Académique Documents
Professionnel Documents
Culture Documents
Viscosity is an intrinsic property of a fluid that When the Reynold’s number is low (< 2000), the
describes that fluid's resistance to flow. Fluids flow is viscous effect dominant which is called
resist the relative motion of objects that have a laminar flow. Laminar flow is also called parallel
different velocity to them. When a fluid flows flow and is characterized by a streamline
through a pipe, the pipe has a velocity of zero motion. When the Reynold’s number is high (>
which creates a negative relative motion to the 4000), the flow is bulk motion dominated which
fluid flow. The fluid will resist this relative is called turbulent flow. Turbulent flow is
motion by developing a velocity gradient. The characterized by chaotic flow with large scale
velocity of a fluid next to the pipe will equal mixing. When the Reynold’s number is
zero and gradually increase to maximum at the intermediate (2000 < Re < 4000), the flow is
center of the flow profile. The viscosity classified as transitional or a mix between
essentially determines how much the fluid will laminar and turbulent properties.
resist the relative motion. If a fluid has high
viscosity, the resistance to flow will be high. SIMULATION THEORY
Viscosity (µ) is defined as the sheer stress (τ) When fluid flows through a non-horizontal pipe,
required to drag one layer of fluid with unit both bulk motion and viscous effects are at
velocity (dv) past another layer a unit distance play. The bulk motion is caused by the driving
away (dy). force of gravity. The gravitational driving force
is equal to the mass of the fluid multiplied by its
µ= τ
dv
acceleration (g). The fluid mass is equal to the
( dy )
fluid density (ρ) multiplied by the h eight of the
Viscosity is one of two major components that fluid (h) and the cross sectional area of the fluid
affect fluid flow. The viscous effect happens at (A).
the molecular scale and describes the
movement of molecules against one another. F g = ρhAg
This molecular movement causes momentum
diffusion through a fluid. The other major
In the case of laminar flow, the viscous effect
component is the bulk flow effect. Bulk flow
translates into a pressure drop which is
happens on the macroscopic scale and
characterized by the Hagen-Poiseuille equation
describes how the bulk of the fluid is moving as
where L is fluid length and Q is fluid flow rate.
a result of an external driving force. Bulk flow
causes a momentum convection through a fluid. 8μLQ
ΔP = πr4
1
Flow Oscillation Simulation Andrews
The pressure drop multiplied by the cross To experimentally model fluid flow in a U-tube,
sectional area of the pipe is representative of a long tube was mounted as a vertical U and
the drag force on the fluid. Drag is the force filled with enough water so the level, at
that resists (or slows) fluid motion. equilibrium, is about halfway between the
curve and the top of the tube. Figure 1 shows a
8μLQA
Fg = πr4
diagram of the experimental setup.
d sign(y(2))*B/A*abs(y(2))^beta
dt |x x′ | = |0 1 − C /A − B /A ||x x′ | + |0 q(t)/A |
A MATLAB function that calculates x and x’ for In the case of a constriction, a 4th term is
single time steps was implemented as the introduced to model the pressure drop due to
following: the constriction:
The output of this function is a time vector, t One trial from each tubing size was graphed in
3
Flow Oscillation Simulation Andrews
SIMULATION RESULTS
Laminar Flow
4
Flow Oscillation Simulation Andrews
The derived ODE was then solved analytically laminar flow model and an experimental
using the initial conditions of the experimental turbulent flow condition. Figure 5 shows a
set up. Derivation of the analytical solution can height versus time comparison between the
be found in Appendix A, Laminar Flow computational models of turbulent and laminar
Derivations. to the experimental model of flow through the
1” U-tube.
The computational and analytical models of
laminar flow were then plotted in MATLAB and
compared to an experimental laminar flow
condition. Figure 4 shows a height versus time
comparison between the computational,
analytical and experimental models of flow
through the 1/16” U-tube.
5
Flow Oscillation Simulation Andrews
6
Flow Oscillation Simulation Andrews
turbulent flow model is more complicated to simulation models follow the same general
quantify, we see a similar inverse relationship trend but show a higher degree of damping
between tube radius and damping force. than the experimental data. Sources of this
error will be discussed in the Approximations
Reynold’s Number Trends
and Simplifications secton.
