Sei sulla pagina 1di 10

ME 450/550 Computational Fluid Dynamics

CFD Lab 1: Simulation of Laminar Pipe Flows


Submitted to: Professor Tao Xing

Name: Cristofer A. Farnetti


University ID: 111-67743
E-mail: farn7259@vandals.uidaho.edu
Department: Mechanical Engineering

Date: 03/08/2016

I.

Problem Description and Simulation Design

Computational Fluid Dynamics (CFD) has the ability to model very complex systems and
in order to get some experience with CFD, Ansys Fluent and Workbench 16.2 will be used to
find a number of flow characteristics for a given system, such as velocity profiles, vector plots,
and pressure gradients. To estimate numerical errors in the solution that are associated with grid
spacing, grid triplet studies will be performed on the solution for friction factors in the system.
The system thats modeled is laminar pipe flow through a 7.62 [m] long pipe that is
0.02619 [m] in radius. Five mesh sizes will be used to assess the numerical errors, and provide
verification for the system. With the pipe radius in mind, in order to maintain laminar flow a
Reynolds number of 655 is chosen which means the average velocity will be set to 0.2 [m/s].
The pipe flow is axisymmetric, and has boundary conditions at the wall, the inlet, the outlet, and
also the axis of symmetry. For ease, the outlet initial condition will be set to zero gauge pressure,
and no slip conditions will be applied at the wall.

II.

CFD Process

The CFD process for this study was not overly complicated; however, it is important to
note the process path in a systematic way along with some helpful description. To start out, the
pipe geometry must be created so that a two dimensional analysis can be performed. The
geometry that is created in the sketching tab is a planar rectangle with a length of 7.62 [m] and a
width of 0.02619 [m]. From here the mesh can be imposed on the surface of the sketch.
It is more conservative to list out characteristics of the CFD process, so we will start with
the mesh. Throughout the lab, five different face meshes will be created on the sketch of the
pipe. Table 1 on the next page gives information on the characteristics of the CFD setup in Ansys
along with the different meshes that are to be used and the different convergence criteria for the
solver that will be used in convergence studies on the solution for the friction factor calculations.
Each of the CFD process characteristics listed in Table 1 are an option in Ansys.

Table 1: CFD Process Setup information


For a process layout see the table below.
1.) Geometry
a. Open geometry editor from Workbench 16.2
i. In the sketch tab select rectangle and draw the pipe from the origin on
the xy plane
ii. Define the length and width of the pipe
iii. Apply surface
2.) Mesh
a. Open mesh editor from Workbench 16.2
i. Define boundaries and apply a face mesh
ii. Define axial and radial grid points

3.) Setup/Physics
a. Open Ansys Fluent by selecting the setup editor in Workbench 16.2
i. Before Fluent opens define double precision, axisymmetric, and Parallel
processing
ii. In the General tab
1. Ensure Viscous-Laminar flow is on
2. BCs: Wall, Inlet, Outlet, Axis
3. Fluid: Air
4. Define density (1.17 kg/m^3) and dynamic viscosity (1.8724E-5)
5. Set velocity at the inlet= 0.2 m/s
iii. Set the reference values and have the solver compute from inlet
4.) Solution
a. Method: 2nd Order Upwind Model
b. Initialization: Standard
c. Residuals Monitoring, set all equal convergence criteria
d. Calculation by iteration, 5000
5.) Results
a. CFD solution for friction factor
b. Plots
i. Centerline velocity profile
ii. Centerline pressure distribution
iii. Residuals
iv. Velocity vectors
v. Radial velocity contours
vi. Velocity profiles for CFD inlet and outlet, AFD for fully developed ,
CFD for developing
vii. Skin friction coeff.

III. Data Analysis and Discussion (65 points)


Mesh Number
Grid Points
(Axial Radial)

A.

113 11

160 16

226 23

320 32

451 45

Figures and tables in CFD Lab 1

Error

FactorCFD FactorAFD
100%
FactorAFD

Mesh
Number

Grid Points
(Axial
Radial)

Friction Factor
(Convergence
tolerance 10-5)

Relative Error
(Convergence
tolerance 10-5)

5
1

113 11
451 45

0.09701
0.10200

0.757%
4.348%

Friction
Factor
(Convergence
tolerance 10-6)
0.09698
0.09818

Relative Error
(Convergence
tolerance 10-6)
0.788%
0.440%

Developing length for mesh 1 and convergence tolerance 10-6 2.8 (m)

Figure 1. Residuals

Figure 2. Centerline pressure distribution

Figure 3. Centerline velocity distribution

Figure 4. Wall Skin friction factor distribution

Figure 6. Contours of radial velocity


Figure 5. profiles of axial velocity at specified
axial locations with AFD data

Figure 7. Velocity vectors in the region showing


transition from developing to developed flow

Figure 8. Normalized developed axial velocity


profile

B. Solution verification study for friction factor in developed region using the
factor of safety method (8 points)
Mesh Number
1,2,3
2,3,4
3,4,5

C.

