Atmospheric and Vacuum Crude Distillation Units

[Book navigation](../index.html) · [Assessment](../assessments/unit-02.html) · [Applied project](../projects/unit-02.html)

Unit target

Connect physical pretreatment and boiling-range separation to real material and energy boundaries.

Engineering context

Atmospheric and vacuum distillation split a feed according to vapor–liquid behavior and side-draw operations. A crude tower cannot create molecules; the yields of distillate and residue must reconcile with feed, slop, losses, and internal reflux. Utility estimates need heat-capacity and heat-integration assumptions. The modeler should decide what constitutes a material stream, what was measured directly, and which quantities follow from an equation. Whenever possible, start with an inlet-and-outlet balance in physical units rather than with an index or a normalized graph. An index can be useful for a classroom sensitivity exercise, but it can also conceal the absolute process magnitude needed to estimate utilities, storage or emissions.

Decision boundary

Identify what can be calculated from a TBP surrogate and what would require tray hydraulics, assays, vapor–liquid data and licensed column models. A good recommendation distinguishes what is established by conservation from what depends on kinetic, thermodynamic or operating assumptions. Record your data source and test method. Identify unknown parameters before asserting any performance. If two alternatives seem very close numerically, compute the range of results that plausible uncertainty could produce rather than picking the one with more decimals.

Technical study sequence

1. [Crude desalting mass balances](../lessons/unit-02-lesson-01.html) — [interactive lab 04](../labs/lab-04.html)
2. [Atmospheric crude distillation](../lessons/unit-02-lesson-02.html) — [interactive lab 05](../labs/lab-05.html)
3. [Vacuum distillation and thermal constraints](../lessons/unit-02-lesson-03.html) — [interactive lab 06](../labs/lab-06.html)
4. [Stabilization and light-end handling](../lessons/unit-02-lesson-04.html) — [interactive lab 06](../labs/lab-06.html)

Laboratory strategy

These three distinct models illuminate different physics or accounting relationships in the unit: Two-stage desalter, Atmospheric cut sensitivity, Vacuum boiling temperature. Take each model as a conditional hypothesis. For one model, reconstruct the default answer from the equation; for a second, find a limiting case; for the third, propose a measurement that could disprove it. A smooth trend is not an indicator of industrial accuracy, and sensitivity does not establish causality when the underlying variables are correlated.

Deliverable and independent scrutiny

Prepare a two-page engineering memo with a material and energy boundary, one calculation table, at least two uncertainty or failure scenarios, and a decision that names the data still needed. Use the [16 question assessment](../assessments/unit-02.html) to identify technical gaps. The [applied project](../projects/unit-02.html) then adds constraints from adjacent units so that a local process improvement can be evaluated for its plant-wide consequence.

Unit 02 · Lesson 01 — Crude desalting mass balances

[Unit map](../units/unit-02.html) · [Open numerical laboratory](../labs/lab-04.html) · [Assessment](../assessments/unit-02.html) · [Applied project](../projects/unit-02.html)

The engineering problem

Desalting mixes water with crude to transfer water-soluble salts into the aqueous phase, then separates brine. Actual performance depends on mixing, wash-water quality, electrostatic coalescence, residence time, emulsion stability and water carryover. Oil contamination of the brine and residual water in the desalted crude are separate outcomes from salt removal.

The practical starting point is not the equation by itself; it is a description of the process boundary. Write down the feed, product, recovered co-product, waste and possible accumulation terms before choosing an indicator. In this lesson, the title crude desalting mass balances indicates a specific decision lens within primary refining. Its purpose is to connect a measured or assumed characteristic to an operational consequence with a statement of what would prove the link. Separate primary evidence, indirect proxy and calculated prediction; those three categories are not interchangeable.

Quantity definition and model reasoning

For a simplified stage, C_out = C_in(1 − η) on a consistent ppmw basis when removal fraction η is specified. Two independent stages give C2 = C0(1 − η1)(1 − η2). Salt leaving with brine equals salt entering minus salt leaving with crude, neglecting additional salt in wash water.

To audit a model, begin with units. Write every mass flow on a common basis such as kg/h, every species flow on a consistent molar basis, and every energy rate as kW or MJ/h. Converting a component fraction from molar to mass basis requires component molar masses. A dimensionless index can be compared across scenarios only if the reference conditions and denominator have not changed. When the result is a percentage, specify percentage of what: total feed mass, converted feed, recovered product, sulfur atoms, or an observed time interval. The wrong denominator can change the interpretation even when the arithmetic is correct.

One useful way to check the numerical relationship is to inspect its limiting behavior. At zero conversion or removal, no product can appear or disappear through that mechanism; at complete removal, the tracked contaminant should be accounted for in another stream; at a cutpoint moved above or below the stated assay range, extrapolation should be rejected. A monotonic curve may be mathematically well behaved yet physically incomplete if secondary reactions, equilibrium or equipment capacity create a turning point outside the model.

Worked investigation: hypothetical data

For 50 ppmw salt entering two model stages with 60% and 65% nominal removal, estimate the outlet concentration and the removal per stage. Explain what wash-water salt would do to the answer.

Worked answer and interpretation

Stage 1 leaves 50(1−0.60)=20 ppmw. Stage 2 leaves 20(1−0.65)=7 ppmw, an overall 86% removal under the independent ideal-stage assumption. A paired inlet and outlet salt analysis can test the apparent effectiveness, but the result alone does not bound residual water or chloride hydrolysis risk.

Make a compact worksheet with four rows: starting assumptions, symbolic equation, numerical substitution and mass/energy/quality consistency check. Always place the units in the substitution row. An answer with four significant figures may be mathematically reproducible but should not be presented as a plant forecast when the assumed feed is only known roughly. Repeat the worksheet with one changed input, quantify the absolute and percentage change in the output, and ask which upstream measurement would be needed to defend that perturbation.

For numerical work, explicitly mark the inputs in this lesson as synthetic instructional values, including any yield factor, catalyst rate, representative boiling temperature or solvent efficiency. They must not be used as operating limits. If you wish to compare with published performance, define an independent, dated and source-specific dataset and evaluate whether its units and stream definitions match those of the exercise.

