7ENT1133 CFD & Applications – Revision Questions with Solutions

Lecture 1: The Nature of Fluids

1. Define the Eulerian and Lagrangian reference frames as they apply to fluid dynamics.
Create a sketch to outline the difference between the two approaches.
An Eulerian reference frame considers variations in the properties of a single fluid
element that is fixed in space.

A Lagrangian reference frame considers the dynamics of each particle and

interactions between them to estimate the response of the fluid as a whole.

Lecture 1, Slides 9-12

2. What are the three conservation laws used to build a continuum mechanics
description of a fluid flow?
Conservation of Mass.
Conservation of Momentum.
Conservation of Energy.
Lecture 1, Slides 15-20

Page 1 of 16
3. What is the mathematical definition for fluid compressibility? How is this defined for
an incompressible fluid?
A mathematical description of fluid compressibility is the rate of change of fluid
density with respect to time:
𝐷𝜌
𝐷𝑡
For an incompressible flow, this is equal to zero:
𝐷𝜌
=0
𝐷𝑡
Lecture 1, Slide 22

4. State Newton’s relationship between shear stress and velocity gradient in a fluid.
What fundamental fluid property does this relate to?
𝑑𝑢
𝜏=𝜇
𝑑𝑦
This relationship relates to fluid viscosity (𝜇 = coefficient of viscosity).
Lecture 1, Slide 24

5. Give a mathematical definition for the Reynolds number of a fluid flow.

𝜌𝑢𝐿
𝑅𝑒 =
𝜇
Lecture 1, Slide 28

6. Describe Reynolds number in words. What does a large Reynolds number imply?
Reynolds number is a dimensionless parameter that gives the ratio of inertial to
viscous forces. A large Reynolds number indicates that inertial forces are dominant
over viscous forces.
Lecture 1, Slide 29

Page 2 of 16
Lecture 2: The Governing Equations
1. Define the 𝑢, 𝑣 and 𝑤 velocity components of a fluid flow in terms of the scalar
potential function 𝜙 and cartesian coordinates 𝑥, 𝑦 and 𝑧.
𝜕𝜙
𝑢=
𝜕𝑥
𝜕𝜙
𝑣=
𝜕𝑦
𝜕𝜙
𝑤=
𝜕𝑧
Lecture 2, Slide 8

2. Define the 𝑢 and 𝑣 velocity components of a fluid flow in terms of the scalar stream
function 𝜓 and cartesian coordinates 𝑥 and 𝑦.
𝜕𝜓
𝑢=
𝜕𝑦
𝜕𝜓
𝑣=−
𝜕𝑥
Lecture 2, Slide 12

3. What is the relationship between the potential function 𝜙 and stream function 𝜓?
The stream function 𝜓 is defined at 90º to lines of equipotential given by the potential
function 𝜙 (i.e. they have an orthogonal relationship).
Lecture 2, Slides 11-14

4. What does the divergence of a vector field describe? Give a mathematical definition
for the divergence of a velocity vector field 𝑣⃗.
The divergence at a point within the vector field is a description of the tendency of the
field to migrate in the direction of the vector at that point.
𝑑𝑖𝑣 𝑣⃗ = 𝛻 ∙ 𝑣⃗
Lecture 2, Slide 16

5. What does the curl of a vector field describe? Give a mathematical definition for the
curl of a velocity vector field 𝑣⃗.
The curl at a point within the vector field is a description of the angular momentum of
the field about another vector at that point.
𝑐𝑢𝑟𝑙 𝑣⃗ = 𝛻 × 𝑣⃗
Lecture 2, Slide 17

Page 3 of 16
6. What are the assumptions and limitations implicit in the derivation of a potential flow
model? How do these compare to a Reynolds-Averaged Navier-Stokes (RANS)
solver?
Incompressible flow (𝑑𝑖𝑣 𝑣⃗ = 𝛻 ∙ 𝑣⃗ = 0).
Irrotational flow (𝑐𝑢𝑟𝑙 𝑣⃗ = 𝛻 × 𝑣⃗ = 0).
Inviscid flow.
The Reynolds-Averaged Navier-Stokes solver is capable of modelling both
compressible, rotational (turbulent) and viscous flows, however the potential flow
model is more computationally efficient.
7. Consider the two-dimensional velocity vector fields:
⃗⃗⃗⃗
𝑉1 = [−1,1,0]
⃗⃗⃗⃗
𝑉2 = [𝑥, 𝑦, 0]

