DOE Big Hill Sweet complete modeled assay slate

DoeBigHillSweetAssay assembles the refinery campaign’s qualified public inputs into one reusable Java/JPype assay definition. The factory configures an OilAssayCharacterisation on a caller-selected mass basis and deliberately requires a separate apply() call before the thermodynamic system is mutated.

Primary public sources

DOE SPR Big Hill Sweet comprehensive assay, reported 24 September 2021;
DOE SPR Big Hill Sweet PIANO workbook.

The comprehensive workbook reports the following non-overlapping mass-basis cuts. The mass yields close to exactly 100.00% at the displayed precision.

Cut	Mass %	SG60/60	Sulfur mass %	Nitrogen mass %	Characterization
C2-C4 gas	1.70	—	0 assumed	0 assumed	DOE PIANO C2-C4 subset normalized over the whole gas slice
C5-175 degF	5.22	0.6731	0.0008	0 assumed	PIANO-derived 0.0791538366563 kg/mol; upper boundary only
175-250 degF	8.32	0.7432	0.0026	0 assumed	finite DOE cut boundaries
250-375 degF	12.55	0.7817	0.019	0 assumed	finite DOE cut boundaries
375-530 degF	16.19	0.8297	0.096	0.0018	finite DOE cut boundaries
530-650 degF	13.18	0.8604	0.313	0.0186	finite DOE cut boundaries
650-850 degF	18.44	0.9039	0.534	0.102	finite DOE cut boundaries
850-1050 degF	12.84	0.9336	0.752	0.234	finite DOE cut boundaries
1050 degF+	11.56	1.0089	1.334	0.501	lower boundary only; DOE Watson factor 11.7

DOE leaves sulfur blank for the gas cut and nitrogen blank through the 250-375 degF cut. The zeros above are the same explicit screening assumptions used by the independently qualified sulfur and nitrogen bookkeeping tests; they are not reported measurements.

Light-end allocation assumption

The PIANO debutanization table reports ethane 0.09, propane 10.38, i-butane 10.21, and n-butane 45.95 wt%, totaling the reported 66.63 wt% C2-C4 subset. The reference factory normalizes those four weights and allocates the full 1.70 mass% gas cut between them. This preserves the published gas-cut mass while supplying real NeqSim standard components, but it is not a measured complete gas analysis.

Usage

SystemInterface crude = new SystemSrkEos(298.15, 1.01325);
OilAssayCharacterisation assay = DoeBigHillSweetAssay.create(crude, 1.0); // kg

// Inspect mass fractions, sulfur/nitrogen, and boundaries before mutation.
double sulfurMassPercent = assay.getBulkSulfurMassPercent();
assay.apply();

The one-kilogram regression requires 12 positive components (four standard light molecules and eight petroleum pseudo-components), reconstructed component mass closure within 1e-10 kg, sulfur 0.40867518 mass%, and nitrogen 0.1095129 mass%. A second construction must produce the identical resolved mass vector.

650 degF+ vacuum-screening feed

DoeBigHillSweetAssay.createVacuumScreeningFeed(...) returns a three-cut heavy feed assembled only from the 650-850 degF, 850-1050 degF, and one-sided 1050 degF+ source rows. Those rows account for 42.84 mass% of whole crude. The factory preserves their relative masses and normalizes them to the screening-feed basis:

\[w_i^{screen}=\frac{w_i^{whole\ crude}}{42.84\%}\]

Source cut	Whole-crude mass %	Screening-feed mass %
650-850 degF	18.44	43.0438842204
850-1050 degF	12.84	29.9719887955
1050 degF+	11.56	26.9841269841

The source-weight accessor returns a defensive copy, while the factory preserves the source specific gravities, sulfur, nitrogen, finite cut boundaries, and the residue’s one-sided lower boundary and Watson factor. The normalized feed reconstructs 0.815211951447 sulfur mass% and 0.249229691877 nitrogen mass%.

This feed is a transparent characterization and numerical-screening basis. It is not a measured atmospheric-column bottoms stream: DOE does not report tower entrainment, cut-point overlap, light-tail carryover, operating pressure, or a matching atmospheric-column material balance. The factory therefore does not infer any of those quantities and does not perform pressure correction or ASTM D1160 conversion.