Plant integration and operating evidence

Account for incoming salt, injected wash-water salt, brine disposal, chemical use and oily wastewater. Validate stage efficiencies with paired inlet/outlet samples; vary mixing and emulsion treatment cautiously in accordance with site procedures.

Place the result in a real flowsheet context. Physical separation concentrates rather than destroys impurities. The simplifying assumption of sharp cuts can help a student audit material balances, but commercial columns show overlaps and side-stream contamination. Include inventories, side draws, overhead systems and energy integration when interpreting performance. Ask where the upstream stream comes from, which downstream unit receives it, and how a change would shift at least one other required resource. A product-quality improvement might raise utility demand; more extensive contaminant removal might increase solvent regeneration; higher conversion may move load to gas compression or wastewater treatment. Draw the additional arrows on the block flow diagram, and do not claim that the local calculation has optimized the full plant.

For an evidence-based evaluation, design a measurement table with columns for variable, unit, sample or instrument location, time basis, quality method, expected repeatability and action on discrepancy. Include one direct process variable, one product property and one integration variable. A useful model-validation report should show residuals or errors for several cases, not only agreement at the nominal case used to choose parameters. Record potential confounding changes in feed, temperature, operator procedure and catalyst condition.

Assumption challenge and failure diagnosis

Perfect salt-transfer or separation assumptions can hide difficult emulsions. A successful salt mass balance does not prove that hydrochloric acid will not form later in the process.

A critical reader should ask which implicit assumptions were required to write the equation: steady state or dynamic accumulation, ideal mixing or plug flow, fixed or changing properties, homogeneous or multiphase composition, constant versus variable selectivity, and independence versus coupling of disturbances. Pick the assumptions that matter here. If a model fails, identify whether the defect originates in incorrect units, an invalid physical assumption, unmeasured variability, or numerical implementation. Merely adding complexity does not fix poor calibration data.

In practical terms, define a stop condition for inference: a measurement outside the sampled composition range, evidence of a new phase, an unexpected hazard, a negative concentration, or an implausible material imbalance invalidates the shortcut. Escalate to a validated model or an expert review before making a design or safety claim. The appropriate corrective action can be to collect data, not to alter equipment.

Numerical laboratory and reproducible evidence

The linked [interactive laboratory, Two-stage desalter](../labs/lab-04.html), provides a separate numerical sensitivity exercise. Its governing relationship is Cout=Cin(1−η1)(1−η2). The laboratory is connected to this lesson through its unit-level concepts, but it may explore a different variable from the preceding worked example. Check its JSON parameter definitions, alter one input at a time and export the sampled curve to CSV. Compare two cases by recomputing the equation manually; document the difference between software correctness and physical validity.

Before declaring completion, articulate these four results in your own words: the stream or specimen under study, the dimension and reporting basis of the output, the reason for the sign of its sensitivity, and a specific observation that could disconfirm the model. Relate the lesson to Contaminants, compatibility, and feed envelopes as prior context and Atmospheric crude distillation as a subsequent design issue. This linkage should be causal, not merely a sequence of chapter names.

Independent checkpoints

1. Sketch the process boundary for crude desalting mass balances and label the stream or specimen basis that is actually measured.
2. Recreate the numerical reasoning from the worked investigation with all units and a limiting-case check.
3. Identify one operation, market specification or safety constraint that the formula does not contain.
4. Propose two measurements that would distinguish this model from a competing physical interpretation.
5. Describe why a correct plot of two-stage desalter does not establish plant-scale validity.

Editorial status: This is a provisional university-level teaching draft. Peer review, equipment-specific validation and current local requirements remain outside this instructional package.

Unit 02 · Lesson 02 — Atmospheric crude distillation

[Unit map](../units/unit-02.html) · [Open numerical laboratory](../labs/lab-05.html) · [Assessment](../assessments/unit-02.html) · [Applied project](../projects/unit-02.html)

The engineering problem

The atmospheric crude unit takes heated, desalted feed into a flash region and fractionation system. Internal reflux and steam stripping influence cut separation, while side strippers help remove lighter constituents from distillates. Pumparounds recover heat and shape tower internal flows. Each product draw has a rate, composition and specification constraint that cannot be deduced from a single set of theoretical cutpoints.

The practical starting point is not the equation by itself; it is a description of the process boundary. Write down the feed, product, recovered co-product, waste and possible accumulation terms before choosing an indicator. In this lesson, the title atmospheric crude distillation indicates a specific decision lens within primary refining. Its purpose is to connect a measured or assumed characteristic to an operational consequence with a statement of what would prove the link. Separate primary evidence, indirect proxy and calculated prediction; those three categories are not interchangeable.

Quantity definition and model reasoning

A material balance is F = ΣPi + losses + measured inventory change on a common time basis. With cumulative assay F(T), a first cut fraction may be estimated by differences at cut temperatures. The real enthalpy balance also requires feed and product enthalpies, furnace duty, pumparound removal and reflux.

To audit a model, begin with units. Write every mass flow on a common basis such as kg/h, every species flow on a consistent molar basis, and every energy rate as kW or MJ/h. Converting a component fraction from molar to mass basis requires component molar masses. A dimensionless index can be compared across scenarios only if the reference conditions and denominator have not changed. When the result is a percentage, specify percentage of what: total feed mass, converted feed, recovered product, sulfur atoms, or an observed time interval. The wrong denominator can change the interpretation even when the arithmetic is correct.

One useful way to check the numerical relationship is to inspect its limiting behavior. At zero conversion or removal, no product can appear or disappear through that mechanism; at complete removal, the tracked contaminant should be accounted for in another stream; at a cutpoint moved above or below the stated assay range, extrapolation should be rejected. A monotonic curve may be mathematically well behaved yet physically incomplete if secondary reactions, equilibrium or equipment capacity create a turning point outside the model.

Worked investigation: hypothetical data

For an idealized 1000 t/d feed with cumulative fractions F(180)=0.20, F(350)=0.62 and F(550)=0.88, estimate three product-range masses and residue; show a reconciled total.

