1. CFD Foundations, Scope and Engineering Workflow

Introduction and Context

Computational Fluid Dynamics (CFD) is the study of fluid flow, heat transfer, mass transfer, and related physical phenomena through numerical solution of governing equations. It is not merely the use of a software package to display coloured contours. CFD is an engineering discipline that combines fluid mechanics, thermodynamics, heat transfer, numerical mathematics, computer science, and physical interpretation.

The central attraction of CFD is that it allows engineers to investigate a flow field in considerable detail without constructing and testing every physical design. Pressure, velocity, temperature, species concentration, vorticity, wall shear stress, and many other quantities can be estimated throughout a computational domain. This capability supports design exploration, diagnosis of failures, performance improvement, and safety assessment.

At the same time, CFD does not eliminate the need for theory, experiments, or engineering judgement. A numerical solution is always based on a mathematical model, a set of assumptions, input data, boundary conditions, a discretisation procedure, and a solution algorithm. Errors or unjustified assumptions at any of these stages can produce a visually convincing but physically misleading result. The essential competence in CFD therefore lies not only in operating a solver but also in deciding what should be solved, how it should be represented, and whether the results are credible.

This chapter establishes the conceptual and practical foundation for subsequent study. It introduces the language of CFD, places the method within the historical development of fluid mechanics, compares analytical, experimental, and computational approaches, and presents an engineering workflow from problem definition to reporting.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Define CFD and explain its relationship with fluid mechanics and numerical methods.

2. Describe the principal stages in the historical development of computational fluid analysis.

3. Distinguish analytical, experimental, and computational approaches to engineering investigation.

4. Explain essential CFD terminology, including domain, mesh, discretisation, solver, model, convergence, verification, and validation.

5. Formulate a CFD problem by identifying objectives, assumptions, governing physics, material properties, and boundary conditions.

6. Structure a defensible workflow for preprocessing, solution, post-processing, interpretation, and reporting.

7. Recognise major sources of error, uncertainty, and misuse in CFD.

8. Evaluate the professional, industrial, and emerging digital roles of CFD.

From Governing Physics to Numerical Prediction

Most CFD analyses begin with conservation laws. Conservation of mass produces the continuity equation; conservation of momentum produces the Navier–Stokes equations; conservation of energy produces an energy equation; and conservation of chemical species produces species transport equations. These equations may be supplemented by constitutive relations, such as Newton’s law of viscosity, equations of state, turbulence models, combustion models, or multiphase-flow closures.

For a Newtonian fluid, the governing equations express relationships among velocity, pressure, density, temperature, viscosity, and external forces. In a simple incompressible flow, the continuity equation may be written as

∇ · u = 0,

where u is the velocity vector. The momentum equation can be represented in vector form as

ρ(∂u/∂t + u · ∇u) = −∇p + ∇ · τ + ρf,

where ρ is density, p is pressure, τ is the viscous stress tensor, and f represents body force per unit mass. The equations are generally nonlinear because of the convective term u · ∇u. Their solution becomes especially difficult when the geometry is complex, the flow is turbulent, or several physical processes are strongly coupled.

A CFD model transforms these continuous equations into a finite set of algebraic equations. The computational domain is divided into small elements or control volumes, and the governing equations are integrated or approximated over these discrete locations. The resulting equations are solved iteratively or, in some cases, directly. The computed values are then reconstructed as fields and interpreted in relation to the engineering objective.

This transformation involves three distinct ideas. First, the physical model determines which equations and constitutive descriptions are appropriate. Second, the numerical method determines how those equations are represented and solved. Third, the engineering interpretation determines whether the results answer the original question. Confusing these levels is a common source of error. A highly accurate numerical solution of an inappropriate physical model is still an inappropriate engineering result.

Historical Development and Changing Scope

The intellectual roots of CFD lie in classical fluid mechanics. The Euler and Navier–Stokes equations were established during the eighteenth and nineteenth centuries, while analytical methods were developed for idealised flows. Exact solutions were obtained for special cases such as fully developed pipe flow, creeping flow, and boundary-layer approximations. These solutions remain valuable because they provide insight and benchmarks, but most industrial geometries do not permit closed-form solutions.

