dataset
Laboratory experiments (3) ConstantMassExpansion, ConstantVolumeDepletion, DifferentialLiberation, MultiStageSeparatorTest, SwellingTest; corresponding tests Core routes implemented; swelling correction is verified below. Equilibrium contacts and gas revaporization need explicit capability/benchmark mapping
EOS (4) SRK, PR, CPA, PC-SAFT, volume correction and nonclassical mixing-rule implementations Broad model surface; fix the complete EOS definition before comparing results. No assertion that different models give identical results
C7+ characterization (5) PlusFractionModel, TBPfractionModel, LumpingModel, PseudoComponentCombiner; PedersenPlusModelSolverTest Implemented, with tracked MW-moment and covolume-override gaps; qualify conservation and lump-count/common-slate sensitivity
Flash and envelopes (6) ThermodynamicOperations, saturation and envelope solvers; saturation tests and envelope guide Numerical regressions exist; require stability, fugacity and EOS-volume closure, critical/retrograde cases and explicit failure status
PVT simulation (7) CCE/CVD/DL/separator arrays and laboratory guide Implemented; validate measured relative volume, dropout, Bo/Bg/Rs, densities, GOR and standard-state conventions
Thermal properties (8) Heat capacities, enthalpy/entropy, PH/PS flashes and JT coefficient Implemented property/flash routes; compare caloric data and derivatives with consistent ideal-gas and formation references
Regression (9) PVTRegression, data classes and PVTRegressionTest CCE/CVD/DLE/separator/viscosity routes exist. Enum values alone do not prove target support; see #4567
Transport (10) PFCT/CSP and LBC viscosity, conductivity models, ViscositySim; PFCTViscosityMethodTest Implemented; model-to-model checks must be supplemented with measured gas and oil viscosity/conductivity over the intended range
Wax (11) WAT, WaxFractionSim, WaxCurveCalculator; WaxModelComparisonTest Model-specific routes and regression repairs exist; require measured appearance/precipitation curves and solid balances
Asphaltenes (12) Characterization, onset flashes and quantitative/screening models; onset and stability suites Partial qualification by method; field-risk classification is not a measured equilibrium onset or precipitated-amount benchmark
Hydrates (13) Formation T/P and inhibitor flashes; hydrate and brine assessment suites Selected literature assessment exists; qualify the actual gas/brine/inhibitor combination and distinguish equilibrium from kinetics
Depth variation (14) TPgradientFlash, TPgradientFlashTest Implemented with regression fixtures; independently validate pressure/composition profiles and gravity/thermal-gradient assumptions
Gas-injection EOR (15) Bubble-branch SwellingTest, SlimTubeSim, MMPCalculator Partial: critical/dew swelling and rigorous conservative MMP remain open (#4565, #4566)
Water and inhibitors (16) CPA/electrolyte phase calculations, inhibitor and mutual-solubility routes Model- and dataset-specific qualification required; check all phase contents, salts, inhibitor basis and balances
Mineral scale (17) Empirical ScalePredictionCalculator, activity models and MultiMineralScaleEquilibrium; ScaleKspLiteratureBenchmarkTest Selected Ksp/solution benchmarks exist; distinguish screening, high-salinity activities and coupled quantitative precipitation

Reproduced swelling defect and correction

For an SRK oil containing 0.2 mol methane and 0.8 mol nC10, constructed at 298.15 K, the previous calculation reported 1.052108003210363 at zero gas addition when the requested experiment temperature was 353.15 K. The reference volume had been calculated before applying that temperature. It also rescaled the caller’s injection template and ignored the temperature unit at later stages. The two original regression tests failed before the implementation change.

The corrected calculation applies one temperature to the reference and every addition, uses density-corrected volumes, clones gas additions, conserves the original-oil mole basis, and starts each run from the constructor’s initial oil snapshot. Invalid schedules are rejected. A failed bubble-point calculation throws with stage/cause information; no completed result arrays are published. See issue #4561.