Worked answer and interpretation

The four synthetic feed fractions are 0.20 below 180 °C, 0.62−0.20=0.42 between 180 and 350 °C, 0.88−0.62=0.26 between 350 and 550 °C, and 1−0.88=0.12 above 550 °C. For 1000 t/d these are 200, 420, 260 and 120 t/d; their sum is 1000 t/d, consistent with no modeled losses.

Make a compact worksheet with four rows: starting assumptions, symbolic equation, numerical substitution and mass/energy/quality consistency check. Always place the units in the substitution row. An answer with four significant figures may be mathematically reproducible but should not be presented as a plant forecast when the assumed feed is only known roughly. Repeat the worksheet with one changed input, quantify the absolute and percentage change in the output, and ask which upstream measurement would be needed to defend that perturbation.

For numerical work, explicitly mark the inputs in this lesson as synthetic instructional values, including any yield factor, catalyst rate, representative boiling temperature or solvent efficiency. They must not be used as operating limits. If you wish to compare with published performance, define an independent, dated and source-specific dataset and evaluate whether its units and stream definitions match those of the exercise.

Plant integration and operating evidence

Compare prospective distillate gain with product quality and column capacity. Check overhead corrosion risks, salt deposition, condenser duties and hydraulic limits, not only apparent fraction yield. Distinguish atmospheric residue from vacuum residue.

Place the result in a real flowsheet context. Physical separation concentrates rather than destroys impurities. The simplifying assumption of sharp cuts can help a student audit material balances, but commercial columns show overlaps and side-stream contamination. Include inventories, side draws, overhead systems and energy integration when interpreting performance. Ask where the upstream stream comes from, which downstream unit receives it, and how a change would shift at least one other required resource. A product-quality improvement might raise utility demand; more extensive contaminant removal might increase solvent regeneration; higher conversion may move load to gas compression or wastewater treatment. Draw the additional arrows on the block flow diagram, and do not claim that the local calculation has optimized the full plant.

For an evidence-based evaluation, design a measurement table with columns for variable, unit, sample or instrument location, time basis, quality method, expected repeatability and action on discrepancy. Include one direct process variable, one product property and one integration variable. A useful model-validation report should show residuals or errors for several cases, not only agreement at the nominal case used to choose parameters. Record potential confounding changes in feed, temperature, operator procedure and catalyst condition.

Assumption challenge and failure diagnosis

A pseudo-TBP model omits vapor–liquid equilibrium, tray efficiency, entrainment and overhead constraints. Do not use its cut yields as an operator setpoint.

A critical reader should ask which implicit assumptions were required to write the equation: steady state or dynamic accumulation, ideal mixing or plug flow, fixed or changing properties, homogeneous or multiphase composition, constant versus variable selectivity, and independence versus coupling of disturbances. Pick the assumptions that matter here. If a model fails, identify whether the defect originates in incorrect units, an invalid physical assumption, unmeasured variability, or numerical implementation. Merely adding complexity does not fix poor calibration data.

In practical terms, define a stop condition for inference: a measurement outside the sampled composition range, evidence of a new phase, an unexpected hazard, a negative concentration, or an implausible material imbalance invalidates the shortcut. Escalate to a validated model or an expert review before making a design or safety claim. The appropriate corrective action can be to collect data, not to alter equipment.

Numerical laboratory and reproducible evidence

The linked [interactive laboratory, Atmospheric cut sensitivity](../labs/lab-05.html), provides a separate numerical sensitivity exercise. Its governing relationship is Ymiddle=F(Tmid)−F(200 °C), Yheavy=F(500 °C)−F(Tmid). The laboratory is connected to this lesson through its unit-level concepts, but it may explore a different variable from the preceding worked example. Check its JSON parameter definitions, alter one input at a time and export the sampled curve to CSV. Compare two cases by recomputing the equation manually; document the difference between software correctness and physical validity.

Before declaring completion, articulate these four results in your own words: the stream or specimen under study, the dimension and reporting basis of the output, the reason for the sign of its sensitivity, and a specific observation that could disconfirm the model. Relate the lesson to Crude desalting mass balances as prior context and Vacuum distillation and thermal constraints as a subsequent design issue. This linkage should be causal, not merely a sequence of chapter names.

Independent checkpoints

1. Sketch the process boundary for atmospheric crude distillation and label the stream or specimen basis that is actually measured.
2. Recreate the numerical reasoning from the worked investigation with all units and a limiting-case check.
3. Identify one operation, market specification or safety constraint that the formula does not contain.
4. Propose two measurements that would distinguish this model from a competing physical interpretation.
5. Describe why a correct plot of atmospheric cut sensitivity does not establish plant-scale validity.

Editorial status: This is a provisional university-level teaching draft. Peer review, equipment-specific validation and current local requirements remain outside this instructional package.

Unit 02 · Lesson 03 — Vacuum distillation and thermal constraints

[Unit map](../units/unit-02.html) · [Open numerical laboratory](../labs/lab-06.html) · [Assessment](../assessments/unit-02.html) · [Applied project](../projects/unit-02.html)

The engineering problem

Vacuum operation lowers the required boiling temperature for a given heavy pseudo-component. This permits recovery of vacuum gas oil while limiting severe thermal exposure of heavier material. Vacuum systems involve ejectors or pumps, steam, condensers and leakage control; lower pressure does not eliminate residence-time and wall-temperature concerns.

The practical starting point is not the equation by itself; it is a description of the process boundary. Write down the feed, product, recovered co-product, waste and possible accumulation terms before choosing an indicator. In this lesson, the title vacuum distillation and thermal constraints indicates a specific decision lens within primary refining. Its purpose is to connect a measured or assumed characteristic to an operational consequence with a statement of what would prove the link. Separate primary evidence, indirect proxy and calculated prediction; those three categories are not interchangeable.

Quantity definition and model reasoning

For a simplified constant enthalpy vaporization approximation, ln(P/P0) = −ΔHvap/R (1/T − 1/T0). This relation describes one representative pseudo-component over a restricted range; a whole crude fraction needs a distribution of components and credible thermodynamic methods.