R
5.85714
3
0.23333
3
0.71428
6

UG (%S1)

-2.5502

N/A

2.0995

0.426

0.4854
3

1.556

Questions in CFD Lab 1 (30 points)


(1) Can you use centerline pressure distribution to determine the developing length? if
yes, how? if not, why? Where are the minimum and maximum pressure locations?
What does the axial velocity profile look like?
Answer:
The developing length can be found from the non-linear pressure drop close to the inlet
of the pipe. The maximum and minimum pressure can be found at the inlet and the outlet,
respectively.
The axial velocity profile looks much like the plot of a square-root function from 0 x
2 with the velocity rapidly increasing from 0.2 to 0.4 in that region; this is the developing region
or developing length in which the flow boundaries from the pipe wall are converging towards
the centerline of the pipe. From 2 < x < 7.62 the velocity is constant because the flow is fully
developed and laminar.
(2) What are the magnitudes of radial velocity components in the developing and

developed region?
Answer:
The radial velocity in the developing region starts negative and gradually increases
throughout the developing region (-1.13E-2 to -1.13E-3), while the radial velocity in the
developed region is constant and unchanging (5.45E-8 [m/s]).
(3) What are the main differences for pressure distribution and axial velocity profile in
the developing and developed regions?
Answer:
In the developing region, the pressure distribution is rather linear regardless of whether it
is in the developing or the developed region, whereas the velocity profile in the developing
region shows a relatively lower velocity at the centerline than that of the fully developed profile.
Essentially, there is a shift in velocity that occurs between the developing and developed regions,
when in the pressure distribution there is no such shift or disturbance.
(4) For meshes 1 and 5, which one is more sensitive to the convergence tolerance
(exercise A)? Why? For a finer grid, should a bigger or smaller convergence tolerance
be used?
Answer:
The results from exercise A show that a finer grid is more sensitive to convergence
tolerance than the course grid. A finer grid is going to be more sensitive because it has more grid
points that can essentially detect the smaller changes that a larger grid would overlook because
of the fewer number of grid points. A smaller convergence tolerance should be used with a finer
grid because of the relative error that occurs between orders of magnitude. For example, in
exercise A the convergence tolerance of 10^-5 on the fine grid produced a relative error of about
4.3%, and that relative error decreases when a convergence tolerance of 10^-6 is used to give a
relative error of about 0.4%.
(5) Did you achieve monotonic convergence for the three grid-triplets in exercise B?
Which grid-triplet has a lower grid uncertainty? Why? Do the solutions approach the
asymptotic range when the grids are refined?
Answer:
Exercise B showed that for the first grid triplet study (1,2,3), it could only achieve
monotonic divergence since R > 1; however, in grid triplet study (2,3,4) and (3,4,5) the R values
are between 0 and 1 which means that they will monotonically converge. Only two of the three
grid triplet studies achieved monotonic convergence. The (2,3,4) grid triplet study had a lower
grid uncertainty because the range of the solutions from 2 to 4 was about 3.7E-4, whereas for
(1,2,3) the range grew to 4.8E-4 and for (3,4,5) the range grew to 7.2E-4. The larger differences
in solutions for the triplet studies produces a larger grid uncertainty in this case.
To determine whether the solutions approach the asymptotic range (AR) it is easy to use the
metric P to determine magnitude away P_re is from P_th. AR is achieved when P_re = P_th, so
as the mesh is refined from (3,4,5) to (2,3,4), P actually increases which indicates that the
solution is closer to the AR range in the coarse mesh than in the finer mesh.

IV. Conclusions (15 points)


7

A.

Conclusions regarding achieving purpose of simulation (3 points)


The purpose of the simulation was to become familiar with using CFD software
and employ some of the course content for analyzing the results. Additionally,
creating an accurate laminar pipe flow model was another objective of the lab. The
solutions for friction factor in the CFD model agreed quite well with the AFD
solutions, and the velocity profiles in the developed region are nearly identical so the
simulation results do provide confidence in attaining correct solutions.
In exercise A, when the grid sensitivity to convergence tolerance was examined,
the results showed that there is a higher sensitivity with the finer grids and this agrees
well with not only intuition, but common CFD practices. The issue in the study arises
when the grid triplet studies are performed. The solution for the (1,2,3) grid triplet
study showed that the solution was monotonically diverging while the solutions for
(2,3,4) and (3,4,5) were monotonically converging. The question then arises: Is it
typical for solutions to diverge as finer meshes are created?

B. What I have learned from CFD Lab1 (6 points)


CFD lab 1 has given me the opportunity to become familiar with using Ansys
Workbench and Fluent and I feel that I would be able to apply boundary conditions
and initial conditions for other similar internal flow geometries. The Workbench
platform and how to navigate the software was once kind of a mystery to me, but
after spending some time with it I have marked important characteristics to remember
with setting BCs and ICs and conducting the solution setup.
Aside from the software, getting to conduct my own grid triplet study was really
informative and gave me some intuition on how possible future grid triplet studies
might turn out. Additionally, in the convergence sensitivity analysis, I saw how
greatly the solutions can change with fine grids compared to coarse grid and it has
taught me to be more cautious of the solutions for fine grids unless they have the
appropriate convergence criteria.

C. Comments on the hands-on experience, the software interface, and


overall lab performance (4 points)
Having hands-on experience has taught me a lot, and its learning by trial-anderror at its best. I believe it is crucial for the class because someone could be taught
every aspect of CFD but if they have never worked with the software they would still
look like a beginner.
The software interface made the CFD process as easy as possible in my opinion
because the other alternative would be to write your own code, which could be a
daunting task. The fact that most of the setup requirements are in a combination of
dropdown menus and check boxes means that running a CFD simulation can be as
easy as checking the correct boxes and setting the right BCs (verification and
validation would still need to be done though).

D. Suggestions and improvements (2 points)

Since I am a beginner at using CFD and Ansys it is hard to say where the tutorial
could improve, because I followed it step-by-step and came to reasonable solutions.
Really the next best way that you could possibly add to the tutorial would be a video
of how to conduct the simulation where the user could just watch once, and then start
and stop the video as they go through the video. I think that having a video might
have the potential to hinder the learning experience though because trouble-shooting
where you might have messed up is part of the process, and a video might kill the
opportunity for students to mess up.

Potrebbero piacerti anche