Sarir atmospheric validation reference
SarirAtmosphericReference exposes a public refinery case to Java and Python/JPype callers while preserving the distinction between measured evidence, simulated results, and missing source data.
Source and license
The source is Hamza E. Omran Almansouri, Simulation of Sarir Crude Oil Refinery Using Aspen HYSYS, Journal of Engineering Research (Libya), issue 33, pages 51-64, published 31 March 2022, DOI 10.66411/jer.v33i.46. The open-access article is licensed CC BY 4.0.
The paper reports a Sarir crude density of 841.5 kg/m3 at 15 degC, API gravity 36.5 at 60 degF, sulfur 0.120 mass%, average molar mass 0.2447 kg/mol, and a public TBP assay originally prepared by the Libyan Petroleum Institute.
Whole-crude property evidence
Table 1 gives the following numeric bulk measurements:
| Property | Published value | Basis |
|---|---|---|
| Density | 841.5 kg/m3 | 15 degC |
| API gravity | 36.5 | 60 degF |
| Total sulfur | 0.120 mass% | Whole crude |
| Asphaltenes | 0.20 mass% | Whole crude |
| Mercaptan sulfur | 8 ppm by mass | Whole crude |
| Water and sediment | 0.05 volume% | Whole crude |
| Cloud point | 48.7-49.6 degC | Published interval |
| Pour point | +21 degC | Whole crude |
| Kinematic viscosity | 10.63 cSt | 100 degF; the source prose gives the rounded equivalent 37.7 degC |
| Average molar mass | 0.2447 kg/mol | Whole crude |
The Java API keeps the cloud point as separate lower and upper endpoints. It also exposes both published viscosity reference-temperature forms: 100 degF from Table 1 and 37.7 degC from the source prose. An exact unit conversion gives 37.777777… degC; the API does not silently replace the reported rounded Celsius value.
These are whole-crude reference measurements. The source does not publish uncertainty bounds or per-cut allocations for these properties, so they are not distributed across pseudo-components and do not qualify a wax, asphaltene, viscosity, water, or sulfur prediction model.
Preserved TBP evidence
| TBP temperature (degC) | 70 | 90 | 110 | 150 | 195 | 215 | 255 | 275 | 295 | 335 | 370 | 400 | 460 | 480 | 500 | 520 | 550 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Cumulative liquid volume (%) | 7.44 | 10.47 | 13.83 | 21.16 | 28.52 | 31.54 | 38.03 | 41.76 | 44.68 | 51.97 | 59.19 | 63.50 | 72.52 | 75.61 | 78.66 | 81.05 | 83.70 |
The source then reports a 550 degC+ terminal residue reaching 100 volume%. NeqSim retains 550 degC as a one-sided lower boundary and computes the implied 16.30 volume% residue; it does not invent a finite 100% endpoint or an upper boiling limit. The source marks light-end hydrocarbons as not determined, so no light-end composition is synthesized.
Product specifications
Table 2 reports ASTM D86 temperatures at 95 liquid volume percent as product specifications:
| Product | Test basis | Published specification |
|---|---|---|
| Light Naphtha | ASTM D86 at 95 volume% | 90 degC |
| Heavy Naphtha | ASTM D86 at 95 volume% | 160 degC |
| Kerosene | ASTM D86 at 95 volume% | 221 degC |
| Diesel | ASTM D86 at 95 volume% | 327 degC |
| Residual | ASTM D86 at 95 volume% | <550+ |
The Java API keeps these five specification rows separate from the Table 5 laboratory/HYSYS results. In particular, the 327 degC diesel specification is not replaced by the 346 degC laboratory result or the 339 degC HYSYS result. The residual value is retained as the source’s nonnumeric, open-ended boundary; requesting a numeric specification temperature for that row fails closed. These rows are source design/reference criteria, not independently reproduced measurements and not evidence that NeqSim meets the specifications.
