Convection in Process Equipment

The engineering problem

Convection couples diffusion near a wall to bulk fluid motion. The film coefficient h is not a material constant. It summarizes the temperature field for a specific flow, geometry and property set. A published correlation can give a precise number while being inapplicable to a developing, fouled or multiphase stream.

Scope and learning outcomes

Select internal, external and natural-convection models by flow regime, geometry and property range, and quantify correlation uncertainty. Students should be able to select a governing equation, construct a measurable boundary condition, calculate a plausible output with its units, and explain a failure case before applying a correlation.

Concept map

4.1 Internal forced convection

For a fully developed turbulent flow, selected Nusselt correlations link h to Reynolds and Prandtl numbers; they are not substitutes for regime verification. A cooled solvent flows through an instrumented smooth tube. Check Re, Pr, heating/cooling exponent and whether entrance lengths invalidate the correlation. Representative relationship: Nu=0.023Re^0.8Pr^n; h=Nu*k/D.

4.2 External flow and boundary layers

An external boundary layer grows along the surface; average and local heat-transfer coefficients are distinct and depend on transition. Air crosses the exterior of a drying-tray cover. Evaluate changes in plate length and flow speed while keeping the laminar applicability limit explicit. Representative relationship: NuL=0.664ReL^0.5Pr^(1/3) (laminar plate).

4.3 Natural convection

Buoyancy drives motion when density variation couples to gravity; Grashof and Rayleigh numbers organize the flow, not an assumed pump velocity. A hot holding tank cools without forced ventilation. Contrast stable vertical-wall buoyancy with an enclosed horizontal gap where stratification matters. Representative relationship: Ra=gbetaabs(dT)L^3/(nualpha).

4.4 Fins and extended surfaces

Fins increase area but their effectiveness is limited by axial temperature gradients and contact resistance; extra area is not uniformly at base temperature. An air-cooled control-panel heat sink is compared with a flat plate. Evaluate fin efficiency, base temperature and whether airflow bypass limits useful area. Representative relationship: eta_fin=tanh(mL)/(mL); m=sqrt(hP/(kAc)).

A disciplined workflow for this unit

Start with a diagram of a process boundary, including fluid inlets, outlets, phase boundaries, hot and cold external surfaces and any energy storage. Record which values are measurements, which are assumed, and which are calculated. Write the conservation statement before inserting a transfer correlation. After obtaining a numerical result, test the limiting case: the heat flow should vanish with zero temperature difference in a passive linear resistance model, while doubling the wall area should increase conduction duty if all other conditions remain fixed. For nonlinear radiation a twofold temperature increase cannot be treated as a twofold radiation increase.

Report results with their uncertainty drivers. For a heated process tube, a thermocouple may measure external wall temperature but not internal wall temperature, and a flow meter may give volumetric rather than mass rate. An apparent discrepancy between measured and predicted duty may be the result of imperfect knowledge of heat capacities, uncontrolled bypass or fouling. If a correlation is used beyond the printed range, label the result an extrapolation instead of assigning spurious precision.

Unit investigation method

Use a blank process sheet with five sections: known properties, governing physics, calculation, validation, and operating decision. At minimum compare two physically plausible operating states and identify what extra measurement would most efficiently distinguish them. Do not rely on the number of digits printed by a simulation to argue that a real process is controlled to that precision.

Unit 4, Lesson 1: Internal forced convection

Unit: Convection in process equipment. Core formula: Nu=0.023Re^0.8Pr^n; h=Nu*k/D. Suggested model: [Laboratory 07 — Turbulent pipe-side convection](../labs/lab-07.html).

Why this problem appears in process plants

A cooled solvent flows through an instrumented smooth tube. For a fully developed turbulent flow, selected Nusselt correlations link h to Reynolds and Prandtl numbers; they are not substitutes for regime verification. This topic is a decision about where energy travels, which mechanism controls its rate, and what evidence is necessary to justify a calculation. The same reported heater duty can correspond to distinct wall temperatures depending on geometry, flow and material properties. A credible engineer must therefore distinguish the energy rate demanded by a process from the thermal resistance that allows the equipment to deliver that rate.

A process operator experiences this distinction as an operating constraint rather than as an equation. When product throughput increases, an unchanged heating utility can be insufficient even when the external wall stays hot. A measurement campaign should record mass flow, inlet and outlet conditions, utility conditions, local wall temperatures when available, and the operating history that may affect fouling. Without these signals, increasing utility temperature may mask the actual cause of poor performance and worsen degradation.

Physical mechanism and assumptions

For a fully developed turbulent flow, selected Nusselt correlations link h to Reynolds and Prandtl numbers; they are not substitutes for regime verification. The equation selected here, Nu=0.023Re^0.8Pr^n; h=Nu*k/D, summarizes a limited physical model. Every term must refer to a declared boundary, an appropriate unit system, and a physically relevant value of temperature or material property. Before using the equation, identify whether the system is steady or transient, whether the fluid undergoes phase change, and whether the heat-transfer coefficient is known from a defensible correlation or simply a fitted number.

The engineer's first drawing should show an inward or outward normal for the relevant surface and a clear sign convention. Heat transport is directed down a temperature driving force in a passive system, although a controlled refrigeration cycle can move heat from a colder source to a hotter sink by supplying work. A temperature sensor located in a mixing chamber is not necessarily the right measurement of the wall temperature immediately adjacent to a process fluid. A fixed temperature boundary also differs from an imposed heat-flux boundary and gives a different mathematical problem.

Derivation and dimensional reasoning

Start from the first-law energy balance, accumulation = net incoming enthalpy + supplied heat + work, with the appropriate sign conventions. Introduce the constitutive expression Nu=0.023Re^0.8Pr^n; h=Nu*k/D only after identifying the relevant energy-transfer mechanism. If a rate equation gives watts, confirm that conductivity is in W/(m K), surface area in m² and temperature difference in kelvin for a planar slab. The analogous exercise for a convection coefficient uses W/(m² K). For radiation, temperatures must be absolute because emission depends on the fourth power of thermodynamic temperature.

Terms that look superficially similar can require fundamentally different models. A cylindrical pipe wall does not have a constant radial area; a nonlinear radiation exchange cannot be replaced by a temperature-independent resistance without specifying a linearization point. Correlations expressed with Reynolds and Prandtl numbers need fluid properties at the proper reference temperature and geometry. In a thermally developing flow, the local Nusselt number may differ from its fully developed average. Write these distinctions directly beside the calculation rather than after the final answer.