The growth of numerical analysis and electronic computing during the twentieth century made approximate solutions of complex differential equations possible. Early computational fluid studies addressed problems such as supersonic flow and atmospheric motion. Finite-difference methods were initially prominent, especially for structured grids. Later, finite-volume methods became widely used in engineering because they enforce conservation naturally over discrete control volumes. Finite-element methods, originally influential in structural mechanics, were adapted to many flow and transport problems, particularly those involving complex geometries and multiphysics.

The development of turbulence modelling substantially expanded CFD capability. Direct Numerical Simulation resolves all relevant turbulent scales but requires computational resources that are generally impractical for industrial Reynolds numbers. Large-Eddy Simulation resolves large turbulent structures and models smaller scales, while Reynolds-Averaged Navier–Stokes methods model the effects of turbulence through averaged equations and closure models. These approaches involve different costs, assumptions, and levels of physical resolution.

Modern CFD has progressed through increased computing power, parallel processing, high-performance computing, improved meshing, advanced linear solvers, adaptive methods, and integrated computer-aided engineering platforms. It now supports external aerodynamics, turbomachinery, internal flows, thermal management, combustion, environmental flows, biomedical flows, microfluidics, additive manufacturing, and fluid–structure interaction. Nevertheless, greater computational power does not remove modelling uncertainty. It makes more detailed calculations possible, but it does not by itself establish that the calculations are correct.

Three Ways of Knowing Flow Behaviour

Analytical methods

Analytical analysis seeks a mathematical solution in closed form or through controlled approximations. Its principal strength is explanatory power. A derivation can reveal how a quantity depends on dimensionless parameters, boundary conditions, or geometry. Analytical results are often inexpensive to obtain and are excellent for checking numerical codes and developing physical intuition.

Its limitation is that real engineering problems frequently involve irregular geometry, separation, turbulence, moving interfaces, or coupled heat and mass transfer. Simplifying assumptions may then be too restrictive. Analytical methods should therefore not be regarded as obsolete; rather, they provide the conceptual and benchmarking foundation against which more complex calculations can be judged.

Experimental methods

Experiments measure the behaviour of a physical system or a scaled model. They include wind-tunnel tests, water-channel experiments, rotating machinery rigs, engine tests, particle-image velocimetry, pressure taps, thermocouples, and laser-based diagnostics. Experiments capture real physical effects, including some phenomena that have not been adequately modelled.

However, experiments can be expensive, time-consuming, and difficult to instrument. Measurements may disturb the flow, and scale models require careful attention to similarity, including Reynolds, Mach, Froude, and other relevant dimensionless numbers. Experimental data also contain uncertainty associated with calibration, spatial resolution, temporal resolution, and repeatability.

Computational methods

CFD occupies an intermediate and complementary position. It can provide complete spatial fields and examine conditions that are difficult or dangerous to reproduce experimentally. Design variants can often be evaluated rapidly once a reliable workflow has been established. The method is particularly powerful for parametric studies and for examining quantities that are difficult to measure.

Its limitations arise from discretisation error, iterative error, round-off error, uncertain input data, incomplete physical models, and inappropriate boundary conditions. A CFD result is therefore a prediction conditioned by a model, not a direct observation of nature. The strongest engineering practice combines all three approaches: analytical reasoning for simplification and checks, experiments for physical evidence, and CFD for detailed prediction and exploration.

Essential CFD Language

A computational domain is the portion of physical space represented in the simulation. It may be an internal passage, such as a pipe or turbine blade channel, or an external region surrounding a vehicle or building. The domain boundaries require boundary conditions.

A mesh or grid is the collection of discrete cells, elements, faces, and nodes used to represent the domain. Mesh quality refers to characteristics such as skewness, non-orthogonality, aspect ratio, smoothness, and alignment with important flow features. Mesh refinement means reducing the characteristic cell size in selected regions.

