TwoFluidPipe Reporting and Validation

This page describes how to extract engineering results from neqsim.process.equipment.pipeline.TwoFluidPipe for long multiphase flowlines and how to compare those results with external simulator or field data.

Current reporting status

TwoFluidPipe exposes detailed profile and summary APIs. The convenience class neqsim.process.equipment.pipeline.twophasepipe.reporting.TwoFluidPipeReport builds standard CSV, text, JSON, event, and benchmark comparison outputs from those APIs.

The recommended production workflow is:

  1. Build and run the TwoFluidPipe model.
  2. Extract spatial profiles from the pipe, or call TwoFluidPipeReport helper methods.
  3. Add summary metrics and flow-assurance indicators.
  4. Export to CSV/JSON for plotting, review, or comparison.
  5. If available, compare against traceable experimental, field, or external-simulator data using the benchmark harness. Keep model-to-model checks distinct from experimental validation.

Steady-state profile reporting

After pipe.run(), the following methods provide one value per pipe section:

API Unit Description
getPositionProfile() m Section midpoint positions
getPressureProfile() Pa Pressure profile
getTemperatureProfile() K Fluid temperature profile
getTemperatureProfile("C") degC Fluid temperature profile in Celsius
getLiquidHoldupProfile() fraction Total liquid holdup
getWaterCutProfile() fraction Water fraction of liquid
getOilHoldupProfile() fraction Oil holdup
getWaterHoldupProfile() fraction Water holdup
getGasVelocityProfile() m/s Gas velocity
getLiquidVelocityProfile() m/s Liquid velocity
getOilVelocityProfile() m/s Oil velocity when oil-water slip is active
getWaterVelocityProfile() m/s Water velocity when oil-water slip is active
getOilWaterSlipProfile() m/s Oil velocity minus water velocity
getFlowRegimeProfile() enum Gas-liquid flow regime by section
getOilWaterFlowRegimeProfile() enum Oil-water flow configuration by section
getWaterWettingProfile() boolean Water-wetting indicator for corrosion screening
getWaterDropoutRiskProfile() boolean Water dropout / accumulation risk
getEntrainmentFractionProfile() fraction Estimated liquid entrainment in annular/mist flow
getEntrainedDropletDiameterProfile() m Characteristic entrained droplet diameter
getInclinedSectionGasCarryoverNumberProfile() dimensionless Local uphill liquid-carryover screen; not a system stability criterion
getInclinedSectionLiquidFallbackPotentialProfile() boolean Local fallback flag from the carryover screen
getSevereSluggingNumberProfile() dimensionless Deprecated alias for the local carryover screen
getSevereSlugPotentialProfile() boolean Result flag from the most recent explicit flowline-riser evaluation
getHeatTransferProfile() W/(m2 K) Heat-transfer coefficient profile, when configured
getSurfaceTemperatureProfile() K Ambient/surface temperature profile, when configured

Example Java extraction:

org.apache.logging.log4j.Logger logger =
    org.apache.logging.log4j.LogManager.getLogger("TwoFluidPipeReporting");
pipe.run();

double[] x = pipe.getPositionProfile();
double[] pressurePa = pipe.getPressureProfile();
double[] temperatureC = pipe.getTemperatureProfile("C");
double[] liquidHoldup = pipe.getLiquidHoldupProfile();
double[] waterCut = pipe.getWaterCutProfile();
double[] gasVelocity = pipe.getGasVelocityProfile();
double[] liquidVelocity = pipe.getLiquidVelocityProfile();
PipeSection.FlowRegime[] regimes = pipe.getFlowRegimeProfile();
double[] entrainment = pipe.getEntrainmentFractionProfile();
double[] gasCarryoverNumber = pipe.getInclinedSectionGasCarryoverNumberProfile();
boolean[] localFallback = pipe.getInclinedSectionLiquidFallbackPotentialProfile();
boolean[] waterWetting = pipe.getWaterWettingProfile();

for (int i = 0; i < x.length; i++) {
  double pBara = pressurePa[i] * 1.0e-5;
  logger.info("{},{},{},{},{},{},{},{}",
      x[i], pBara, temperatureC[i], liquidHoldup[i], waterCut[i],
      gasVelocity[i], liquidVelocity[i], regimes[i]);
}

The same steady-state results can be exported directly:

