Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Pipe Flow Lab Report

Guy Paterson
02226440

Abstract
This report aims to compare three research methods used to analyse the flow through a pipe, the
methods being: experimental, computational and theoretical. The experimental method physically
measures the flow in a real pipe, it is an accurate method but is often expensive and time consuming.
The theoretical method uses mathematical models to produce results but to keep solutions analytical,
simplifications have to be made, reducing the accuracy of results. However it is extremely cost and
time effective. The computational method uses numerical methods to produce many equations solved
that is then solved by computers , this method is often a good balance between cost and accuracy.
The ultimate objective of this report was to find the best method to analyse the flow through a pipe
and the research found...

26/02/23
Contents
List of Figures 1

List of Tables 1

List of Symbols 2

1 Introduction 3

2 Theory and Background 3

3 Verification and CFD Model Setup 4


3.1 Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Model Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

4 Results and Discussions 5


4.1 Comparison of Cp between the Experimental and Computational Method . . . . . . . . . 5
4.2 Calculation of Cf for CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.3 Comparison of Cf for different methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

5 Conclusions 8

References 8

Appendix 8
5.1 Appendix A - Raw Data/MATLAB code . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.2 Appendix B - Error Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.3 Appendix C - Formula Derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

List of Figures
1 Velocity vs. Radial Displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Coefficient of Pressure vs. Distance Comparison . . . . . . . . . . . . . . . . . . . . . . . 5
3 Velocity vs. Radial Distance for CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4 Skin Friction Coefficient vs. Reynolds Number Comparison, with Logarithmic Axes, Base 2 7

List of Tables
1 Comparison of Average Centre Line Velocity in fully developed flow for different Mesh Sizes 4
2 Absolute errors in measured values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1
List of Symbols
ReD Reynolds number in the pipe
ReD,crit critical Reynolds number in the pipe
x displacement in x-direction from inlet
y displacement in y-direction from centre line
r displacement in the radial-direction from centre line
u x component of velocity
y y component of velocity
U average velocity in a cross section of the fully developed flow
Uinlet inlet velocity
Cf skin friction coefficient
Cp pressure coefficient
τw shear stress
p pressure
pa atmospheric pressure of air
ρ density of air
µ dynamic viscosity of air
ν kinematic viscosity of air
R radius of pipe
D diameter of pipe
h height of manometer
ha height of manometer at atmospheric pressure
∆h0 difference in manometer height between orifice plate
k correlation factor
θ angle of manometer
ρm density of alcohol
T atmospheric temperature
µref reference dynamic viscosity
S Sutherland’s constant
g gravitational acceleration
t time

2
1 Introduction
This report aims to compare the three research methods, experimental, computational (CFD) and the-
oretical to analyse the laminar flow, at ReD « 1000, in a circular pipe. To achieve this two different
trends will be compared. The first been plotting skin friction coefficient, Cf , against radial velocity, uy ,
and the second been plotting coefficient of pressure, Cp , against distance along the pipe, x. Each method
will be compared in these plots and the error between methods will be found, where experimental data
is accepted to be true.
Another aim of this report is to find the critical Reynolds number, ReD,crit for the experimental method
and compare this to literature and explain why they do not agree. This research was conducted to guide
professionals in the field of fluid mechanics, to the strengths and limitations of each research method
stated, and encourage future study in this area.

2 Theory and Background


The fully developed flow in a pipe, is of great importance since it is simple enough to be modelled by
solvable equations and is one of the only analytical solutions to the Naiver-Stokes equations. The fully
developed state is achieved at some displacement from the inlet, when the boundary layer growth from
all surfaces converge in the centre, such that the entire flow in a cross section of the pipe is located
within the boundary layer. The fluid is in equilibrium as the shear forces is counteracted by the constant
pressure gradient. This effect causes the axial momentum to remain constant along the pipe.[1]
The governing equations are the set of 2D Naiver-Stokes equations, one for mass conservation and two
for momentum conservation. The 3D pipe, is modelled in 2D since it is axisymmetric. These equations
are solved by the CFD and is important to understand the aerodynamics:
ˆ 2
B u B2 u
ˆ 2
B2 v
˙ ˙
Bu Bv Bu Bu 1 Bp Bv Bv 1 Bp B v
` “ 0 , u `v “ `ν ` 2 , u `v “ `ν ` 2 (1,2,3)
Bx By Bx By ρ Bx Bx2 By Bx By ρ By Bx2 By

