Aerospace and Mechanical Engineering

1st Exam of Computational Fluid Dynamics

January 19th, 2024, 8:00 AM
Rooms: V1.06, V1.07, V1.08
Closed Book Exam
Duration: 2 hours
Name: ________________________________________________________________________ Number: ______________

We are all expected to abide by the IST standards for academic excellence and integrity.

Problem 1 (2.0+1.0+3.0 = 6.0 Marks)

Consider the homogeneous conditions in a room with Volume 𝑉, ventilated with a Flow rate 𝐹̇ ,
which has a constant release of carbon monoxide at a Rate 𝑅̇ . The carbon monoxide’s
concentration in the room is 𝐶𝑅 , changing with time, and the concentration of carbon monoxide
in the incoming flow is zero, 𝐶𝐼𝑁 = 0. For the initial condition, consider 𝐶𝑅 (𝑡 = 0) = 0.

Under this conditions the ODE governing the process is:

𝑑𝐶𝑅 𝑅̇ 𝐹̇ 𝑑𝐶𝑅
= − 𝐶𝑅 → = 300 − 0.02𝐶𝑅
𝑑𝑡 𝑉 𝑉 𝑑𝑡
where 𝐶𝑅 is the carbon monoxide concentration in parts per million (ppm), 300 is the release per
unit volume in ppm, and 0.02 is the air changes per second.

a) Discretize the equation using:

i) First order Euler explicit

ii) First order Euler implicit

iii) Leapfrog. Also refer how to discretize the first time step

iv) Runge-Kutta second order. Present the final result as a single expression like
𝐶𝑅𝑛+1 = 𝐹 (𝐶𝑅𝑛 )

1
 Formulae

b) A set of 8 simulations with a

sequence of different time
steps – different time grids –
is presented on the RHS.
From one time grid to
another the time step was
reduced twice.
The table below clarifies the
sequence as well as the
numerical and exact values of
the carbon monoxide’s
concentration (in ppm) for
two different time stages: 75
s and 300 s.
The values were rounded to
the first decimal place.
#TimeGrid Colour 𝑪𝑹 (𝟕𝟓 𝒔) 𝑪𝑹 (𝟑𝟎𝟎 𝒔)
1 Yellow 22500.0 14062.5
2 Orange 14062.5 14999.8
3 Red 12711.2 14991.9
4 Green 12151.1 14980.5
5 Blue 11895.0 14972.5
6 Purple 11772.3 14967.8
7 Gray 11712.3 14965.4
8 Black 11682.6 14964.1
Exact Dashed Pink 11653.0 14962.8
i) Chose one time stage and conclude about the order of accuracy of the numerical
method used for the time integration

2
ii) For the same time stage, apply Richardson extrapolation in two time grids, inside
the asymptotic range of convergence, and conclude about its accuracy

Formulae
𝜖 = 𝜙 (ℎ ) − 𝜙𝐸𝑥𝑎𝑐𝑡 = 𝛼ℎ 𝑝 + 𝐻. 𝑂. 𝑇. 𝐿𝑜𝑔(𝜖) = 𝐿𝑜𝑔(𝛼 ) + 𝑝𝐿𝑜𝑔(ℎ)
𝜙3 − 𝜙2 𝜙𝐸𝑥𝑎𝑐𝑡 (𝑥) = 𝜙ℎ (𝑥) + 𝛼ℎ 𝑝 + 𝑂(ℎ 𝑝+1 )
𝑝 = 𝐿𝑜𝑔 ( ) /𝐿𝑜𝑔(𝑟)
𝜙2 − 𝜙1
𝜙ℎ − 𝜙ℎ1 𝑝
|𝜙𝐸𝑥𝑎𝑐𝑡 − 𝜙ℎ1 | = |[ 𝑝2 ] ℎ1 | + 𝑂(ℎ2𝑝+1 ) + 𝑂(ℎ1𝑝+1 )
ℎ2 − ℎ1𝑝
ℎ𝑖+1 𝐹𝜖 |𝜖| 𝐹𝜖 |𝜖|𝑟 𝑝
𝑟= 𝐺𝐶𝐼𝑓𝑖𝑛𝑒 = 𝑝 𝐺𝐶𝐼𝑐𝑜𝑎𝑟𝑠𝑒 = 𝑝
ℎ𝑖 𝑟 −1 𝑟 −1
𝜙1 − 𝜙2 𝐹𝜖 |𝜖|
𝜙ℎ=0 = 𝜙𝑓𝑖𝑛𝑒 + 𝑝 𝐺𝐶𝐼 = 𝑝
𝑟 −1 𝑟 −1
𝐺𝐶𝐼2,3
𝐴𝑠𝑦𝑚𝑝𝑡𝑜𝑡𝑖𝑐 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑐𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑛𝑔𝑒: 𝑝 𝑠ℎ𝑜𝑢𝑙𝑑 𝑏𝑒 ≅ 1
𝑟 × 𝐺𝐶𝐼1,2