Worked investigation: turbulent pipe-side convection

We use a numerical teaching model with the following documented inputs: bulk velocity = 2; D = 0.04; rho = 990; mu = 0.001; k = 0.6; cp = 4100. The primary control bulk velocity is measured in m/s. Other parameters are D [m]; rho [kg/m³]; mu [Pa s]; k [W/(m K)]; cp [J/(kg K)]. Using the model equation, the predicted estimated h is 6175.22 W/(m² K). This is an arithmetic result for assumed properties, not a measured industrial duty or a certified equipment rating.

Increase the primary control from 2 to 2.48 m/s while holding the other inputs fixed. The same model predicts 7334.82 W/(m² K). Compare the absolute difference (1159.61 W/(m² K)) with the percentage difference, and interpret the sign using the physical mechanism. Does this change chiefly alter the driving force, the thermal resistance, the material storage, or the rate at which the process moves energy? Explain why interpreting the curve without checking the other controlled inputs could be misleading.

The control range is deliberately finite. At very large or small values a simplified correlation can violate its validity conditions even if the numerical program continues to return finite numbers. For this lab, the important limitation is: Dittus–Boelter for smooth, fully developed turbulent internal flow (use only Re≥10000 and Pr≈0.7–160). Decide whether the default state is inside that limitation and state what kind of additional property data, geometry measurement, or experimental coefficient would be needed to increase confidence in the estimate.

Measurement strategy and uncertainty

Check Re, Pr, heating/cooling exponent and whether entrance lengths invalidate the correlation. Select one variable from the worked model that could realistically be measured on a plant floor, one that would normally be inferred, and one that is liable to drift over months of operation. Suggest how each could be independently checked. Thermal couples need installation and dynamic-response consideration, while flow meters require correct fluid properties and calibration. A sensor can be accurate at its measurement point yet irrelevant to the heat-transfer boundary being modeled.

Develop a first-order sensitivity calculation. If the output Y depends on parameter p, estimate dY/dp from two nearby runs of the simulator. If the measured parameter has uncertainty u(p), its approximate output contribution is abs(dY/dp)*u(p). This is a local linear approximation; it does not characterize poorly known geometry, phase-change instability or systematic installation bias. Keep uncertainty in the output's physical units until presenting any final percentage. Where a temperature difference is small, uncertainty can become the dominant proportion of the inferred transfer rate.

Process decision and limitation

Check Re, Pr, heating/cooling exponent and whether entrance lengths invalidate the correlation. An excellent model is still a screening tool if it ignores details that control real equipment performance. A preliminary calculation may support an experiment or identify the direction of a sensitivity, but it cannot substitute for a vendor curve, verified thermophysical properties, design margins, pressure-equipment requirements or an installed sensor audit. Present the conclusion as an evidence-backed statement with a named range of validity. Avoid claiming that a temperature or duty is safe solely because one steady-state expression has a positive numerical solution.

Independent worked case

[Read a fully contextualized numerical case: Turbulent pipe convection regime](../worked-cases/unit-04-case-01.html).

Exercises

1. Starting with Nu=0.023Re^0.8Pr^n; h=Nu*k/D, identify the independent variables and produce a dimensionally consistent derivation or explanation of each term. Give an example where using the wrong unit would change the answer by at least a factor of 100.
2. Repeat the worked experiment with the control variable reduced by 10%. Compute the new output, the percent change, and compare the result to a first-order sensitivity estimate. Explain whether a linear approximation is justified.
3. Check Re, Pr, heating/cooling exponent and whether entrance lengths invalidate the correlation. Describe the minimum sensor set and a method for recording uncertainty in the field.
4. Construct an alternative physical model for the same plant scenario by changing one boundary condition. State which observation could distinguish the two models.
5. Design a sensitivity chart comparing two material-property assumptions. What input dominates the uncertainty at the baseline and why?

Reproducible laboratory

Run [Laboratory 07: Turbulent pipe-side convection](../labs/lab-07.html), export 151 modeled points to CSV, and annotate where the correlation is appropriate. Cite the chosen property source and write an explicit statement about the limits of extrapolation. The laboratory provides a schematic response curve and numerical readout but cannot validate a design without measured plant data.

Unit 4, Lesson 2: External flow and boundary layers

Unit: Convection in process equipment. Core formula: NuL=0.664ReL^0.5Pr^(1/3) (laminar plate). Suggested model: [Laboratory 08 — Laminar external plate convection](../labs/lab-08.html).

Why this problem appears in process plants

Air crosses the exterior of a drying-tray cover. An external boundary layer grows along the surface; average and local heat-transfer coefficients are distinct and depend on transition. This topic is a decision about where energy travels, which mechanism controls its rate, and what evidence is necessary to justify a calculation. The same reported heater duty can correspond to distinct wall temperatures depending on geometry, flow and material properties. A credible engineer must therefore distinguish the energy rate demanded by a process from the thermal resistance that allows the equipment to deliver that rate.

A process operator experiences this distinction as an operating constraint rather than as an equation. When product throughput increases, an unchanged heating utility can be insufficient even when the external wall stays hot. A measurement campaign should record mass flow, inlet and outlet conditions, utility conditions, local wall temperatures when available, and the operating history that may affect fouling. Without these signals, increasing utility temperature may mask the actual cause of poor performance and worsen degradation.

Physical mechanism and assumptions

An external boundary layer grows along the surface; average and local heat-transfer coefficients are distinct and depend on transition. The equation selected here, NuL=0.664ReL^0.5Pr^(1/3) (laminar plate), summarizes a limited physical model. Every term must refer to a declared boundary, an appropriate unit system, and a physically relevant value of temperature or material property. Before using the equation, identify whether the system is steady or transient, whether the fluid undergoes phase change, and whether the heat-transfer coefficient is known from a defensible correlation or simply a fitted number.

The engineer's first drawing should show an inward or outward normal for the relevant surface and a clear sign convention. Heat transport is directed down a temperature driving force in a passive system, although a controlled refrigeration cycle can move heat from a colder source to a hotter sink by supplying work. A temperature sensor located in a mixing chamber is not necessarily the right measurement of the wall temperature immediately adjacent to a process fluid. A fixed temperature boundary also differs from an imposed heat-flux boundary and gives a different mathematical problem.