The radial velocity distribution to the governing Eqs. 1,2,3 , for the boundary conditions and sim-
plifications of the problem can be solved for Eq. 4. This demonstrates that velocity, u, follows a second
order polynomial relationship with radial distance, r.
1 ∆p ` 2 ˘
uprq “ ´ R ´ r2 (4)
4µ ∆x
Skin friction coefficient, Cf , is the dimensionalized coefficient of shear stress, τw . It is defined by Eq. 5
and is a function of the Reynolds number.
τw du
Cf “ “ f pReq where τw “ µ | (5,6)
1 2 dr r“R
2 ρU
Therefore in the laminar regime, an analytical solution for the skin friction coefficient can be derived
in terms of the Reynolds number, and follows a simple inverse law, known as the Poiseulle equation
(1810). However for the turbulent regime, since the flow is unsteady an analytical solution can not be
derived. Therefore the use of an empirical equation must be used. Eq. 8, was first used by Prandtl
(1935) is one of many empirical equations used but will be used in analysis in this report.
16 1 a
Laminar : Cf “ Turbulent : a “ ´0.4 ` 4 logpReD Cf q (7,8)
Re Cf
The Reynolds’ Experiment, carried out by Osborne Reynolds in the late 1800s, was a major break-
through in fluid mechanics. Reynolds was investigating the transition from laminar to turbulent flows by
carefully changing the conditions of fluid in pipe. His observations lead him to realise the flow character-
istics depended on the non-dimensional quantity, the Reynolds number, ReD , which is defined as ratio of
inertial to viscous forces. From his experiment he defined a critical Reynolds number, ReD,crit , and refers
to the Reynolds number for transition to occur. His findings lead him to believe that ReD,crit “ 2300.
An in depth study of Reynolds’ experiment can seen at [3].

3
3 Verification and CFD Model Setup
3.1 Verification
A critical part of the verification process, is to carry out a Mesh Convergence Study. This involves
comparing the percentage error of a certain quantity, for meshes of different base sizes. In this experiment
the quantity compared was the average centre line velocity in the fully developed region between x
positions, 1m and 4.95m. This was done by finding velocities at 100 equidistant points between the
positions stated.
The aim of this study is to find successive meshes, where the increased resolution has little affect on the
results of the CFD by comparing the chosen quantity. If this is shown Mesh Independence is proved
verifying the results. The data from this study has been presented in Table 1, to prove Mesh Independence
for the Coarse Mesh stated.

Table 1: Comparison of Average Centre Line Velocity in fully developed flow for different Mesh Sizes
Number of Average Velocity Percentage Error between the Mesh
Cells [m/s] and the Coarse Mesh [10´6 %]
Fine Mesh 65744 3.2538763 11.5
Medium Mesh 54784 3.2538762 7.1
Coarse Mesh 37932 3.2538760 N/A

As shown in Table 1, there is a negligible percentage error between the coarse mesh and the two
meshes of higher resolution. This shows increasing the resolution has an insignificant affect on the
results, therefore proving Mesh Independence.
Also from theory it is expected that the centre line velocity remains constant in the fully developed
regime. Therefore the standard deviation (SD), in the velocity of the 100 equidistant points should be
very small. The SD calculated was 0.00027 and since 0.00027 « 0 the CFD agrees with theory.

To further verify the results of the sim-


ulation, radial velocity in fully developed
flow, at x “ 2.5m, was compared with the-
oretical results from Eq. 4. As shown in
Figure 1, both methods show very similar
trends and have very small errors between
values, with a maximum percentage error
of 1.81%. Also all residuals for the mesh
run converged at an order of magnitude
of less than 10´5 , which is considered well
converged.[2] Therefore since Mesh Inde-
pendence was proved and results are in
line with theory, the simulation results are
verified.
Figure 1: Velocity vs. Radial Displacement
3.2 Model Setup
Star-CCM was the software used to run the CFD. The model assumptions were the flow was: 2D-
axisymmetric, incompressible, steady, uniformly viscous. and the no-slip condition The segregated flow
model was chosen as the numerical solution approach.
The boundary conditions/initial conditions was chosen to be consistent to those measured in the exper-
imental method, and are as follows: density, ρ “ 1.19kgm´3 , viscosity, µ “ 1.82 ˆ 10´5 kgm´1 s´2 , inlet
velocity, Uinlet “ 1.657ms´1 and atmospheric pressure, pa “ 100560Pa. And importantly the Reynolds
number in the fully developed regime, was equal to, ReD “ 1032.
The mesh used has the following properties: quadrilateral mesher, base size=0.0006m, minimum and
target surface size equal to 100% of the base, surface growth rate=1.1 and 37932 cells. Since there are
three governing equations, there is a total 113796 equations to be approximated per iteration.