3
c) Consider a ‘FULLY IMPICITICY’ within the Finite Volume framework, where the time line is
discretized as usual for the space line.

i) Apply the Finite Volume Method, using backward differences (upwind) for the first
derivative and conclude about the discrete equation as below:
𝑑𝐶𝑅
= 300 − 0.02𝐶𝑅 → 𝐶𝑅𝑖 − 𝐶𝑅𝑖 −1 = Δ𝑡300 − 0.02Δ𝑡𝐶𝑅𝑖
𝑑𝑡

ii) Prove that the first and second lines are as shown below, with 𝛽 = 1 + 0.02Δ𝑡.

𝜷 𝑪𝑹𝟏 𝟑𝟎𝟎𝚫𝒕
𝟐

−𝟏 𝜷 𝑪𝑹𝟑 𝟑𝟎𝟎𝚫𝒕
𝟐

−𝟏 𝜷 𝑪𝑹? = 𝟑𝟎𝟎𝚫𝒕

⋱ ⋱ ⋮ ⋮

−𝟏 𝜷 𝑪𝑹 𝟏 𝟑𝟎𝟎𝚫𝒕
𝑵−
𝟐

4
 iii) Perform 2 iterations using the Jacobi and the Gauss-Seidel methods for the case a 3x3
matrix. Comment the results obtained.

Jacobi: Gauss-Seidel:
Formulae

5
iv) The flux limiters are non-linear schemes,
depending on the current field.
Assume that the final numerical field
corresponds to the exact field as depicted on
the RHS.
If you apply the MINMOD under that field,
express the generic Finite Volume equation
and fill the Yellow bands on the system below.

𝑪𝑹𝟏 𝟑𝟎𝟎𝚫𝒕
𝟐

𝑪𝑹𝟑 𝟑𝟎𝟎𝚫𝒕
𝟐

⋱ ⋱ 𝑪𝑹? = 𝟑𝟎𝟎𝚫𝒕

⋱ ⋱ ⋮ ⋮

𝑪𝑹 𝟏 𝟑𝟎𝟎𝚫𝒕
𝑵−
𝟐

6
Formulae

7
8
Name: ________________________________________________________________________ Number: ______________

Problem 2 (1.0+1.0+1.0 = 3.0 Marks)

Consider the Black-Red duct network bellow, for an incompressible fluid, feeding an industrial
facility where each section should receive the same flow rate 𝑉̇ = 100.
The duct network comprises 4 nodes with momentum equation, in between each section, given
by:
−Δ𝑃 = 𝛽𝑉̇
where local pressure losses were neglected as well as losses on the local Blue duct network. 𝑃 is
the pressure, 𝑉̇ is the volumetric flow rate and β = 3.
The continuity equation can be taken as:
∑𝑉 ⃗ ̇ ∙ ⃗⃗⃗
𝑛𝑖 = 0

The pressure is imposed in the chicken boundary and the flow rates are imposed for all sections.

a) Sketch a possible ‘minimal’ staggered mesh to illustrate the solution of the problem and
derive the ‘SIMPLE’ algorithm to obtain an equation for the pressure correction.
b) Define an initial guess for the pressure and estimate the velocity field.
c) Correct the pressure and velocities after the solution of the equation. Justify how many
iterations do you estimate as needed to conclude the process.