We also observe a trend between the tube
EFFECT OF TURBULENCE ON DAMPING
diameter and Reynold’s number. As shown in
Table 1, as tube diameter increases, so does the As discussed in the Damping Trends section, the
Reynold’s number. The specific relationship turbulent flow model produces a different
between tube diameter and Reynold’s number damping relationship than the laminar model.
depends on whether the Reynold’s number is This difference can be observed in Figure 5
within the laminar or turbulent flow range. In which shows a height versus time comparison
the laminar range, the Reynold’s number between the computational models of
increases linearly with the tube radius. This can turbulent and laminar flow to the experimental
be explained by the direct proportionality of model of flow through the 1” U-tube. As shown
tube diameter to Reynold’s number in the in the figure, the turbulent model is more
Reynold’s number formula and the fact that the damped than the laminar model. Since a 1”
laminar flow model is linear with respect to tube would be expected to have turbulent flow,
velocity. the experimental results show that the
turbulent model is indeed more accurate than
In the turbulent range however, the Reynold’s
the laminar model.
number increases non-linearly with tube radius.
This can be explained by the non-linear The difference in damping between turbulent
behavior of velocity to tube radius in the flow and laminar flow can be explained by the
turbulent flow model. non-linear damping term in the turbulent flow
ODE model. The non-linearity results from the
relationship between the Fanning friction factor
and Reynold’s number (derivation in Appendix
3 TYPES OF MODELS FOR LAMINAR FLOW A, Turbulent Flow Derivations). The turbulent
drag force term is derived from the Darcy flow
In an attempt to model laminar flow, equation
experimental data was collected in conjunction
dP
with both computational and analytical dL = f ( 2ρ 2
D )v ,
simulations of flow. Figure 4 shows the height
versus time comparison between the where f is the Fanning friction factor equation.
computational, analytical and experimental After taking the derivative and multiplying by
models of flow through the 1/16” U-tube. The the cross sectional area (derivation in Appendix
analytical and computational models produce A, Turbulent Flow Derivations) the drag force
nearly identical results. The small error between term becomes
the two simulation curves is resultant of the F = 0.025πLρ0.8 µ0.2 D0.8 v 1.8 .
fact that they were ploted using different time
steps. The fact that the computational and Since water is a Newtonian, incompressible
analytical models are nearly identical is due to fluid, desity (ρ) and viscosity (µ) are constant.
the fact that they represent two ways of solving Therefore, it can be observed that both
the same differential equation. In comparison Reynold’s number and the drag force are
to the actual experimental results, the exclusively dependant on diameter (D) and
7
Flow Oscillation Simulation Andrews
velocity (v) and are both increased by increasing This effect can be explained by taking into
diameter and velocity. That means that at high consideration the relative magnitudes of
Reynold’s numbers, the diameter and velocity damping due to viscous/frictional effects
of the fluid will be greater, therefore increasing compared to pressure loss from the tube
the damping force as shown in the turbulent constriction. Even at first observation, it can be
model of flow. seen that the damping force due to the
constriction is proportionate to the square of
DAMPING EFFECTS IN CONSTRICTED U-TUBE
the velocity. Therefore, the damping term
Small Diameter Tube resulting from the constriction will have a
greater effect on the overall damping than the
The damping effects observed in a small open flow laminar or turbulent damping terms.
diameter constricted U-tube contradict the
aforementioned effect that turbulence In addition, as the size of the constriction
increases damping. As observed in Figure 6, the decreases relative to the size of the tube, the
turbulent model is actually less damped than constrictive damping force is multiplied by an
the laminar model for a small diameter exponentially increasing number. This effect can
constricted U-tube. be shown by the derivation of the constrictive
damping force (full derivation in Appendix A,
A general explanation for why this is the case is Constricted Flow Derivations). The resulting
that in a 1/16” diameter tube, we would expect drag force equation is
laminar flow, therefore, the turbulent model is
2 2
already an inaccurate representation of the A
F = (( AT ) − 1)ρv ,
O
flow.
where AT is the cross sectional area of the tube
The actual damping effect can be explained by
and AO is the cross sectional area of the orifice.
the quantitative behavior of the turbulent flow
If the orifice is a factor of x smaller than the
model. In the damping term of the turbulent
tube, the ratio between the areas becomes
flow ODE, velocity is raised to the power of 1.8.