The independent reference generator uses Python’s standard library without importing NeqSim. It solves SRK cubic volume roots analytically, iterates incipient-vapor fugacity coefficients and bisects pressure to enforce equilibrium. It rejects equal-composition trivial roots and checks the pressure equation for both phases. This is an original synthetic numerical benchmark with frozen inputs, not measured petroleum data.

Input Frozen value
EOS / mixing / shifts Original Soave SRK alpha; classical quadratic/linear mixing; all kij = 0; volume translation disabled
Initial oil / injection 0.2 mol methane + 0.8 mol nC10 / pure methane
Temperature 353.15 K
Methane Tc / Pc / acentric factor 190.56 K / 45.99 bara / 0.0115
nC10 Tc / Pc / acentric factor 617.70 K / 21.1 bara / 0.4923
Gas constant 8.3144621 J/(mol K)
Addition basis Cumulative mol% of the original oil, not final-mixture mol%
Cumulative addition, mol% Independent bubble pressure, bara Independent volume ratio
0 51.2837097379 1.0000000000
10 73.2010992783 1.0301020972
20 93.1415075418 1.0602317416

SwellingTestTest requires agreement within 0.0001 bara and 0.000001 in volume ratio. The generator’s maximum absolute log-fugacity residual is below 2e-13; its maximum relative EOS-pressure residual is below 2e-14. The Java tests also cover Celsius/Kelvin/default temperature, repeat runs, schedule copying and duplicate stages, input preservation, invalid inputs and explicit saturation failure. Mixed methane/ethane injection tests check every component and total mass for both SRK and PR, including ethane absent from the initial oil.

A separate failing regression exposed omitted liquid-density volume shifts. For constant methane/nC10 shifts of 0.00001/0.00003 m3/mol, the independently shifted 20 mol% volume ratio is 1.0578891507452128; the former implementation still returned the unshifted 1.0602317415732487. The corrected API includes these shifts and the test also verifies that saturation pressures are unchanged.

These tolerances verify numerical agreement for identical model inputs. They are not universal accuracy tolerances for measured reservoir-fluid properties.

Open implementation work and existing dependencies

Work Tracking
Swelling temperature, reference volume and reusable input repair #4561
Critical transition and dew-branch swelling, stage CCE/dropout #4565
Conservative displacement, complete injection components and qualified MMP methods #4566
Unsupported regression experiments currently dispatch to zero #4567
Plus-fraction molecular-weight moment conservation #4549
Persistent per-component PR covolume overrides #4548
Named oil density/viscosity in CCE #4547, active PR #4556
Heavy-TBP validity and SRK alpha inversion PR #4540
Flash/stability and electrolyte qualification #2937, #3144

Closed wax/onset defects #3709, #3914 and #4083 are historical regression evidence, not newly open blockers. Retain their tests while adding measured-data coverage.

How to establish comparable results

Freeze the full model before comparing calculations: component properties, composition and normalization, heavy-fraction distribution/lumping, EOS and alpha function, mixing rule and BIPs, volume translation, transport model, enabled phases, units, standard temperature/pressure and experiment procedure. Otherwise an output difference may reflect different physical inputs or definitions.

For each public benchmark, record source and reuse permission, measured values and uncertainty, exact NeqSim revision, complete fluid inputs, untuned residuals, parameters and bounds used for fitting, and separate hold-out residuals. Fix property-specific tolerances before fitting. Use both absolute and relative errors, including absolute dropout/precipitation tolerances near zero. Validate component/total balances, equilibrium closure and nearby operating points.

Start with measured oil and condensate CCE/CVD/DL/separator data, then complete gas-injection and heavy-fraction dependencies. Extend the same traceable approach to caloric/transport data, depth profiles, aqueous partitioning and solid-phase equilibria. Only claim accuracy for the cases and ranges actually qualified.

References