Atmospheric operating case
The published HYSYS case uses 34 valve trays and feeds crude to tray 31 counted from the top. Table 3 publishes the complete ADU inlet/outlet stream set:
| Source label | Direction | Temperature (degC) | Pressure (kPa) | Mass flow (kg/h) |
|---|---|---|---|---|
| Crude oil tower | Inlet | 350 | 233 | 54,420 |
| Steam | Inlet | 150 | 476 | 340.2 |
| Kerosene steam | Inlet | 150 | 476 | 68.04 |
| Diesel steam | Inlet | 150 | 476 | 226.8 |
| Gas To Flare | Outlet | 49 | 140 | 6.985e-6 |
| Naphtha | Outlet | 49 | 140 | 8,706 |
| Kerosene product | Outlet | 126.3 | 210 | 952.2 |
| Diesel product | Outlet | 214.8 | 219.1 | 17,709.24 |
| Residual | Outlet | 341.9 | 230 | 26,937.99 |
| Water draw | Outlet | 49 | 140 | 745.5 |
The four published inlet flows sum to 55,055.04 kg/h and the six outlet flows sum to
55,050.930006985 kg/h. Their 4.109993015 kg/h difference is less than 1e-4 of the inlet flow.
This is a transcription and source-rounding check, not an independent plant conservation claim.
The three steam rows total 635.04 kg/h. Table 3 explicitly reports 150 degC and 476 kPa for each steam service, correcting the earlier incomplete guide statement. The source does not report steam quality, enthalpy, or injection tray locations, so temperature and pressure alone are not treated as a complete thermodynamic state. The API exposes the rows as evidence and does not create executable water streams or side-stripper topology.
Table 4 gives the complete numeric pump-around rows:
| Source label | Draw tray | Return tray | Flow (kg/h) | Draw temperature (degC) | Return temperature (degC) | Derived drop (K) |
|---|---|---|---|---|---|---|
| Top pump around (TPA) | 3 | 1 | 29,777.64 | 143.9 | 80.99 | 62.91 |
| Bottom pump around (BPA) | 22 | 19 | 60,423.66 | 232.4 | 173.99 | 58.41 |
The temperature drops are direct differences of the published draw and return temperatures. Table 4 does not explicitly state whether its tray labels are counted from the top or bottom. The Java API therefore exposes them as raw source tray numbers and reports the numbering basis as unresolved. They are not automatically converted to NeqSim’s bottom-up indices or used to configure a column.
Four products have numeric laboratory and simulated ASTM D86 T5/T95 evidence:
| Product | Lab T5 (degC) | Lab T95 (degC) | HYSYS T5 (degC) | HYSYS T95 (degC) |
|---|---|---|---|---|
| Light naphtha | 42 | 90 | -9 | 97 |
| Heavy naphtha | 96 | 160 | 83 | 153 |
| Kerosene | 185 | 221 | 159 | 214 |
| Diesel | 262 | 346 | 235 | 339 |
The residual laboratory curve is explicitly excluded from the numeric API because the paper states that it was unavailable and presents a non-numeric 550+ limit.
Strict NeqSim T95 comparison
A calculated product standard can be compared with the three published Sarir T95 references through the strict 95 liquid-volume-percent Riazi-Daubert path:
Standard_ASTM_D86 d86 = new Standard_ASTM_D86(productFluid);
d86.calculate();
SarirD86ProductComparison.Result comparison =
SarirD86ProductComparison.compareT95(d86, "Diesel");
double neqsimT95C = comparison.getNeqsimT95Celsius();
double laboratoryErrorPercent =
comparison.getNeqsimAbsoluteRelativeErrorPercent();
double specificationMarginC = comparison.getSpecificationMarginCelsius();
The workflow delegates the NeqSim value to
Standard_ASTM_D86.getQualifiedD86Temperature(95.0). It retains the laboratory, HYSYS, and
numeric specification values as read-only comparison evidence. Neither the source HYSYS value nor
the specification is a tuning target, and no laboratory-agreement threshold is imposed.
Sarir T5 values are not converted because 5 volume% is outside the qualified Riazi-Daubert
recovery set. The residual 550+ specification also remains nonnumeric and fails closed. A
specification margin is only arithmetic against the published reference; it is not an ASTM
procedure or compliance determination.
Independent product-rate evidence
| Product | Plant (metric t/day) | HYSYS (metric t/day) | Absolute error (%) |
|---|---|---|---|
| Total naphtha | 208.95 | 208.2 | 0.359 |
| Kerosene | 22.85 | 20.0 | 12.473 |
| Diesel | 425.018 | 393.0 | 7.533 |
| Residual | 646.5 | 706.1 | 9.219 |
Unlike the Al-Diwiniya reference, the Sarir paper does not say these HYSYS rates were imposed. It describes calculated production rates compared with actual refinery data, so the rows are retained as independent validation targets. The API recomputes absolute relative errors from the raw published values instead of copying rounded percentages.