String profileCsv = TwoFluidPipeReport.toSteadyStateProfileCsv(pipe);
String summaryText = TwoFluidPipeReport.toSummaryText(pipe);
String summaryJson = TwoFluidPipeReport.toSummaryJson(pipe);
String eventsCsv = TwoFluidPipeReport.toSlugAndFlowAssuranceCsv(pipe);

Transient reporting

A successful runTransient(dt, id) completes the requested interval to floating-point precision. getTime(), getSimulationTime(), and the mass report’s elapsed time count only accepted substeps. setMaximumTransientSubsteps(...) limits attempted substeps; exhausting it throws instead of increasing the CFL step or reporting an incomplete interval as successful. An exception can leave an accepted partial interval: inspect the report and clock before resuming. Rejected hydrodynamic trials restore cell pressure, density, conservative state, configured oil-water closures and their last accepted diagnostics. Stiff source splitting must preserve an invalid raw trial for rejection before primitive recovery can repair phase masses. Raw negative phase mass or nonfinite state is rejected even when adaptive retry is disabled; in that mode the call fails at the last accepted state. A stable step below clock resolution reports that limitation separately from exhausting the attempt budget. The public integrator CFL helpers also retain the stable upper bound when it is below the preferred minimum step. This is substep recovery, not transactional rollback of an entire connected process. setCflNumber(...) honors finite values strictly between zero and one, including values below 0.1 used for refinement studies. Invalid values throw rather than silently selecting another CFL.

For transient cases, call runTransient(dt, id) repeatedly and store a snapshot at the reporting interval required by the study. Do not store every internal sub-step for long pipelines unless high-frequency pressure waves or slug arrivals are being investigated.

double dt = 1.0;
int reportEvery = 10;

for (int step = 0; step < 3600; step++) {
  pipe.runTransient(dt, UUID.randomUUID());

  if (step % reportEvery == 0) {
    double time = pipe.getSimulationTime();
    double[] x = pipe.getPositionProfile();
    double[] p = pipe.getPressureProfile();
    double[] hL = pipe.getLiquidHoldupProfile();
    // Write one row per position with this time stamp.
  }
}

Recommended transient CSV columns:

time_s,position_m,pressure_bara,temperature_C,liquid_holdup,water_cut,
gas_velocity_m_s,liquid_velocity_m_s,oil_velocity_m_s,water_velocity_m_s,
flow_regime,oil_water_flow_regime,water_wetting,water_dropout_risk,
entrainment_fraction,entrained_droplet_diameter_m,
inclined_section_gas_carryover_number,inclined_section_liquid_fallback_potential,
severe_slugging_number,severe_slug_potential

Thermal-energy validation

For a closed-boundary thermal runTransient(...), call getLastThermalEnergyBalanceReport() to validate the same post-step sensible-energy model that changed the fluid and wall temperatures. Its discrete balance is

\[\Delta E_f+\Delta E_w=E_{adv}+E_{JT}-E_{amb}\]

where positive $E_{adv}$ and $E_{JT}$ add energy and positive $E_{amb}$ removes energy. CLOSED external faces contribute zero sensible advection, but internal face transport remains active. In the multilayer model, the fluid and first wall layer use the same instantaneous heat rate; the reported ambient loss is the last-layer flux from the same explicit update.

pipe.setHeatTransferCoefficient(25.0); // W/(m2 K); the report is null when heat transfer is disabled
pipe.runTransient(0.001, UUID.randomUUID());
TwoFluidThermalEnergyBalanceReport thermal = pipe.getLastThermalEnergyBalanceReport();
boolean closes = thermal.isWithinTolerance(1.0e-5, 1.0e-10);

The report is null when heat transfer is disabled. It covers fluid sensible energy and simple-wall or radial-layer storage. Strict domain-level closure applies only when both external mass boundaries are CLOSED and phase transfer does not change material inventory. Open boundaries require explicit boundary-enthalpy terms; phase-changing cases additionally require compositional and latent-energy terms. In those cases this report is an internal post-step temperature-model diagnostic, not a full domain energy audit. For a closed cooldown, also verify zero boundary mass/enthalpy transport, monotonic all-cell cooling without ambient undershoot, repeatability, serialization/copy behavior, both explicit and IMEX paths, and mesh/time-step refinement.