To audit a model, begin with units. Write every mass flow on a common basis such as kg/h, every species flow on a consistent molar basis, and every energy rate as kW or MJ/h. Converting a component fraction from molar to mass basis requires component molar masses. A dimensionless index can be compared across scenarios only if the reference conditions and denominator have not changed. When the result is a percentage, specify percentage of what: total feed mass, converted feed, recovered product, sulfur atoms, or an observed time interval. The wrong denominator can change the interpretation even when the arithmetic is correct.

One useful way to check the numerical relationship is to inspect its limiting behavior. At zero conversion or removal, no product can appear or disappear through that mechanism; at complete removal, the tracked contaminant should be accounted for in another stream; at a cutpoint moved above or below the stated assay range, extrapolation should be rejected. A monotonic curve may be mathematically well behaved yet physically incomplete if secondary reactions, equilibrium or equipment capacity create a turning point outside the model.

Worked investigation: hypothetical data

For a pseudo-component boiling at 640 K near 101 kPa with ΔHvap 40 kJ/mol, estimate the boiling temperature near 15 kPa. Compare with the atmospheric reference temperature.

Worked answer and interpretation

For pressure 15 kPa, ln(15/101)=−1.907; with ΔHvap=40,000 J/mol, 1/T=1/640−(8.314/40000)(−1.907)=0.001959 K⁻¹, so T≈510.5 K. The lower boiling temperature illustrates the vacuum principle but is not a furnace or column thermal-design prediction.

Make a compact worksheet with four rows: starting assumptions, symbolic equation, numerical substitution and mass/energy/quality consistency check. Always place the units in the substitution row. An answer with four significant figures may be mathematically reproducible but should not be presented as a plant forecast when the assumed feed is only known roughly. Repeat the worksheet with one changed input, quantify the absolute and percentage change in the output, and ask which upstream measurement would be needed to defend that perturbation.

For numerical work, explicitly mark the inputs in this lesson as synthetic instructional values, including any yield factor, catalyst rate, representative boiling temperature or solvent efficiency. They must not be used as operating limits. If you wish to compare with published performance, define an independent, dated and source-specific dataset and evaluate whether its units and stream definitions match those of the exercise.

Plant integration and operating evidence

Use pressure-temperature sensitivity to explain the purpose of vacuum. Real design requires flash-zone pressure, stripping requirements, liquid residence, heater film temperature, and hydraulic validation. Document what basis is used for volume and temperature.

Place the result in a real flowsheet context. Physical separation concentrates rather than destroys impurities. The simplifying assumption of sharp cuts can help a student audit material balances, but commercial columns show overlaps and side-stream contamination. Include inventories, side draws, overhead systems and energy integration when interpreting performance. Ask where the upstream stream comes from, which downstream unit receives it, and how a change would shift at least one other required resource. A product-quality improvement might raise utility demand; more extensive contaminant removal might increase solvent regeneration; higher conversion may move load to gas compression or wastewater treatment. Draw the additional arrows on the block flow diagram, and do not claim that the local calculation has optimized the full plant.

For an evidence-based evaluation, design a measurement table with columns for variable, unit, sample or instrument location, time basis, quality method, expected repeatability and action on discrepancy. Include one direct process variable, one product property and one integration variable. A useful model-validation report should show residuals or errors for several cases, not only agreement at the nominal case used to choose parameters. Record potential confounding changes in feed, temperature, operator procedure and catalyst condition.

Assumption challenge and failure diagnosis

Applying one latent heat to a broad residue is a deliberately poor approximation at large ranges. Cracking and coke formation are chemistry, not predicted by Clausius–Clapeyron.

A critical reader should ask which implicit assumptions were required to write the equation: steady state or dynamic accumulation, ideal mixing or plug flow, fixed or changing properties, homogeneous or multiphase composition, constant versus variable selectivity, and independence versus coupling of disturbances. Pick the assumptions that matter here. If a model fails, identify whether the defect originates in incorrect units, an invalid physical assumption, unmeasured variability, or numerical implementation. Merely adding complexity does not fix poor calibration data.

In practical terms, define a stop condition for inference: a measurement outside the sampled composition range, evidence of a new phase, an unexpected hazard, a negative concentration, or an implausible material imbalance invalidates the shortcut. Escalate to a validated model or an expert review before making a design or safety claim. The appropriate corrective action can be to collect data, not to alter equipment.

Numerical laboratory and reproducible evidence

The linked [interactive laboratory, Vacuum boiling temperature](../labs/lab-06.html), provides a separate numerical sensitivity exercise. Its governing relationship is 1/T = 1/T0 − (R/ΔHvap)ln(P/P0). The laboratory is connected to this lesson through its unit-level concepts, but it may explore a different variable from the preceding worked example. Check its JSON parameter definitions, alter one input at a time and export the sampled curve to CSV. Compare two cases by recomputing the equation manually; document the difference between software correctness and physical validity.

Before declaring completion, articulate these four results in your own words: the stream or specimen under study, the dimension and reporting basis of the output, the reason for the sign of its sensitivity, and a specific observation that could disconfirm the model. Relate the lesson to Atmospheric crude distillation as prior context and Stabilization and light-end handling as a subsequent design issue. This linkage should be causal, not merely a sequence of chapter names.

Independent checkpoints

1. Sketch the process boundary for vacuum distillation and thermal constraints and label the stream or specimen basis that is actually measured.
2. Recreate the numerical reasoning from the worked investigation with all units and a limiting-case check.
3. Identify one operation, market specification or safety constraint that the formula does not contain.
4. Propose two measurements that would distinguish this model from a competing physical interpretation.
5. Describe why a correct plot of vacuum boiling temperature does not establish plant-scale validity.

Editorial status: This is a provisional university-level teaching draft. Peer review, equipment-specific validation and current local requirements remain outside this instructional package.

Unit 02 · Lesson 04 — Stabilization and light-end handling

[Unit map](../units/unit-02.html) · [Open numerical laboratory](../labs/lab-06.html) · [Assessment](../assessments/unit-02.html) · [Applied project](../projects/unit-02.html)

The engineering problem

Stabilization removes light volatile compounds from a hydrocarbon liquid to improve downstream handling and product requirements. Naphtha stabilization, LPG fractionation and gas recovery rely on multicomponent phase equilibrium and pressure-dependent separation. The overhead stream must be accounted for as a product or fuel component, never as a disappearance.