Derivation and dimensional reasoning

Start from the first-law energy balance, accumulation = net incoming enthalpy + supplied heat + work, with the appropriate sign conventions. Introduce the constitutive expression NuL=0.664ReL^0.5Pr^(1/3) (laminar plate) only after identifying the relevant energy-transfer mechanism. If a rate equation gives watts, confirm that conductivity is in W/(m K), surface area in m² and temperature difference in kelvin for a planar slab. The analogous exercise for a convection coefficient uses W/(m² K). For radiation, temperatures must be absolute because emission depends on the fourth power of thermodynamic temperature.

Terms that look superficially similar can require fundamentally different models. A cylindrical pipe wall does not have a constant radial area; a nonlinear radiation exchange cannot be replaced by a temperature-independent resistance without specifying a linearization point. Correlations expressed with Reynolds and Prandtl numbers need fluid properties at the proper reference temperature and geometry. In a thermally developing flow, the local Nusselt number may differ from its fully developed average. Write these distinctions directly beside the calculation rather than after the final answer.

Worked investigation: laminar external plate convection

We use a numerical teaching model with the following documented inputs: air speed = 2.4; length = 0.3; nu = 1.6e-05; Pr = 0.71; k = 0.026. The primary control air speed is measured in m/s. Other parameters are length [m]; nu [m²/s]; Pr [fraction]; k [W/(m K)]. Using the model equation, the predicted average h is 10.8905 W/(m² K). This is an arithmetic result for assumed properties, not a measured industrial duty or a certified equipment rating.

Increase the primary control from 2.4 to 3.936 m/s while holding the other inputs fixed. The same model predicts 13.9466 W/(m² K). Compare the absolute difference (3.05613 W/(m² K)) with the percentage difference, and interpret the sign using the physical mechanism. Does this change chiefly alter the driving force, the thermal resistance, the material storage, or the rate at which the process moves energy? Explain why interpreting the curve without checking the other controlled inputs could be misleading.

The control range is deliberately finite. At very large or small values a simplified correlation can violate its validity conditions even if the numerical program continues to return finite numbers. For this lab, the important limitation is: Laminar isothermal flat plate; use approximate ReL<500000; leading edge and transition excluded. Decide whether the default state is inside that limitation and state what kind of additional property data, geometry measurement, or experimental coefficient would be needed to increase confidence in the estimate.

Measurement strategy and uncertainty

Evaluate changes in plate length and flow speed while keeping the laminar applicability limit explicit. Select one variable from the worked model that could realistically be measured on a plant floor, one that would normally be inferred, and one that is liable to drift over months of operation. Suggest how each could be independently checked. Thermal couples need installation and dynamic-response consideration, while flow meters require correct fluid properties and calibration. A sensor can be accurate at its measurement point yet irrelevant to the heat-transfer boundary being modeled.

Develop a first-order sensitivity calculation. If the output Y depends on parameter p, estimate dY/dp from two nearby runs of the simulator. If the measured parameter has uncertainty u(p), its approximate output contribution is abs(dY/dp)*u(p). This is a local linear approximation; it does not characterize poorly known geometry, phase-change instability or systematic installation bias. Keep uncertainty in the output's physical units until presenting any final percentage. Where a temperature difference is small, uncertainty can become the dominant proportion of the inferred transfer rate.

Process decision and limitation

Evaluate changes in plate length and flow speed while keeping the laminar applicability limit explicit. An excellent model is still a screening tool if it ignores details that control real equipment performance. A preliminary calculation may support an experiment or identify the direction of a sensitivity, but it cannot substitute for a vendor curve, verified thermophysical properties, design margins, pressure-equipment requirements or an installed sensor audit. Present the conclusion as an evidence-backed statement with a named range of validity. Avoid claiming that a temperature or duty is safe solely because one steady-state expression has a positive numerical solution.

Independent worked case

[Read a fully contextualized numerical case: External plate length and coefficient](../worked-cases/unit-04-case-02.html).

Exercises

1. Starting with NuL=0.664ReL^0.5Pr^(1/3) (laminar plate), identify the independent variables and produce a dimensionally consistent derivation or explanation of each term. Give an example where using the wrong unit would change the answer by at least a factor of 100.
2. Repeat the worked experiment with the control variable reduced by 10%. Compute the new output, the percent change, and compare the result to a first-order sensitivity estimate. Explain whether a linear approximation is justified.
3. Evaluate changes in plate length and flow speed while keeping the laminar applicability limit explicit. Describe the minimum sensor set and a method for recording uncertainty in the field.
4. Construct an alternative physical model for the same plant scenario by changing one boundary condition. State which observation could distinguish the two models.
5. Design a sensitivity chart comparing two material-property assumptions. What input dominates the uncertainty at the baseline and why?

Reproducible laboratory

Run [Laboratory 08: Laminar external plate convection](../labs/lab-08.html), export 151 modeled points to CSV, and annotate where the correlation is appropriate. Cite the chosen property source and write an explicit statement about the limits of extrapolation. The laboratory provides a schematic response curve and numerical readout but cannot validate a design without measured plant data.

Unit 4, Lesson 3: Natural convection

Unit: Convection in process equipment. Core formula: Ra=gbetaabs(dT)L^3/(nualpha). Suggested model: [Laboratory 09 — Natural convection vertical plate](../labs/lab-09.html).

Why this problem appears in process plants

A hot holding tank cools without forced ventilation. Buoyancy drives motion when density variation couples to gravity; Grashof and Rayleigh numbers organize the flow, not an assumed pump velocity. This topic is a decision about where energy travels, which mechanism controls its rate, and what evidence is necessary to justify a calculation. The same reported heater duty can correspond to distinct wall temperatures depending on geometry, flow and material properties. A credible engineer must therefore distinguish the energy rate demanded by a process from the thermal resistance that allows the equipment to deliver that rate.

A process operator experiences this distinction as an operating constraint rather than as an equation. When product throughput increases, an unchanged heating utility can be insufficient even when the external wall stays hot. A measurement campaign should record mass flow, inlet and outlet conditions, utility conditions, local wall temperatures when available, and the operating history that may affect fouling. Without these signals, increasing utility temperature may mask the actual cause of poor performance and worsen degradation.

Physical mechanism and assumptions

Buoyancy drives motion when density variation couples to gravity; Grashof and Rayleigh numbers organize the flow, not an assumed pump velocity. The equation selected here, Ra=gbetaabs(dT)L^3/(nualpha), summarizes a limited physical model. Every term must refer to a declared boundary, an appropriate unit system, and a physically relevant value of temperature or material property. Before using the equation, identify whether the system is steady or transient, whether the fluid undergoes phase change, and whether the heat-transfer coefficient is known from a defensible correlation or simply a fitted number.