Phase-transfer validation

When setIncludeMassTransfer(true) is enabled, validate gas, oil, and water separately rather than checking total mass alone. Condensation is assigned from the equilibrium oil/aqueous mass split, whereas evaporation is withdrawn from the actual donor inventories. The transfer-only requirements are:

\[\Gamma_G+\Gamma_O+\Gamma_W=0\]

and, using donor velocity for transferred momentum,

\[S_{p,G}+S_{p,O}+S_{p,W}=0\]

Use TwoFluidMassBalanceReport to check GAS, OIL, WATER, LIQUID, and TOTAL. A useful phase-transition test starts from a cell with no liquid seed, crosses the SRK/CPA dew point in both directions, and sweeps at least three nearby temperatures on each side. Report the EOS, mixing rule, composition, absolute pressure, temperature, relaxation time, time step, mesh, and phase inventories. Repeat the run to verify deterministic behavior and compare a refined time step and mesh. Serialize a condensed-state copy and require the original and copy to follow the same reheating trajectory. As a negative control, repeat the cooldown with setIncludeMassTransfer(false) and require every phase source and inventory change to remain zero even though the temperature crosses the dew point.

For an aqueous-first transition, the first condensation source must be water even though the gas-only hydrodynamic water cut defaults to zero. For an oil-first transition, the water source must remain zero. In a gas + oil + aqueous flash, the reported equilibrium liquid mass fractions must both be included and sum to one. FlashTable and rigorous-flash runs should give the same phase identity; use sufficiently fine tables near phase boundaries.

The default phase-resolved closure tracks bulk phase inventories. Optional component transport, enabled with setComponentTransportEnabled(true) before initialization, also carries named component inventories by phase and cell and exposes component-conservation reports. Its boundary and phase-transfer limitations must be checked separately. Neither total-mass closure nor component closure establishes commercial-simulator equivalence or experimental transient accuracy.

The severe_slugging_number header is retained as a deprecated duplicate for CSV compatibility. It contains the local inclined-section gas-carryover number, not the explicit system stability result. New consumers should use inclined_section_gas_carryover_number. Call evaluateSevereSluggingSystem(...) before exporting if severe_slug_potential should contain a system-level classification.

Summary metrics

Use these methods for an executive summary or design report:

API Description
getInletPressure() Inlet pressure in bara
getOutletPressure() Outlet pressure in bara
getAverageLiquidHoldup() Volume-weighted average liquid holdup
getDominantFlowRegime() Most frequent flow regime
getAverageSuperficialGasVelocity() Average superficial gas velocity
getAverageSuperficialLiquidVelocity() Average superficial liquid velocity
getAverageMixtureDensity() Volume-weighted mixture density
getMaxMixtureVelocity() Maximum mixture velocity
getErosionalVelocity() API 14E erosional velocity
getErosionalVelocityMargin(double) Maximum velocity divided by erosional velocity
getFlowAnalysisSummary() Mid-pipe dimensionless flow summary
getThermalSummary() Thermal model summary
getSlugStatisticsSummary() Slug-tracking summary
getHydrateRiskSections() Sections below configured hydrate temperature
getWaxRiskSections() Sections below configured wax appearance temperature

Closure diagnostics

The two-fluid closure pass updates additional diagnostics that are useful for model review and validation. Each value is available both on TwoFluidSection and as a top-level TwoFluidPipe profile:

Section API Pipe profile API Description
getOilWaterFlowRegime() getOilWaterFlowRegimeProfile() Oil-water flow configuration
isWaterWetting() getWaterWettingProfile() Water-wetting indicator for corrosion screening
isWaterDropoutRisk() getWaterDropoutRiskProfile() Water dropout / accumulation risk
getEntrainmentFraction() getEntrainmentFractionProfile() Estimated liquid entrainment fraction
getEntrainedDropletDiameter() getEntrainedDropletDiameterProfile() Entrained droplet diameter
getInclinedSectionGasCarryoverNumber() getInclinedSectionGasCarryoverNumberProfile() Local uphill liquid-carryover screen
isInclinedSectionLiquidFallbackPotential() getInclinedSectionLiquidFallbackPotentialProfile() Local fallback flag
getSevereSluggingNumber() getSevereSluggingNumberProfile() Deprecated aliases for the same local screen
isSevereSlugPotential() getSevereSlugPotentialProfile() Last explicit flowline-riser system result

The steady-state and transient profile CSV exporters include these diagnostics. Boolean values are written as true or false; a missing oil-water regime is written as an empty field. The system-result profile is meaningful only after evaluateSevereSluggingSystem(...) and is cleared when the next transient step changes the solved state.

