10. Turbulence Physics and Modeling

Introduction and Context

Turbulence is one of the central difficulties in computational fluid dynamics (CFD). Laminar flows are often governed by relatively orderly motion, whereas turbulent flows contain a broad range of interacting eddies, strong fluctuations, rapid mixing, and sensitivity to geometry and boundary conditions. In engineering systems, turbulence influences skin friction, pressure loss, heat transfer, combustion, pollutant dispersion, aerodynamic loading, noise, and flow separation.

The Navier–Stokes equations contain the physics required to describe turbulence, but direct numerical solution of all turbulent scales is generally impractical for industrial Reynolds numbers. CFD therefore relies on models that represent unresolved motions. The principal approaches are Reynolds-averaged Navier–Stokes (RANS), large-eddy simulation (LES), hybrid methods such as detached-eddy simulation (DES), and direct numerical simulation (DNS). Each method involves a different balance between physical resolution, computational cost, numerical requirements, and modeling uncertainty.

This chapter develops the physical basis of turbulence, introduces statistical descriptions and turbulence scales, explains near-wall behavior, and examines the selection and use of turbulence models. Particular attention is given to the distinction between model uncertainty and numerical error, because a solution may be numerically converged while still being physically unreliable.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain the physical origin and characteristic scales of turbulence.

2. Distinguish mean flow quantities, turbulent fluctuations, and statistical moments.

3. Describe the energy cascade and the significance of Kolmogorov scales.

4. Interpret wall units, boundary-layer regions, and near-wall turbulence behavior.

5. Formulate the principal ideas behind RANS, Reynolds-stress, LES, DES, and DNS approaches.

6. Select an appropriate wall treatment for a given mesh and flow problem.

7. Assess the strengths, limitations, and uncertainty of turbulence models in industrial CFD.

8. Design more credible CFD studies through verification, validation, sensitivity analysis, and best-practice procedures.

The Physical Character of Turbulence

Turbulence is not simply a flow with a large Reynolds number. It is an irregular, three-dimensional, rotational, and strongly coupled state of motion containing fluctuations over a wide range of length and time scales. The Reynolds number,

\[

Re=\frac{\rho UL}{\mu}=\frac{UL}{\nu},

\]

compares inertial and viscous effects. As Reynolds number increases, inertial transport becomes increasingly important relative to viscous diffusion, and disturbances can develop into a hierarchy of eddies. Reynolds number is therefore a useful indicator of the potential for turbulence, but transition also depends on geometry, surface roughness, pressure gradients, free-stream disturbances, and boundary conditions.

A turbulent velocity can be expressed using Reynolds decomposition:

\[

u_i=\overline{u_i}+u_i',

\]

where \(\overline{u_i}\) is a time or ensemble average and \(u_i'\) is the fluctuation about that average. Pressure and other variables may be decomposed similarly. Averaging the Navier–Stokes equations produces the RANS equations. The nonlinear convective term generates additional correlations such as \(\overline{u_i' u_j'}\), known as Reynolds stresses. These stresses represent the momentum transport caused by turbulent fluctuations and create the closure problem.

Turbulence is statistically irregular but not necessarily random in every sense. Coherent structures, such as hairpin vortices, shear-layer rollers, streamwise streaks, and large separated eddies, can dominate transport in particular flows. Consequently, a model that performs well for an attached boundary layer may perform poorly for swirling flow, strong separation, curvature, rotation, or highly transient jets.

Scales of Motion and the Energy Cascade

The largest turbulent structures are constrained by the geometry and the mean-flow organization. Their length scale is often comparable to a boundary-layer thickness, shear-layer thickness, channel dimension, or obstacle size. These large eddies contain much of the turbulent kinetic energy and are strongly affected by the mean flow and boundaries.

Energy is transferred from large eddies to progressively smaller eddies through nonlinear interactions. This process is called the energy cascade. At sufficiently small scales, viscous effects dominate and convert kinetic energy into thermal energy. The smallest scales are characterized by the Kolmogorov length, time, and velocity scales:

\[

\eta=\left(\frac{\nu^3}{\varepsilon}\right)^{1/4},\qquad

\tau_\eta=\left(\frac{\nu}{\varepsilon}\right)^{1/2},\qquad

u_\eta=(\nu\varepsilon)^{1/4},

\]

where \(\varepsilon\) is the turbulent kinetic-energy dissipation rate. The Kolmogorov scales are approximately universal for high-Reynolds-number turbulence, even though the large scales depend strongly on the flow configuration.