9
10
Name: ________________________________________________________________________ Number: ______________

Problem 3 (2.0 Marks)

Consider the three figures below.

The first one is a domain where the boundary was fully defined and one single point was included
in the interior to construct a mesh.
The second and third figures represent a mesh built using the Advanced Front Method.

Express, step by step, the process followed in these mess generation processes. Clearly
identify the parts of the process that was similar in both meshes as well as the parts of the
process that was different from one mesh to another.

11
12
Name: ________________________________________________________________________ Number: ______________

Problem 4 (1.0+2.0+2.0 = 5.0 Marks)

Consider the unstructured mesh oh the right where the

central control volume is to be taken as P.
The pertinent points on the surrounding were properly
labeled.
The thin colored lines may help in the relative positions
of the pertinent points.
A local Coordinate system, presented in red color, may
be the support for all computations.
All segments in the Control Volume 𝑷 has a length of 1.
All non vertical lines are oblique lines (with 45° with
some vertical or horizontal line)

As a methodology, the subscript PA, for instance, refers to the entities at face separating control
volumes P and A, or to the distance vector of A relative to P.
For the sake of simplicity, some quantities/entities were previously computed and are
presented below.
Again, to clarify, if 𝜙𝑃𝐴 is the interpolation of 𝜙 to the left green face, then:
𝜙𝑃𝐴 = 𝑔𝑃𝐴 𝜙𝑃 + (1 − 𝑔𝑃𝐴 )𝜙𝐴

Normal Centroid
𝑛⃗𝑟 = (1; 0) √2 1 √2
𝑛⃗𝑙 = (−1; 0) 𝐶𝑃 = ( ; + )
2 2 4
√2 √2 √2 1 √2
𝑛⃗𝑡𝑟 = ( ; ) 𝐶𝐵 = ( ; − + )
2 2 2 2 4
√2 √2 √2 3 √2
𝑛⃗𝑡𝑙 = (− ; ) 𝐶𝑇 = ( ; + )
2 2 2 2 4
𝑛⃗𝑏𝑟
√2 1 √2
√2 √2 𝐶𝐿 = (− ; − )
= (− ;− ) 2 2 4
2 2
3√2 1 √2
√2 √2 𝐶𝑅 = ( ; − )
𝑛⃗𝑏𝑙 = ( ; − ) 2 2 4
2 2

Distance vector Interpolation factor

𝑟𝑃𝑇 = (0; 1) 1 √2
𝑟𝑃𝐵 = (0; −1) 𝑔𝑃𝑇 = −
2 4
√2 1 √2
𝑟𝑃𝐿 = (−√2; − ) 𝑔𝑃𝐵 = +
2 2 4
√𝟐 𝑔𝑃𝐿 = 0.5
⃗𝒓𝑷𝑹 = (√𝟐; − ) 𝑔𝑃𝑅 = 0.5
𝟐

13
a) Regarding the four blocks above with geometric entities, chose one quantity per block and
conclude, with all steps, about the value in the table.

b) Determine the Cell ‘average’ Gradient for cell P. Verify that:

̅̅̅̅̅̅̅̅
∇(𝜙)𝑃 |𝑥 = 𝛼(𝜙𝑅 − 𝜙𝐿 )
̅̅̅̅̅̅̅̅
∇ (𝜙)𝑃 | = (𝛾 + 𝛽)𝜙𝐵 + (−2𝛽)𝜙𝑃 + (−𝛾 + 𝛽)𝜙 𝑇
𝑦
Using the generic function 𝜙(𝑥; 𝑦) = 𝑥 + 𝑦, conclude about the order of accuracy of the
cell gradient.