Benchmark comparison format

The validation harness reads external simulator or field data in this CSV format:

case,time_s,position_m,variable,value,abs_tolerance,rel_tolerance,source

Supported captured variables include:

pressure_pa
pressure_bara
temperature_k
liquid_holdup
water_cut
oil_holdup
water_holdup
gas_velocity_m_s
liquid_velocity_m_s
oil_velocity_m_s
water_velocity_m_s
entrainment_fraction
entrained_droplet_diameter_m
inclined_section_gas_carryover_number
inclined_section_liquid_fallback_flag
severe_slugging_number
water_wetting_flag
water_dropout_risk_flag
severe_slug_potential_flag

severe_slugging_number is the deprecated benchmark key for inclined_section_gas_carryover_number; it is retained only for existing comparison files.

Example use:

Path reference = java.nio.file.Paths.get("reference_export.csv");
List<TwoFluidBenchmarkHarness.BenchmarkPoint> points =
    TwoFluidBenchmarkHarness.readCsv(reference);

TwoFluidBenchmarkHarness.Snapshot snapshot = TwoFluidBenchmarkHarness.capture(pipe);
TwoFluidBenchmarkHarness.Comparison comparison =
    TwoFluidBenchmarkHarness.compare(snapshot, points);

if (!comparison.isPassed()) {
  throw new AssertionError(comparison.failureSummary());
}

For transient comparisons, capture and pass a list of snapshots. The harness uses linear interpolation in time and position for continuous profiles. Variables ending in _flag use nearest-neighbour sampling so boolean diagnostics remain 0.0 or 1.0. Intervals containing a non-finite diagnostic sentinel also use the nearest endpoint instead of producing NaN. Comparison results can be exported as CSV:

String comparisonCsv = TwoFluidPipeReport.toComparisonCsv(comparison);

Reporting recommendations for long flowlines

For long oil and gas flowlines, report at least:

Public severe-slugging evidence

SevereSluggingBenchmarkHarness reads the digitized public Tengesdal (2002) -3-degree velocity map and reports a confusion matrix without forcing transition observations into a binary class. Across 26 severe and 15 stable observations, the current Taitel system screen has 22 true positives, 4 false negatives, 8 false positives, and 7 true negatives (70.7% accuracy). The 14 transition observations are reported separately as 6 predicted severe and 8 predicted stable.

SevereSluggingExperimentalBenchmarkTest contains the large-facility Test 3 geometry, but is currently disabled at class level: the existing transient routes do not produce a valid unclamped solution (issues #2909 and #2911). The previously reported ensemble numbers are historical exploratory output, not current passing validation. Sensitivity to tiny perturbations alone does not establish physical chaos; numerical instability must first be excluded through mesh and timestep studies. The active flow-map screen above does not time-march the pipe.

The opt-in coupled-pressure progress and long-horizon tests check numerical progress, residuals and boundedness. They do not establish measured severe-slugging period, amplitude or outlet slug distributions. Liquid-rich fixed-point and pressure-runaway acceptance tests also remain disabled. The 600-second liquid-outlet/inventory acceptance case has been re-enabled after phase-force corrections. This is a narrower result than steady-to-transient consistency: the separate 1800-second fixed-point test still exceeds its original 5% inventory-drift limit.

The coupled solver applies donor-bounded phase transfers and a common damping factor for pressure corrections. Its positive numerical pressure bound is configurable in Pa and defaults to 1 Pa; an atmospheric outlet is not an interior lower bound. The standalone solver retains the affine sound-speed-based density response for compatibility. The coupled pipe defaults to a local polytropic gas response, anchored in the supplied pressure, density and sound speed. It also enables implicit face momentum interpolation to remove alternating cell-pressure modes. An inadmissible correction returns nonconvergence; the same applied pressure increment must drive density, mass flux and momentum. Convergence requires both cell-volume closure and the fixed outlet-pressure residual to meet the configured relative tolerance. Reset and failed full steps clear previous correction results and phase-transfer ledgers. Coarse and refined Tengesdal progress cases now complete their five-second intervals under the original 24-iteration, 1e-7 volume and 1e-9 mass-residual gates. Their assertions remain active and unchanged. This does not qualify measured severe-slugging period or amplitude. General phase-transition dynamics require a qualified thermodynamic pressure closure; the local polytropic approximation is not a full real-fluid energy/EOS solve.