The engineer's first drawing should show an inward or outward normal for the relevant surface and a clear sign convention. Heat transport is directed down a temperature driving force in a passive system, although a controlled refrigeration cycle can move heat from a colder source to a hotter sink by supplying work. A temperature sensor located in a mixing chamber is not necessarily the right measurement of the wall temperature immediately adjacent to a process fluid. A fixed temperature boundary also differs from an imposed heat-flux boundary and gives a different mathematical problem.

Derivation and dimensional reasoning

Start from the first-law energy balance, accumulation = net incoming enthalpy + supplied heat + work, with the appropriate sign conventions. Introduce the constitutive expression Ra=gbetaabs(dT)L^3/(nualpha) only after identifying the relevant energy-transfer mechanism. If a rate equation gives watts, confirm that conductivity is in W/(m K), surface area in m² and temperature difference in kelvin for a planar slab. The analogous exercise for a convection coefficient uses W/(m² K). For radiation, temperatures must be absolute because emission depends on the fourth power of thermodynamic temperature.

Terms that look superficially similar can require fundamentally different models. A cylindrical pipe wall does not have a constant radial area; a nonlinear radiation exchange cannot be replaced by a temperature-independent resistance without specifying a linearization point. Correlations expressed with Reynolds and Prandtl numbers need fluid properties at the proper reference temperature and geometry. In a thermally developing flow, the local Nusselt number may differ from its fully developed average. Write these distinctions directly beside the calculation rather than after the final answer.

Worked investigation: natural convection vertical plate

We use a numerical teaching model with the following documented inputs: surface ΔT = 40; length = 0.65; Tfilm = 325; nu = 1.9e-05; alpha = 2.7e-05; k = 0.028; Pr = 0.71. The primary control surface ΔT is measured in K. Other parameters are length [m]; Tfilm [K]; nu [m²/s]; alpha [m²/s]; k [W/(m K)]; Pr [fraction]. Using the model equation, the predicted average h is 4.62684 W/(m² K). This is an arithmetic result for assumed properties, not a measured industrial duty or a certified equipment rating.

Increase the primary control from 40 to 54.4 K while holding the other inputs fixed. The same model predicts 5.08553 W/(m² K). Compare the absolute difference (0.458692 W/(m² K)) with the percentage difference, and interpret the sign using the physical mechanism. Does this change chiefly alter the driving force, the thermal resistance, the material storage, or the rate at which the process moves energy? Explain why interpreting the curve without checking the other controlled inputs could be misleading.

The control range is deliberately finite. At very large or small values a simplified correlation can violate its validity conditions even if the numerical program continues to return finite numbers. For this lab, the important limitation is: Ideal vertical isothermal plate; buoyancy Boussinesq β≈1/Tfilm. Decide whether the default state is inside that limitation and state what kind of additional property data, geometry measurement, or experimental coefficient would be needed to increase confidence in the estimate.

Measurement strategy and uncertainty

Contrast stable vertical-wall buoyancy with an enclosed horizontal gap where stratification matters. Select one variable from the worked model that could realistically be measured on a plant floor, one that would normally be inferred, and one that is liable to drift over months of operation. Suggest how each could be independently checked. Thermal couples need installation and dynamic-response consideration, while flow meters require correct fluid properties and calibration. A sensor can be accurate at its measurement point yet irrelevant to the heat-transfer boundary being modeled.

Develop a first-order sensitivity calculation. If the output Y depends on parameter p, estimate dY/dp from two nearby runs of the simulator. If the measured parameter has uncertainty u(p), its approximate output contribution is abs(dY/dp)*u(p). This is a local linear approximation; it does not characterize poorly known geometry, phase-change instability or systematic installation bias. Keep uncertainty in the output's physical units until presenting any final percentage. Where a temperature difference is small, uncertainty can become the dominant proportion of the inferred transfer rate.

Process decision and limitation

Contrast stable vertical-wall buoyancy with an enclosed horizontal gap where stratification matters. An excellent model is still a screening tool if it ignores details that control real equipment performance. A preliminary calculation may support an experiment or identify the direction of a sensitivity, but it cannot substitute for a vendor curve, verified thermophysical properties, design margins, pressure-equipment requirements or an installed sensor audit. Present the conclusion as an evidence-backed statement with a named range of validity. Avoid claiming that a temperature or duty is safe solely because one steady-state expression has a positive numerical solution.

Independent worked case

[Read a fully contextualized numerical case: Buoyancy versus imposed airflow](../worked-cases/unit-04-case-03.html).

Exercises

1. Starting with Ra=gbetaabs(dT)L^3/(nualpha), identify the independent variables and produce a dimensionally consistent derivation or explanation of each term. Give an example where using the wrong unit would change the answer by at least a factor of 100.
2. Repeat the worked experiment with the control variable reduced by 10%. Compute the new output, the percent change, and compare the result to a first-order sensitivity estimate. Explain whether a linear approximation is justified.
3. Contrast stable vertical-wall buoyancy with an enclosed horizontal gap where stratification matters. Describe the minimum sensor set and a method for recording uncertainty in the field.
4. Construct an alternative physical model for the same plant scenario by changing one boundary condition. State which observation could distinguish the two models.
5. Design a sensitivity chart comparing two material-property assumptions. What input dominates the uncertainty at the baseline and why?

Reproducible laboratory

Run [Laboratory 09: Natural convection vertical plate](../labs/lab-09.html), export 151 modeled points to CSV, and annotate where the correlation is appropriate. Cite the chosen property source and write an explicit statement about the limits of extrapolation. The laboratory provides a schematic response curve and numerical readout but cannot validate a design without measured plant data.

Unit 4, Lesson 4: Fins and extended surfaces

Unit: Convection in process equipment. Core formula: eta_fin=tanh(mL)/(mL); m=sqrt(hP/(kAc)). Suggested model: [Laboratory 04 — Straight-fin efficiency](../labs/lab-04.html).

Why this problem appears in process plants

An air-cooled control-panel heat sink is compared with a flat plate. Fins increase area but their effectiveness is limited by axial temperature gradients and contact resistance; extra area is not uniformly at base temperature. This topic is a decision about where energy travels, which mechanism controls its rate, and what evidence is necessary to justify a calculation. The same reported heater duty can correspond to distinct wall temperatures depending on geometry, flow and material properties. A credible engineer must therefore distinguish the energy rate demanded by a process from the thermal resistance that allows the equipment to deliver that rate.