The separation between the largest and smallest scales increases rapidly with Reynolds number. For homogeneous isotropic turbulence, the ratio of an outer scale to a Kolmogorov scale grows approximately as a power of Reynolds number. This explains why DNS becomes prohibitively expensive at engineering Reynolds numbers: the computational mesh and time step must resolve the smallest scales throughout the domain.

In the inertial subrange, where production and viscous effects are relatively less important, the energy spectrum is often represented by the Kolmogorov form

\[

E(k)=C_K\varepsilon^{2/3}k^{-5/3},

\]

where \(k\) is wavenumber and \(C_K\) is a constant. The relation is an idealization; real engineering flows may have limited inertial ranges, anisotropy, strong inhomogeneity, or significant wall influence.

The turbulent kinetic energy is

\[

k=\frac{1}{2}\overline{u_i'u_i'}=\frac{1}{2}\left(\overline{u'^2}+\overline{v'^2}+\overline{w'^2}\right).

\]

Its transport equation contains production, diffusion, dissipation, and, where relevant, buoyancy or compressibility terms. Production commonly results from interaction between Reynolds stresses and mean velocity gradients. In a simple shear flow, turbulent fluctuations extract energy from the mean flow, while the cascade transfers that energy toward dissipative scales.

Statistical Description of Turbulent Flow

Engineering predictions rarely require the exact instantaneous field everywhere. They often require mean velocity, pressure drop, wall shear stress, heat-transfer coefficient, rms fluctuations, or probability distributions. These quantities are statistical, and the averaging procedure must match the physical question.

The mean of a variable is not sufficient to characterize turbulence. Variance and rms values measure fluctuation intensity; covariance measures correlation between fluctuations. The Reynolds stress tensor is symmetric and has six independent components. Its diagonal terms represent normal turbulent stresses, while off-diagonal terms represent turbulent momentum transport.

The turbulent kinetic-energy production term in incompressible flow is commonly written as

\[

P_k=-\overline{u_i'u_j'}\frac{\partial \overline{u_i}}{\partial x_j}.

\]

This expression shows why turbulence models must account for the orientation and magnitude of turbulent stresses relative to the mean strain field. The Boussinesq eddy-viscosity hypothesis assumes

\[

-\rho\overline{u_i'u_j'}=2\mu_t S_{ij}-\frac{2}{3}\rho k\delta_{ij},

\]

where \(\mu_t\) is turbulent viscosity and \(S_{ij}\) is the mean strain-rate tensor. This approximation is efficient but imposes an alignment between stress and strain that is not universally valid.

Turbulence intensity is frequently defined as an rms fluctuation divided by a reference velocity. It is useful for specifying inlet conditions, but it does not uniquely define a turbulent field. Two inlets may have the same intensity and different integral length scales, spectra, anisotropy, and spatial correlations, producing different downstream behavior.

Near-Wall Turbulence and Wall Units

Near solid surfaces, turbulence is strongly affected by the no-slip condition. The mean velocity is zero at the wall, and viscous stresses dominate immediately adjacent to it. A useful velocity scale is the friction velocity,

\[

u_\tau=\sqrt{\frac{\tau_w}{\rho}},

\]

and wall units are defined by

\[

y^+=\frac{yu_\tau}{\nu},\qquad u^+=\frac{\overline{u}}{u_\tau}.

\]

The near-wall region is commonly divided into several zones. In the viscous sublayer, approximately \(y^+<5\), the mean velocity follows the linear relation \(u^+=y^+\). The buffer layer, approximately \(5<y^+<30\), contains both viscous and turbulent effects. Farther from the wall, the logarithmic region is often represented by

\[

u^+=\frac{1}{\kappa}\ln y^+ + B,

\]

where \(\kappa\) is the von Kármán constant and \(B\) is an empirical intercept. The exact limits and constants depend on pressure gradient, roughness, separation, compressibility, and other effects.

