Lecture 51 : Boundary Conditions I

IIT Kanpur July 201
20 followers
2 Video Views·Oct 5, 2024
Hello everyone, welcome back. We are discussing about Blench models, particularly the adiabatic
class of models, where we try to basically find out all the Reynolds system by using
a single variable which is adiabatic specific. And we use the Boussinesq hypothesis and we
relate the stress with the adiabatic viscosity and the strain rate.
So, in the last class we discussed about the k epsilon turbulence model. So, the k is the
turbulence kinetic energy and epsilon is the dissipation rate. And we found that by using
these two variables, by basically finding the transport equations for kinetic energy,
turbulence kinetic energy and the dissipation rate, we can find out the adiabatic viscosity
which was, let us say if I can write it here. So, nu t was like equal to C mu k square by
epsilon, right. So, k has the unit of meter per second. So, this was meter per second
square, this is meter square per second cube and C mu is a constant as well, this typically
taken as 0.09. And that can be basically found out from the
experimental data. So, that gives you the unit, if you take the entire thing, that gives
you the adiabatic viscosity which is of the unit of meter square per second, kinetic adiabatic
viscosity, right. And from this you can also, if you think in terms of nu t equal to u 0
l 0, then also the u 0 is like basically square root of k and then l 0 will become k to the
power 3 by 2 by epsilon. So, this is your u 0 and this is your l 0 and then some multiplying
constant. And we saw that this k epsilon turbulence model does not do very good, very close to
the wall, right. So, we actually discussed about the low Reynolds
number model also. Basically, we had Reynolds number, and we saw that very close to the
wall you need to do some special things, use some kind of damping to get the behavior right.
But even with that damping, k epsilon turbulence model is not considered to be very good, very
close to the wall. Especially, let us say you have very high curvature as you see in
the turbine blades or in many different configurations, k epsilon turbulence model does not do very
well. So, that led to development of another class of model called the k omega model.
So, the k epsilon model as you know that is family of models, your standard k epsilon,
your barangay k epsilon, your realizable k epsilon. And that was like I think from 1970s
several versions have come starting from the original work of Spalding. So, k omega also
has several versions, and like we have Wilcox k omega, we have Mentos k omega, and there
are many different versions in between. Wilcox himself has many versions. So, when you are
using a k omega model, you should basically know that which version you are using.
Nowadays, k omega model by itself is not used, what is used is SST k omega model. So, these
two models are different, k omega standard k omega and SST k omega, both are different.
We will discuss about those models in this class, and we will also find out why this
k omega model is not used, standard k omega is not used in practice in CFD simulations.
Let us look at this standard k omega model. So, as we know that we need two quantities
to basically model this eddy viscosity or the turbulent viscosity. So, in the k epsilon
we had this turbulent, I mean k was the turbulent kinetic energy, which is given as half u prime
square plus v prime square plus w prime square. And then we had the dissipation epsilon, which
was the turbulence dissipation that is given here, unit was meter square per second, which
was something like del u i j by del x j del u i by del x j, something like that, there
is a quantity. So, this represents that how the dissipation is happening.
So, what we do in this k omega model, that instead of using a turbulent dissipation rate,
we define another quantity, which also basically characterizes how the dissipation is happening
in the flow. Always remember that original diagram that we had the energy cascade, that
the famous Richardson's poem, that big holes have little holes and they feed on velocity,
little holes have even lesser holes, which feed on viscosity. So, you have something
which is basically the kinetic energy is coming in and this is going dissipated here at the
smallest scale. So, you need some equation, which tells you
that how the kinetic energy being produced, transported, how the kinetic energy changes
between the different eddies and something, which represents your dissipation. So, in
the k epsilon we had this k and epsilon two things, but now in this k omega model, it
is a slightly different model from k epsilon, instead of epsilon we are going to use a quantity
called omega. And omega is called the specific turbulence dissipation rate and this unit
is 1 by second. So, in a way it represents that at what frequency the eddies are basically
getting formed and getting destroyed. So, that is the major what omega gives you.