The conversion to kg/h uses exactly 1000 kg per metric tonne and 24 hours per day. Each Table 5 plant row is also bound to its matching Table 3 ADU hydrocarbon outlet:
| Product | Table 3 outlet | Plant (kg/h) | Table 3 (kg/h) |
|---|---|---|---|
| Total naphtha | Naphtha | 8706.25 | 8706.00 |
| Kerosene | Kerosene product | 952.083333 | 952.20 |
| Diesel | Diesel product | 17709.083333 | 17709.24 |
| Residual | Residual | 26937.50 | 26937.99 |
| Total | 54304.916667 | 54305.43 |
The absolute difference between those independently published totals is 0.513333 kg/h, less than (10^{-5}) of the plant total. This is a transcription and unit-basis reconciliation between source tables, not proof of plant mass conservation or model agreement. The gas-to-flare, water, and steam rows are therefore excluded from this hydrocarbon-product comparison.
Java and Python access
double[] volumePercent = SarirAtmosphericReference.getTbpCumulativeVolumePercent();
double residuePercent = SarirAtmosphericReference.getTerminalResidueVolumePercent();
double cloudPointLowerCelsius = SarirAtmosphericReference.getCrudeCloudPointLowerCelsius();
double cloudPointUpperCelsius = SarirAtmosphericReference.getCrudeCloudPointUpperCelsius();
double viscosityCst = SarirAtmosphericReference.getCrudeKinematicViscosityAt100FCst();
SarirAtmosphericReference.AduStreamReference crudeFeed =
SarirAtmosphericReference.getAduStream("Crude oil tower");
double feedRateKgPerHour = crudeFeed.getMassFlowRateKgPerHour();
double tableClosure =
SarirAtmosphericReference.calculatePublishedAduMassBalanceErrorFraction();
SarirAtmosphericReference.ProductSpecificationReference dieselSpecification =
SarirAtmosphericReference.getProductSpecification("Diesel");
double dieselSpecificationCelsius =
dieselSpecification.getSpecificationTemperatureCelsius();
SarirAtmosphericReference.ProductSpecificationReference residualSpecification =
SarirAtmosphericReference.getProductSpecification("Residual");
boolean residualSpecificationIsNumeric =
residualSpecification.hasNumericSpecificationTemperature();
SarirAtmosphericReference.ProductYieldReference diesel =
SarirAtmosphericReference.getProductYield("Diesel");
double plantRateKgPerHour = diesel.getPlantMassFlowRateKgPerHour();
double hysysRateKgPerHour = diesel.getSimulationMassFlowRateKgPerHour();
String tableThreeOutlet = diesel.getAduStreamName();
double calculatedErrorPercent =
diesel.calculateAbsoluteRelativeErrorPercentForMassFlowKgPerHour(
calculatedDieselRateKgPerHour);
SarirAtmosphericReference.PumparoundReference topPumparound =
SarirAtmosphericReference.getPumparound("Top pump around (TPA)");
double sourceFlowKgPerHour = topPumparound.getMassFlowRateKgPerHour();
int rawSourceDrawTray = topPumparound.getSourceDrawTrayNumber();
boolean trayBasisIsExplicit =
SarirAtmosphericReference.hasExplicitPumparoundTrayNumberingBasis();
SarirAtmosphericReference.SteamInjectionReference keroseneSteam =
SarirAtmosphericReference.getSteamInjection("Kerosene side stripper");
double steamRateKgPerHour = keroseneSteam.getMassFlowRateKgPerHour();
double totalSteamRateKgPerHour = SarirAtmosphericReference.getTotalSteamRateKgPerHour();
boolean steamTemperatureAndPressureAreExplicit =
SarirAtmosphericReference.hasExplicitSteamTemperatureAndPressure();
boolean steamQualityIsExplicit = SarirAtmosphericReference.hasExplicitSteamQuality();
boolean steamStateIsExplicit =
SarirAtmosphericReference.hasExplicitSteamThermodynamicState();
Static methods are directly accessible through JPype. Arrays are defensive copies, and unknown product labels, negative calculated rates, and non-finite error inputs fail closed. A zero calculated rate remains a valid comparison value. Plant and HYSYS rates are read-only evidence: the comparison API neither changes a stream nor imposes an accuracy acceptance threshold.
Constrained pseudo-component input
SarirAtmosphericAssay converts the published cumulative TBP coordinates into 18 liquid-volume
cuts: a 7.44 vol% 70 degC- cut, 16 bounded intervals, and the 16.30 vol% 550 degC+
residue. The first and last cuts remain one-sided; the factory does not invent a numeric
initial boiling point, residue endpoint, light-end composition, or per-cut property.
The article reports only whole-crude density and average molar mass. Callers must therefore supply 18 cut specific gravities and 18 cut molar masses. Before changing the attached assay, the factory requires the volume-weighted density profile to reproduce 841.5 kg/m3 within 1.0 kg/m3 and the mass-weighted number-average molar mass to reproduce 0.2447 kg/mol within 0.001 kg/mol. These tolerances are deterministic model-consistency gates, not estimates of experimental uncertainty.
double[] cutSpecificGravity = engineeringDensityProfile;
double[] cutMolarMassKgPerMol = engineeringMolarMassProfile;
OilAssayCharacterisation assay = SarirAtmosphericAssay.create(
system, 1000.0, cutSpecificGravity, cutMolarMassKgPerMol);
assay.apply();
calculateBulkSpecificGravity(...), calculateBulkMolarMassKgPerMol(...), and
getCutVolumePercent() are available to Java and Python/JPype callers for preflight and
traceability. Profile validation completes before existing assay data or thermodynamic
components are modified.
Reusable 34-tray case factory
SarirAtmosphericFractionationCase.create(...) promotes the qualified sensitivity setup into a
reusable Java and Python/JPype entry point. It returns the configured crude feed and an unsolved
DistillationColumn, allowing callers to execute, inspect, or extend the case without copying the
source-to-model tray mapping.
The factory reads only the published 34-valve-tray count, tray-31-from-the-top feed location,
54,420 kg/h crude rate, 350 degC feed temperature, and 233 kPa absolute feed pressure from
SarirAtmosphericReference. The source tray maps to NeqSim simple-tray index 4 because the
library numbers simple trays from the bottom and reserves index 0 for the reboiler.
Every unreported property or operating choice remains explicit. Callers must supply 18 cut specific
gravities, 18 cut molar masses, top and bottom pressures, reboiler temperature, condenser reflux
ratio, and both liquid side-draw tray indices and fractions. The operating-input object rejects
non-finite pressures, temperatures, reflux, or fractions; an inverted pressure profile; fractions
outside [0, 1); the feed tray as a side draw; and a kerosene screen draw at or below the heavier
diesel screen draw. The existing SarirAtmosphericAssay bulk-density and average-molar-mass gates
remain authoritative for the property profiles.
The returned column uses the qualified MESH-residual numerical controls, but it is intentionally unsolved. Steam services, side strippers, pump-arounds, tray efficiencies, and source product rates are not configured. In particular, the factory does not tune either side-draw fraction to the published plant or HYSYS yields, and creating or solving the case is not evidence of product-yield or D86 agreement.
Solved-case product and balance summary
After the caller runs the configured column, SarirAtmosphericFractionationResult.evaluate(case)
creates an immutable Java/JPype summary of the rigorous result. The four rows remain in the
established top-to-bottom comparison order: Total Naphtha, Kerosene, Diesel, and Residual. Each row
reports calculated mass flow in kg/h, mass fraction of feed, mole-weighted mean normal boiling point,
the independently published plant rate, and the absolute relative rate difference.
The evaluator requires a solved non-fallback MESH-residual column. It rejects non-finite or negative flows, mass or energy errors above the qualified five-percent integration tolerance, feed/product mass-closure error above that tolerance, fewer than two material products, and material products whose mean normal boiling points do not increase from overhead to bottoms. Product arrays are defensive copies, and exact-label lookup fails closed for unsupported labels.
The calculated mass fractions and mean normal boiling points depend on the caller-supplied cut
properties and unreported column controls. They are model outputs, not source measurements, ASTM
D86 curves, or evidence that NeqSim reproduces the plant. The plant rates remain read-only
comparators: no calculated-versus-plant error is used as a solver control, tuning target, or pass
threshold. Java callers use the same evaluator directly; Python callers access the factory,
getColumn().run(...), and evaluator through the existing JPype bridge.
Shared discrete product boiling-range diagnostics
Material products returned by SarirAtmosphericFractionationResult expose the same immutable
ProductBoilingPointDistribution used by the qualified Big Hill vacuum result. This shared
implementation avoids separate atmospheric and vacuum quantile algorithms. It derives an ascending
normal-boiling-point support from positive-composition pseudo-components, normalizes their
cumulative product mole fractions to one, and reports the mole-weighted mean normal boiling point.
Java and Python/JPype callers can inspect a material row as follows:
SarirAtmosphericFractionationResult.ProductResult diesel =
result.getProduct("Diesel");
boolean distributionAvailable = diesel.hasBoilingPointDistribution();
double dieselT10Kelvin = diesel.getNormalBoilingPointQuantileKelvin(0.10);
double dieselT50Celsius = diesel.getNormalBoilingPointQuantileCelsius(0.50);
double dieselT90Kelvin = diesel.getNormalBoilingPointQuantileKelvin(0.90);
double[] supportKelvin = diesel.getBoilingPointTemperaturesKelvin();
double[] cumulativeMoleFractions = diesel.getCumulativeMoleFractions();
The support and cumulative arrays are defensive copies. Requests fail closed unless the cumulative
mole fraction is finite and in (0, 1]. A source-table row whose calculated stream is below the
qualified material-flow threshold reports hasBoilingPointDistribution() == false; attempting to
read its distribution or quantiles fails closed instead of presenting an undefined curve.
The focused 34-tray integration requires finite positive support temperatures, strictly increasing normalized cumulative fractions, exact cumulative closure to one, and T10 <= T50 <= T90 for every material product. Existing convergence, conservation, product-ordering, and read-only plant-rate comparisons remain unchanged.
These values are discrete cumulative-mole pseudo-component diagnostics. They are not continuous simulated-distillation or TBP curves, ASTM D86/D1160 temperatures, pressure-corrected laboratory measurements, specification-compliance results, or evidence that NeqSim reproduces the Sarir plant yields. The cut density and molar-mass profiles and unreported column controls remain explicit engineering inputs.
Scientific boundary
The source-derived volumes, boiling boundaries, and product-specification rows are reproducible, while every supplied cut density and molar mass remains an explicit engineering input. Product specifications are not laboratory results, independent validation targets, or a claim of NeqSim product compliance. The residual specification remains nonnumeric and open-ended. The factory does not resolve the missing light-end composition, distribute the published whole-crude sulfur among cuts, or claim that NeqSim reproduces the plant yields. A follow-on 34-tray comparison must preserve the plant rates as untouched acceptance targets and avoid tuning draw rates to the published products.
34-tray property-profile sensitivity gate
SarirAtmosphericFractionationSensitivityTest carries the constrained assay through 34 NeqSim
simple trays, matching the published valve-tray count. The source feed location, tray 31 counted
from the top, maps to NeqSim internal index 4 because NeqSim numbers simple trays from the bottom
and reserves index 0 for the reboiler. The test also uses the published feed rate, temperature, and
pressure.
Two separate cut-density and molar-mass profiles are exercised. Both profiles are engineering inputs because the article does not publish cut properties, and both independently reproduce the published 841.5 kg/m3 bulk density and 0.2447 kg/mol average molar mass before the assay is applied. The bounded perturbation demonstrates that the unreported profile changes the calculated split or product boiling metrics rather than being numerically inert.
The top pressure, condenser/reboiler settings, reflux ratio, side-draw trays, and side-draw fractions are fixed engineering test controls. They are not values reported by the article. The qualification uses the repository’s MESH-residual refinery-audit solver with a 0.20 K temperature tolerance and requires rigorous convergence without fallback, external and per-component conservation, energy and per-tray material closure, finite non-negative candidate-stream rates, at least two material products, and increasing mean normal boiling point across material products in column order. A zero-flow candidate is permitted and excluded from composition and boiling-order evaluation because the source reports light ends as not determined; the test does not synthesize inventory merely to force every screen stream positive.
The published plant rates remain untouched read-only evidence. They are not used as product-flow specifications, tuning targets, or numerical acceptance thresholds. This sensitivity gate does not qualify plant-yield or D86 reproduction and does not model the published steam, pump-around, side-stripper, tray-hydraulic, or efficiency details. The separate pump-around reference rows are source evidence only; unresolved tray-numbering direction prevents direct model configuration. Likewise, the steam rows retain the published service allocation, rates, temperatures, and pressures but do not infer quality or enthalpy or configure water, injection locations, side-stripper topology, or heat duties.