Discretisation is the approximation of differential or integral equations on the discrete mesh. Common approaches include finite difference, finite volume, and finite element methods. A discretisation scheme may be first-order or higher-order and may have different properties with respect to accuracy, boundedness, numerical diffusion, and stability.

A solver is the numerical algorithm that assembles and solves the discrete equations. It may be pressure-based, density-based, coupled, segregated, steady, or transient, depending on the physical problem and implementation.

A model is a mathematical representation of a physical process. Examples include turbulence, radiation, combustion, cavitation, porous media, and multiphase models. A model introduces assumptions and often empirical or semi-empirical constants.

Boundary conditions specify information at the limits of the domain. Typical types include velocity inlet, pressure inlet, pressure outlet, wall, symmetry, periodic, and far-field conditions. Boundary conditions must represent the physical environment rather than simply satisfy software requirements.

Initial conditions define the state of a transient simulation at its starting time. They also provide the initial field from which many steady solvers begin their iterations.

Convergence describes the reduction of residuals and the stabilisation of relevant solution quantities as iterations proceed. Low residuals are useful but do not alone guarantee a correct solution. Mass imbalance, forces, temperatures, and other engineering monitors must also be examined.

Verification asks whether the equations have been solved correctly. It includes code verification, comparison with manufactured or benchmark solutions, iterative convergence checks, and mesh or time-step independence studies.

Validation asks whether the chosen mathematical model represents the real physical system adequately for its intended use. Validation requires comparison with suitable experimental or trusted reference data. Verification and validation are distinct; a verified solution may still be invalid if its physical model is unsuitable.

Uncertainty quantification concerns the effect of uncertain inputs, model parameters, numerical choices, and measurements on predicted outputs. Sensitivity analysis identifies which inputs most strongly influence an output.

The Engineering Workflow

A credible CFD study begins before geometry is imported into a software package. The workflow is iterative rather than strictly linear, but its stages can be organised as follows.

Define the engineering question

The first task is to state what decision the simulation must support. “Analyse the flow” is not an adequate objective. A useful statement might be: determine whether the pressure loss through a valve meets a specified limit at three operating conditions; estimate the temperature uniformity at the outlet of a heat exchanger; or compare two diffuser geometries using pressure recovery and separation indicators.

The objective determines the required physics, outputs, accuracy, computational cost, and validation strategy. It also prevents the study from becoming an unfocused collection of attractive plots.

Establish the physical idealisation

The engineer next identifies the relevant physics and selects assumptions. Questions include whether the flow is steady or transient, compressible or incompressible, laminar or turbulent, single-phase or multiphase, isothermal or conjugate heat transfer, and whether gravity, radiation, chemical reaction, or rotation is important.

Dimensional analysis and characteristic scales support these decisions. Reynolds number indicates the relative importance of inertia and viscosity; Mach number indicates compressibility effects; Prandtl number relates momentum and thermal diffusion; and other nondimensional groups may govern buoyancy, surface tension, or reaction. Assumptions should be documented with their rationale and expected consequences.

Prepare geometry and computational domain

The geometry must represent the flow-relevant features while excluding details that have no meaningful influence on the objective. Small gaps, sharp edges, fillets, roughness, leakage paths, and component interfaces may be decisive in one study and negligible in another. Simplification is not automatically beneficial: eliminating a feature can change separation, pressure loss, or heat transfer.

For external flows, the outer domain must be sufficiently distant that artificial boundaries do not distort the solution. For internal flows, inlet and outlet extensions may be required to obtain physically appropriate development. The computational domain should be checked for gaps, intersections, non-manifold surfaces, and unintended openings.

Generate and assess the mesh

Mesh generation is a modelling decision, not merely a geometric operation. Cells should be refined where gradients, curvature, boundary layers, wakes, shocks, interfaces, or recirculation are expected. Near-wall treatment must be consistent with the turbulence model and wall-resolution strategy. Excessive refinement in unimportant regions wastes resources, whereas insufficient refinement can suppress important physics.