So, because this unit is 1 per second, so it is a kind of frequency and it can be related
to your epsilon, because this is also representing dissipation rate by epsilon by k. So, omega
is 1 by second, k is meter per second, meter square per second, meter cube per second,
meter per second square and you have epsilon is meter square per second cube. So, the quantity
which is epsilon by k, that will become your, that will be proportional to your value omega.
Because this is meter square per second cube and this is meter square per second, so that
will give you 1 by second. So, you will take this, I mean if you go. So, this c mu term
you can, it depends on many models. In many models you do not use c mu, in many models
you use that. So, that also you have to be very careful, but if you are using the c mu
term omega is equal to epsilon by c mu k. And if you go the, for the definition of nu
t, nu t by c mu k epsilon by epsilon earlier. So, this will become c mu k square by, you
put the value of epsilon, that will be your omega. I am putting the value of epsilon from
here, omega c mu into k. So, the c mu, c mu gets cancelled out and this becomes k by omega.
So, now our task is to find a transport equation for omega. Equation is as we have discussed,
this can be derived from the first principle. Although you can basically make some of the
assumptions about how certain terms which are not very important, like we saw that viscous
transport of the kinetic energy or certain stresses, that you can basically omit. So,
there are some modeling aspect there, but it can be derived from the first principle.
And omega model like the epsilon equation, omega equation also has to be basically completely
modeled, the equation will be completely modeled. So, we need to find out the equation for k
and omega. Once you find k and omega at every time step, we can take the ratio of these
two and get you this nu t. So, let us look at those equations. So, the k epsilon we have
already here. So, this is basically p k and we know meaning of each term. This I am showing
for the incompressible flow and you have sigma k and sigma omega. So, earlier we had sigma
epsilon here. So, let me look at this picture for the compressible flow first. So, we had
in turbulent kinetic energy, we had this temporal term, advection term, the diffusion term and
many terms in the diffusion we had ignored. For example, the diffusion due to the pressure
or the viscous effects, we have ignored that and then we have this sources and sinks.
So, taking the inspiration from this equation, we will get this omega, replace the k with
omega and make sure that you have basically all the terms that you require. So, the time
dependent term of omega, how the omega is getting invected in the flow, how it is getting
diffused and like the way you had sources and sinks in the k equation, we will have
sources and sinks in the epsilon equation also. So, for example, the production of kinetic
energy by the source and the dissipation. Here, it will be basically there would not
be any gamma term here. So, this is the modified term for the omega equation here. So, p k
omega by nu t and there is another term for I mean production of the dissipation, a specific
dissipation rate and the destruction of the dissipation rate or the basically dissipation
of the dissipation rate. So, this will be omega again. So, we got the equation for the
k and omega. If you look at in a simpler form, I will come to this gamma. This is basically
taken for a transitional model. Right now, you can just think of it 1 by nu t. So, this
is for the incompressible flow. That is why you see there is no rho here. So, I am giving
equation for both compressible and incompressible flow and this is how the equation looks like
in the incompressible flow. So, what we note in this is that the coefficient
that we had earlier, we had this c epsilon 1 and c epsilon 2 in your dissipation equation
in the k epsilon model. Now, the terms model coefficients we have are like this is kind
of beta. Now, this sigma w and there is a certain difference between the epsilon equation
and omega equation. It is not that if you put epsilon equal to or omega equal to epsilon
by s mu k, you will get the omega equation. That is not the case. So, there are certain
terms which are missing. It is just that inspired by the kinetic equation, we modeled a dissipation
kinetic equation and we got some coefficients. So, these coefficients are basically different
from the k epsilon model, but the beauty of this model is that it basically solves all
the way to the wall without any damping function or anything. In fact, this model is very good
for the wall. So, whatever damping that you need near the wall, you want to basically
model the viscous sub layer or the laminar sub layer, the log layer and all those things.