Evidence and validity boundary

This class is a source-specific reproducible reference composition. It does not mix the older 1998 five-cut table with the 2021 nine-cut workbook. It also does not validate the normalized gas allocation, generated critical properties, acentric factors, vapor-liquid equilibrium, atmospheric product yields, or conversion-unit performance. DOE publishes the workbook for information purposes with no warranty of accuracy or completeness; users remain responsible for its application.

The complete slate has separately passed atmospheric-column integration and the public Sarir case now provides the campaign’s operating-case fractionation evidence. The normalized 650 degF+ slice is the characterization handoff for a later low-pressure column benchmark; it does not itself validate vacuum fractionation or product yields.

Vacuum-column case handoff

DoeBigHillVacuumFractionationCase.create(...) turns the normalized screening feed into a composable Java/JPype process handoff. It applies the three public heavy cuts to an SRK system, sets the requested mass flow and feed state, and configures a partial-condenser, reboiler-equipped DistillationColumn with the MESH-residual solver.

Every operating value is explicit because the DOE assay does not report a matching refinery vacuum-column case:

OperatingInputs inputs = new OperatingInputs(
    12,    // simple trays
    4,     // bottom-up feed-tray index
    640.0, // feed temperature, K
    0.12,  // feed pressure, bara
    0.08,  // top pressure, bara
    0.16,  // bottom pressure, bara
    700.0, // reboiler outlet temperature, K
    0.5);  // condenser reflux ratio

DoeBigHillVacuumFractionationCase model =
    DoeBigHillVacuumFractionationCase.create("Big Hill vacuum screen", 1000.0, inputs);
DistillationColumn column = model.getColumn();

The factory requires the pressure topology

\[0<P_{top}\leq P_{feed}\leq P_{bottom}<1.01325\;\mathrm{bara}\]

with a strictly positive pressure rise from top to bottom, a valid internal feed tray, and a reboiler temperature above the feed temperature. Invalid, non-finite, atmospheric, or non-positive inputs fail before a case is created.

Factory creation runs only the feed flash and deliberately returns an unsolved column. The example values above are transparent engineering assumptions for API and low-pressure handoff qualification; they are not DOE measurements or recommended design conditions. A caller must run the column and independently qualify convergence, mass and energy closure, product ordering, operating sensitivity, and suitability of the chosen thermodynamic model.

This handoff does not identify the screening feed as measured atmospheric bottoms, perform ASTM D1160 or pressure correction, or claim vacuum-gas-oil/residue yields, product quality, equipment design, or plant agreement. Those remain separate solved-case and public-benchmark gates.

Solved screening-point result

After a caller explicitly runs the configured column, DoeBigHillVacuumFractionationResult.evaluate(model) provides a fail-closed Java/JPype summary:

DoeBigHillVacuumFractionationCase model =
    DoeBigHillVacuumFractionationCase.create("Big Hill vacuum screen", 1000.0, inputs);
model.getColumn().run();

DoeBigHillVacuumFractionationResult result =
    DoeBigHillVacuumFractionationResult.evaluate(model);
double overheadMassFraction = result.getProduct("Overhead").getMassFractionOfFeed();
double bottomsMassFraction = result.getProduct("Bottoms").getMassFractionOfFeed();

Evaluation accepts only a converged MESH-residual solve without failed or fallback products. The external mass closure, column mass balance, column energy balance, maximum tray material balance, final MESH residual, and every component molar balance must each satisfy the configured acceptance limit. Both overhead and bottoms must have positive material flow, and their mole-weighted mean normal boiling points must increase from overhead to bottoms. The result also records iteration count, solve time, and convergence diagnostics, and returns defensive product arrays.

The regression executes the explicit 12-tray point shown above twice from independently constructed cases. It requires the three exact DOE pseudo-component identities, no more than 5% mass, energy, tray, or component closure error, and product flow and mean-boiling-point repeatability within 1%.

These gates qualify numerical conservation and separation direction for one transparent engineering screening point. No public DOE measurement defines a matching vacuum-column product split, so the reported streams are deliberately labeled Overhead and Bottoms, not validated VGO or vacuum residue. The calculation does not establish pressure correction, ASTM D1160 behavior, equipment design, a general SRK accuracy envelope, or plant agreement.