A mesh-quality report should be accompanied by a mesh-sensitivity study. At least two or three systematically refined meshes are often compared using quantities of engineering interest, such as pressure drop, lift, drag, heat-transfer coefficient, or peak temperature. Mesh independence is not absolute; it means that further refinement produces changes small enough for the decision being made.

Specify materials, models, and boundary conditions

Material properties should correspond to the relevant temperature, pressure, and composition ranges. Constant properties may be appropriate for a narrow operating range but unreliable in strongly heated or compressible flows.

Boundary conditions require particular care. A specified inlet velocity must be consistent with the desired mass flow and turbulence quantities. Outlet conditions should be placed where the flow is sufficiently developed, when possible. Wall conditions must identify no-slip or slip behaviour, thermal condition, roughness, motion, and possible mass transfer. Symmetry and periodic boundaries reduce cost but are justified only when the physical problem possesses the corresponding symmetry.

Select solver controls and obtain a solution

The solver type, discretisation schemes, relaxation factors, time step, and convergence criteria affect robustness and accuracy. A difficult nonlinear problem may be approached through staged initialisation, gradual activation of physics, lower-order schemes during startup, or continuation in Reynolds number or heat input. Such strategies improve stability but should not be confused with physical accuracy.

For transient simulations, the time step must resolve relevant physical time scales and should be assessed through time-step sensitivity. A nominally steady solution may actually be an average of unsteady behaviour or may fail to exist because the physical flow is inherently unsteady. Residuals should be monitored together with conservation errors and integral quantities.

Post-process quantities that answer the question

Post-processing should begin with conservation and plausibility checks. Examine mass flow balance, pressure levels, temperature ranges, velocity vectors, streamlines, and regions of recirculation. Then calculate objective-specific outputs such as pressure loss, efficiency, force coefficients, heat-transfer rate, uniformity index, or residence-time distribution.

Contours are useful for locating features but can be misleading if colour scales, interpolation, or plane placement conceal the actual values. A plot should communicate a defined quantity, unit, location, and operating condition. Integrated quantities and line or area averages are often more defensible than isolated extreme values.

Interpret, verify, validate, and report

Interpretation connects numerical patterns to fluid-mechanical mechanisms. For example, a pressure loss may result from separation, friction, turning, mixing, or acceleration rather than from a single generic “flow resistance.” Alternative explanations should be tested with additional probes, balance calculations, or simplified analyses.

The final report should state the objective, geometry, assumptions, equations, material properties, mesh, boundary and initial conditions, solver settings, convergence evidence, verification studies, validation data, uncertainty, results, and limitations. Reproducibility improves when software version, mesh files, input parameters, scripts, and post-processing definitions are preserved. A conclusion should distinguish calculated evidence from engineering judgement and should identify the operating range over which the findings apply.

An Illustrative Applied Case: Cooling Flow in an Electronic Enclosure

Background

Consider an illustrative industrial case in which a design team must evaluate airflow through a ventilated electronic enclosure. Several circuit boards generate heat, and two fans draw air through inlet vents and discharge it through an outlet grille. The design objective is to keep component temperatures below allowable limits while avoiding unnecessary fan power.

Situation and problem definition

The initial proposal uses a steady, incompressible turbulent model with conjugate heat transfer between air and solid components. The key outputs are maximum component temperature, temperature uniformity, total pressure drop, and flow rate through each important region. The team must compare two grille designs.

Analysis and approach

The engineers first simplify screws and small surface details that do not affect the flow, but retain the board spacing, heat-generating regions, fan locations, and grille openings. They estimate Reynolds numbers to justify a turbulent model and assess whether buoyancy is significant relative to forced convection. Because the fans are represented by performance curves rather than detailed rotating blades, the fan model is documented as an approximation.