This model can do very well. The way the coefficients have been tuned in this particular model that
you do not require any damping function. It is in by default. It is a low Reynolds number
stabilized model. So, it works better very close to the wall and it is better because
of that reason only because whenever you have this adverse pressure gradient or there is
a separation. Those things effects of those things are seen in the boundary layer and
because this model works very well near the boundary layer. This is very good for this
kind of flows, but the problem in this model is that it is very sensitive to the free stream
turbulence. So, once you when you try to do a range simulation,
you specify certain values. For example, you give the turbulence intensity which is given
as I. This is basically in a way you can call it that this is u prime by u. What is your
fluctuation? Estimated fluctuation with respect to your free stream velocity and then sometimes
you give the length scale. Sometimes you give the ratio of the eddy viscosity to the dynamic
viscosity. All those kind of things you can write. So, one is very important which is
the turbulence intensity. Now, what problem with this particular model is that let us
say you do a simulation with a turbulence intensity, free stream turbulence intensity
of around 10 power minus 5 and get certain results. Now, you change it to let us say
something like 5 into 10 power minus 5. There will be drastic change in your results in
terms of let us say prediction of separation or the skin friction and all other things.
So, this model has found to be very sensitive to the free stream turbulence. So, because
of this sensitivity, this model is not very good because you know when you do an experiment
and you try to measure the turbulence, there is always some uncertainty. If the model does
not account for that kind of uncertainty and it gives you very drastically different results,
then the model is not good for your CFD. So, that is the reason this model is not very
good. It does not work well for a wide class of flows. So, to improve this situation, we
use some kind of model which is called the k omega system model. So, what it does? So,
you have this k omega model. So, let me draw a picture here. So, let us say this is a boundary
layer and this is your free stream. So, you have a model which works very well here which
is k omega model. Then, there is a model which works very well which is the k epsilon
model which is away from the boundary layer. So, can we basically blend this model and
propose a new model which has the advantage of both of them? It is not very sensitive
to the free stream turbulence. That is what the k omega epsilon is and it solves the flow
accurately very close to the wall. That is the k omega. So, that blending is what is
called the s s t k omega or k omega s t. So, s s t is shear stress transport and k
omega is k omega. So, let us look at that model now. So, once again k omega epsilon
model is very good very close to the wall, but it is extremely sensitive to the free
stream turbulence. That is why it is not very useful because you will always have uncertainty
in your measurements or you do not want the thing to change drastically. If your turbulence
intensity shifts by let us say 5 percent, I mean the intensity shifts by certain value
like 5 percent, 10 percent. You want your model to be robust and that is what you see
in your s s t k omega turbulence model. So, to learn about this s s t k omega or k
omega s s t model, let us review the epsilon equation in the k epsilon model. We want to
compare this dissipation term in the k omega model and the k epsilon model. That is the
purpose. So, this is the equation that we have for the k epsilon model for the epsilon
transport equation. Now, what we do is that we take this epsilon. We know that omega is
equal to epsilon by c mu k. So, why do not we just put it here and can we derive a omega
equation? Of course, we can do that. Just replace this epsilon by omega c mu k and try
to derive an equation. So, what we will see is that you will get the equation which looks
like this. So, remember we are using the epsilon equation from the k epsilon model and putting
epsilon equal to c mu k omega. So, this gives you an equation. Of course,
you have to do some manipulation there and take the help of the k equation that you already
have. So, this is the equation that you are going to get. Now, this term is there in your
k equation also. This is also there in k equation. This is also there in k equation. This is
also there in k equation, but this is the term which is not there in k equation. So,
this is a kind of cross diffusion term between the turbulence kinetic energy and the space
to dissipation rate. So, if you take the epsilon equation and try to derive your k equation
from that. So, this is the equation omega equation from that. This is the equation that