Discrete product boiling-range diagnostics

Each calculated product also exposes the ascending normal-boiling-point support of its positive pseudo-components and the corresponding normalized cumulative product mole fractions:

ProductResult overhead = result.getProduct("Overhead");
double overheadT10Kelvin = overhead.getNormalBoilingPointQuantileKelvin(0.10);
double overheadT50Kelvin = overhead.getNormalBoilingPointQuantileKelvin(0.50);
double overheadT90Kelvin = overhead.getNormalBoilingPointQuantileKelvin(0.90);
double[] supportKelvin = overhead.getBoilingPointTemperaturesKelvin();
double[] cumulativeMoleFractions = overhead.getCumulativeMoleFractions();

For a requested cumulative mole fraction (q) in ((0,1]), the quantile is the first discrete pseudo-component normal boiling point whose normalized cumulative product mole fraction reaches (q). The support and cumulative arrays are defensive copies, are ordered by increasing normal boiling point, and close at a cumulative fraction of one. Invalid or non-finite quantile requests fail closed.

The focused regression requires T10 <= T50 <= T90 for both products, an overhead T50 below the bottoms T50, and 1% repeatability of all three diagnostics across independently constructed and solved cases. These temperatures make the modeled separation direction and broad boiling range inspectable on the exact three-pseudo-component DOE basis.

The curve is deliberately discrete and molar-basis. It is not a continuous simulated-distillation curve, a TBP curve, an ASTM D86 or ASTM D1160 result, or a pressure-corrected laboratory measurement. With only three heavy pseudo-components it must not be used to infer unreported cut tails, detailed product quality, or validated VGO/residue yields.

Absolute-pressure sensitivity screening

DoeBigHillVacuumPressureSensitivity.run(...) independently rebuilds, solves, and evaluates multiple Big Hill vacuum cases while scaling the feed, top, and bottom absolute pressures together. Tray count, feed tray, feed and reboiler temperatures, reflux ratio, feed composition, and feed mass flow remain fixed. Pressure factors must be finite, positive, unique, strictly increasing, and retain the qualified sub-atmospheric pressure topology.

The documented three-point screen uses factors close to the qualified base point:

OperatingInputs baseline =
    new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.5);
double[] pressureFactors = {0.98, 1.00, 1.02};

DoeBigHillVacuumPressureSensitivity sensitivity =
    DoeBigHillVacuumPressureSensitivity.run(
        "Big Hill vacuum pressure screen", 1000.0, baseline, pressureFactors);

for (DoeBigHillVacuumPressureSensitivity.PointResult point : sensitivity.getPoints()) {
  double factor = point.getPressureScaleFactor();
  double topPressureBara = point.getOperatingInputs().getTopPressureBara();
  double overheadMassFraction = point.getOverheadMassFraction();
  double overheadT50Kelvin = point.getOverheadBoilingPointQuantileKelvin(0.50);
}

Every point must pass the already qualified MESH-residual, fallback, mass, component, energy, material-product, and boiling-point-order gates. The summary returns a defensive point array, exact applied operating inputs, immutable per-point fractionation results, the observed overhead-yield bounds, and the worst external mass closure, component closure, column energy error, and final MESH residual. A failed point aborts the complete sensitivity instead of returning a partial envelope.

The factors scale absolute pressure, not vacuum gauge or pressure drop. This keeps the relative pressure profile fixed and isolates one numerical operating variable; it does not represent an optimized or vendor-recommended pressure profile. The narrow 0.98/1.00/1.02 regression is a convergence and conservation test around the documented screening point. It does not establish a measured pressure response, require a monotonic yield trend, define an uncertainty distribution, validate product quality, or demonstrate equipment turndown.

All DOE assay provenance and the three-cut 650 degF+ normalization remain unchanged. The calculation does not add ASTM D1160 or TBP pressure correction, measured vacuum-column data, fitted parameters, calibrated VGO/residue yields, equipment design, or plant-agreement evidence.

Condenser-reflux sensitivity screening

DoeBigHillVacuumRefluxSensitivity.run(...) independently rebuilds, solves, and evaluates multiple Big Hill vacuum cases while varying only the condenser reflux ratio. Tray count, feed tray, feed and reboiler temperatures, all three absolute pressures, feed composition, and feed mass flow remain fixed. Reflux ratios must be finite, non-negative, unique, and strictly increasing.