The practical starting point is not the equation by itself; it is a description of the process boundary. Write down the feed, product, recovered co-product, waste and possible accumulation terms before choosing an indicator. In this lesson, the title stabilization and light-end handling indicates a specific decision lens within primary refining. Its purpose is to connect a measured or assumed characteristic to an operational consequence with a statement of what would prove the link. Separate primary evidence, indirect proxy and calculated prediction; those three categories are not interchangeable.

Quantity definition and model reasoning

For a simplified binary flash, a component split depends on K values, feed composition and vapor fraction. Rachford–Rice expresses Σzi(Ki−1)/(1+β(Ki−1))=0, with admissible vapor fraction β between zero and one when a two-phase region exists.

To audit a model, begin with units. Write every mass flow on a common basis such as kg/h, every species flow on a consistent molar basis, and every energy rate as kW or MJ/h. Converting a component fraction from molar to mass basis requires component molar masses. A dimensionless index can be compared across scenarios only if the reference conditions and denominator have not changed. When the result is a percentage, specify percentage of what: total feed mass, converted feed, recovered product, sulfur atoms, or an observed time interval. The wrong denominator can change the interpretation even when the arithmetic is correct.

One useful way to check the numerical relationship is to inspect its limiting behavior. At zero conversion or removal, no product can appear or disappear through that mechanism; at complete removal, the tracked contaminant should be accounted for in another stream; at a cutpoint moved above or below the stated assay range, extrapolation should be rejected. A monotonic curve may be mathematically well behaved yet physically incomplete if secondary reactions, equilibrium or equipment capacity create a turning point outside the model.

Worked investigation: hypothetical data

A 100 kmol/h synthetic liquid contains 5 mol% lumped light ends. If a model removes 80% of those light ends to overhead, compute the overhead light-end molar rate and the residual in liquid, before considering entrained heavy material.

Worked answer and interpretation

The feed has 100×0.05=5 kmol/h notional light ends. At 80% transfer, overhead receives 4 kmol/h light ends and the liquid retains 1 kmol/h. Total overhead flow cannot be deduced without knowing how much heavier material also enters the vapor; claiming the overhead total is exactly 4 kmol/h would be an unjustified assumption.

Make a compact worksheet with four rows: starting assumptions, symbolic equation, numerical substitution and mass/energy/quality consistency check. Always place the units in the substitution row. An answer with four significant figures may be mathematically reproducible but should not be presented as a plant forecast when the assumed feed is only known roughly. Repeat the worksheet with one changed input, quantify the absolute and percentage change in the output, and ask which upstream measurement would be needed to defend that perturbation.

For numerical work, explicitly mark the inputs in this lesson as synthetic instructional values, including any yield factor, catalyst rate, representative boiling temperature or solvent efficiency. They must not be used as operating limits. If you wish to compare with published performance, define an independent, dated and source-specific dataset and evaluate whether its units and stream definitions match those of the exercise.

Plant integration and operating evidence

Check total material and component balances around stabilizers, reflux drums, reboilers and off-gas treatment. Measure volatility using a specified test method and include gas composition and containment in the operating discussion.

Place the result in a real flowsheet context. Physical separation concentrates rather than destroys impurities. The simplifying assumption of sharp cuts can help a student audit material balances, but commercial columns show overlaps and side-stream contamination. Include inventories, side draws, overhead systems and energy integration when interpreting performance. Ask where the upstream stream comes from, which downstream unit receives it, and how a change would shift at least one other required resource. A product-quality improvement might raise utility demand; more extensive contaminant removal might increase solvent regeneration; higher conversion may move load to gas compression or wastewater treatment. Draw the additional arrows on the block flow diagram, and do not claim that the local calculation has optimized the full plant.

For an evidence-based evaluation, design a measurement table with columns for variable, unit, sample or instrument location, time basis, quality method, expected repeatability and action on discrepancy. Include one direct process variable, one product property and one integration variable. A useful model-validation report should show residuals or errors for several cases, not only agreement at the nominal case used to choose parameters. Record potential confounding changes in feed, temperature, operator procedure and catalyst condition.

Assumption challenge and failure diagnosis

A constant relative volatility or single binary pair may be useful pedagogically but does not represent gasoline volatility. High volatile content changes hazard classification and loss accounting.

A critical reader should ask which implicit assumptions were required to write the equation: steady state or dynamic accumulation, ideal mixing or plug flow, fixed or changing properties, homogeneous or multiphase composition, constant versus variable selectivity, and independence versus coupling of disturbances. Pick the assumptions that matter here. If a model fails, identify whether the defect originates in incorrect units, an invalid physical assumption, unmeasured variability, or numerical implementation. Merely adding complexity does not fix poor calibration data.

In practical terms, define a stop condition for inference: a measurement outside the sampled composition range, evidence of a new phase, an unexpected hazard, a negative concentration, or an implausible material imbalance invalidates the shortcut. Escalate to a validated model or an expert review before making a design or safety claim. The appropriate corrective action can be to collect data, not to alter equipment.

Numerical laboratory and reproducible evidence

The linked [interactive laboratory, Vacuum boiling temperature](../labs/lab-06.html), provides a separate numerical sensitivity exercise. Its governing relationship is 1/T = 1/T0 − (R/ΔHvap)ln(P/P0). The laboratory is connected to this lesson through its unit-level concepts, but it may explore a different variable from the preceding worked example. Check its JSON parameter definitions, alter one input at a time and export the sampled curve to CSV. Compare two cases by recomputing the equation manually; document the difference between software correctness and physical validity.

Before declaring completion, articulate these four results in your own words: the stream or specimen under study, the dimension and reporting basis of the output, the reason for the sign of its sensitivity, and a specific observation that could disconfirm the model. Relate the lesson to Vacuum distillation and thermal constraints as prior context and Fluid catalytic cracking yield networks as a subsequent design issue. This linkage should be causal, not merely a sequence of chapter names.