we get and this is different from the omega equation that we use in the standard k omega
which is basically just taking the expression from the turbulence kinetic energy equation
and using some model coefficients. So, this equation, I mean this term may also
be responsible for this very weird behavior of the k omega model that it is very sensitive
to the friction turbulence. So, what we want to do that we want to somehow retain this
term away from the wall. If you retain this term away from the wall, that means you are
essentially solving the k epsilon model because this is nothing but the k epsilon model and
make this term disappear very close to the wall. So, that it is able to give the damping
that you get from the k omega model. So, if you remove this term, what you are getting
is the standard k omega model, the k equation for the standard k omega model. So, that is
how this term will become. So, this is your k equation and this is your omega equation
and then you will have additional term, but remember here it was 2 rho sigma omega 2 again
sigma omega 2 is a coefficient model coefficient omega delta. So, this is basically term which
is basically multiply each term of delta k and delta w. So, this will be k by delta x
into w by delta x plus k by delta y plus delta w by delta y and k by delta z into w by delta
z. That is the term that is represented here. So, we want to retain this term. So, instead
of basically solving 2 different equations, 3 different equations which will be expensive,
you can do it may be you can solve k epsilon and omega 3 equations and then basically use
k omega near the wall and k epsilon away from the wall, but instead of doing that why not
we use this k omega only equation which is this omega equation, but use some kind of
blending. So, that this term goes to 0 very close to
the wall and becomes 1 away from the wall, but the blending has to be done in such a
way that there is no abrupt jump. So, that is what is done in a s h t k omega model.
So, you can see this term which is basically 1 minus f 1 term. So, what does this 1 minus
f 1 term does is when f 1 equal to 1, I will show this behavior of this term. This depends
on the distance from the wall. So, when f 1 is equal to 1, you can see that this term
goes to 0 and then that will become k omega. So, that means we want f 1 equal to 1 very
close to the wall or 1 minus f 1 to be 0 and when f 1 equal to 0, the equation is k epsilon.
So, one thing I am not explaining. So, I have been using this term gamma. So, for now you
can take gamma equal to 1 for all this flow. Gamma is a representation of intermittency
which is I will talk about it very soon may be today or in the next class. It is like
if you have a boundary layer which is basically laminar and that becomes transitional and
then become completely turbulent. So, then this value of gamma is equal to 0 here and
in the fully turbulent region, this is equal to 1 and this is basically 0 and 1. So, these
are fully turbulent models. We take gamma equal to 1 by default, but this gamma term
is very useful. Let us say if you want to do for the transitional flow where the flow
can be laminar in some region and turbulent in some region and in between it can be transitional.
So, for now you can assume this gamma to be equal to 1 or you can remove this term also
in the SST k omega model, but what I am trying to say that we want to retain the behavior
of the k omega model very close to the wall. So, that will be given by this function 1
minus f 1. So, if f 1 equal to 1, the equation is k omega f 1 equal to 0 equation is k epsilon
and also we can not only do the blending here, but we can use this parameter f 1 which is
the blending parameter to get this behavior everywhere. I mean you let us say you have
c epsilon and alpha somewhere c epsilon is the coefficient which is there in k epsilon
and alpha is the coefficient in k omega. So, you can use this kind of blending in those
all the parameters that you want for the omega and for the epsilon. So, whenever f 1 equal
to 1, this is behavior of omega f 1 equal to 0 is behavior of epsilon and this f 1 depends
on the distance from the wall. So, if you plot this f 1, so here I am plotting 1 minus
f 1. So, because I want to show how this thing works about between the k omega and k epsilon
is more natural to understand. So, when f 1 minus f 1, so when f 1 equal to 0, 1 minus
f 1 equal to 1. So, that is the point which is very close to the wall. So, let us say
this is my wall, this is my y. So, this is very close to the wall. So, the value 1 minus
f 1 is equal to 1. So, no. So, if it is if 1 minus f 1 term is equal to 1, then f 1 equal
to 0. So, that will be k epsilon. So, this is actually away from the wall. So, this is
may be away from the wall and this is may be close to the wall because I have plotted
in 1 minus f 1 and you can think of it is the distance y is may be some kind of y inverse.
