Introduction to Finite Volume Methods-I

Prof. Ashoke De
Department of Aerospace Engineering
Indian Institute of Technology, Kanpur

Lecture – 14

(Refer Slide Time: 00:14)

So, welcome to the lecture of this finite volume method. Now, we can actually
approximate how many number of points you can have.

(Refer Slide Time: 00:28)

So, this is a particular surface, let us say which is connected with F 1 and I have 1
integration points. So, that means, in this case i p is 1 and also w ip would be 1. So, I get
a nice calculation for the flux calculations. Now, one can have multiple integration
points. If you have multiple integration points, then what happens?

(Refer Slide Time: 01:37)

Let us say, you have 2 integration points along the surfaces ok. So, every connecting
surface will have 2 integration points 1 and 2. So, in this case, i p is 2. So, if i p is 2, now
my weighting functions are going to be different and then, I can get 2 different numbers
x i 1 and x i 2. These are the and waiting function w 1 equals to w 2 equals to half.

So, the previous case I had one integration point. So, w 1 was 1 this case I have 2
integration point. So, the waiting function would be half and half and this xi 1 and xi 2
dictate the distance from one particular point to f 1 and then other side f 2. So, they are
the distance.
(Refer Slide Time: 02:45)

Similarly, you can have 3 integration points. So, that case, my i p would be 3. So, if you
have 3 integration points, i p would be 3. Then, obviously, my waiting function w 1
would be 5 by 18 w 2 would be 4 by 9 w 3 is again 5 by 18.

So, if you look at that waiting function, they are equally sort of distributed then the
distances would be x i 1 is 5 minus root 15 divided by 10 x i 2 is half and x i 3 is 5 plus
root 15 divided by 10. So, these are the distance from one point to the other points ok.
So, essentially what I got is that, rho v phi dot ds. It get me back the summation of f this
is faces over v. So, i p over i p f w i p rho v phi i p dot s f.

So, that is what you get and similarly, for gamma del phi dot ds, that is my diffusive
term. So, you get f faces over v and i p i p over f w i p minus gamma del phi ip dot s f.
So, this is what you end up getting and the source term you can always calculate by
doing the integration like, the source term you can always get q dv equals to summation
of i p over v q i p w i p v. So, that is the way you get the source term integrated. So, then
you can and i p and all these will depend how many integration points you have over the
surface, ok.
(Refer Slide Time: 05:53)

Now, you can get the similarly, the volume integration points with 1 integration point.
So, basically the volume integration of source term.

(Refer Slide Time: 06:10)

Similarly, you can have one point you can essentially, one point means here, you could
have 4 integration points. Now, other case you had the points along the surface. Now,
when you try to get the volume integration, you need multiple points inside the element
of the volume. So, you could have 1 point to integrate, you could have 4 point to
integrate ok. So, or you can have more. So, if you have it, then accordingly, your
calculation of these x i and w i p they will be different. So, and you can also, I think you
can have the 9 integration points.

(Refer Slide Time: 06:56)

So, it depends so all these points. So, i p can actually vary accordingly your w i p will
also vary and. So, your integration that you have got here like for the source term that Q
dv, this is your volume integration of that. So, that will have the i p v q ip w i p v. So, this
actually represents that how many points I have inside that volume. And, that is why this
loop goes from number of points inside the volume. So, the difference remains between
flux and the volume is that in the flux calculation. The number of points, they are varied
along the faces, when you come down to the source term integration which is a volume
integration, the number of points varying within the cell ok.
(Refer Slide Time: 08:09)

So, after getting all these done, if you look at that particular element C and you have all
these neighboring elements and they are connected with the faces. So, my transform
equation, transformed equation become obviously, over element C that would become
like f nb of c rho V phi minus gamma del phi dot S f equals to Q dot Q c V c ok. So, that
is my transformed equation.

So, what I started off? I started off a differential equation of steady state equation of the
scalar transport equation. Now, I have written algebraic system. This is nothing, but the
integration over the faces because, these are the fluxes. So, that means, it will go from 1
2 3 4 5 6. So, all these fluxes to be integrated. Now, another thing that one may wish to
do is the linearization ok. So, you can linearize the system and so, how do you linearize
the fluxes? You had this total flux J f dot S f. So, you can actually, you can say, this is the
total flux T f, that is total flux and that can be realized as the flux C f plus flux F f plus
flux V f.

What do they stand for? So, this essentially means, my total flux for face f if I consider, f
1, then my total flux should be this, is essentially my J f phi, that is the that includes both
the component ok. So, that has both convection and diffusion. So, this represents total
flux. What is this? This, this term actually represent the flux linearization coefficient for
element C; that means, this is the element which I am interested in and trying to get the
equation integrated over that. So, when I do the linearization because, this is the total
flux along this face and this face is connected with face F 1.