Formulae
1 1
̅̅̅̅̅̅
∇𝜙𝑃 = ∫ ∇𝜙 𝑑𝑉 ̅̅̅̅̅̅
∇𝜙𝑃 = ∫ 𝜙 𝑑𝑺
𝑉𝑃 𝑉𝑃 𝑉𝑃 𝜕𝑉𝑃
1
̅̅̅̅̅̅
∇𝜙𝑃 𝑉𝑃 = ∑ ∫ 𝜙 𝑑𝑺 ̅̅̅̅̅̅
∇𝜙𝑃 = ∑ ̅̅𝜙̅̅𝑓 𝑺𝒇
𝑓𝑎𝑐𝑒 𝑉𝑃
𝜕𝑉𝑃 𝑓=𝑛𝑏(𝑃)

14
 ∂ϕ
c) Considering the transport equation ∂t = ∇ ∙ (∇(𝜙)), apply the Finite Volume method
taking in mind the cell P. Discretize the transport equation in the control volume P. For
that do not consider gradient corrections. Comment the expected order of accuracy of the
diffusive term.

15
16
Name: ________________________________________________________________________ Number: ______________

Problem 5 (20 x 0.2 = 4.0 Marks) Wrong answers discount 50% (-0.1 Mark)

Indicate if the statements below are True or False.

T F
1 The transient Euler equations are elliptic for subsonic and hyperbolic for
supersonic flows.
When discretizing the linear one-dimensional convection equation with an
2 implicit scheme, the application of a flux limiter makes the solution procedure
non-linear.
3 When discretizing the Poisson equation with second order central differences,
the error is higher where the gradient is higher.
4 Approximating the second spatial derivative with three-point central-difference
method, becomes first-order accurate when applied to nonuniform grids.
5 The SOR method with an under-relaxation factor 0 <𝜆 < 1 can speed up
convergence when the solutions oscillate about the converged solution.
6 For the linear one-dimensional advection equation, the analytical amplification
factor is unitary.
7 PWIM method calculates the face velocity for continuity equation by a weighted
interpolation by the pressure, not the pressure gradient.
8 A consistent linear numerical scheme for the linear wave equation that is
monotonicity-preserving can at most be first-order accurate.
9 The inflow boundary conditions needed in the simulation of 1D Euler
compressible flows are the same for subsonic and supersonic flow.
When applying the finite volume method in a one-dimensional problem with N
10 cells with Dirichlet boundary conditions, the resulting system of equations has a
matrix with size (N-2)*(N-2) because the boundary values are known.
11 Godunov’s method despite the use of an exact solution is first-order accurate in
space and time because a piecewise-constant reconstruction is selected.
12 When using the Grid Convergence Index (GCI) method, a mesh refinement ratio
(r) equal to 1.15 is recommended to ensure that the finer mesh is solved with the
available computational resources.
In non-orthogonal meshes, if the line connecting the centroids passes in the face
13 centroid but is not colinear with the normal to the face, a correction of the
gradient based on Taylor series is needed to keep the overall second order
accuracy.
14 Thomas algorithm is highly recommended for systems of equations with high
density matrices.
15 A set of three points always generate a Delaunay triangle.
16 One of the limitations of the finite difference method is the application to
unstructured grids.
17 Five possible wave patterns exist in the solution of the Riemann problem for the
inviscid 1D Burgers equation, when evaluating the Godunov flux.
18 In a density-based solver, the pressure is obtained from the ideal gas law while
density comes from the continuity equation.
19 In the SIMPLE algorithm, the solution obtained from the momentum equations
satisfy the continuity equation.
20 The SIMPLEC algorithm belongs to the class of coupled solvers.

17