Under laminar flow conditions (such as those in AT π(xRO )2
AO = ,
the 1/16” tube), we would expect low π(RO )2
8
Flow Oscillation Simulation Andrews
frictional effects that those effects aren’t extracting points from the Moody diagram.
visualized on the graph. Since we were interested in Reynolds numbers
on the order of magnitude of 103 to 104 The
In order to visualize the effect of both the
approximation of a smooth tube probably had a
constrictive drag and the difference between
small overall effect because the slopes through
turbulent and laminar flow, the size of the
the Reynold’s numbers in our range of interest
constriction can be increased (decreasing the
are roughly the same.
constrictive drag). Figure 8 shows the height
versus time plot for turbulent and laminar flow The second approximation resulted from the
simulations through a 1” U-tube with a 7/8” linearization of the Moody diagram. This
constriction. With a 7/8” constriction, the approximation also probably had minimal effect
multiplication factor on the constrictive drag on our calculations because the Moody diagram
force is only 0.7 therefore the effect can be is very close to linear in the Reynold’s number
visualized without overpowering the turbulent range that we were interested in.
vs laminar drag effects. As can be seen in Figure
The total effect of the Fanning friction factor
8, the overall damping is greater than in Figure
approximations and simplification to 2 external
5 (no constriction) due to the constrictive drag
forces can be visualized in Figure 5 by the
force, but the difference between turbulent and
difference between the experimental and
laminar flow can still be observed.
turbulent flow model.
APPROXIMATIONS AND SIMPLIFICATIONS
Approximation of K-value
Simplification to 2 External Forces
In the derivation of the drag force due to the
For all of the derivations, we took into account constriction, we estimated a K value of 2. Since
acceleration by gravity and drag by the K value depends on the geometry of the
visous/friction effects and constrictive pressure orrifice, and we were doing a theoretical
drop. While those are the major factors simulation, technically our result has no error
affecting drag, other factors include do to the K value we chose as long as we agree
hydrophobic/hydrophylic interactions with with the geometry that K value respresents. The
tubing, surface tension effects, tube roughness, range of K values goes from 1.01 (for a smooth
imperfect curvature of tubing, loss due to Venturi contraction) to as high as 2.7 (for a
angular velocty component in bend, etc. The sharp orrifice), so the greatest possible effect of
effect of these neglections can be visualized in K value choice would be a multiplication in the
figure 4 by the difference between the drage force by a factor of 2.67.
experimental and simulated results. The
ACCOUNTING FOR SIGN ERRORS IN MATLAB
magnitude of error by these neglections would
be different in every tubing case with the In the case of the turbulent and constrictive
greatest error percentage likely being in the differential equations, the drag force term is
smaller diameter tubing. non-linear. Two problems arise when handeling
non-linear terms in MATLAB.