Independent checkpoints

1. Sketch the process boundary for stabilization and light-end handling and label the stream or specimen basis that is actually measured.
2. Recreate the numerical reasoning from the worked investigation with all units and a limiting-case check.
3. Identify one operation, market specification or safety constraint that the formula does not contain.
4. Propose two measurements that would distinguish this model from a competing physical interpretation.
5. Describe why a correct plot of vacuum boiling temperature does not establish plant-scale validity.

Editorial status: This is a provisional university-level teaching draft. Peer review, equipment-specific validation and current local requirements remain outside this instructional package.

Unit 02 applied project — Primary refining

[Unit 02](../units/unit-02.html) · [Book projects](../index.html)

Commissioning question

A fictional integrated refinery-and-petrochemicals teaching facility is reviewing its primary refining decisions after receiving evidence that the assumed nominal feed or operating condition is no longer reliable. Your task is to produce an engineering screening memo, not an equipment datasheet or operating instruction. The central problem is: Connect physical pretreatment and boiling-range separation to real material and energy boundaries.

Atmospheric and vacuum distillation split a feed according to vapor–liquid behavior and side-draw operations. A crude tower cannot create molecules; the yields of distillate and residue must reconcile with feed, slop, losses, and internal reflux. Utility estimates need heat-capacity and heat-integration assumptions.

Synthetic case boundary

Start from a nominal notional feed of 1,000 mass units per day for a single material balance; choose its actual chemical meaning and reporting basis before calculations. Allocate an assumed 10% uncertainty to the most sensitive unmeasured characteristic. Do not silently convert the same number into tonnes, cubic metres, kmol and standard gas volume. If a composition is needed, invent and label a test vector whose fractions sum exactly to unity. Keep the upstream and downstream units visible in a block flow diagram.

For this unit, the four required technical workstreams are Crude desalting mass balances, Atmospheric crude distillation, Vacuum distillation and thermal constraints, Stabilization and light-end handling. Use the lesson descriptions to select what evidence exists and what has to be assumed. If your assumed inputs are internally inconsistent—for example, a product yield above converted material, more recovered hydrogen than contained hydrogen, or a heat exchanger duty exceeding its thermodynamic ceiling—correct the conceptual error rather than tuning an output.

Computational experiment

Run three independent unit laboratories: [lab 04: Two-stage desalter](../labs/lab-04.html), [lab 05: Atmospheric cut sensitivity](../labs/lab-05.html), [lab 06: Vacuum boiling temperature](../labs/lab-06.html). For each lab, define a base case, one credible perturbation and a stress case near the end of the control range. Record the JSON model identifier, all parameter values, model output, units and qualitative interpretation. Export CSV. Reconstruct at least two displayed values with an independent worksheet or script, and plot both the baseline and a sensitivity trajectory.

Then write an uncertainty section: distinguish numerical precision, parameter uncertainty, structural model uncertainty and sampling uncertainty. An exact identity can be extremely precise but may be irrelevant to a poorly sampled feed. Conversely, a fitted kinetic coefficient can capture a short operating period and fail when catalyst activity changes. Explain which uncertainties could reverse the ranking of proposed alternatives.

Whole-plant coordination

Identify what can be calculated from a TBP surrogate and what would require tray hydraulics, assays, vapor–liquid data and licensed column models. Name at least one impact on a different department: sulfur or wastewater handling, hydrogen consumption, off-gas, steam and power, product specifications, instrumentation, emergency response or carbon accounting. A nominally favorable local change is not a positive overall result until its resource and product interactions are reconciled. Identify an affected stakeholder who owns the required measurement or approval.

Deliverables and scoring

1. Boundary diagram and data dictionary (20%): readable stream labels, units, basis, data provenance and calculation assumptions.

2. Calculation appendix (25%): auditable mass or energy balance, three lab experiments and independent numerical recomputation.

3. Validation design (20%): at least two measurements, instrument/assay location, measurement frequency and acceptance criteria.

4. Trade-off memorandum (20%): alternative scenarios, constraints, uncertainty range and a reasoned but conditional comparison.

5. Responsible practice (15%): safety and environmental exclusions, uncertainty about regulations, and an explicit statement that the work is not plant-commissioning advice.

Challenge extension

Suppose the feed characteristic that controls the main model suddenly moves beyond the laboratory range. Write a plan for collecting the minimum data that would permit responsible recalibration without risking an unsupported process change. Your memo should explain why extrapolation is not a substitute for evidence and should name a decision point where engineering review is mandatory. The strongest result may be a refusal to infer a safe or profitable configuration from insufficient information.

Interactive laboratory 03: Natural-gas heating value

[Open lab in the browser](lab-03.html) · [Unit 01](../units/unit-01.html)

Engineering purpose

Blend methane and ethane with a fixed CO₂ fraction, then report approximate molar LHV. The selected input is intentionally a one-variable exploratory slice. Move one setting at a time, check whether the displayed rate or fraction remains physically meaningful, and compare the curve against the conservation requirements of the process. The exercise is designed to show sensitivity and falsifiability, not to choose equipment sizes or plant operating conditions.

Equations and units

LHVmix = yCH4·802.3 + yC2H6·1428 MJ/kmol

The result is reported as MJ/kmol. Entered quantities are documented in the associated JSON and visible controls. Distinguish this calculated quantity from plant measurements and from a standardized certificate-of-analysis property. Whenever the formula contains a fitted coefficient or lumped index, describe what experimental dataset would be needed for calibration.

Procedure and recorded evidence

1. Start from the displayed nominal controls and record the output with all units.
2. Change the primary parameter to one end of its range. Use the plot to identify the sign and approximate local slope; record both the input and output.
3. Return to nominal settings, vary the second parameter when available, and explain why the primary trend changes—or why the one-dimensional model prevents a conclusion.
4. Export the 81-point dataset. Plot it independently in a spreadsheet or numerical notebook; check the endpoint values by hand using the equation.
5. Test a limiting or boundary case. State what would make the equation physically inappropriate or numerically invalid, even when the software shows a smooth graph.

Investigation

Vary methane at fixed CO₂ and discuss why LHV per kilogram behaves differently. Compare your result with an independent arithmetic calculation. Report a mass-, species-, or energy-balance audit appropriate to the process rather than relying on numerical agreement alone.

Scientific boundary

Dry, ideal-gas molar mixture of only CH4/C2H6/CO2; synthetic approximate component LHVs. The graph is not evidence of a measured plant result. The model omits material characteristics, uncertainty sources and coupled constraints that would be required in a licensed process simulation. A useful lab report has separate sections for model prediction, measurement requirements, likely failure modes and decisions that must be deferred until validation.

Interactive laboratory 04: Two-stage desalter

[Open lab in the browser](lab-04.html) · [Unit 02](../units/unit-02.html)

Engineering purpose

Model water-soluble salt removal across two hypothetical independent desalter stages. The selected input is intentionally a one-variable exploratory slice. Move one setting at a time, check whether the displayed rate or fraction remains physically meaningful, and compare the curve against the conservation requirements of the process. The exercise is designed to show sensitivity and falsifiability, not to choose equipment sizes or plant operating conditions.

Equations and units

Cout=Cin(1−η1)(1−η2)

The result is reported as ppmw. Entered quantities are documented in the associated JSON and visible controls. Distinguish this calculated quantity from plant measurements and from a standardized certificate-of-analysis property. Whenever the formula contains a fitted coefficient or lumped index, describe what experimental dataset would be needed for calibration.

Procedure and recorded evidence

1. Start from the displayed nominal controls and record the output with all units.
2. Change the primary parameter to one end of its range. Use the plot to identify the sign and approximate local slope; record both the input and output.
3. Return to nominal settings, vary the second parameter when available, and explain why the primary trend changes—or why the one-dimensional model prevents a conclusion.
4. Export the 81-point dataset. Plot it independently in a spreadsheet or numerical notebook; check the endpoint values by hand using the equation.
5. Test a limiting or boundary case. State what would make the equation physically inappropriate or numerically invalid, even when the software shows a smooth graph.

Investigation

How much does improving stage 1 compare with improving stage 2? Compare your result with an independent arithmetic calculation. Report a mass-, species-, or energy-balance audit appropriate to the process rather than relying on numerical agreement alone.

Scientific boundary

Does not represent wash-water salinity, carryover or emulsion stability. The graph is not evidence of a measured plant result. The model omits material characteristics, uncertainty sources and coupled constraints that would be required in a licensed process simulation. A useful lab report has separate sections for model prediction, measurement requirements, likely failure modes and decisions that must be deferred until validation.

Interactive laboratory 18: Lubricant viscosity blending proxy

[Open lab in the browser](lab-18.html) · [Unit 06](../units/unit-06.html)

Engineering purpose

Compare logarithmic blending of two positive dynamic viscosities at one fixed temperature. The selected input is intentionally a one-variable exploratory slice. Move one setting at a time, check whether the displayed rate or fraction remains physically meaningful, and compare the curve against the conservation requirements of the process. The exercise is designed to show sensitivity and falsifiability, not to choose equipment sizes or plant operating conditions.

Equations and units

ln μ=x ln μA+(1−x)ln μB

The result is reported as mPa·s. Entered quantities are documented in the associated JSON and visible controls. Distinguish this calculated quantity from plant measurements and from a standardized certificate-of-analysis property. Whenever the formula contains a fitted coefficient or lumped index, describe what experimental dataset would be needed for calibration.

Procedure and recorded evidence

1. Start from the displayed nominal controls and record the output with all units.
2. Change the primary parameter to one end of its range. Use the plot to identify the sign and approximate local slope; record both the input and output.
3. Return to nominal settings, vary the second parameter when available, and explain why the primary trend changes—or why the one-dimensional model prevents a conclusion.
4. Export the 81-point dataset. Plot it independently in a spreadsheet or numerical notebook; check the endpoint values by hand using the equation.
5. Test a limiting or boundary case. State what would make the equation physically inappropriate or numerically invalid, even when the software shows a smooth graph.

Investigation

Compare with an arithmetic mean and explain which is a hypothesis requiring calibration. Compare your result with an independent arithmetic calculation. Report a mass-, species-, or energy-balance audit appropriate to the process rather than relying on numerical agreement alone.

Scientific boundary

Not a universal lubricant blending rule or viscosity–temperature model. The graph is not evidence of a measured plant result. The model omits material characteristics, uncertainty sources and coupled constraints that would be required in a licensed process simulation. A useful lab report has separate sections for model prediction, measurement requirements, likely failure modes and decisions that must be deferred until validation.

Unit 02 assessment — Primary refining

[Return to unit](../units/unit-02.html) · [Solutions rubric](../ASSESSMENT-RUBRIC.html)

Instructions. Answer all sixteen tasks. Use hand calculations or exported simulator data where appropriate. State your assumptions, units and the difference between data and hypothetical values. Results are not checked by automated grading.

Crude desalting mass balances

Question 01 — Reconstruct the quantitative link

Given the synthetic case posed in the lesson — For 50 ppmw salt entering two model stages with 60% and 65% nominal removal, estimate the outlet concentration and the removal per stage. Explain what wash-water salt would do to the answer. — write the defining equation from first principles or its stated empirical assumptions, substitute the quantities with explicit units, and compute a result and reasonable significant figures. Explain what precise physical quantity the answer represents. Then change one input by 10% and determine the approximate sensitivity, specifying whether a percentage or a percentage-point change is being discussed.

Question 02 — Test the conservation boundary

Draw an inlet–outlet boundary around the relevant operation for crude desalting mass balances. Include feed, saleable product, recycle, off-gas, waste, accumulation and energy transfer where applicable. State a component or elemental balance that must be satisfied. Create a simple inconsistent dataset, quantify its imbalance and propose two concrete measurements that would help identify the error.

Question 03 — Challenge the physical model

Consider the assumption in the lesson that Perfect salt-transfer or separation assumptions can hide difficult emulsions. A successful salt mass balance does not prove that hydrochloric acid will not form later in the process. Construct a plausible case in which its limitation becomes important. Predict the direction of model bias if defensible, identify any sign that cannot be inferred, and design a laboratory or process-data check to distinguish an inadequate equation from a sensor or sampling error.