So, one contribution comes from the linearization coefficient of this particular element
along this face. So, the second is that flux linearization coefficient for f or you can put
element f for element f. So, if you are talking about this flux. So, mind it, that we are
talking about total flux for a particular face.

So, the total flux linearization at f 1 must be different from f 2. But if you look at only
one particular face, again if you look at this equation where you have the total system
which is integrated to the linear system, I mean algebraic system you have a loop over all
the faces around that particular element. For this particular example, you have to have
this loop go over from 1 to 6 because you have 6 faces.

Now, once you go down or come down to each of these faces now if you can look at
each fluxes. So, along one particular face it needs to be linearized between this element
and this element. So, one coefficient come from this another coefficient come from
element F and then you have a component which is called the non linearized part so, you
get 3 component ok.

So, the total flux along a particular face is decomposed in essentially 3 components; one
component come from the element itself, the coefficient of the linearization. Second
component come from the linearization coefficient of the neighboring or the connecting
cell along that face and third coefficient is the non-linear part.
(Refer Slide Time: 14:48)

So, if you look at all this 6 faces, you will have this components ok. Now, if I essentially
write down that equation, which is going for this element dot S f ok. So, this would be
nothing, but my f the total flux ok. So, that means, this equation left hand side of the
equation, I am writing at the total flux. So, this nothing, but the f over that element you
have flux C f phi C plus flux F phi F plus flux V f; that means, if I come down to face 1,
this would be flux C 1, I mean the coefficient of C 1 phi C coefficient of f 1 phi f flux V
1 ok.

So, for all the faces you can write down and then, I have the source term Q c and V c
equals to essentially the total flux. So, if you write down the flux C phi C the flux F phi f.
So, it is a linearized equation ok. Now, if you have let us say, constant source term, what
happen? The constant source term So, I am again writing the source term. In this kind of
fashion the contribution coming from the C element contribution coming from the
neighboring element. Constant source term means flux C would be 0. So, sorry this
should be this is flux V. So, my Q c V c is equal to flux V only ok.

Now, if you substitute everything back in this particular equation ok, if you substitute
everything back there, you get phi c a c plus summation of this integration a F phi F
equals to b c. So, essentially this expression and this expression both of them are, if you
collect them together, you get this and what a c stands here? A c is f n b c flux C f minus
flux C and a F would be flux F f and so, b c is nothing, but summation of f over this
element flux V f plus flux V. So, what you have here? And here, if you put them together,
then you get this. So, this is nothing, but your linear system and if you put them in a
matrix form this is going to get you back the A X equals to B ok.

So, the thing is that over a particular element C you integrate your equation. So, the
integral equation will look like that and once you compute all the fluxes and put them
together, so you get back this system. Now, here you get a linear system like this, you
cannot solve without having the.

(Refer Slide Time: 20:06)

So, once you need to solve this linear system, the important ingredients are to have the
boundary condition ok. One is the boundary condition which will fit to the system and
then, the linear solver. Here, you can have direct solver, you can have iterative solver ok.

So, we have already discussed in a, I mean very quickly what kind of solvers. But, in
details, we will do later on in a due course of time. But now, let us look at the boundary
condition. How do you implement that? So, first, one important condition is that your
have a the value is specified; that means, in other words, you say it is a dirichlet
boundary condition ok. So, if you have a cell like this, what you had like and then you
have a point here. Now, this is your c center and this is connecting and this would be the
normal ok. And this is where your phi b this is the boundary surface this is t, this is
specified ok. This is the point b and this is e b ok.
So, if phi b is specified, then I should have phi b equals to phi b specified ok. Then, at
that particular face, this is the boundary face, this is the boundary face my J b phi dot S
b. This is the, if you look at this integration, so, at that particular face, this will become J
b phi dot S b. So, which would be if you just say the convective part rho V phi b dot. So,
this is flux C b phi c plus flux V b. So, this will get you back rho b V b dot S b phi b
which is nothing, but m dot f phi b specified. So, you can have flux C b is 0 and flux V b
is m dot f phi b specified. So, you get that value.

(Refer Slide Time: 24:11)

So, once you have that value, then now, the second type of boundary condition would be
specified flux ok; that means, called the Neumann boundary condition. So, if you look at
the same element here, this is the center. Then you come this is the normal this is e b, this
is point b, this is t, this is where q b specified ok. So, once you do that, then your J b phi
dot S b would be J b phi dot n b S b, ok.

So, this is your specified flux. So, which is nothing, but q b specified S b ok. So, my flux
C b is 0 and flux V b equals to q b specified S b. So, these are the 2 type of conditions
that you can specify at the face. So, now, we will continue from here in the next lecture.

Thank you.