Fanning Friction Factor Approximations
Creation of Imaginary Components
In the derivation of the Fanning friction factor
vs Reynold’s number formula, two major The first problem results from having a
approximations were made. First, we non-integer exponent. When a negative
approximated a tube roughness of zero when number is raised to a non-integer exponent,
9
Flow Oscillation Simulation Andrews
Loss of Directionality
10
Flow Oscillation Simulation Andrews
APPENDIX A: DERIVATIONS
11
Flow Oscillation Simulation Andrews
12
Flow Oscillation Simulation Andrews
rho = 1000;
LAMINAR FLOW SIMULATION
r = .001588/2;
ODE Solver Function
L = .3;
function dy = secondODE(t,y)
% function to be integrated g = 9.81;
u = 8.9*10^(-4); h0 = .0324979;
p = 1000;
h= @(t)
r = .001588/2;
exp(((-4*mu)/(rho*r^2))*t)*(h0*cos(t
L = .3;
g = 9.81; *sqrt((2*g/L)-(16*mu^2)/((rho^2)*(r^
4))))+((4*mu*h0)/((rho*r^2)*sqrt((2*
A=p*pi*r^2*L; g/L)-(16*mu^2)/((rho^2)*(r^4)))))*si
B=8*u*L*pi; n(t*sqrt((2*g/L)-(16*mu^2)/((rho^2)*
C=2*p*g*pi*r^2; (r^4)))));
%analytical A=p*pi*r^2*L;
B=.1*pi*u^(.2)*p^(.8)*(2*r)^(.8)*L/4
13
Flow Oscillation Simulation Andrews
14
Flow Oscillation Simulation Andrews
tspan, y0);
A=p*pi*r^2*L;
B=.1*pi*u^(.2)*p^(.8)*(2*r)^(.8)*L/4 y_compLam = yLam(1:end,1);
;
C=2*p*g*pi*r^2; plot(t_compTurb,y_compTurb,'b',t_com
D=15*pi*r^2*p; pLam,y_compLam,'y','linewidth',1.5)
legend('Constricted
dy = zeros(2,1); Turbulent','Constricted Laminar')
dy(1) = y(2); title('2 Models of Flow Through a
dy(2) = -C/A*y(1)- 1/16" U-Tube With 1/32"
sign(y(2))*B/A*abs(y(2))^1.8 - Constriction')
D/A*abs(y(2))*y(2); xlabel('Time from Release (s)')
ylabel('Fluid Height in One Arm of
end U-Tube (m)')
set(gcf,'color','white')
Laminar ODE Solver Function
CONSTRICTED FLOW SIMULATION: LARGE
function dy = TUBE
secondODEconstLamSmall(t,y)
% function to be integrated Turbulent ODE Solver Function
u = 8.9*10^(-4); function dy =
p = 1000; secondODEconstTurbLarge(t,y)
r = .001588/2; % function to be integrated
L = .3;
g = 9.81; u = 8.9*10^(-4);
p = 1000;
A=p*pi*r^2*L; r = .0254/2;
B=8*u*L*pi; L = 2;
C=2*p*g*pi*r^2; g = 9.81;
D=15*pi*r^2*p;
A=p*pi*r^2*L;
dy = zeros(2,1); B=.1*pi*u^(.2)*p^(.8)*(2*r)^(.8)*L/4
dy(1) = y(2); ;
dy(2) = -C/A*y(1)- B/A*y(2)- C=2*p*g*pi*r^2;
D/A*abs(y(2))*y(2); D=15*pi*r^2*p;
end dy = zeros(2,1);
dy(1) = y(2);
Execute dy(2) = -C/A*y(1)-
sign(y(2))*B/A*abs(y(2))^1.8 -
%computational turbulent large D/A*abs(y(2))*y(2);
y0 = [.0324979; 0];
tspan = [0 1.88]; Laminar ODE Solver Function
[t_compTurb, yTurb] = function dy =
ode45(@secondODEconstTurbSmall, secondODEconstLamLarge(t,y)
tspan, y0); % function to be integrated
15
Flow Oscillation Simulation Andrews
C=2*p*g*pi*r^2;
D=15*pi*r^2*p;
dy = zeros(2,1);
dy(1) = y(2);
dy(2) = -C/A*y(1)- B/A*y(2) -
D/A*abs(y(2))*y(2);
Execute
%computational turbulent large
y0 = [.705-.51; 0];
tspan = [0 22.83];
[t_compTurb, yTurb] =
ode45(@secondODEconstTurbLarge,
tspan, y0);
y_compTurb = yTurb(1:end,1);
[t_compLam, yLam] =
ode45(@secondODEconstLamLarge,
tspan, y0);
y_compLam = yLam(1:end,1);
plot(t_compTurb,y_compTurb,'b',t_com
pLam,y_compLam,'y','linewidth',1.5)
legend('Constricted
Turbulent','Constricted Laminar')
title('2 Models of Flow Through a 1"
U-Tube With 1/2" Constriction')
xlabel('Time from Release (s)')
ylabel('Fluid Height in One Arm of
U-Tube (m)')
set(gcf,'color','white')
16