Three meshes are generated, with refinement near vents, board surfaces, and narrow passages. The near-wall treatment is selected consistently with the turbulence model. Inlet and outlet conditions are defined using measured fan operating data, while the boards are assigned heat-generation rates from electrical power estimates. The study monitors mass imbalance, fan flow rate, maximum temperature, and pressure drop.

The coarse mesh predicts a maximum component temperature of 84 °C, the intermediate mesh 80 °C, and the fine mesh 79 °C. The design decision requires temperature accuracy within approximately 3 °C, so the intermediate and fine results are sufficiently close for preliminary selection, subject to validation. The first grille produces a high-velocity jet that bypasses one board and creates a recirculation region. The second grille distributes flow more evenly, reducing the predicted maximum temperature and slightly increasing pressure drop.

Validation and outcome

A prototype enclosure is instrumented with thermocouples at selected component locations. Measurements show that the second design reduces the hottest measured location by approximately the same qualitative amount predicted by CFD, although the absolute temperatures differ. Investigation identifies uncertainty in contact resistance between components and heat sinks, fan performance at the installed operating point, and unmodelled cable obstructions.

The team therefore does not claim that CFD has exactly predicted every temperature. Instead, it concludes that the second grille is the more robust design within the tested operating range, while recommending improved thermal-contact data and a further simulation including cable blockage.

Learning points

The case demonstrates that CFD value depends on a clearly defined decision, appropriate abstraction, mesh and model checks, validation against measurements, and transparent treatment of uncertainty. The most useful result is not necessarily the most visually detailed contour; it is the result that supports a justified engineering choice.

Limitations, Risks, and Professional Responsibilities

CFD can fail through incorrect physics, inadequate domain size, poor mesh quality, inappropriate boundary conditions, insufficient convergence, or careless interpretation. Common mistakes include treating residual reduction as the sole convergence criterion, comparing designs on different mesh quality, reporting maximum values at singular corners without examining mesh dependence, and using default turbulence settings without understanding wall treatment.

Another risk is false precision. Reporting a pressure drop to many decimal places does not imply equivalent physical accuracy. Model-form uncertainty, uncertain material properties, manufacturing variation, and operating-condition variability may dominate numerical error. Engineers should report appropriate significant figures and, where possible, ranges or confidence intervals.

Professional responsibility also includes traceability and honest communication. Simulation results used in safety-critical, environmental, medical, or regulated applications require documented assumptions, independent review, appropriate validation, and adherence to applicable standards and organisational procedures. A contour should never be used to conceal an adverse result, and limitations should be stated alongside benefits.

Digital Practice and Emerging Directions

CFD is increasingly integrated with computer-aided design, product-lifecycle management, optimisation, and automated design exploration. Scripting enables repeatable geometry changes, batch simulations, automated mesh generation, and extraction of standard metrics. Reduced-order models and surrogate models can approximate expensive CFD results for rapid design-space exploration.

Machine learning is being applied to turbulence closures, field reconstruction, surrogate modelling, anomaly detection, and adaptive mesh refinement. These applications can accelerate analysis, but they introduce new concerns about training-data coverage, extrapolation, interpretability, and physical consistency. A data-driven model should be evaluated against conservation laws and independent cases rather than accepted solely because it predicts a training set accurately.

Cloud computing and high-performance parallel solvers make large parametric studies more accessible. Digital twins may combine CFD-based models with sensor data to monitor and predict the behaviour of operating assets. The continuing trend is toward integrated simulation workflows, but engineering judgement remains essential for selecting the physical model, assessing credibility, and making decisions under uncertainty.

Best Practices for a Defensible CFD Study

A reliable study generally follows these principles:

• Define a measurable engineering question before choosing software settings.
• Use dimensional analysis and physical reasoning to select relevant models.
• Simplify geometry deliberately and document every consequential assumption.
• Match mesh resolution and wall treatment to the selected physical models.
• Check conservation, residuals, monitored outputs, mesh sensitivity, and time-step sensitivity where relevant.
• Validate important outputs against analytical, experimental, or trusted benchmark data.
• Use consistent methods when comparing design alternatives.
• Preserve inputs, meshes, solver settings, scripts, and post-processing definitions.
• Report uncertainty, applicability range, and limitations together with conclusions.
• Treat CFD as evidence supporting engineering judgement, not as a substitute for it.