4
4 Results and Discussions
4.1 Comparison of Cp between the Experimental and Computational Method

Figure 2: Coefficient of Pressure vs. Distance Comparison

In Figure 2 both plots show strong correlation in linear regression, this is to be expected as literature
states, in the fully developed regime of the pipe, the pressure gradient is a constant.[1] The correlation
is significantly weaker in the experimental data due to the random errors which are not present in
computational method.
The regression line of the CFD, fits through all error bars. Therefore it is shown the CFD lies within the
experimental uncertainty and so is a good representation of the real aerodynamics. However, the data
seems to be shifted in the y-direction which implies there is a systematic error between the data sets.
This systematic error could be due to a number of factors but may include: Model errors, where the
CFD involves simplifications of the real physical phenomenon such as heat transfer or gusts, Boundary
Condition errors, where the boundary conditions are not well defined and do not match those of the
experimental data.
Refer to Appendix B, for Eq. z used to calculate the error bars in Figure 2. There is only error bars
shown in the vertical direction because there is negligible error in the position of the tappings. In Eq.
z, the only variable that is a function of x, is h ´ ha , therefore the error in Cp was calculated for each
data point through MATLAB. In doing this, the absolute error in Cp remained almost constant through
each data point at UCp “ ˘6.2 with negligible difference. Therefore the error bars were chosen to have
to have the fixed error stated, as this best represented the data.

5
4.2 Calculation of Cf for CFD

Figure 3: Velocity vs. Radial Distance for CFD

Figure 3 was used to calculate the skin friction coefficient for the computation method at Re=1032.
First wall shear stress, τw , needs to be calculated via Eq. 6. Since Eq. 6, requires a derivative of
velocity, u, with respect to radial distance, r; a function uprq must be approximated from the CFD
data first. This was done using the built in ’polyfit’ function in MATLAB, setting the order of the
polynomial to 2, since this is expected from theory, Eq. 4. This method resulted in the polynomial:
uprq “ ´112917r2 ´ 64.78r ` 3.2561. Figure 3 shows the polynomial is a great fit to the data set.
Differentiating the function gives:
Bu
“ ´225834r ´ 64.78 6 τw “ µp´225834r ´ 64.78q |r“R “ ´0.0210N m´2
Br
It is worth noting that the negative sign implies the wall shear stress acts in the opposite direction to
that of velocity (since it is resistive), and doesn’t affect our skin friction coefficient calculations. Next to
non-dimensionalize τw apply Eq. 5, by dividing by dynamic pressure. Therefore at Re=1032, Cf =0.0128
in the CFD at the fully developed regime.

6
4.3 Comparison of Cf for different methods

Figure 4: Skin Friction Coefficient vs. Reynolds Number Comparison, with Logarithmic Axes, Base 2

In Figure 4 the theoretical trends were plotted for both laminar and transitional regimes using Eqs.
7,8. The Laminar Regime Theoretical was easily plotted by finding 50 data points in the range of
the experimental data points; then fitting a power trendline exact to Eq 7. The Turbulent Regime
Theoretical was more difficult to plot, as the empirical formula required a numerical methods solver,
done in MATLAB.
Based on our results, it is apparent that in the laminar regime, the theoretical is a great fit to the
experimental data as it is well within in the error bars. In the turbulent regime, the data shows a
vertical shift between experimental and theoretical data, this is most likely due to the large percentage
uncertainty of 5% in the correlation factor, k. However the data still fits within the experimental
uncertainty, so is also a good fit to the data.
The Cf data point of Cf =0.0128 calculated from the CFD at Re “ 1032, is shown on Figure 4. The
percentage uncertainty in this value between the true experimental value is 21.4%. This is greater than
the experimental uncertainty of 19.6% and therefore this lies outside of the error bar. This result cannot
either validate or disprove whether the CFD is a good model for the laminar regime of the pipe.
Ideally a range of CFD data points throughout the laminar region would be necessary to calculate a
trendline to see whether it passes through all error bars. This would be possible for different Re values
in the laminar regime, with the same CFD model setup but would require many more simulations and
even more to prove Mesh Independence for each Re. However it would require a different model setup if
data was to be compared in the turbulent regime since the flow is no longer steady.
As shown in Figure 4, the critical Reynolds number, ReD,crit is located at the transitional data point
with the minimum Re, with a numerical value of ReD,crit =4392. This data point has been classified to
be in the transitional regime since it does not follow the trend for the laminar regime or the trend for the
turbulent regime. It can be concluded from the data with 100% confidence ReD,crit can not be greater
than the value stated, however it is not whether ReD,crit is less than our stated value, due to lack of
data points in the experiment.
From Reynolds’ Experiment, he quotes a figure of ReD,crit « 2300 for flow in a pipe.[3] 4392 ą 2300,
this is because in the Reynolds Experiment he did not consider all external factors, because ReD,crit is
a function of many different variables that is impossible to quantify. Nowadays the value of 2300 is seen
as a minimum ReD,crit , meaning ReD,crit should be quoted as a range of « 2300 ´ 4000.[4]
The reason to why there is substantial increase in ReD,crit is due to the bell mouth inlet of the pipe.