The coupled IMEX predictor uses centered pressure flux to avoid retaining AUSM’s explicit acoustic velocity-diffusion timestep restriction. Standalone AUSM and the legacy pipe path keep their original split. Analytical pressure-mode damping and a constant-gradient interpolation check do not establish general nonuniform-grid gravity well-balancing or shock-capturing accuracy.

The experimental source is Tengesdal’s 2002 BSEE report.

Dynamic qualification scope

Numerical regression coverage includes phase-identity preservation at water-cut endpoints, separate phase momenta, a stationary closed three-phase pressure check, accepted-time accounting, slug merge inventory, and tracker geometry at cell midpoints. These checks address specific code defects; they do not qualify a general operating envelope.

The legacy SIMPLIFIED and LAGRANGIAN modes remain empirical overlays. Their diagnostic inventory accounting is separate from the Eulerian phase balance: removing a tracked slug does not inject liquid into cells from which it was never withdrawn.

setSlugTrackingMode(TwoFluidPipe.SlugTrackingMode.CONSERVATIVE_LAGRANGIAN) selects predictive liquid-continuity interface motion and conservative subcell slug/film reconstruction. It also enables coupled pressure, interfacial pressure and adaptive stepping. For each partly occupied cell, SlugFilmCoupling constructs body and film states whose weighted average equals all seven stored conservative variables. Only the existing shared phase face flux moves physical inventory. The tracker reads the accepted film state; geometry feeds the next face reconstruction. Oil and water use independent donors when slip is enabled. Available phase mass, momentum and sensible energy bound reconstruction; a marker cannot manufacture a high-holdup liquid body. Closed boundaries constrain the external flux explicitly, and rejected steps cannot advance markers. The timestep resolves both reconstructed velocities and phase-inventory draining: a body face can have much greater liquid holdup than the cell average. Pressure-correction phase transfers also enter accepted component and sensible-temperature face ledgers.

In this mode the Taylor-bubble closure supplies tail velocity and liquid continuity determines front velocity and pickup/shedding. Empirical initiation and bubble-velocity closures remain. Tracked phase mass/momentum/energy fields describe a partition of existing Eulerian inventory; do not add them to pipe inventory. Legacy geometric outlet volumes remain marker diagnostics; actual delivery is the accepted phase outlet-flux integral in TwoFluidMassBalanceReport. The jump closure uses total liquid continuity; independently predictive selective oil/water pickup still requires separate qualified closures.

Conservative interface velocities are evaluated for all markers before any marker moves, so they use a common accepted geometry. Complete reverse exits at an open inlet remove markers and are counted separately by getTotalSlugsExitedAtInlet() and getTotalMassExitedAtInlet(). These are overlay diagnostics; they do not add or remove Eulerian phase inventory and do not change downstream outlet-arrival statistics.

Register a fixed-position observation after initialization with SlugProbe probe = pipe.getLagrangianSlugTracker().addProbe(positionMetres). probe.drainEvents() returns accepted front/tail passages with crossing time, direction, speed and connected liquid-union length. Crossings are collected before merge/exit handling; endpoints inside overlapping bodies are suppressed. Instantaneous birth and merge geometry do not invent passages. The history is bounded, so inspect getDroppedEventCount() before treating it as complete. This measures tracked marker geometry, not an independent Eulerian holdup detector.

MohmmedSlugFlowBenchmark reads all 75 tabulated observations from the CC BY 4.0 Mohmmed et al. experimental dataset: 15 translation speeds, 30 lengths and 30 frequencies. The fixture records original spreadsheet cells, missing pointwise uncertainty, the conflicting upstream 54D/58D labels, and unspecified speed stations. Comparison targets are fixed engineering tolerances, not measurement uncertainties. The separate benchmark-tagged marker experiment advances geometry in prescribed flow and tests the tail/translation closure under an explicit source-definition assumption. It does not validate pipe pressure, film thickness, slug generation or independently measured three-phase dynamics.