A process operator experiences this distinction as an operating constraint rather than as an equation. When product throughput increases, an unchanged heating utility can be insufficient even when the external wall stays hot. A measurement campaign should record mass flow, inlet and outlet conditions, utility conditions, local wall temperatures when available, and the operating history that may affect fouling. Without these signals, increasing utility temperature may mask the actual cause of poor performance and worsen degradation.

Physical mechanism and assumptions

Fins increase area but their effectiveness is limited by axial temperature gradients and contact resistance; extra area is not uniformly at base temperature. The equation selected here, eta_fin=tanh(mL)/(mL); m=sqrt(hP/(kAc)), summarizes a limited physical model. Every term must refer to a declared boundary, an appropriate unit system, and a physically relevant value of temperature or material property. Before using the equation, identify whether the system is steady or transient, whether the fluid undergoes phase change, and whether the heat-transfer coefficient is known from a defensible correlation or simply a fitted number.

The engineer's first drawing should show an inward or outward normal for the relevant surface and a clear sign convention. Heat transport is directed down a temperature driving force in a passive system, although a controlled refrigeration cycle can move heat from a colder source to a hotter sink by supplying work. A temperature sensor located in a mixing chamber is not necessarily the right measurement of the wall temperature immediately adjacent to a process fluid. A fixed temperature boundary also differs from an imposed heat-flux boundary and gives a different mathematical problem.

Derivation and dimensional reasoning

Start from the first-law energy balance, accumulation = net incoming enthalpy + supplied heat + work, with the appropriate sign conventions. Introduce the constitutive expression eta_fin=tanh(mL)/(mL); m=sqrt(hP/(kAc)) only after identifying the relevant energy-transfer mechanism. If a rate equation gives watts, confirm that conductivity is in W/(m K), surface area in m² and temperature difference in kelvin for a planar slab. The analogous exercise for a convection coefficient uses W/(m² K). For radiation, temperatures must be absolute because emission depends on the fourth power of thermodynamic temperature.

Terms that look superficially similar can require fundamentally different models. A cylindrical pipe wall does not have a constant radial area; a nonlinear radiation exchange cannot be replaced by a temperature-independent resistance without specifying a linearization point. Correlations expressed with Reynolds and Prandtl numbers need fluid properties at the proper reference temperature and geometry. In a thermally developing flow, the local Nusselt number may differ from its fully developed average. Write these distinctions directly beside the calculation rather than after the final answer.

Worked investigation: straight-fin efficiency

We use a numerical teaching model with the following documented inputs: fin length = 0.045; h = 35; k = 185; width = 0.06; thick = 0.003. The primary control fin length is measured in m. Other parameters are h [W/(m² K)]; k [W/(m K)]; width [m]; thick [m]. Using the model equation, the predicted fin efficiency is 0.919258 fraction. This is an arithmetic result for assumed properties, not a measured industrial duty or a certified equipment rating.

Increase the primary control from 0.045 to 0.057 m while holding the other inputs fixed. The same model predicts 0.8776 fraction. Compare the absolute difference (-0.0416588 fraction) with the percentage difference, and interpret the sign using the physical mechanism. Does this change chiefly alter the driving force, the thermal resistance, the material storage, or the rate at which the process moves energy? Explain why interpreting the curve without checking the other controlled inputs could be misleading.

The control range is deliberately finite. At very large or small values a simplified correlation can violate its validity conditions even if the numerical program continues to return finite numbers. For this lab, the important limitation is: Constant properties; adiabatic fin tip; uniform rectangular cross-section. Decide whether the default state is inside that limitation and state what kind of additional property data, geometry measurement, or experimental coefficient would be needed to increase confidence in the estimate.

Measurement strategy and uncertainty

Evaluate fin efficiency, base temperature and whether airflow bypass limits useful area. Select one variable from the worked model that could realistically be measured on a plant floor, one that would normally be inferred, and one that is liable to drift over months of operation. Suggest how each could be independently checked. Thermal couples need installation and dynamic-response consideration, while flow meters require correct fluid properties and calibration. A sensor can be accurate at its measurement point yet irrelevant to the heat-transfer boundary being modeled.

Develop a first-order sensitivity calculation. If the output Y depends on parameter p, estimate dY/dp from two nearby runs of the simulator. If the measured parameter has uncertainty u(p), its approximate output contribution is abs(dY/dp)*u(p). This is a local linear approximation; it does not characterize poorly known geometry, phase-change instability or systematic installation bias. Keep uncertainty in the output's physical units until presenting any final percentage. Where a temperature difference is small, uncertainty can become the dominant proportion of the inferred transfer rate.

Process decision and limitation

Evaluate fin efficiency, base temperature and whether airflow bypass limits useful area. An excellent model is still a screening tool if it ignores details that control real equipment performance. A preliminary calculation may support an experiment or identify the direction of a sensitivity, but it cannot substitute for a vendor curve, verified thermophysical properties, design margins, pressure-equipment requirements or an installed sensor audit. Present the conclusion as an evidence-backed statement with a named range of validity. Avoid claiming that a temperature or duty is safe solely because one steady-state expression has a positive numerical solution.

Independent worked case

[Read a fully contextualized numerical case: Practical fin efficiency](../worked-cases/unit-04-case-04.html).

Exercises

1. Starting with eta_fin=tanh(mL)/(mL); m=sqrt(hP/(kAc)), identify the independent variables and produce a dimensionally consistent derivation or explanation of each term. Give an example where using the wrong unit would change the answer by at least a factor of 100.
2. Repeat the worked experiment with the control variable reduced by 10%. Compute the new output, the percent change, and compare the result to a first-order sensitivity estimate. Explain whether a linear approximation is justified.
3. Evaluate fin efficiency, base temperature and whether airflow bypass limits useful area. Describe the minimum sensor set and a method for recording uncertainty in the field.
4. Construct an alternative physical model for the same plant scenario by changing one boundary condition. State which observation could distinguish the two models.
5. Design a sensitivity chart comparing two material-property assumptions. What input dominates the uncertainty at the baseline and why?

Reproducible laboratory

Run [Laboratory 04: Straight-fin efficiency](../labs/lab-04.html), export 151 modeled points to CSV, and annotate where the correlation is appropriate. Cite the chosen property source and write an explicit statement about the limits of extrapolation. The laboratory provides a schematic response curve and numerical readout but cannot validate a design without measured plant data.

Applied project 04: Convection in process equipment

Project brief

Convection couples diffusion near a wall to bulk fluid motion. The film coefficient h is not a material constant. It summarizes the temperature field for a specific flow, geometry and property set. A published correlation can give a precise number while being inapplicable to a developing, fouled or multiphase stream. Build a small but defensible engineering evaluation for a hypothetical process facility. The target is not an equipment purchase specification: it is a traceable analysis of what would need to be known before design or procurement. Select internal, external and natural-convection models by flow regime, geometry and property range, and quantify correlation uncertainty.

Process case and data plan

The scenario begins with incomplete information. A process technician has measured bulk inlet and outlet temperatures, approximate flow and a surface temperature, while operations has an estimate of annual hours. Choose one of the four lesson topics below and define a physically plausible dataset with at least eight rows. Use SI units. You may use representative approximate fluid properties if you label them as assumptions; do not describe fabricated samples as field measurements.

• Option 1: Internal forced convection. A cooled solvent flows through an instrumented smooth tube. Check Re, Pr, heating/cooling exponent and whether entrance lengths invalidate the correlation. Begin with Nu=0.023Re^0.8Pr^n; h=Nu*k/D and then examine which neglected effect might dominate.
• Option 2: External flow and boundary layers. Air crosses the exterior of a drying-tray cover. Evaluate changes in plate length and flow speed while keeping the laminar applicability limit explicit. Begin with NuL=0.664ReL^0.5Pr^(1/3) (laminar plate) and then examine which neglected effect might dominate.
• Option 3: Natural convection. A hot holding tank cools without forced ventilation. Contrast stable vertical-wall buoyancy with an enclosed horizontal gap where stratification matters. Begin with Ra=gbetaabs(dT)L^3/(nualpha) and then examine which neglected effect might dominate.
• Option 4: Fins and extended surfaces. An air-cooled control-panel heat sink is compared with a flat plate. Evaluate fin efficiency, base temperature and whether airflow bypass limits useful area. Begin with eta_fin=tanh(mL)/(mL); m=sqrt(hP/(kAc)) and then examine which neglected effect might dominate.

Work packages

A — Model the thermal circuit. Produce a diagram of thermal boundaries and flow directions. Record each resistance, energy-storage term, or imposed heat flux. Separate first-law calculations from empirical closure equations. Verify one result using an independent algebraic or numerical method.

B — Establish a baseline. Choose reasonable conditions, calculate heat duty in watts or kilowatts and identify one required material property that is not supplied. Define how a student would obtain that property from an authoritative source and which temperature or pressure governs the lookup.

C — Experiment and plot. Run two relevant interactive laboratories and export the curves. Compare at least three operating points, keeping all but one input fixed at a time. Include a table with the model assumption, the changed parameter and the resulting percentage change. Explain why some apparent trends would fail in a full system because stream capacity rates, pressure drop or properties co-vary.

D — Measurement and risk. Write a list of five instruments or process observations necessary to validate the simplified model. Include likely systematic and random errors, a rough range of plausible input uncertainty, and a note about unavailable information such as surface condition or contact resistance. Do not infer any pressure-vessel or worker-safety adequacy from the simulator.

E — Decision memorandum. Provide a two-page recommendation that ranks measurement, retrofit and operating-control options by estimated benefit and uncertainty. For energy economics, use explicit unit price, operating hours and equipment efficiency assumptions. Highlight which conclusions would change under a realistic alternative case.

Acceptance criteria

Deliver a readable process schematic, a table of inputs and sources, equation derivations with units, one independent verification, at least two experimental response plots, a measurement strategy and a decision memo. A technically unsupported precise number is not acceptable. The model is a teaching artifact, not a certified plant design.

Worked case 4.1: Turbulent pipe convection regime

Process situation and calculation

A waterlike stream enters a 25 mm diameter tube at 1.5 m/s with density 998 kg/m³ and viscosity 0.001 Pa·s. Reynolds number is 998×1.5×0.025/0.001≈37,425; Prandtl number with cp=4180 J/(kg K) and k=0.6 W/(m K) is 6.97. Dittus–Boelter gives a heating Nusselt estimate around 0.023×Re^0.8×Pr^0.4 ≈ 228.

Physical interpretation and possible error

That value implies h≈Nu k/D≈5470 W/(m² K). The correlation still requires an adequately developed turbulent flow, appropriate property basis and smooth-tube geometry. In an industrial exchanger, shell-side leakage or fouling may dominate the total thermal resistance.

Verification tasks

1. Annotate a sketch with heat-flow directions and temperature measurement positions. Explain whether the quoted numerical value is a rate, a transport coefficient, a stored-energy scale or a transient timescale.
2. Repeat the calculation after changing one physical parameter by 10%. Identify whether the result responds linearly, logarithmically or with a power law, and derive that dependence rather than relying only on a graph.
3. State which property or correlation requires independent external verification. What uncertainty would a surface thermocouple, flow meter or geometry measurement introduce?
4. Describe the limiting condition under which the numerical relationship is not meaningful. Do not generalize beyond its stated geometry or thermal regime.

Scope: This is an illustrative, parameter-specific calculation for education, not an equipment rating or validation of actual process data.

Worked case 4.2: External plate length and coefficient

Process situation and calculation

An air stream at 3 m/s flows over an approximately isothermal plate 0.3 m long. For nu=1.6×10⁻⁵ m²/s, Reynolds number is 56,250, within a laminar flat-plate screening regime. With Pr=0.71 and k=0.026 W/(m K), average Nu≈0.664√56250×0.71^(1/3)≈141 and h≈12.2 W/(m² K).

Physical interpretation and possible error

If the plate length doubles at the same speed, average h decreases roughly as L^(-1/2) in this regime, even while total contact area increases. A single average coefficient should not be used as though every point on the plate has identical local flux.

Verification tasks

1. Annotate a sketch with heat-flow directions and temperature measurement positions. Explain whether the quoted numerical value is a rate, a transport coefficient, a stored-energy scale or a transient timescale.
2. Repeat the calculation after changing one physical parameter by 10%. Identify whether the result responds linearly, logarithmically or with a power law, and derive that dependence rather than relying only on a graph.
3. State which property or correlation requires independent external verification. What uncertainty would a surface thermocouple, flow meter or geometry measurement introduce?
4. Describe the limiting condition under which the numerical relationship is not meaningful. Do not generalize beyond its stated geometry or thermal regime.