7
5 Conclusions
Here, you draw conclusions based on the presented results. In conclusions, you should discuss the success
of the experiment and address aims and objectives of the experiment

References
[1] labsheet
[2] Kuron M. 3 criteria for assessing CFD Convergence [Internet]. Engineering.com. 2015 [cited
2023Feb6]. Available from: https://www.engineering.com/story/3-criteria-for-assessing-
cfd-convergence
[3] Rott N. Note on the history of the Reynolds number. Annual Review of Fluid Mechanics,
221(1):112, 1990
[4] Cadence CFD. Overview of the critical Reynolds number [Internet]. Cadence. 2022 [cited
2023Feb26]. Available from: https://resources.system-analysis.cadence.com/blog/msa2022-
overview-of-the-critical-reynolds-number
[5] White FM. Fluid mechanics. New York, NY: McGraw Hill; 2011.

Appendix
5.1 Appendix A - Raw Data/MATLAB code
Raw data collected from Pipe Flow experiment located at:

C:\Users\guymp\OneDrive\Aero\Labs\PipeFlowDataRaw.xlsx

Raw data collected from CFD located at:

C:\Users\guymp\OneDrive\Aero\Labs\PIPEFLOW SIMULATIONS\CFDRawData.xlsx

All MATLAB code used in analysis located at:

C:\Users\guymp\OneDrive\Aero\Labs\PIPEFLOW SIMULATIONS\MATLAB

5.2 Appendix B - Error Analysis

Table 2: Absolute errors in measured values


Measured Value Absolute Error
in Measurement
Atmospheric Pressure 100560Pa ˘5Pa
Atmospheric Temperature 294K ˘0.5K
Correlation Factor 0.10 ˘0.005
Density of Alcohol 800kgm´3 ˘0.5kgm´3
Height of Manometer N/A ˘0.025cm
Angle of Manometer N/A « 0˝

Combined Uncertainty:

1.Height of Manometer, h:

8
Error in precision: Error due to electrical noise:
#
Uh1 “ 0.025cm Laminar/Turbulent 0.5cm
Uh2 “
Transitional 3cm

for laminar/turbulent regime:


g
f n
fÿ
6 Uh “ e Uhi “ 0.501cm
i“1

Error propagation:

1.Density, ρ

dˆ ˙2 ˆ ˙2
p Bf Bf
ρ“ “ f pp, T q 6 Uρ “ Up ` UT , PercentageError “ 0.170% (x)
RT Bp BT

2. Viscosity, µ:
dˆ ˙2
T 3{2 Bf
µ “ µref “ f pT q 6 Uµ “ UT , PercentageError “ 0.00550% (y)
T `S BT
3. Pressure Coefficient, Cp
dˆ ˙2 ˆ ˙2 ˆ ˙2
h ´ ha Bf Bf Bf
Cp “ “ f ph, k, ∆h0 q 6 Uρ “ Uh ` Uk ` 2Uh , (z)
k∆h0 Bph ´ ha q Bk B∆h0

UCp « ˘6.2
4. Velocity, U

d dˆ ˙2 ˆ ˙2 ˆ ˙2
2kρm g∆h0 sin θ Bf Bf Bf
U“ “ f pk, ∆h0 , ρa q 6 UU “ Uk ` U∆h0 ` Uρ ,
ρ Bk B∆h0 Bρ
(zz)
Percentage Error « 18%
5. Reynolds Number, Re
dˆ ˙2 ˆ ˙2 ˆ ˙2
ρa U D Bf Bf Bf
Re “ “ f pρa , U , µq 6 URe “ Uρa ` UU ` Uµ , (zy)
µ Bρa BU Bµ

5.3 Appendix C - Formula Derivations


Derivation of Equation x:
Naiver Stokes equation in polar form:
ˆ ˙
Bu Bu Bu 1 Bp ν B Bu
`u `v “´ ` r
Bt Bx By ρ Bx r Br Br

Bu Bu Bv
Simplifications: flow is steady, Bt “ 0, flow is fully developed, Bx “ 0, continuity applies, By “ 0 so
Bp ∆p
v “ 0, constant pressure gradient, Bx “ ∆x
ˆ ˙
1 ∆p ν B Bu
“` r
ρ ∆x r Br Br

9
Bu
Integrating for uprq and applying boundary conditions: upRq “ 0 and Br p0q “0

1 ∆p ` 2 ˘
uprq “ ´ R ´ r2
4µ ∆x
Derivation of Equation x:
żR
1 R2 ∆p
U“ 2πruprqdr “ ´
πR2 0 8µ ∆x

10

You might also like