The documented three-point screen stays close to the qualified base point:

OperatingInputs baseline =
    new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.5);
double[] refluxRatios = {0.49, 0.50, 0.51};

DoeBigHillVacuumRefluxSensitivity sensitivity =
    DoeBigHillVacuumRefluxSensitivity.run(
        "Big Hill vacuum reflux screen", 1000.0, baseline, refluxRatios);

for (DoeBigHillVacuumRefluxSensitivity.PointResult point : sensitivity.getPoints()) {
  double refluxRatio = point.getCondenserRefluxRatio();
  double overheadMassFraction = point.getOverheadMassFraction();
  double overheadT50Kelvin = point.getOverheadBoilingPointQuantileKelvin(0.50);
}

The condenser reflux ratio is a dimensionless DistillationColumn input. This screen isolates that single numerical operating variable; it does not represent a measured, optimized, or vendor-recommended reflux policy. The narrow 0.49/0.50/0.51 regression is a convergence and conservation test around the documented screening point. It does not establish a measured reflux response, require a monotonic yield trend, quantify condenser or reboiler duty, define utilities, size equipment, validate product quality, or demonstrate turndown.

Reboiler-temperature sensitivity screening

DoeBigHillVacuumReboilerTemperatureSensitivity.run(...) independently rebuilds, solves, and evaluates multiple Big Hill vacuum cases while varying only the reboiler outlet temperature. Tray count, feed tray, feed temperature, all three absolute pressures, condenser reflux ratio, feed composition, and feed mass flow remain fixed. Temperatures must be finite, positive, unique, strictly increasing, and above the fixed feed temperature.

The documented three-point screen stays close to the qualified base point:

OperatingInputs baseline =
    new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.5);
double[] reboilerTemperaturesKelvin = {698.0, 700.0, 702.0};

DoeBigHillVacuumReboilerTemperatureSensitivity sensitivity =
    DoeBigHillVacuumReboilerTemperatureSensitivity.run(
        "Big Hill vacuum reboiler-temperature screen",
        1000.0,
        baseline,
        reboilerTemperaturesKelvin);

for (DoeBigHillVacuumReboilerTemperatureSensitivity.PointResult point :
    sensitivity.getPoints()) {
  double reboilerTemperatureKelvin = point.getReboilerTemperatureKelvin();
  double overheadMassFraction = point.getOverheadMassFraction();
  double overheadT50Kelvin = point.getOverheadBoilingPointQuantileKelvin(0.50);
}

The temperature is the specified reboiler outlet temperature, not a measured tray profile, boiling curve, heat duty, or utility demand. This screen isolates that single numerical operating variable; it does not represent a measured, optimized, or vendor-recommended temperature policy. The narrow 698/700/702 K regression is a convergence and conservation test around the documented screening point. It does not establish a measured temperature response, require a monotonic yield trend, quantify heat-transfer performance, size equipment, validate product quality, or demonstrate turndown.

Feed-temperature sensitivity screening

DoeBigHillVacuumFeedTemperatureSensitivity.run(...) independently rebuilds, solves, and evaluates multiple Big Hill vacuum cases while varying only the feed temperature. Tray count, feed tray, all three absolute pressures, reboiler outlet temperature, condenser reflux ratio, feed composition, and feed mass flow remain fixed. Temperatures must be finite, positive, unique, strictly increasing, and below the fixed reboiler temperature.

The documented three-point screen stays close to the qualified base point:

OperatingInputs baseline =
    new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.5);
double[] feedTemperaturesKelvin = {638.0, 640.0, 642.0};

DoeBigHillVacuumFeedTemperatureSensitivity sensitivity =
    DoeBigHillVacuumFeedTemperatureSensitivity.run(
        "Big Hill vacuum feed-temperature screen",
        1000.0,
        baseline,
        feedTemperaturesKelvin);

for (DoeBigHillVacuumFeedTemperatureSensitivity.PointResult point :
    sensitivity.getPoints()) {
  double feedTemperatureKelvin = point.getFeedTemperatureKelvin();
  double overheadMassFraction = point.getOverheadMassFraction();
  double overheadT50Kelvin = point.getOverheadBoilingPointQuantileKelvin(0.50);
}

The temperature is the specified feed-stream temperature at the fixed feed pressure, not a measured preheat-train profile, furnace outlet temperature, flash-zone temperature, or heat duty. This screen isolates that single numerical operating variable; it does not represent a measured, optimized, or vendor-recommended preheat policy. The narrow 638/640/642 K regression is a convergence and conservation test around the documented screening point. It does not establish a measured temperature response, require a monotonic yield trend, quantify furnace or exchanger performance, define heat integration, size equipment, validate product quality, or demonstrate turndown.