Scope: This is an illustrative, parameter-specific calculation for education, not an equipment rating or validation of actual process data.

Worked case 4.3: Buoyancy versus imposed airflow

Process situation and calculation

A warm enclosure wall transfers heat to still room air; a worker opens a ventilation door and induces a crossflow. In the first case a natural-convection correlation based on Rayleigh number may be appropriate. In the second, a mixed or forced-convection treatment may be required.

Physical interpretation and possible error

Do not add natural- and forced-convection coefficients without an explicitly justified mixed-convection method. Flow direction relative to buoyancy, wall orientation and temperature differences affect which mechanism controls the local boundary layer.

Verification tasks

1. Annotate a sketch with heat-flow directions and temperature measurement positions. Explain whether the quoted numerical value is a rate, a transport coefficient, a stored-energy scale or a transient timescale.
2. Repeat the calculation after changing one physical parameter by 10%. Identify whether the result responds linearly, logarithmically or with a power law, and derive that dependence rather than relying only on a graph.
3. State which property or correlation requires independent external verification. What uncertainty would a surface thermocouple, flow meter or geometry measurement introduce?
4. Describe the limiting condition under which the numerical relationship is not meaningful. Do not generalize beyond its stated geometry or thermal regime.

Scope: This is an illustrative, parameter-specific calculation for education, not an equipment rating or validation of actual process data.

Worked case 4.4: Practical fin efficiency

Process situation and calculation

For a slender rectangular aluminum fin, increased length initially increases heat-transfer area but also produces a larger axial temperature drop. At high h, a much longer fin can have lower efficiency because its tip approaches ambient temperature. The effective additional area is efficiency multiplied by fin area, not geometric fin area alone.

Physical interpretation and possible error

Compare material cost, fan power and manufacturing tolerance when choosing fin pitch. Closely spaced fins may obstruct airflow and invalidate the original assumed film coefficient, despite appearing to offer more geometric area.

Verification tasks

1. Annotate a sketch with heat-flow directions and temperature measurement positions. Explain whether the quoted numerical value is a rate, a transport coefficient, a stored-energy scale or a transient timescale.
2. Repeat the calculation after changing one physical parameter by 10%. Identify whether the result responds linearly, logarithmically or with a power law, and derive that dependence rather than relying only on a graph.
3. State which property or correlation requires independent external verification. What uncertainty would a surface thermocouple, flow meter or geometry measurement introduce?
4. Describe the limiting condition under which the numerical relationship is not meaningful. Do not generalize beyond its stated geometry or thermal regime.

Scope: This is an illustrative, parameter-specific calculation for education, not an equipment rating or validation of actual process data.

Laboratory 07 — Turbulent pipe-side convection

Linked unit: 4. Measured output: Estimated h (W/(m² K)). Equation: Nu=0.023 Re^0.8 Pr^0.4; h=Nu k/D.

Experiment objective and theory

This interactive experiment isolates one thermal relationship so that parameter sensitivity can be observed without the distraction of a full process model. It does not solve an entire plant; the objective is to learn which inputs the mathematical expression regards as controlling, which inputs the engineer would have to obtain from measurements, and which unmodeled effects could reverse an apparent benefit. Start by writing the direction of expected heat transfer. For an entirely passive system, check that the response is compatible with the stated hot and cold temperatures and the sign convention.

Validity boundary: Dittus–Boelter for smooth, fully developed turbulent internal flow (use only Re≥10000 and Pr≈0.7–160). The simplified correlation treats the remaining conditions as fixed even if that assumption would not survive an actual change in flow, phase or material temperature. If the laboratory displays a finite number at an invalid combination, do not call that number a validated prediction. Temperature-dependent properties should be evaluated at a defensible reference state; this model deliberately does not call a property database.

Default test and sensitivity measurements

Primary variable: bulk velocity, unit m/s. Initial value: 2 m/s. Predicted output: 6175.218 W/(m² K). Your result may round on the web page. Confirm the orders of magnitude independently with the printed equation.

At the first selected test value 1.24 m/s, the reference model gives 4212.749 W/(m² K). At the second test value 4.06 m/s, it gives 10880.51 W/(m² K). Do not draw a straight-line conclusion from these two numbers without checking the full response curve. Export the CSV, calculate a central finite difference near the default value, and compare it with the curvature of the graph. Keep the parameter-change interval small enough for a sensitivity calculation but large enough to avoid rounding effects.

Controlled-variable experiment

1. Record the baseline output and plot before touching any slider. Explain whether a temperature difference, resistance, heat capacity or dimensionless correlation establishes the result.
2. Move the primary slider to its low and high limits. Identify the graph region where an extrapolation of the physics would be most dangerous.
3. Restore the baseline. Change one secondary material or geometry parameter by approximately 10%. State the expected direction before observing it; then describe the difference between your prediction and the numerical response.
4. Use reset to reproduce the default experiment, and export the CSV. Check the exported data for units, sign and monotonicity. A curve with a sharp numerical discontinuity should be investigated rather than smoothed away automatically.
5. Prepare a four-column comparison of baseline, adjusted input, adjusted output and physical interpretation. Finish with a statement of the lab model's limitations and what one real measurement would validate it.

Interpretation and field extension

A computational curve is a relationship under controlled assumptions, not a set of experimental observations. In a plant, thermal conductivity, viscosity and heat capacity change with temperature; fouling deposits grow in time; and equipment area can be effectively reduced by flow maldistribution. Explain whether the chosen experiment can distinguish these effects. If not, propose a paired measurement, such as wall temperatures and stream flow, that creates additional evidence. Never claim improved process safety or energy economics solely from a single slider result.

Questions: (a) Which input appears with the highest exponent in the local law? (b) Could an unmeasured contact resistance mimic the same trend? (c) Does the graph represent a local coefficient, an area-average value or a whole-unit thermal duty? (d) What would an alternative geometry change? (e) What independent instrument or property reference would be required before industrial scale-up?

Laboratory 08 — Laminar external plate convection

Linked unit: 4. Measured output: Average h (W/(m² K)). Equation: Nu=0.664 sqrt(ReL) Pr^(1/3); h=Nu k/L.