So, what I am saying that when minus 1 minus f 1 is equal to 1, this is the behavior of
k epsilon and this is the behavior of k omega. Same thing is shown here. So, when f 1 is
equal to 1, you can see that 1 minus f 1 will become 0. So, this term will go to 0 and you
have the pure omega equation that you had in the standard k epsilon k omega model and
this will be your k omega model. When you go away from the wall, so your f 1 equal to
0. So, 1 minus f 1 is equal to 1. So, this term remains and this term actually
becomes this term becomes 1. So, then finally, you have the pure k epsilon. There may be
typo here. The two may be a typo. I will basically correct it and send you in the lecture notes.
In between, you have to make sure that the blending is perfect. So, you use some kind
of function here, so that it is not an abrupt jump from 1 to 0. It is not like at one grid
point, you have the k omega and other point is k epsilon. So, k omega, this is k epsilon.
It will be a problem because you are basically making drastic change and a drastic change
is not good in any CFD simulation, whether it is a drastic change in the grid, whether
it is a drastic change in the flow parameters. The solver is always a difficulty if you have
a very drastic change. So, in order to avoid that, we will use some kind of a blending
function and that is how you will get it. This is basically given as I think some kind
of tan inverse function. So, what is the advantage of the s s t
omega model? So, you basically use the best properties of the k epsilon and k omega in
the respective regions. This is a very important thing in respective regions. So, let us say
you use k omega away from the wall and k epsilon close to the wall. Then, you are just doing
the worst because these models are not going to work there. So, you are basically doing
k omega near the wall and k epsilon away from the wall. So, it is not tan inverse, but tan
h. You use f 1 equal to tan h argument 1 to the 4, where argument 1 depends on the distance
from the wall. It is a function of distance from the wall. So, you take some value of
the distance from the wall, take the tan hyperbolic profile, draw it and then based on the value
of this f 1, the equation may be k epsilon k omega or may be a mixture of the k epsilon
k omega. That depends on the value of f 1. So, I hope you understood what is the main
principle behind this s s t omega model.
Now, there are versions of s s t k omega model also. So, there is this b s t model, which
is called a baseline shear transport model. So, as we know in the k omega model that your
mu t is equal to k by omega or mu t is equal to rho by omega. This is how you will get
it from the k omega model, but mentor is a famous modeler. You will see many models from
him. You may go back and read literature about the models that he has proposed. They are
very successful models. So, he proposed later that we need to introduce some kind of viscosity
limiter. So, you can see here is a value called a 1 and then there is another function
called s f 2. So, what it says that I mean in certain regions, you can get the value
of k by omega, but in some flows that viscosity that you are getting from the k by omega is
very high. It is so high that it is basically let us say you have a expansion fan.
In some region, you expect that your viscosity should eddy viscosity should go down, but
this model if you use the baseline turbulence model, it is still going to produce I mean
very high level of k by omega. So, what mentors assistive does is basically uses some kind
of limiter here. You can see the function. This is the max function. So, it will take
a max of this and because this is denominator, your value of this k by omega or basically
effective k by omega will go down. So, this mentors assistive is the most popular model
now with the viscosity limiter. You can actually change this value of limiter. I think the
default value of limiter is a 1 equal to 0.31. You see that wherever this a 1 omega is greater
term, you will get back to this value mu t equal to rho k by omega, but when this s f
2 is greater than this a 1 omega, this term s s is the structure parameter and f 2 is
another function. So, you will have this term dominating which is a 1 rho k by s f 2. So,
because denominator becomes higher, it is lesser than k by omega. That is the idea.
We want to reduce the eddy viscosity in certain regions. So, this model basically gives prediction
better prediction for separated flows. In fact, this is the one of the best model for
the flow with separations. So, this is very popular in your turbo machinery flows or wherever
flows you have got high curvature or you want to you are expecting separation in those flows.
You would like to use this mentors assistive model. So, I think we will stop here. In the
next class, we are going to discuss the transitional model. Then, we will briefly look at the some
of the advanced turbulence model. Then, we will stop the discussion about the trans module.
Thank you very much.