Evaluate the vector components of ⃗⃗⃗⃗

𝑉1 and ⃗⃗⃗⃗
𝑉2 using at least four points per field.
Sketch the streamlines of each flow and describe them briefly in words.

Firstly, evaluate the velocity vector ⃗⃗⃗⃗

𝑉1 at four (𝑥, 𝑦) co-ordinate combinations. Since
the vector components are constants (i.e. they do not depend on 𝑥 or 𝑦), we already
have values for these:

X Y Vx Vy Vz
1 1 -1 1 0
-1 1 -1 1 0
-1 -1 -1 1 0
1 -1 -1 1 0

This makes it trivial to create a sketch of the flow streamlines (which are tangential to
the velocity vectors), revealing a uniform flow travelling in a north-westerly direction.:

Page 4 of 16
 Secondly, evaluate the velocity vector ⃗⃗⃗⃗
𝑉2 at four (𝑥, 𝑦) co-ordinate combinations.
Since the vector components do now depend on 𝑥 and 𝑦, we must calculate values
for these:

X Y Vx Vy Vz
1 1 1 1 0
-1 1 -1 1 0
-1 -1 -1 -1 0
1 -1 1 -1 0

Which, when sketched, approximates a divergent flow away from the origin:

8. The divergence of a velocity vector field is given by the following identity:

𝜕𝑉𝑥 𝜕𝑉𝑦 𝜕𝑉𝑧
⃗⃗ =
𝛻∙𝑉 + +
𝜕𝑥 𝜕𝑦 𝜕𝑧

Show that ⃗⃗⃗⃗

𝑉2 cannot describe a potential flow.
The vector identity would be provided in an exam situation. To evaluate the
divergence of the vector field, we must take the specified partial derivative for each
vector component and summate them. For ⃗⃗⃗⃗ 𝑉2 :
𝜕𝑉𝑥 𝜕𝑉𝑦 𝜕𝑉𝑧 𝜕(𝑥) 𝜕(𝑦) 𝜕(0)
+ + = + + =1+1+0=2
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

If ⃗⃗⃗⃗ ⃗⃗ = 0, which is not the

𝑉2 described a potential flow, then we would expect 𝛻 ∙ 𝑉
case. Thus, ⃗⃗⃗⃗
𝑉2 cannot describe a potential flow.

Page 5 of 16
Lecture 3: Towards Modern CFD

1. Fill in the blanks:

𝐷𝑽 Conservation of ?
𝜌 = −𝛻𝑝 + 𝛻 ∙ 𝜏 + 𝜌𝒈
𝐷𝑡
𝜕𝜌 𝐷𝜌 Conservation of ?
+ 𝛻 ∙ (𝜌𝑽) =
𝜕𝑡 𝐷𝑡
𝐷ℎ 𝐷𝑝 Conservation of ?
𝜌 = + 𝛻 ∙ (𝑘𝛻𝑇) + 𝜙
𝐷𝑡 𝐷𝑡

Momentum, Mass, Energy.

Lecture 3, Slide 9

2. What assumptions are necessary to simplify the Navier-Stokes equations for

compressible, Newtonian fluids to the Navier-Stokes equations for incompressible,
Newtonian fluids?
Constant density and constant viscosity.
Lecture 3, Slide 10

3. Starting from the conservation of momentum term shown in Question 1, derive

Euler’s momentum equation for an incompressible, Newtonian fluid:
𝑑𝑝
+ 𝑣𝑑𝑣 + 𝑔𝑑𝑧 = 0
𝜌
List any assumptions made as part of the derivation.
The conservation of momentum term is the first equation in the set:
⃗⃗
𝐷𝑉
𝜌 = −𝛻𝑝 + 𝛻 ∙ 𝜏 + 𝜌𝑔
𝐷𝑡
⃗⃗
𝐷𝑉
The total derivative 𝐷𝑡
should be expanded to give:

𝜕𝑉 𝜕𝑝
( ⃗⃗ ∙ 𝛻𝑉
+𝑉 ⃗⃗ ) = − +𝑔
𝜕𝑡 𝜌
Assuming steady flow (i.e. 𝜕𝑉⁄𝜕𝑡 = 0), we thus obtain the exact differential
equation:
𝑑𝑝
𝑣𝑑𝑣 = − + 𝑔𝑑𝑧
𝜌
𝑑𝑝
+ 𝑣𝑑𝑣 + 𝑔𝑑𝑧 = 0
𝜌
Lecture 3, Slides 12-15

Page 6 of 16
4. State Bernoulli’s equation, defining each of the terms.
1
𝑝 + 𝜌𝑣 2 + 𝜌𝑔𝑧 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
2
1
‘𝑝’ ‘ 𝜌𝑣 2’ ‘𝜌𝑔𝑧’
2

Static Pressure Dynamic Pressure Gravitational Potential

Lecture 3, Slide 17

5. How do static and dynamic pressure relate to total pressure?

Total pressure is the sum of the static and dynamic pressures.
Lecture 3, Slide 18

6. A differential pressure sensor mounted to a wind tunnel model reads a surface static
pressure of 268.75Pa. A Pitot tube mounted upstream of the model reads a total
pressure of 1800Pa. Calculate the surface flow velocity at the static pressure
measurement point and the flow Reynolds number, assuming an aerofoil wind tunnel
model with chord length 0.5m.
Firstly, we can find the dynamic pressure component by subtracting static pressure
from total pressure:
𝑞 = 1800 − 268.75 = 1531.25𝑃𝑎
Rearranging the equation for dynamic pressure allows calculation of the
corresponding flow velocity:
1
𝑞 = 𝜌𝑣 2
2

2𝑞 2 × 1531.25
𝑣=√ =√ = 50𝑚/𝑠
𝜌 1.225

Assuming a model chord length of 𝑙 = 0.5𝑚, the corresponding Reynolds number is:
𝜌𝑣𝐿 1.225 × 50 × 0.5
𝑅𝑒 = = = 1.692 × 106
𝜇 0.0000181

Page 7 of 16
Lecture 4: Boundary Layer Models and Turbulence
1. What assumptions can be made about flow viscosity outside of the boundary layer?
The freestream flow can be assumed inviscid outside of the boundary layer (i.e. the
effects of viscosity can be neglected) only if compressibility effects are also
negligible.
Lecture 4, Slide 6

2. What is the ‘Law of the Wall’? State an equation for this relationship and sketch the
velocity profile within the boundary layer adjacent to a solid boundary.
‘The Law of the Wall’ states that the streamwise flow velocity near a wall boundary
varies logarithmically with increasing distance away from the wall.
1
𝑢+ = ln(𝑦 + ) + 𝐵
𝑘
or
1
𝑢+ = ln(𝐸𝑦 + )
𝑘

3. State an equation for wall shear velocity, 𝑢𝜏 .

𝜏𝑤
𝑢𝜏 = √
𝜌

Lecture 4, Slide 8

Page 8 of 16
4. Define 𝑢+ and 𝑦 + .
𝑢+ is the dimensionless velocity (i.e. the velocity parallel to the wall as a function of
the distance to the wall, ‘𝑦’):
𝑈
𝑢+ =
𝑢𝜏
𝑦 + is the dimensionless distance to the wall:
𝜌𝑢𝜏 𝑦
𝑦+ =
𝜇
Lecture 4, Slide 9

5. In which region of the boundary layer is dimensionless wall distance 𝑦 + = 100?

The log-law layer.
6. A uniform flow of air passes a smooth flat plate at 𝑈 = 17𝑚/𝑠. Calculate the
minimum cell size 𝑦 required to fully resolve the boundary layer flow to 𝑦 + = 2, given
a dimensionless wall velocity 𝑢+ = 20. Give your answer in millimetres.
Since the local flow velocity is known, we can work towards calculating the near-wall
cell size 𝑦 in terms of 𝑦 + :
𝜌𝑢𝜏 𝑦
𝑦+ =
𝜇
𝑦+𝜇
𝑦=
𝜌𝑢𝜏
We first need to calculate 𝑢𝜏 :
𝑈
𝑢+ =
𝑢𝜏
𝑈 17
𝑢𝜏 = +
= = 0.85𝑚/𝑠
𝑢 20
Giving:
𝑦 + 𝜇 2 × 0.0000181
𝑦= = = 0.0000348𝑚
𝜌𝑢𝜏 1.225 × 0.85
𝑦 = 0.0348𝑚𝑚

Page 9 of 16
7. State three defining physical characteristics of a turbulent flow.
Turbulence is random.
Turbulence decays without energy input.
Turbulence displays scale-free behaviour.
Lecture 4, Slide 17

8. Why are turbulent flows challenging to simulate computationally?

A turbulence-resolving simulation must always be three-dimensional, time accurate
and have extremely fine computational grids. This implies a large overhead in terms
of computational memory and wall clock time to run the simulation.
Lecture 4, Slide 18

9. How is turbulence generated in a fluid flow?

Turbulence is initially generated by instabilities in the flow caused by mean velocity
gradients. The resulting rotational flows (eddies) in turn breed new instabilities, and
hence progressively smaller eddies.
Lecture 4, Slide 19

10. How is turbulence dissipated in a fluid flow?

The process of turbulence generation continues until the eddies become sufficiently
small (and velocity fluctuations sufficiently large) that viscous effects become
important and dissipate the turbulent kinetic energy as heat.
Lecture 4, Slide 19

11. Create a graph to outline the turbulent energy cascade, using appropriate axis labels.

Lecture 4, Slide 22

Page 10 of 16
Lecture 5: Turbulence Modelling
1. Is it necessary to resolve the detail of turbulent fluctuations within a fluid flow? How
can the turbulent fluctuations be isolated from the mean effect of the flow?
It depends on the application, however for most engineering applications it is not
necessary to resolve the detail of the turbulent fluctuations – solving for the ‘mean
flow’ is good enough. To do this, the turbulent velocity is split into mean and
fluctuating components via a process known as Reynolds decomposition.
𝑢 = 𝑢 + 𝑢′
Lecture 5, Slides 11-12

2. Substitute an expression for the mean and fluctuating turbulent velocity components
into the continuity equation to show that:
𝜕𝑢 𝜕𝑣 𝜕𝑤
+ + =0
𝜕𝑥 𝜕𝑦 𝜕𝑧
Starting with the continuity equation:
𝜕𝑢 𝜕𝑣 𝜕𝑤
+ + =0
𝜕𝑥 𝜕𝑦 𝜕𝑧
Substituting mean and fluctuating turbulent velocity components gives:
𝜕𝑢 + 𝑢′ 𝜕𝑣 + 𝑣 ′ 𝜕𝑤 + 𝑤 ′
+ + =0
𝜕𝑥 𝜕𝑦 𝜕𝑧
Time-averaging this expression gives:
̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
𝜕𝑢 + 𝑢′ 𝜕𝑣 + 𝑣 ′ 𝜕𝑤 + 𝑤 ′
+ + =0
𝜕𝑥 𝜕𝑦 𝜕𝑧
Given that the time-average of the fluctuating component tends to zero, only the
time-average of the mean component remains, and thus:
𝜕𝑢 𝜕𝑣 𝜕𝑤
+ + =0
𝜕𝑥 𝜕𝑦 𝜕𝑧
Lecture 5, Slide 13

3. Give examples of zero-equation, one-equation and two-equation eddy viscosity

models. How is the two-equation model different from the zero-equation model?
Zero-equation model could be mixing length.
One-equation model could be Spalart-Allmaras.
Two-equation model could be k-epsilon (𝑘-𝜀) or k-omega (𝑘-𝜔).
The two-equation model is different from the zero-equation model in that it permits
turbulence generation/dissipation to be specified using localised rates. This accounts
for variations in turbulent kinetic energy within the flow field.
Lecture 5, Slide 19 and 23

Page 11 of 16
4. State an advantage and a disadvantage for the zero-equation model.
The advantages of mixing length models are that they are easy to implement,
computationally cheap and give good predictions for thin shear layers (jets, wakes
and boundary layers).
The disadvantages are that mixing length models are incapable of describing flows
with separation and recirculation; turbulence is described using a constant length
scale (isotropy), and only mean flow properties are calculated.
Lecture 5, Slide 22

5. State an advantage and a disadvantage for the two-equation model.

The advantages of a two-equation model are that it is the most simple turbulence
model for which only initial and boundary conditions must be specified, and that it
provides excellent performance for most industrially-relevant flows.
The disadvantages of a two-equation model are that it is more expensive to
implement that the mixing length model (two additional partial differential equations)
and gives poor performance in specific flow regimes (e.g. unconfined flows, rotating
flows, flows with large strain due to curved boundary layers).
Lecture 5, Slide 27

6. How is the 𝑘-𝜔 turbulence model derived from the 𝑘-𝜀 turbulence model (i.e. what
does 𝜔 define)?
The 𝑘-𝜔 model is derived from the standard 𝑘-𝜀 model by normalising viscous
dissipation per unit turbulent kinetic energy:
𝜀
𝜔=
𝑘
Lecture 5, Slide 28

7. Give an expression for turbulent kinetic energy production.

The kinetic energy of a turbulent eddy can be defined:
1 ̅̅̅̅
′2 + ̅̅̅̅
𝑘 = (𝑢 𝑣 ′2 + ̅̅̅̅̅
𝑤 ′2 )
2
Assuming isotropic turbulence, this simplifies to:
3 ̅̅̅̅
′2 )
𝑘 = (𝑢
2
The units for 𝑘 are 𝑚2 ⁄𝑠 2 or 𝐽⁄𝑘𝑔.
Lecture 5, Slide 24

Page 12 of 16
8. Give an expression for turbulent kinetic energy dissipation.
The dissipation rate of the turbulent kinetic energy can be defined:
3
𝑘2
𝜀 = 𝐶𝜇
𝑙
The units for 𝜀 are 𝑚2 ⁄𝑠 3 or 𝐽⁄𝑘𝑔𝑠 or 𝑊 ⁄𝑘𝑔.
Lecture 5, Slide 25

9. Give an expression for turbulence intensity.

′
𝑢𝑟𝑚𝑠
𝑇𝑢𝑟𝑏𝑢𝑙𝑒𝑛𝑐𝑒 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 = 𝐼𝑢 =
𝑈
Lecture 5, Slide 12

10. A flow has a turbulence intensity of 0.5%, a length scale of 0.0125m and a velocity of
120m/s. Calculate the turbulent kinetic energy 𝑘 and the turbulent dissipation rate 𝜀 at
the inlet boundary. Recall that:

′
𝑢𝑟𝑚𝑠 = √(𝑢′2 )

Assuming isotropic turbulence, turbulent kinetic energy can be defined:

3 ̅̅̅̅
′2 )
𝑘 = (𝑢
2
Turbulent kinetic energy dissipation is defined:
3
𝑘2
𝜀 = 𝐶𝜇
𝑙
We are also told that:

′
𝑢𝑟𝑚𝑠 = √(𝑢′2 )

Therefore:
′ 2
𝑢𝑟𝑚𝑠 = 𝑢′2
P.T.O.

SOLUTION CONTINUED OVERLEAF

Page 13 of 16
From the definition for turbulence intensity:
𝑦𝑖𝑒𝑙𝑑𝑠 2
′ ′
𝑢𝑟𝑚𝑠 = 𝐼𝑢 × 𝑈 → 𝑢𝑟𝑚𝑠 = (𝐼𝑢 × 𝑈)2
Substituting this back into the expression for turbulent kinetic energy:
3 ̅̅̅̅ 3
′2 ) = (𝑢 ′ 2 3 2
𝑘 = (𝑢 𝑟𝑚𝑠 ) = (𝐼𝑢 × 𝑈)
2 2 2
3 3
𝑘 = (𝐼𝑢 × 𝑈)2 = (0.005 × 120)2 = 0.54𝐽/𝑘𝑔
2 2
Knowing that 𝐶𝜇 = 0.09 (Lecture 5, Slide 20), we obtain:
3
𝑘2 0.541.5
𝜀 = 𝐶𝜇 = 0.09 × = 2.86𝑊/𝑘𝑔
𝑙 0.0125

Page 14 of 16
Lectures 6 and 7: The CFD Process, Domain Dimensions and Discretisation
1. How do you distinguish between laminar and turbulent flow regimes?
Calculating the dimensionless Reynolds number gives you an indication of the flow
regime. Values larger than 2300 indicate a turbulent flow regime.
Lecture 6, Slides 11-12

2. What are the characteristic zones of development in Poiseuille flow according to

theoretical models?
This question refers to the hydrodynamic entrance region and fully-developed
regions for Poiseuille flow, taken from the work of Cengel, Y. A. (2010). Fluid
Mechanics: Fundamentals and Applications (SI Units). Tata McGraw Hill Education
Private Limited.
Lecture 6, Slide 11 and Lecture 7, Slide 11

3. What are the errors that can occur in computational domain sizing and what is the
most practical way to avoid them?
Wind-blocking error that is caused by insufficient upstream length of the domain.
Lecture 7, Slides 23 and 27-29

Flow-recirculation error associated with the downstream length.

Lecture 7, Slides 23 and 30

Global Venturi Effects (GVE) due to large blockage ratios and Local Venturi Effects
(LVE) caused by insufficient clearance between the object and top and lateral
domain boundaries.
Lecture 7, Slide 31

Given enough computational power, one could use the recommended domain
dimensions.
Lecture 7, Slide 33

4. In STAR-CCM+, what mesh quality characteristics can you recall?

It depends on the physical problem that you are solving, e.g. convection, diffusion
(L:7, S:45-46). The quality characteristics were covered in L:7, S:47-54, including
face validity, cell quality, volume change, skewness angle, etc.
5. What is the meant by a mesh-independent solution in numerical modelling
(CFD/FEA)? How would you prove that the solution of your simulation is independent
of the mesh?
A mesh-independent solution is a solution which has consistent results that do not
change after mesh refinement. To demonstrate mesh independence, one would need
to run an initial simulation to extract the value from key parameter(s). Secondly the
mesh quality is varied, normally by increasing the number of elements and re-running
simulation. The third step is to compare key parameter values between runs. Steps
two and three are repeated until there is no significant change in the value of the key
parameter(s).
Lecture 7, Slides 63-66

Page 15 of 16
6. What are the main steps in the ‘CFD Process’? Can you select one step and explain
it?
Pre-processing, analysis, post-processing.
Lecture 6

7. How could you validate your numerical findings? Why do you need to validate them?
Experimentally, analytically or through the use of historical data (e.g. academic
publications). Validation is required to determine whether the correct model was used
to characterise the investigated physical problem.
Lecture 6, Slides 18 and 27

8. How could you verify your numerical findings? Why do you need verification step(s)?
Residual error should be below a certain threshold value (specified by the engineer),
and the simulation must be converging. Also, verification can be achieved by simply
observing the results and analysing the data. Verification is needed to show that the
model was accurately constructed in terms of the assumptions, geometry, boundary
conditions and discretisation (meshing).
Lecture 6, Slides 18 and 27

9. What stage(s) of your simulation would require mesh quality analyses and
improvement? Can you elaborate your answer(s)?
Discretisation is part of the mathematical modelling stage in the analysis step.
However, it could also be necessary in the consecutive stages and the post-
processing step.
Lecture 6, Slide 15

Page 16 of 16