Chapter Summary

CFD is the numerical study of fluid flow and associated transport phenomena through discretised governing equations. Its development has been enabled by advances in numerical methods and computing, but its credibility still depends on physical modelling, numerical quality, and validation. Analytical, experimental, and computational approaches are complementary: analysis provides insight and benchmarks, experiments provide physical evidence, and CFD provides detailed field predictions and efficient design exploration.

A CFD investigation should begin with a precise engineering question. The engineer then selects the relevant physics, idealises the geometry, creates and assesses a mesh, specifies material properties and boundary conditions, chooses solver controls, obtains a converged solution, and extracts quantities relevant to the decision. Verification addresses whether the equations were solved correctly, whereas validation addresses whether the model represents reality adequately. Mesh and time-step studies, conservation checks, uncertainty assessment, and transparent reporting are essential elements of professional practice. Emerging automation, high-performance computing, surrogate modelling, and machine learning expand CFD capability but do not remove the need for judgement and accountability.

Key Terms

CFD: Numerical solution of fluid-flow and related transport problems.

Computational domain: The physical region represented in a simulation.

Mesh: Discrete cells, elements, faces, or nodes used to represent a domain.

Discretisation: Approximation of continuous governing equations on a discrete mesh.

Finite-volume method: A method that integrates conservation equations over control volumes.

Finite-element method: A method that approximates fields using functions defined over elements.

Solver: Numerical algorithm that solves the discrete algebraic equations.

Boundary condition: Mathematical specification of variables or fluxes at a domain boundary.

Initial condition: The field state specified at the beginning of a transient calculation.

Turbulence model: A closure model representing the effects of unresolved turbulent motion.

Convergence: Stabilisation of the numerical solution and reduction of relevant residuals and imbalances.

Residual: A measure of the remaining algebraic error in a discretised equation.

Verification: Assessment of whether the equations were solved correctly.

Validation: Assessment of whether the model represents the physical system adequately.

Mesh sensitivity: Dependence of a predicted result on mesh resolution or quality.

Uncertainty quantification: Evaluation of how uncertain inputs and models affect outputs.

Post-processing: Examination and calculation of useful quantities from computed fields.

Model-form uncertainty: Uncertainty caused by imperfect representation of physical processes.

Dimensionless number: A scale-independent ratio used to characterise competing physical effects.

Surrogate model: An inexpensive approximation of a more computationally expensive simulation.

Review Questions

Short Answer Questions

1. What is the central distinction between CFD and experimental fluid mechanics?

2. Define verification and validation in the context of CFD.

3. Why is mesh generation considered a modelling decision?

4. What is meant by a boundary condition?

5. State two reasons why low residuals alone do not establish a credible CFD result.

Long Answer Questions

1. Explain the historical development of CFD and relate advances in numerical methods to present industrial capability.

2. Compare analytical, experimental, and computational approaches with respect to insight, cost, uncertainty, and applicability.

3. Describe a complete CFD workflow from engineering-question definition to final reporting.

4. Discuss the role of physical idealisation, dimensional analysis, and model selection in CFD.

5. Explain how verification, validation, mesh sensitivity, and uncertainty assessment contribute to trustworthy simulation.

Analytical / Application Questions

1. A simulation of flow through a valve predicts a pressure drop that changes substantially when the mesh is refined. Identify possible causes and propose a systematic investigation.

2. A heat exchanger model gives stable residuals but violates mass conservation by a significant amount and predicts unrealistic outlet temperatures. How should the engineer diagnose and respond to the result?

3. Two enclosure designs have similar predicted temperatures, but one result uses a detailed fan model and the other uses a simplified fan curve. Design a comparison and validation plan before recommending either design.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes CFD from a physical experiment?