Feed-mass-flow sensitivity screening

DoeBigHillVacuumFeedMassFlowSensitivity.run(...) independently rebuilds, solves, and evaluates multiple Big Hill vacuum cases while varying only the feed mass flow. Tray count, feed tray, feed and reboiler temperatures, all three absolute pressures, condenser reflux ratio, and feed composition remain fixed. Mass flows must be finite, positive, unique, and strictly increasing.

The documented three-point screen stays close to the qualified 1000 kg/h base point:

OperatingInputs baseline =
    new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.5);
double[] feedMassFlowsKgPerHour = {980.0, 1000.0, 1020.0};

DoeBigHillVacuumFeedMassFlowSensitivity sensitivity =
    DoeBigHillVacuumFeedMassFlowSensitivity.run(
        "Big Hill vacuum feed-mass-flow screen",
        baseline,
        feedMassFlowsKgPerHour);

for (DoeBigHillVacuumFeedMassFlowSensitivity.PointResult point :
    sensitivity.getPoints()) {
  double feedMassFlowKgPerHour = point.getFeedMassFlowKgPerHour();
  double overheadMassFraction = point.getOverheadMassFraction();
  double overheadT50Kelvin = point.getOverheadBoilingPointQuantileKelvin(0.50);
}

Every point must pass the already qualified MESH-residual, fallback, mass, component, energy, material-product, and boiling-point-order gates. The summary returns a defensive point array, exact applied feed mass flows and operating inputs, immutable per-point fractionation results, the observed overhead-yield bounds, and the worst external mass closure, component closure, column energy error, and final MESH residual. A failed point aborts the complete sensitivity instead of returning a partial envelope.

The mass flow is the specified feed-stream throughput in kg/h, not a measured or design column capacity. This screen isolates that single numerical input while keeping feed composition and every column operating input fixed. The narrow 980/1000/1020 kg/h regression is a convergence and conservation test around the documented screening point. It does not establish a measured throughput response, require a monotonic yield trend, demonstrate scale-up, quantify flooding, weeping, entrainment, pressure drop, heat duty, or utilities, size equipment, validate product quality, or demonstrate turndown.

Combined operating-scenario screening

DoeBigHillVacuumScenarioScreen.run(...) independently rebuilds, solves, and evaluates complete caller-defined operating scenarios. This is the integration step after the qualified one-factor pressure, reflux, feed-temperature, reboiler-temperature, and feed-mass-flow screens. Scenario order is preserved; names must be unique, and every scenario supplies its own validated operating inputs and positive feed mass flow.

The documented low/base/high scenarios combine only the narrow ranges already exercised by those one-factor screens:

DoeBigHillVacuumScenarioScreen.Scenario[] scenarios = {
    new DoeBigHillVacuumScenarioScreen.Scenario(
        "low",
        980.0,
        new OperatingInputs(12, 4, 638.0, 0.1176, 0.0784, 0.1568, 698.0, 0.49)),
    new DoeBigHillVacuumScenarioScreen.Scenario(
        "base",
        1000.0,
        new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.50)),
    new DoeBigHillVacuumScenarioScreen.Scenario(
        "high",
        1020.0,
        new OperatingInputs(12, 4, 642.0, 0.1224, 0.0816, 0.1632, 702.0, 0.51))
};