Near-wall modeling is a principal CFD decision. A wall-resolved approach places the first cell center within the viscous sublayer and resolves steep velocity gradients. It generally requires fine spacing normal to the wall and sufficient streamwise and spanwise resolution. A wall-function approach does not resolve the viscous sublayer directly; instead, it imposes an empirical relationship between wall shear and the velocity at the first computational point, normally located in the logarithmic region.

Wall functions are economical, but their assumptions can become unreliable under strong adverse pressure gradients, separation, reattachment, strong curvature, rotation, roughness transitions, heat-transfer variation, or very low-Reynolds-number conditions. Enhanced or all-y+ treatments attempt to operate over a broader range, but they do not remove the underlying modeling limitations.

The first-cell height should therefore be chosen from the intended wall treatment, not as an afterthought. A nominal target such as \(y^+\approx 1\) is meaningful for wall-resolved modeling, whereas a wall-function mesh requires the first point to lie in an appropriate logarithmic region. Intermediate values may be especially problematic if neither the viscous sublayer nor the log law is adequately represented.

RANS Modeling: Efficient Prediction of Mean Flow

RANS models solve equations for mean flow quantities and model the statistical effect of turbulence. They remain the dominant approach for many industrial applications because they require substantially less computational cost than LES or DNS.

Two-Equation Eddy-Viscosity Models

The standard \(k\)-\(\varepsilon\) model solves transport equations for turbulent kinetic energy \(k\) and dissipation rate \(\varepsilon\). It is robust and often effective for free shear flows, jets, mixing layers, and fully developed turbulent regions. Its traditional wall-function formulation is less suitable for strongly separated flows or regions where the near-wall length scale is important.

The \(k\)-\(\omega\) family replaces \(\varepsilon\) with the specific dissipation rate \(\omega\). It generally offers better near-wall behavior and sensitivity to adverse pressure gradients. The shear-stress-transport (SST) formulation blends near-wall \(k\)-\(\omega\) behavior with an outer-region formulation related to \(k\)-\(\varepsilon\), and limits excessive turbulent shear stress through a stress limiter. It is widely used for boundary layers with pressure gradients and moderate separation.

Eddy-viscosity models are compact, stable, and inexpensive. However, the scalar turbulent viscosity may not represent anisotropy, streamline curvature, rapid strain, rotation, secondary flows, or multiple interacting time scales accurately. Model constants and damping functions are calibrated for classes of flows rather than derived universally.

Reynolds-Stress Models

Reynolds-stress models (RSMs) solve transport equations for individual components of the Reynolds-stress tensor, together with a scale equation such as one for \(\varepsilon\) or \(\omega\). They can represent stress anisotropy and non-equilibrium effects more directly than eddy-viscosity models. This makes them attractive for cyclones, swirling flows, curved ducts, rotating machinery, and flows with strong secondary motion.

The additional equations increase computational cost, numerical stiffness, and sensitivity to boundary conditions. Pressure–strain correlations, turbulent diffusion, dissipation, and wall-reflection terms still require modeling. An RSM is therefore not a model-free solution. It can be superior when anisotropy is central to the physics, but it may not improve results for every problem.

LES, DES, and DNS

Large-Eddy Simulation

LES resolves the larger, energy-containing eddies and models the effects of unresolved subgrid scales. Spatial filtering separates resolved and subgrid motions. The filtered momentum equation contains a subgrid stress tensor that must be modeled. Classical models include the Smagorinsky model, while dynamic procedures determine coefficients from the resolved field. Wall-adapting and localized formulations seek to reduce excessive eddy viscosity near walls and in transitional regions.

LES captures unsteady coherent structures more directly than RANS and can provide valuable information about fluctuating loads, vortex shedding, mixing, and separated flows. Its cost is dominated by the requirement to resolve energetic structures, especially near walls. At high Reynolds number, wall-resolved LES may become impractical for industrial geometries. Wall-modeled LES reduces cost by modeling the near-wall region but introduces additional uncertainty in wall stress and heat transfer.