A. CFD directly measures pressure and velocity in the real system.

B. CFD predicts behaviour by numerically solving a mathematical model of the system.

C. CFD eliminates uncertainty associated with physical assumptions.

D. CFD is applicable only when an analytical solution exists.

Correct Answer: B. CFD predicts behaviour by numerically solving a mathematical model of the system.

Question 2

A CFD study has very low residuals, but the predicted pressure drop changes greatly with mesh refinement. What is the most appropriate conclusion?

A. The solution is automatically validated because the residuals are low.

B. The solver should be stopped because further refinement is unnecessary.

C. The result requires further verification because it is mesh-sensitive.

D. The pressure drop should be replaced by the maximum velocity.

Correct Answer: C. The result requires further verification because it is mesh-sensitive.

Question 3

Which activity is primarily a validation activity?

A. Comparing a computed solution with measured pressure data from a suitable experiment.

B. Reducing the residual tolerance by two orders of magnitude.

C. Checking cell skewness and aspect ratio.

D. Comparing first-order and second-order discretisation schemes.

Correct Answer: A. Comparing a computed solution with measured pressure data from a suitable experiment.

Question 4

Why might an engineer use a finite-volume method for an industrial flow problem?

A. It guarantees an exact analytical solution.

B. It naturally applies conservation equations over discrete control volumes.

C. It removes the need for boundary conditions.

D. It resolves all turbulent scales without modelling.

Correct Answer: B. It naturally applies conservation equations over discrete control volumes.

Question 5

Which sequence most appropriately represents the beginning of a defensible CFD workflow?

A. Create contours, select colours, then define the engineering objective.

B. Define the engineering question, identify relevant physics, then idealise the domain.

C. Refine every region uniformly, then decide what outputs are needed.

D. Select a turbulence model first, then determine whether flow is present.

Correct Answer: B. Define the engineering question, identify relevant physics, then idealise the domain.

Question 6

A designer removes a small gap from a flow passage to simplify geometry. What must be considered before accepting this simplification?

A. Whether the gap affects leakage, pressure loss, or other quantities relevant to the objective.

B. Whether the software can display the original CAD colour.

C. Whether removing the gap always improves numerical accuracy.

D. Whether the resulting model requires no mesh study.

Correct Answer: A. Whether the gap affects leakage, pressure loss, or other quantities relevant to the objective.

Question 7

Which statement correctly describes convergence?

A. It proves that the physical model is valid.

B. It indicates stabilisation of the numerical solution, but should be assessed with residuals and engineering monitors.

C. It means that experimental uncertainty has been eliminated.

D. It occurs only in transient simulations.

Correct Answer: B. It indicates stabilisation of the numerical solution, but should be assessed with residuals and engineering monitors.

Question 8

In the enclosure case, why was prototype measurement still necessary after CFD analysis?

A. CFD cannot calculate any temperature field.

B. Measurements help assess whether modelling assumptions represent the real enclosure adequately.

C. Experiments are required only when the mesh contains triangles.

D. Validation replaces the need to define boundary conditions.

Correct Answer: B. Measurements help assess whether modelling assumptions represent the real enclosure adequately.

Question 9

Which practice is most likely to create false precision in a CFD report?

A. Reporting the operating range and uncertainty of the prediction.

B. Comparing integrated forces on consistently refined meshes.

C. Reporting many decimal places despite substantial model and input uncertainty.

D. Including the software version and solver settings.

Correct Answer: C. Reporting many decimal places despite substantial model and input uncertainty.

Question 10

What is a major concern when machine learning is used as a CFD surrogate?

A. It may extrapolate poorly outside its training-data range and violate physical consistency.

B. It cannot be used for any parametric study.

C. It always requires more computational effort than the original CFD model.

D. It automatically provides experimental validation.

Correct Answer: A. It may extrapolate poorly outside its training-data range and violate physical consistency.