MohmmedTwoFluidPipeExperimentalBenchmarkTest runs all 15 velocity operating points through the actual coupled pipe and compares 42 observations: 15 speeds, 13 lengths and 14 frequencies. The downstream probe is at 81D; upstream 54D/58D observations and three unmatched downstream operating points are explicitly outside the simulated grid. The default 40-cell calculation uses a five-second warm-up and ten-second observation window. -Dneqsim.mohmmed.cells and -Dneqsim.mohmmed.outerStep are independent so a study can hold one axis fixed. -Dneqsim.mohmmed.pointIds accepts exact comma-separated translation-velocity source-cell IDs for a declared subset; every matching downstream 81D length/frequency observation is then retained automatically. The result reports signed error, absolute error, mean absolute relative error, RMS relative error, maximum error, missing predictions, conservation and boundary diagnostics. The source has no pointwise uncertainty, so none of these errors is an uncertainty-normalized score. Keep physical inputs, duration, random seed and targets fixed. Run deliberately with -DexcludedTestGroups= -Dtest=MohmmedTwoFluidPipeExperimentalBenchmarkTest. The log retains every failed/missing prediction, probe event and boundary limitation. A numerically complete calculation is not necessarily a boundary-qualified experimental match.

An independent-axis screen on 2026-09-07 selected three low/middle/high public points (mmc2_sheet1_F5, mmc2_sheet1_F21, and mmc2_sheet1_F37), giving nine exact measured comparisons per run. The source-cell IDs correspond to $(j_G,j_L)$ = (1.047, 0.7), (2.094, 0.86), and (2.792, 1.0) m/s. All runs used the same 15 s duration, 5 s warm-up, seed 20260905 and fixed gates:

Cells Outer step (s) Change from baseline Passed MARE RMSRE Maximum absolute relative error
40 0.050 baseline 3/9 1.0402 1.5291 3.2147
80 0.050 spatial only 2/9 0.9970 1.3648 2.8869
40 0.025 outer-step only 1/9 1.6195 2.3100 5.3748
80 0.025 combined interaction 2/9 2.4082 3.8349 9.0574

All nine predictions were finite at each level. At baseline, all three cases completed with a maximum accepted-step phase-mass residual of $1.65\times10^{-14}$ kg, but all three steady initializations were unconverged and pressure-floor limited, and all three transient cases used outlet-backflow clamping. Neither the error aggregates nor individual predictions approach a stable refined limit. This is explicit failed experimental and numerical qualification, not a model validation claim; it also supplies a bounded public reproducer for the separate gas-oil-water/convergence stage.

The earlier 2026-09-05 joint refinement study met 11/42, 3/42 and 1/42 fixed targets at 40, 80 and 160 cells respectively; 15/15, 14/15 and 13/15 cases completed the 15-second interval. All cases reported pressure-floor-limited steady initialization and outlet phase-backflow clamping. No length target passed at any level. The study therefore failed experimental qualification and did not establish convergence. The prescribed-flow marker experiment met 14/15 speed targets, but that narrower result does not qualify the coupled pipe. These outcomes use one seed and joint spatial/outer-step refinement; they do not isolate spatial or temporal error. After the subsequent final code review, the unchanged 40-cell screen completed all 15 cases and met 5/42 targets (2/15 speeds, 0/13 lengths, 3/14 frequencies), with the same reported boundary limitations. The controlled marker result remained 14/15. The review fixes therefore do not demonstrate improved measured accuracy; the 80- and 160-cell levels have not been repeated on that revision.

The seven-equation explicit total-energy flux includes kinetic enthalpy, signed gravitational work and wall heat. However, the coupled pressure corrector changes mass/momenta without a fully coupled local enthalpy/pressure-work redistribution, and the separate sensible-temperature update is not an inversion of conservative total energy. Exact reconstructed or closed global energy budgets therefore do not establish general nonisothermal acoustic qualification.

Simplified outlet statistics count the first front crossing at the physical pipe end, with the timestamp rounded to the accepted substep end. Lagrangian outlet statistics record completed tail exits; their reported maximum length and arrival time refer only to those exit events. These are different event definitions and should not be mixed in slug-frequency or arrival comparisons.

Before claiming dynamic qualification, require independent pressure, holdup, phase-flow and slug data for single-phase transients, stratified flow, hydrodynamic and terrain slugging, shutdown and restart, and three-phase water-cut changes. Report measured uncertainty and at least three mesh and timestep levels, together with every failed or disabled acceptance gate. An OLGA API name or a synthetic benchmark fixture is not evidence of parity with a commercial simulator.

Gaps and planned improvements

The current API is adequate for engineering studies and benchmark development. A polished industrial report workflow should still add: