Lecture 43

IIT Kanpur July 201
20 followers
Sep 13, 2024
This is lecture number 43, hello everyone, welcome back, we have got almost all the gradients
that is required in CFD to basically solve a fluid flow problem, we have looked at the
governing equations, we have looked at different discretization approaches, flux schemes, the
properties of the flux schemes or discretization schemes, then we also looked at the finite
difference approach, finite volume approach, we solved some small problems and also we
solved the soft tube problem, right. So, now the only thing that is ideally required
is the boundary condition and we will have the discussion about the boundary conditions
very soon, but there is one piece which is very puzzling and that needs to be solved
is the aspect of the turbulent flows. As you might be knowing that most of the flows that
are around us are basically turbulent flows, right. You look at a smoke that is coming
out of a chimney, flow is turbulent, right. You can look at the flow behind an aircraft
or flow behind a car, the flow is turbulent, right. So, it is not that the equation that
we have learnt that is the Navier-Stokes equations cannot solve that problem, right. It can solve
the problem. We have the scheme, we know the schemes, we know the equations, we can discretize
it to a grid and try to solve the problem, but when you try to solve that problem, any
turbulent flow problem at a sufficiently higher Reynolds number, I will come to that number
Reynolds number again. We have discussed or we have seen that Reynolds
number while non-dimensionalizing the Navier-Stokes equations, but that is the number. I am just
jumping ahead here that dictates whether the flow is going to be laminar or turbulent,
right. If the Reynolds number is sufficiently high, let us say if we are talking about a
wing or an aircraft, right. So, the Reynolds number typically is around a million, right
or several millions, right. At that high Reynolds number, if you try to use the Navier-Stokes
equation, put a grid and try to solve it, you will run into trouble and the reason for
that is if the Reynolds number is very, very high, you will have several scales in your
flow, okay. So, if you have like let us say 1 million Reynolds number, you have separation
of the scale that is in other way, the separation of the grid size is of a million or the order
of a million. So, if a smallest scale is let us say 1 millimeter, then the largest scale
will be 10 power 6 millimeter, right. So, that is the kind of scale that we are talking
about, right. So, you will quickly see that your grid size will become very, very large.
So, I talked in one dimension, 1 millimeter to 10 power 6 millimeter, it can go to in
the three dimension, it will go to 10 power 6 power 3, right. So, your grid size will
become humongous and it can become so big that for a practical flow problem, you just
run out of the computing capability. Even if you use the largest x of top machine,
beta flop machine, x of top machine that is available now. So, flop is basically floating
point operations per second, right. So, that is a measure of how much computing power one
computer has, right. So, one typically it is talked in terms of the supercomputers,
right. So, even if you use the fastest supercomputer or the most capable supercomputer in the world,
you cannot solve the problem directly if the flow is turbulent, right.
Then, how do you solve that? In fact, in most cases, you might be using some kind of commercial
software or you know software. You must be getting the answers and there are many aircraft
that have been designed using these softwares, right. The cars, aircraft, wind turbine, turbine
blade, anything you name it and we have been using CFD to basically predict the flows on
these geometries at realistic design conditions. So, how are you doing it? Where is the catch?
So, that is the puzzle that we are going to look at today. So, basically we want to know
how the turbulent flow works, right. If you understand the Cartesian turbulent flow, we
may probably answer this puzzle or basically try to find out a way so that we do not need
to use that kind of computing power as we have in the direct numerical simulation.
That means using the numerical equations as it is to solve a complex turbulent flow problem
at sufficiently high Reynolds number. So, let us, so the topic today is the simulation
of turbulent flows, right. So, before we go to the turbulent flows, I want to emphasize
the difference between the laminar flow and the turbulent flow, right. So, this goes back
to the very old experiment by Osborne Reynolds, I think done around 1893. The Reynolds number
is after this person. So, his name is Osborne Reynolds. So, Reynolds number is behind this
person. So, what he did? He designed a very intelligent experiment and you can see the
setup of the experiment on the right hand side here, right. So, what he did? He took
a water tank and connected a glass tube to that, right. So, glass tube is so that you
can see what is happening inside that flow and then he put a valve here. You can see
the valve here. So, this is the valve that he has put in the tube. So, if you open the
valve more, then the flow rate will become higher and then he introduced the die in between
the tube so that it does not disturb the flow. So, it is basically sufficiently away from
the valve and then what he did? He started opening the valve slowly. So, initially when
you open the valve, what he saw like let us say this is my tube. So, what he saw is basically
if the flow, the die is like this is a red colour die. If it goes in a line, it is actually
going in a line. It is not deviating. So, you can have different lines and they are
not changing. So, in other words, the flow is smooth, right. The flow is smooth, no disturbances.
Even if there is slight disturbances, they get damped out and he called this kind of
flows as the laminar flows. Remember, this happened when he did not open the valve fully,
right. So, the flow speed was quite low. Then what he did? He basically increased the speed,
right. Now, this is second time when we increase the speed, valve same, same size of the tube
may be around here and he found that now it is slightly wavy. So, it comes like this and
then it becomes slightly wavy. This is when the speed is increased slightly and then what
he did is he further increased the speed and then what he found is basically now it
is like there is so much zigzag, so much disturbances all over. They are not basically staying in
line. So, this you can see that now they are no longer very organized as we had earlier.
So, this was an organized flow on the top and this is the flow at the bottom is basically
not organized, very violent. There is mixing all over everywhere, right. So, he termed
this flow as the turbulent flow because it is turbulent. This is laminar, this is turbulent
and in between you can have flow which you can call as the transistor flow.
Now, what he did that now he basically changed the liquid. So, earlier he did with water
then he made little bit more viscous and tried seeing that what is happening. He of course
observed the same phenomena. It was initially laminar and it become turbulent, but it happened
at different speed, different flow speed. So, then he experimented with different liquids
with different density, different viscosity and he tried finding out that what is that
parameter which governs whether the flow is going to be organized or flow is going
to be disorganized, right. So, then he came up with a Reynolds number
which is now known as Reynolds number which is a non-dimensional parameter which is given
by rho u L by nu. In this case because it is a tube, so L is the length scale. So, that
becomes d. So, this is basically in this particular case this rho u d by nu. So, it does not depend
on just the velocity. It also depends on your density, your viscosity and the shape and
the size of your instrument that you are using, right. So, in this case it is tube, right.
So, this number I am sure you are already aware of. You have done the fluid mechanics
course. So, you can see this as basically ratio of inertial forces by viscous forces
because viscosity is at the denominator. So, as you can see if Reynolds number is low,
so let us say Reynolds number is low. Typically, let us say Reynolds number is less than 2000
for a pipe flow. As I told it depends on what kind of instrument that you are using or what
kind of vessel that you have where the flow is happening. So, here it is tube. Let us
say it was a channel or something else. Then, that number will be different. It can be flat
plate, right. So, let us say if Reynolds number is less than 2000 in the case of pipe flow,
then he observed that the flow remains laminar. Then, when the Reynolds number was greater
than 4000 for the same flow, he observed the flow was becoming turbulent, right. So, remember
this 2000 and 4000, it is not a sacrosanct number. Number can change if you can make
the environment very quiet, right. So, let us say that you can make the pipe completely
smooth. There is no friction there and the environment is also very quiet. Then, you
can basically increase this number to a very high value. There are super pipe experiments
in Princeton where they have observed that they can maintain the flow to be laminar to
a very high Reynolds number, but typically for our understanding, we take that in general
situation. If the Reynolds number is less than 2000, the flow is laminar and if the
Reynolds number is more than 4000, the flow becomes turbulent. In a pipe flow, in a boundary
layer flow, it will be a different number. In a channel flow, it will be different, right,
but what we understand is that we need a basically combination of all three, all four to basically
get the value of the Reynolds number and that will dictate whether the flow is laminar or
turbulent, right. So, you can also write it in terms of you can take this rho at the bottom
and this becomes u d by mu, where mu is now the kinematic viscosity.
So, this is what I was saying that this is a typical picture of the laminar flow. If
the flow is happening in the laminar region, if it goes like this, it will continue to
move in that line, but in the turbulent flow, there will be a lot of mixing, right. There
will be, I will use the term called eddy. You can see these shapes here. So, these are
basically kind of your eddy shapes. So, what they do? Basically, they do a lot of mixing
and depending on Reynolds number, there will be a violent mixing from top to bottom and
that actually develops additional stress also. Typically, you would hear that the turbulent
flow has higher drag, right. If there is a wing and the flow is turbulent, the drag is
higher. Why? Because it develops higher stress that is because of this, the momentum that
is being transferred by this eddies and this is a picture in the same pipe, right.
So, if this is my pipe and if the flow was laminar, your velocity profile will look something
like this. It will be a parabolic profile. You all have solved this problem, okay. In
a quiet flow, poiseuille flow, right, it looks parabolic. So, you have the maximum velocity
here and there will be no slip condition here. Basically, the velocity goes to 0 at the boundary,
okay. So, this is my wall on top and bottom, okay and the velocity goes to 0 at the wall,
right. On the other hand, if the flow is turbulent, you would see that it becomes something like
this, okay. It is now more flatter, more fuller, okay.
If you compare the region, let us say if I zoom in this region, say if I take this region
and zoom that in, okay, so how will it look? So, this is my wall, okay and your laminar
boundary layer is going like this, okay and your turbulent boundary layer is this. This
is your turbulent and this is your laminar, okay. So, what is this? This is my velocity
profile. I am plotting the velocity here. Again, you may ask that I just told that lot
of mixing, then how can the profile may look so smooth like what I have drawn here, right.
So, again there is a point here that what I am drawing here is not the instantaneous
flow profile. That means that it is not a measurement at a given time, but let us say
I measured it over some time and then I take an average of that, okay or what I can do?
I can repeat the same experiment many times and take an average of that. That will make
the profile smooth and that is what is called the mean flow profile. So, that is denoted
by a bar sign. So, here in the laminar case, anyway there are no fluctuations, no eddies.
So, then it is a smooth profile. On the other hand, if you look at the turbulent profile
here, for example, you can see there is a lot of waviness. If you measure at any stand
and this change every time, okay, let us say you took at one time, it looks like this.
Another time it may look like this and then another time it may look like this, but when
you take an average of that, then it becomes a smooth profile, but it is different from
the laminar profile. That is the point I am trying to make, right. The slope here in this
case, the slope near the wall in the turbulent flow is much higher. So, right now this is
u and this is my, you can call it y. This is x axis and this is y axis. In this case
also this is x axis, this is y axis. So, this is if you call it d u by d y, okay. So, this
is in turbulent flow is higher than the d u by d y in the laminar flow, right. So, this
is the laminar flow. You can see the, so what is happening? So, let us say this is my profile
in the free stream. This is the velocity that is getting in the free stream. So, it goes
to that free stream velocity in a much quicker time, right. Let us say it is reaching the
free stream velocity somewhere here or may be here. It reaches the free stream velocity
in this particular figure around here, right. So, this becomes like this. So, it reaches
somewhere here and here you can see that in order to get to that velocity, it basically
takes a very long length in this direction, right. So, if this is my u and this is my
y, so this d u by d y is smaller here, right. Why? Because it has to go to 0 velocity. The
velocity is 0 here, right. It went to u, but it took a very long, very long distance to
get to that u. Here in this case, it took a very small distance to get to the u and
why it is happening? Because of the presence of the eddies, it quickly generates momentum
and going from the no slip to the free stream condition, it takes much smaller distance,
but it has to maintain the no slip condition. So, it has to go to 0 at this point. So, to
go to this free stream, it will take a smaller distance and that is why this d u by d y will
be higher in the turbulent flow. In the laminar flow, it will take a longer distance. This
is the distance it is taking. So, that is why your d u by d y is smaller in the case
of laminar flow and that can be related to your wall shear stress, right.
I made a statement that the wall shear stress in the turbulent flow is higher. You can see
that how. So, the wall shear stress is defined tau w is equal to mu d u by d y. You can write
a mu y at y equal to 0, right. So, let us say this is your y equal to 0. You can take
either side this or that side or if you are doing a flat plate. So, this is your profile.
So, this is your y equal to. So, because your d u by d y is 0 is higher at y equal to 0
in the case of turbulent flow. So, that is why the wall shear stress in the turbulent
flow is higher than wall shear stress in the laminar flow. Remember, this viscosity is
the property of the fluid. It does not change because of the change in gradient. You are
experiencing a higher stress and that gives you the drag. So, that is why this wall shear
stress is related to the skin friction drag. As you might know, there are mainly two kinds
of drag. One is the form drag and other is the skin friction drag. So, the form drag
is because of the shape and the skin friction drag is because how the flow is sticking to
the wall, right. So, that is why the drag in the case of turbulent flow is higher than
the drag in the laminar flow. Of course, there are other considerations that you should keep
in mind that if the flow is turbulent, then there is a chance that the separation may
be delayed, right. Let us see if you are trying to simulate flow
over a cylinder or a steady flow over a cylinder, if the flow is laminar, it may separate early.
It may separate here if the flow is laminar. On the other hand, if the flow is turbulent,
it may be delayed and it may separate here. So, the separation introduces another drag
because it changes the effective shape, the form drag changes. So, there are other considerations
also, right. So, that is why you might be knowing about the flow control techniques
on the wing. So, let us say if the wing, the flow is separating, you put some kind
of vortex generator. What it does? It makes the flow turbulent and that is why the separation
is avoided, but that is not something that we are going today. We want to first emphasize
that how the different boundary layers look like the laminar boundary layer and the turbulent
boundary layer, but we have still not gone to the problem that why the turbulent flow
is difficult to simulate. So, let us look at some of the flow pictures of the turbulent
flows. So, this is for example, flow in a jet. What you see? You can see that there
are so many kinds of scale here. You can have one scale. You can think of this as this is
a very big scale here and then there are very very small, I mean shapes that you can see
here. They are very very small ladies, right, but if you take a mean of that, if you basically
time average it for a long time, then you will see that all these small small shapes
disappeared and what you look at is a very nice mean flow.
Similarly, if you look at a boundary layer, you have like so many scales and so many phenomena
that is happening here. You can see at the bottom picture and if you average that out,
it looks quite smooth and this is how your dualistic boundary layer will look like. Now,
let us delve little bit more into the turbulent boundary layer. So, the turbulent flow can
happen in two situations. One is called free shear flow. So, in the free shear flow, there
is no wall shear. It can come out of a wall for example, in a jet, but after it comes
out, it is doubling on its own. So, this one free shear flow, you can have jet, you can
have wake and the other is basically wall wounded flows.
Now, wall wounded flows are another situations, which makes the thing difficult, right. Why?
Because whenever you have wall, you have to apply the no slip boundary condition there
and when you apply the no slip boundary condition, let us say you have developed some algorithm
which nicely works. It gives you a nice structure in the form of matrix, but then you have to
put the values u and v and w equal to 0 on those walls, right. So, then all your algorithms
need to be basically somehow account for those boundary conditions, but most of the situation
that we are dealing with are the walls, right. You want to design a car, you want to design
an aircraft, you want to design a turbine blade, you want to design anything, it has
got a wall, right. The flow will stick to the wall. So, no slip condition has to be
satisfied, right. So, to discuss the wall wounded flows, we need to know that how the
boundary layer gets developed on the wall, right. So, that is why we will discuss little
bit about the turbulent boundary layer that how the flow looks like on a flat plate.
So, flat plate is just an example, but you can generalize it to anything, right. You
can generalize it to a let us say a wing or a turbine blade or an airfoil or cylinder,
whatever thing you want. So, this is a typical picture of the boundary layer. So, you can
see that initially you have a laminar boundary layer, the flow looks nice and smooth and
then in between you have a transition region. So, flow starts getting unsteady. You may
have some spots of the turbulence getting developed here and finally, it becomes fully
turbulent in this region, right. So, this is how typically flow develops. The flow is
coming from left to right and this is the edge of the plate.
Now, if you look at any layer, so this is my typical profile. This is the profile that
you have. You can divide that profile into two parts. One is the inner layer and other
is the outer layer. So, inner layer is the region where the viscous effects are taking
place. The viscous effects are important or the viscosity is important. So, I will discuss
about those inner layer, but you can further divide this inner layer into two parts and
that typically accounts for like 50 percent of your 15 percent of your boundary layer
thickness. This is the inner layer. You can divide into two parts. One is very close to
the wall and that is called the laminar sub layer or viscous sub layer. So, I will use
the word viscous sub layer because the viscosity plays the most important role here. Let me
draw a picture here. May be it will make the things clear. So, this is the boundary layer
made for let us say flow over a flat plate. So, what I am saying here is that we can divide
into two parts. This is my inner layer and then you will have the outer layer. So, this
here the viscosity is important or viscous effects are important and they are not very
important outside because the flow is turbulent here. So, this accounts for around 15 percent
of your total region. So, it can be further divided into a layer very close to the wall
which is called the viscous sub layer and then we will have a laminar sub layer.
So, this is the buffer zone. This region is called the buffer zone and then we will have
something called the log layer. So, all these things come under the law of the wall and
then you have the outer layer. So, there will be something called overlap layer. So, in
between you will have there will be layer which is we call as the overlap layer and
then after that we will have something called law of the wave.
Now, it so happens that in this region. So, in the inner layer if you go to very close
to the wall, the viscous dominates that is in the viscous sub layer. In the outer layer
your turbulent stress dominates as we have already discussed and in the overlap layer
we have both stresses which are important. So, Prandtl and Karman did some study and
mostly on the understanding of the physical picture and also about the dimension analysis
and they found out that in the inner layer. For example, the flow is characterized by
your wall shear stress. Let me write as w and rho and u and the distance from the wall
and in the outer layer it is basically characterized by your pressure gradient. There is no the
pressure gradient does not play a role in the inner layer. The free stream pressure
gradient pressure gradient I will write as delta p. This is the free stream and then
we will have wall shear stress and then boundary layer thickness. These are the things that
will dictate the outer layer and based on that what they did? They formulated the non
dimensional numbers and studied how the flow behaves.
So, what I will do in the next class is I will formulate this non dimensional numbers
equate it different terms and try to derive the laws that we have discussed basically
the law of the wall. I will not be discussing the law of the wake because that is not terribly
important in terms of modeling of this turbulence, but the law of the wall will be very important
when we talk about the different turbulence models. So, let us meet in the next class.