LES also requires physically meaningful unsteady boundary conditions. A steady uniform inlet may produce an unrealistically quiet flow that develops too slowly or incorrectly. Synthetic turbulence, precursor simulations, recycling–rescaling methods, or measured inflow data may be necessary.

Detached-Eddy and Hybrid Methods

Hybrid RANS–LES approaches use RANS in attached boundary layers and LES-like treatment in separated regions. DES and its improved variants are examples. They seek to capture large unsteady structures without paying the cost of wall-resolved LES everywhere.

Hybrid methods require careful grid design. If the grid permits an unintended switch from RANS behavior to resolved eddies inside an attached boundary layer, modeled-stress depletion or grid-induced separation can occur. Shielding functions and delayed formulations reduce this risk, but users must still inspect the solution, length scales, and sensitivity to mesh refinement.

Direct Numerical Simulation

DNS resolves the full range of turbulent scales without a turbulence model. It provides highly detailed data for fundamental research, model development, and validation. Its computational requirements grow rapidly with Reynolds number, making full-scale industrial DNS generally infeasible. DNS should therefore be viewed as a reference methodology for canonical or moderate-Reynolds-number flows, not as a routine design tool.

Choosing a Turbulence Approach

Model selection should follow the dominant physical questions rather than habit. A steady pressure-drop calculation in a long, attached duct may be adequately served by a robust RANS model. Prediction of a highly separated diffuser, fluctuating aerodynamic loads, or unsteady mixing may justify URANS, hybrid RANS–LES, or LES. Strong swirl and anisotropy may motivate an RSM. A model should not be selected solely because it is more advanced; complexity is valuable only when it addresses a known deficiency.

A practical selection process includes the following reasoning:

• Identify whether the required output is a mean quantity, an instantaneous quantity, a spectrum, or a distribution.
• Determine whether separation, rotation, buoyancy, compressibility, transition, or strong curvature controls the result.
• Estimate the available computational resources and permissible turnaround time.
• Define the wall treatment and mesh strategy before selecting the turbulence model.
• Examine the availability and quality of validation data.
• Perform sensitivity studies using at least one credible alternative where the result is design-critical.

The distinction between steady RANS, unsteady RANS, LES, and hybrid methods is not only temporal. A steady RANS solution may represent a time average of a physically unsteady flow, but it cannot provide the correct resolved fluctuation field. URANS resolves only selected large-scale unsteadiness while modeling the remaining turbulence. The interpretation of the computed field must match the method.

Model-Form Uncertainty and CFD Credibility

Model-form uncertainty arises because the governing equations used in a simulation contain approximations about unresolved physics. It is distinct from discretization error, iterative error, uncertainty in material properties, and uncertainty in operating conditions.

A credible assessment combines verification and validation. Verification asks whether the equations have been solved correctly. It includes residual monitoring, iterative convergence, discretization studies, conservation checks, and time-step sensitivity. Validation asks whether the mathematical model represents the physical system adequately, using experimental data or a trusted reference.

For turbulence simulations, validation should compare quantities relevant to the engineering decision. A model may predict mean velocity acceptably while mispredicting rms fluctuations, separation length, wall heat flux, or pressure spectra. Agreement with one integral quantity does not establish universal validity.

Useful uncertainty practices include:

1. Mesh refinement focused on gradients, separation zones, wakes, and walls.

2. Time-step refinement for unsteady simulations.

3. Sensitivity to inlet turbulence intensity and length scale.

4. Comparison of at least two plausible turbulence models.

5. Sensitivity to wall treatment and first-cell location.

6. Comparison against independent measurements at multiple stations.

7. Reporting ranges or confidence intervals rather than a single overly precise value.

Model-form uncertainty can be propagated through an ensemble of models or through calibrated discrepancy approaches. In industrial work, a small set of physically justified models is often more informative than a large collection of arbitrary alternatives. Engineering judgment remains necessary when interpreting differences among models.

Illustrative Case Study: Turbulent Flow Through an Automotive Diffuser

Background

Consider an illustrative CFD study of an automotive underbody diffuser. The objective is to estimate mean pressure recovery and aerodynamic downforce while identifying whether flow separation occurs near the diffuser expansion.