Experiment objective and theory

This interactive experiment isolates one thermal relationship so that parameter sensitivity can be observed without the distraction of a full process model. It does not solve an entire plant; the objective is to learn which inputs the mathematical expression regards as controlling, which inputs the engineer would have to obtain from measurements, and which unmodeled effects could reverse an apparent benefit. Start by writing the direction of expected heat transfer. For an entirely passive system, check that the response is compatible with the stated hot and cold temperatures and the sign convention.

Validity boundary: Laminar isothermal flat plate; use approximate ReL<500000; leading edge and transition excluded. The simplified correlation treats the remaining conditions as fixed even if that assumption would not survive an actual change in flow, phase or material temperature. If the laboratory displays a finite number at an invalid combination, do not call that number a validated prediction. Temperature-dependent properties should be evaluated at a defensible reference state; this model deliberately does not call a property database.

Default test and sensitivity measurements

Primary variable: air speed, unit m/s. Initial value: 2.4 m/s. Predicted output: 10.89045 W/(m² K). Your result may round on the web page. Confirm the orders of magnitude independently with the printed equation.

At the first selected test value 2.56 m/s, the reference model gives 11.24761 W/(m² K). At the second test value 9.64 m/s, it gives 21.82623 W/(m² K). Do not draw a straight-line conclusion from these two numbers without checking the full response curve. Export the CSV, calculate a central finite difference near the default value, and compare it with the curvature of the graph. Keep the parameter-change interval small enough for a sensitivity calculation but large enough to avoid rounding effects.

Controlled-variable experiment

1. Record the baseline output and plot before touching any slider. Explain whether a temperature difference, resistance, heat capacity or dimensionless correlation establishes the result.
2. Move the primary slider to its low and high limits. Identify the graph region where an extrapolation of the physics would be most dangerous.
3. Restore the baseline. Change one secondary material or geometry parameter by approximately 10%. State the expected direction before observing it; then describe the difference between your prediction and the numerical response.
4. Use reset to reproduce the default experiment, and export the CSV. Check the exported data for units, sign and monotonicity. A curve with a sharp numerical discontinuity should be investigated rather than smoothed away automatically.
5. Prepare a four-column comparison of baseline, adjusted input, adjusted output and physical interpretation. Finish with a statement of the lab model's limitations and what one real measurement would validate it.

Interpretation and field extension

A computational curve is a relationship under controlled assumptions, not a set of experimental observations. In a plant, thermal conductivity, viscosity and heat capacity change with temperature; fouling deposits grow in time; and equipment area can be effectively reduced by flow maldistribution. Explain whether the chosen experiment can distinguish these effects. If not, propose a paired measurement, such as wall temperatures and stream flow, that creates additional evidence. Never claim improved process safety or energy economics solely from a single slider result.

Questions: (a) Which input appears with the highest exponent in the local law? (b) Could an unmeasured contact resistance mimic the same trend? (c) Does the graph represent a local coefficient, an area-average value or a whole-unit thermal duty? (d) What would an alternative geometry change? (e) What independent instrument or property reference would be required before industrial scale-up?

Unit 04 — Assessment: Convection in process equipment

Instructions: Work at least three problems with complete units, a diagram, assumptions and a sensitivity check. These questions emphasize independent explanation rather than recognizing a repeated numerical template.

Topic 4.1: Internal forced convection

1. Derivation: Starting with `Nu=0.023*Re^0.8*Pr^n; h=Nu*k/D`, derive the expression from stated physical assumptions or explain the empirical closure that prevents a first-principles derivation. Define the system boundary and identify which values are measurable.

2. Engineering investigation: A cooled solvent flows through an instrumented smooth tube. Produce two competing hypotheses that could explain a 12% deviation between the calculated heat duty and the measured duty; for each give an instrument or a controlled test.

3. Critical evaluation: Check Re, Pr, heating/cooling exponent and whether entrance lengths invalidate the correlation. State how the result would change for a tenfold increase in one relevant physical length, and whether the original correlation remains valid.

Topic 4.2: External flow and boundary layers

1. Derivation: Starting with `NuL=0.664*ReL^0.5*Pr^(1/3) (laminar plate)`, derive the expression from stated physical assumptions or explain the empirical closure that prevents a first-principles derivation. Define the system boundary and identify which values are measurable.

2. Engineering investigation: Air crosses the exterior of a drying-tray cover. Produce two competing hypotheses that could explain a 12% deviation between the calculated heat duty and the measured duty; for each give an instrument or a controlled test.

3. Critical evaluation: Evaluate changes in plate length and flow speed while keeping the laminar applicability limit explicit. State how the result would change for a tenfold increase in one relevant physical length, and whether the original correlation remains valid.

Topic 4.3: Natural convection

1. Derivation: Starting with `Ra=g*beta*abs(dT)*L^3/(nu*alpha)`, derive the expression from stated physical assumptions or explain the empirical closure that prevents a first-principles derivation. Define the system boundary and identify which values are measurable.

2. Engineering investigation: A hot holding tank cools without forced ventilation. Produce two competing hypotheses that could explain a 12% deviation between the calculated heat duty and the measured duty; for each give an instrument or a controlled test.

3. Critical evaluation: Contrast stable vertical-wall buoyancy with an enclosed horizontal gap where stratification matters. State how the result would change for a tenfold increase in one relevant physical length, and whether the original correlation remains valid.

Topic 4.4: Fins and extended surfaces

1. Derivation: Starting with `eta_fin=tanh(m*L)/(m*L); m=sqrt(h*P/(k*Ac))`, derive the expression from stated physical assumptions or explain the empirical closure that prevents a first-principles derivation. Define the system boundary and identify which values are measurable.

2. Engineering investigation: An air-cooled control-panel heat sink is compared with a flat plate. Produce two competing hypotheses that could explain a 12% deviation between the calculated heat duty and the measured duty; for each give an instrument or a controlled test.

3. Critical evaluation: Evaluate fin efficiency, base temperature and whether airflow bypass limits useful area. State how the result would change for a tenfold increase in one relevant physical length, and whether the original correlation remains valid.

Assessment rubric

Conceptual accuracy (30%): state the correct physics, not just a numerical recipe. Traceable units and sign conventions (20%): make every conversion explicit. Sensitivity and validity (25%): quantify a change and identify non-applicable regimes. Engineering communication (25%): diagrams, model limits and measurements must support the recommendation.

The exercises are deliberately open; different property assumptions can produce different numerical outcomes. Clearly distinguish supplied values, literature property values and measurements. A plausible order of magnitude with transparent limitations is more useful than an unjustifiably precise number.