DoeBigHillVacuumScenarioScreen screen =
    DoeBigHillVacuumScenarioScreen.run("Big Hill vacuum combined screen", scenarios);

for (DoeBigHillVacuumScenarioScreen.PointResult point : screen.getPoints()) {
  String scenarioName = point.getScenario().getName();
  double feedMassFlowKgPerHour = point.getScenario().getFeedMassFlowKgPerHour();
  double overheadMassFraction = point.getOverheadMassFraction();
  double overheadT50Kelvin = point.getOverheadBoilingPointQuantileKelvin(0.50);
  DoeBigHillVacuumComponentRecovery recovery = point.getComponentRecovery();
  double overheadHeavyRecovery =
      recovery.getProduct("Overhead").getComponentMolarRecovery("DOE_BH_1050_PLUS_PC");
}

All three absolute pressures move together by factors 0.98, 1.00, and 1.02 relative to the base 0.12/0.08/0.16 bara values. The feed and reboiler temperatures, condenser reflux ratio, and feed mass flow simultaneously use the low/base/high values shown above. Tray topology and feed composition remain fixed.

Every scenario must pass the qualified MESH-residual, fallback, mass, component, energy, material-product, and boiling-point-order gates. Each point also evaluates DoeBigHillVacuumComponentRecovery on the same solved model, without rebuilding or solving the scenario a second time. It exposes the exact DOE pseudo-component order, positive feed component flows in mol/h, product component flows in mol/h, and dimensionless overhead and bottoms recoveries. Every product recovery must remain finite and non-negative, and each component’s two product recoveries must close to unity within 5%.

The screen returns defensive scenario-point arrays, the exact immutable scenario definitions, fractionation results, and recovery results, overhead-yield bounds, and the worst external mass closure, component closure, component-recovery closure, column energy error, and final MESH residual. Any failed fractionation or recovery gate aborts the complete screen.

These three discrete calculations are numerical robustness and interaction-screening evidence only. The component recoveries are numerical pseudo-component partition bookkeeping, not measured or calibrated yields. They do not define a continuous or measured operating envelope, response surface, interaction correlation, probability distribution, contaminant distribution, or optimization model. No monotonic trend is required. The screen does not establish hydraulic capacity, flooding, weeping, entrainment, pressure drop, scale-up, turndown, heat duty, utilities, equipment sizing, ASTM D1160 or TBP pressure correction, product-specification compliance, or plant agreement. All DOE assay provenance, three-cut 650 degF+ normalization, pseudo-component properties, and source-unreported engineering-input limitations remain unchanged.

Pseudo-component recovery diagnostics

DoeBigHillVacuumComponentRecovery.evaluate(...) adds a molar component-partition audit to an already solved and qualified Big Hill vacuum case. It first reuses DoeBigHillVacuumFractionationResult.evaluate(...), so the MESH, fallback, mass, energy, material-product, boiling-range, and aggregate component-closure gates must pass before recovery evidence is returned.

For product $p$ and pseudo-component $c$, the dimensionless recovery is $R_{p,c} = \dot n_{p,c} / \dot n_{feed,c}$. The overhead and bottoms recoveries must sum to unity within the existing 5% screening tolerance for every component. Component order and exact NeqSim component names are preserved from the feed.

OperatingInputs inputs =
    new OperatingInputs(12, 4, 640.0, 0.12, 0.08, 0.16, 700.0, 0.5);
DoeBigHillVacuumFractionationCase model =
    DoeBigHillVacuumFractionationCase.create(
        "Big Hill vacuum component recovery", 1000.0, inputs);
model.getColumn().run(UUID.randomUUID());

DoeBigHillVacuumComponentRecovery recovery =
    DoeBigHillVacuumComponentRecovery.evaluate(model);
for (String componentName : recovery.getComponentNames()) {
  double overheadRecovery =
      recovery.getProduct("Overhead").getComponentMolarRecovery(componentName);
  double bottomsRecovery =
      recovery.getProduct("Bottoms").getComponentMolarRecovery(componentName);
}

The returned feed and product quantities use mol/h; recovery fractions are dimensionless. Name, flow, recovery, and product arrays are defensive. The result is numerical partition bookkeeping for the public DOE-derived three-cut synthetic screening feed. It is not measured cut recovery, calibrated yield, ASTM D1160/TBP or simulated-distillation evidence, a sulfur/nitrogen or contaminant split, a hydraulic-capacity result, an interaction model, optimization, product-specification compliance, or plant agreement. Feed provenance, pseudo-component properties, and every source-unreported operating assumption remain unchanged.