Situation and Problem

An initial steady RANS calculation uses a wall-function mesh, a two-equation eddy-viscosity model, and a uniform inlet turbulence specification. The predicted pressure recovery is favorable, but wind-tunnel comparison shows excessive downforce and a separation region that is too short. The design team must determine whether the discrepancy is caused by geometry, mesh, inlet conditions, or turbulence modeling.

Analysis

The diffuser contains an adverse pressure gradient and three-dimensional separated flow. These features challenge simple equilibrium wall functions and isotropic eddy-viscosity assumptions. Underbody roughness and rotating wheels also generate turbulence and unsteady vortical structures. A steady solution may suppress the separated-flow dynamics that influence the mean pressure field.

The first diagnostic step is verification. The mesh is refined near the diffuser shoulder and wheel wakes, and the time step is reduced in an unsteady calculation. The first-cell locations are checked against the intended wall treatment. Conservation errors and residual histories are reviewed.

The second step examines physical inputs. Inlet turbulence intensity and length scale are varied within experimentally plausible ranges. Wheel rotation and moving-ground conditions are included because a stationary ground can alter the underbody boundary layer and separation pattern.

Approach and Strategy

Three simulations are compared. The first uses SST RANS with a wall-resolved near-wall mesh. The second uses an RSM to examine the effect of stress anisotropy. The third uses a hybrid RANS–LES method with sufficient refinement in the separated shear layer and wheel wakes. Unsteady statistics are collected only after initial transients have decayed, and averaging continues for several flow-through times.

The RANS solutions provide economical estimates of mean pressure distribution. The RSM changes the predicted secondary motion but does not eliminate the separation discrepancy. The hybrid method resolves large-scale shedding and produces a longer time-averaged separated region, closer to the experimental observation. However, its result is sensitive to inlet turbulence generation and local grid resolution.

Outcome

The design team reports a range of pressure recovery rather than a single value. The hybrid calculation is selected for final design comparisons because the primary decision concerns unsteady separation, while SST RANS is retained for rapid screening. The study also identifies moving-ground treatment and inlet turbulence as significant sources of uncertainty.

Learning Points

The case illustrates that a more elaborate model is not automatically correct. Credibility comes from matching the method to the physics, verifying numerical quality, validating against relevant data, and quantifying sensitivity. It also shows that near-wall treatment, boundary conditions, geometry motion, and turbulence model choice are coupled decisions.

Common Failure Modes and Professional Best Practice

A frequent mistake is treating turbulence-model selection as a universal ranking exercise. No model is best for all flows. Another mistake is reporting residual convergence without checking integral quantities, statistical stationarity, or mesh sensitivity. A residual decrease may coexist with an incorrect or physically unstable solution.

Users should avoid imposing arbitrary turbulence intensity and length scale at inlets. These quantities should be based on measurements, upstream geometry, correlations, or sensitivity ranges. They should also avoid interpreting instantaneous RANS fluctuations as resolved turbulence; in RANS, fluctuations are represented statistically rather than directly computed.

Wall functions should not be applied blindly to cells lying in the buffer layer. Conversely, a very fine near-wall mesh is not beneficial if the turbulence model, numerical scheme, or time step cannot use the additional resolution. For LES and hybrid methods, excessive numerical dissipation can remove the very structures the method is intended to resolve.

Professional practice includes documenting model equations, constants, wall treatment, mesh statistics, inlet turbulence, convergence criteria, averaging duration, and validation evidence. Design decisions should identify which outputs are robust and which depend strongly on modeling assumptions.

Emerging Directions

Current turbulence research is developing scale-resolving methods that combine improved wall modeling, adaptive meshes, and more efficient numerical algorithms. High-performance computing enables larger LES databases and more realistic multiphysics simulations, although data management and statistical convergence remain important limitations.

Machine learning is being investigated for closure modeling, subgrid-stress prediction, transition detection, and reduced-order modeling. Data-driven models may represent complex relationships more flexibly than traditional closures, but they can fail outside the training domain, violate physical constraints, or obscure uncertainty. Conservation, realizability, Galilean invariance, interpretability, and rigorous validation are essential requirements.