Question 04 — Prepare an operating evidence decision

Prepare a one-page decision memo linked to crude desalting mass balances. Use this practice context: Account for incoming salt, injected wash-water salt, brine disposal, chemical use and oily wastewater. Validate stage efficiencies with paired inlet/outlet samples; vary mixing and emulsion treatment cautiously in accordance with site procedures. State one measurable target, two competing operational or design choices, one external constraint, and three data sources or instruments. Recommend only the next evidence-collection step; do not convert the classroom model into a real plant setpoint.

Atmospheric crude distillation

Question 05 — Reconstruct the quantitative link

Given the synthetic case posed in the lesson — For an idealized 1000 t/d feed with cumulative fractions F(180)=0.20, F(350)=0.62 and F(550)=0.88, estimate three product-range masses and residue; show a reconciled total. — write the defining equation from first principles or its stated empirical assumptions, substitute the quantities with explicit units, and compute a result and reasonable significant figures. Explain what precise physical quantity the answer represents. Then change one input by 10% and determine the approximate sensitivity, specifying whether a percentage or a percentage-point change is being discussed.

Question 06 — Test the conservation boundary

Draw an inlet–outlet boundary around the relevant operation for atmospheric crude distillation. Include feed, saleable product, recycle, off-gas, waste, accumulation and energy transfer where applicable. State a component or elemental balance that must be satisfied. Create a simple inconsistent dataset, quantify its imbalance and propose two concrete measurements that would help identify the error.

Question 07 — Challenge the physical model

Consider the assumption in the lesson that A pseudo-TBP model omits vapor–liquid equilibrium, tray efficiency, entrainment and overhead constraints. Do not use its cut yields as an operator setpoint. Construct a plausible case in which its limitation becomes important. Predict the direction of model bias if defensible, identify any sign that cannot be inferred, and design a laboratory or process-data check to distinguish an inadequate equation from a sensor or sampling error.

Question 08 — Prepare an operating evidence decision

Prepare a one-page decision memo linked to atmospheric crude distillation. Use this practice context: Compare prospective distillate gain with product quality and column capacity. Check overhead corrosion risks, salt deposition, condenser duties and hydraulic limits, not only apparent fraction yield. Distinguish atmospheric residue from vacuum residue. State one measurable target, two competing operational or design choices, one external constraint, and three data sources or instruments. Recommend only the next evidence-collection step; do not convert the classroom model into a real plant setpoint.

Vacuum distillation and thermal constraints

Question 09 — Reconstruct the quantitative link

Given the synthetic case posed in the lesson — For a pseudo-component boiling at 640 K near 101 kPa with ΔHvap 40 kJ/mol, estimate the boiling temperature near 15 kPa. Compare with the atmospheric reference temperature. — write the defining equation from first principles or its stated empirical assumptions, substitute the quantities with explicit units, and compute a result and reasonable significant figures. Explain what precise physical quantity the answer represents. Then change one input by 10% and determine the approximate sensitivity, specifying whether a percentage or a percentage-point change is being discussed.

Question 10 — Test the conservation boundary

Draw an inlet–outlet boundary around the relevant operation for vacuum distillation and thermal constraints. Include feed, saleable product, recycle, off-gas, waste, accumulation and energy transfer where applicable. State a component or elemental balance that must be satisfied. Create a simple inconsistent dataset, quantify its imbalance and propose two concrete measurements that would help identify the error.

Question 11 — Challenge the physical model

Consider the assumption in the lesson that Applying one latent heat to a broad residue is a deliberately poor approximation at large ranges. Cracking and coke formation are chemistry, not predicted by Clausius–Clapeyron. Construct a plausible case in which its limitation becomes important. Predict the direction of model bias if defensible, identify any sign that cannot be inferred, and design a laboratory or process-data check to distinguish an inadequate equation from a sensor or sampling error.

Question 12 — Prepare an operating evidence decision

Prepare a one-page decision memo linked to vacuum distillation and thermal constraints. Use this practice context: Use pressure-temperature sensitivity to explain the purpose of vacuum. Real design requires flash-zone pressure, stripping requirements, liquid residence, heater film temperature, and hydraulic validation. Document what basis is used for volume and temperature. State one measurable target, two competing operational or design choices, one external constraint, and three data sources or instruments. Recommend only the next evidence-collection step; do not convert the classroom model into a real plant setpoint.

Stabilization and light-end handling

Question 13 — Reconstruct the quantitative link

Given the synthetic case posed in the lesson — A 100 kmol/h synthetic liquid contains 5 mol% lumped light ends. If a model removes 80% of those light ends to overhead, compute the overhead light-end molar rate and the residual in liquid, before considering entrained heavy material. — write the defining equation from first principles or its stated empirical assumptions, substitute the quantities with explicit units, and compute a result and reasonable significant figures. Explain what precise physical quantity the answer represents. Then change one input by 10% and determine the approximate sensitivity, specifying whether a percentage or a percentage-point change is being discussed.

Question 14 — Test the conservation boundary

Draw an inlet–outlet boundary around the relevant operation for stabilization and light-end handling. Include feed, saleable product, recycle, off-gas, waste, accumulation and energy transfer where applicable. State a component or elemental balance that must be satisfied. Create a simple inconsistent dataset, quantify its imbalance and propose two concrete measurements that would help identify the error.

Question 15 — Challenge the physical model

Consider the assumption in the lesson that A constant relative volatility or single binary pair may be useful pedagogically but does not represent gasoline volatility. High volatile content changes hazard classification and loss accounting. Construct a plausible case in which its limitation becomes important. Predict the direction of model bias if defensible, identify any sign that cannot be inferred, and design a laboratory or process-data check to distinguish an inadequate equation from a sensor or sampling error.

Question 16 — Prepare an operating evidence decision

Prepare a one-page decision memo linked to stabilization and light-end handling. Use this practice context: Check total material and component balances around stabilizers, reflux drums, reboilers and off-gas treatment. Measure volatility using a specified test method and include gas composition and containment in the operating discussion. State one measurable target, two competing operational or design choices, one external constraint, and three data sources or instruments. Recommend only the next evidence-collection step; do not convert the classroom model into a real plant setpoint.