Digital engineering workflows increasingly combine CFD with automated mesh generation, parameter sweeps, surrogate models, and uncertainty quantification. Automation improves consistency, but it does not replace physical reasoning. A rapidly generated ensemble of poorly posed simulations is not a substitute for a carefully designed validation study.

Chapter Summary

Turbulence consists of interacting motions spanning a wide range of scales. Energy is transferred from large, geometry-dependent eddies through an inertial cascade to small Kolmogorov scales where viscosity dissipates it. Reynolds decomposition and statistical moments provide the basis for engineering descriptions, while the Reynolds stresses create the closure problem in RANS equations.

RANS models efficiently predict mean behavior but rely on assumptions about turbulent viscosity or stress transport. Eddy-viscosity models are economical; Reynolds-stress models address anisotropy more directly. LES resolves larger eddies and models subgrid motions, while DES and related hybrid methods combine RANS and scale-resolving behavior. DNS resolves all scales but is generally restricted to research-scale problems.

Near-wall treatment must be selected together with the mesh. Wall-resolved methods target the viscous sublayer, whereas wall functions represent the near-wall region through empirical laws. Model choice should reflect the required outputs, dominant flow physics, computational resources, and validation evidence. Reliable CFD requires separating model-form uncertainty from numerical error and reporting sensitivity rather than presenting unjustified precision.

Key Terms

Reynolds number: Ratio of inertial to viscous effects in a flow.

Turbulence: Irregular, three-dimensional fluid motion containing interacting fluctuations over a range of scales.

Reynolds decomposition: Representation of a flow variable as a mean value plus a fluctuating component.

Reynolds stress: Momentum-transport correlation produced by turbulent velocity fluctuations.

Turbulent kinetic energy: Half the sum of the mean-square velocity fluctuations per unit mass.

Energy cascade: Transfer of turbulent kinetic energy from large eddies to smaller eddies.

Kolmogorov scale: Smallest characteristic turbulence scale at which viscous dissipation dominates.

Dissipation rate: Rate at which turbulent kinetic energy is converted into internal energy by viscosity.

Eddy viscosity: Modeled turbulent viscosity used to relate Reynolds stresses to mean strain.

RANS: Reynolds-averaged Navier–Stokes approach for computing mean flow quantities.

RSM: Reynolds-stress model that solves transport equations for individual Reynolds-stress components.

LES: Large-eddy simulation that resolves large structures and models subgrid-scale motions.

DES: Hybrid method using RANS treatment in selected regions and LES-like resolution elsewhere.

DNS: Direct numerical simulation resolving all relevant turbulent scales without a turbulence closure model.

Friction velocity: Velocity scale defined from wall shear stress, \(u_\tau=\sqrt{\tau_w/\rho}\).

Wall units: Nondimensional near-wall coordinates such as \(y^+=yu_\tau/\nu\).

Wall function: Empirical near-wall relation used instead of directly resolving the viscous sublayer.

Model-form uncertainty: Uncertainty caused by approximations in the physical or turbulence model.

Verification: Evaluation of whether the numerical equations have been solved correctly.

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

Review Questions

Short Answer Questions

1. What physical process is represented by the turbulent energy cascade?

2. Define Reynolds stress and explain its origin.

3. What is the significance of the friction velocity in near-wall modeling?

4. State one principal difference between RANS and LES.

5. Why is DNS computationally expensive at high Reynolds numbers?

Long Answer Questions

1. Explain the hierarchy of turbulent scales and derive the significance of the Kolmogorov scales.

2. Describe the viscous sublayer, buffer layer, and logarithmic region using wall coordinates.

3. Compare eddy-viscosity RANS models with Reynolds-stress models.

4. Explain the operating principles, strengths, and limitations of LES and hybrid RANS–LES methods.

5. Discuss how verification, validation, and sensitivity analysis should be combined in an industrial turbulence simulation.

Analytical / Application Questions

1. A CFD model uses wall functions, but the first cell center lies at \(y^+=8\). Analyze the likely consequences and recommend a corrective strategy.

2. A swirling cyclone flow is poorly predicted by a standard eddy-viscosity model. Determine whether an RSM, LES, or hybrid method would be appropriate and justify your choice.

3. Two turbulence models predict nearly identical mean pressure drop but substantially different separation lengths. Explain how you would investigate the discrepancy and report the engineering uncertainty.

22. Multiple Choice Questions with Answers

Question 1

Which statement best describes the turbulence energy cascade?

A. Viscous forces transfer energy directly from the wall to the largest eddies

B. Energy is transferred from large, energy-containing eddies toward smaller dissipative scales

C. All turbulent eddies have approximately the same size and energy

D. Turbulent energy is created only at the Kolmogorov scale

Correct Answer: B. Energy is transferred from large, energy-containing eddies toward smaller dissipative scales

Question 2

What is the principal closure problem produced when the Navier–Stokes equations are Reynolds averaged?

A. The pressure becomes independent of velocity

B. The viscosity becomes exactly zero

C. Additional unknown correlations such as Reynolds stresses appear

D. The continuity equation can no longer be satisfied

Correct Answer: C. Additional unknown correlations such as Reynolds stresses appear

Question 3

A first-cell center with y+ approximately equal to 1 is normally associated with which treatment?

A. Wall-resolved modeling of the near-wall region

B. A wall function requiring the logarithmic region

C. A mesh that intentionally avoids the boundary layer

D. DNS in every industrial application

Correct Answer: A. Wall-resolved modeling of the near-wall region

Question 4

Which approach directly solves transport equations for individual Reynolds-stress components?

A. Standard k-epsilon eddy-viscosity model

B. Reynolds-stress model

C. Classical wall function

D. Kolmogorov spectrum model

Correct Answer: B. Reynolds-stress model

Question 5

Why is wall-resolved LES particularly expensive at high Reynolds number?

A. It requires resolution of energetic and near-wall turbulent structures over a large domain

B. It does not require a time step

C. It solves only the mean flow and therefore needs excessive averaging

D. It eliminates all viscous effects from the equations

Correct Answer: A. It requires resolution of energetic and near-wall turbulent structures over a large domain

Question 6

A steady RANS calculation of a strongly separated diffuser predicts an unrealistically short separation zone. Which response is most technically appropriate?

A. Assume residual convergence proves the prediction is correct

B. Refine only the far-field mesh and retain all physical assumptions

C. Investigate wall treatment, unsteady methods, inlet turbulence, and model sensitivity

D. Set turbulent viscosity to zero throughout the domain

Correct Answer: C. Investigate wall treatment, unsteady methods, inlet turbulence, and model sensitivity

Question 7

What distinguishes model-form uncertainty from discretization error?

A. Model-form uncertainty arises from approximations in representing unresolved physics

B. Model-form uncertainty is caused only by insufficient linear solver iterations

C. Discretization error is caused by experimental measurement uncertainty

D. There is no distinction between the two

Correct Answer: A. Model-form uncertainty arises from approximations in representing unresolved physics

Question 8

Which flow characteristic most strongly motivates consideration of an RSM rather than a simple isotropic eddy-viscosity model?

A. Uniform, fully developed straight-pipe flow only

B. Strong anisotropy, swirl, curvature, or secondary motion

C. A flow with no velocity gradients

D. A purely inviscid uniform flow

Correct Answer: B. Strong anisotropy, swirl, curvature, or secondary motion

Question 9

Why can a uniform, steady inlet be problematic for LES?

A. It may provide insufficient fluctuations and spatial correlations for realistic downstream development

B. It always produces excessive turbulent kinetic energy

C. It prevents the use of any spatial discretization

D. It automatically resolves all Kolmogorov scales

Correct Answer: A. It may provide insufficient fluctuations and spatial correlations for realistic downstream development

Question 10

Which statement correctly distinguishes verification from validation?

A. Verification compares only two turbulence models, while validation checks mesh quality

B. Verification asks whether the equations were solved correctly; validation asks whether the model represents reality adequately

C. Verification requires experiments, while validation requires no physical data

D. They are identical procedures with different names

Correct Answer: B. Verification asks whether the equations were solved correctly; validation asks whether the model represents reality adequately