Advanced Finite Element Analysis

Prof. R. KrishnaKumar
Department of Mechanical Engineering
Indian Institute of Technology, Madras

Lecture - 18

Phi, a variable, the scalar field was given and that the observer was supposed to move
with the velocity v.

(Refer Slide Time: 1:01)

He said it is equivalent to calculating the material derivative. It is a relative term, whether

observer moves or the material particle moves, it is all the same.

(Refer Slide Time: 1:17)

What you do is to calculate dow phi by dow t, then calculate dow phi by dow x or in
other words, grad of phi. Then, calculate phi dot by, first term being dow phi by dow t
was grad of phi dot v.

(Refer Slide Time: 1:36)

The answer happens to be x into 1 plus, so, that is the material time derivative of phi.
Now, let us move to the next problem. Next problem is, same way we will do it.

(Refer Slide Time: 1:59)

Prove that e dot is equal to d minus l transpose e minus e l. I shall give you couple of
minutes, let us see how you work out this problem.

(Refer Slide Time: 2:22)

Remember, what was, what was our e? e is the Eulerian; note that it is the Eulerian strain
rate or strain rate which is, sorry, strain which is for Almansi strain. That is in other
words, it refers to the current coordinate system and it was written as half into I minus F

inverse transpose F inverse. I am interested to determine what e dot is. Let us see,
continue with that. Obviously, the first term will not be there. So, it will be half of F
transpose T F; whole is to be differentiated with respect to time. How do you write that?
What is the relationship between l and F and F dot? l is equal to, F dot F inverse keep that
in mind; so, substitute that and see what you get. What will happen to the first term? Do
that and you will ultimately get l transpose into I minus 2 e plus I minus 2 e l. Note that d
is half of l transpose that is l.

(Refer Slide Time: 4:59)

From this, you can write this down as d minus l transpose e minus e l. Fill up the
remaining steps, we will move to the next problem; just substitution and then just
working it backwards. Both the problems are very simple problems. Only thing is that we
have to manipulate a bit. Let us go over to the next problem. Consider a deformation
gradient that is F, given by what I am going to write now.

(Refer Slide Time: 5:55)

Consider the deformation gradient, F given by C minus as ac minus s s plus ac as plus c.

Of course, C is equal to cos theta and s is equal to sin theta. Take for this problem a is
equal to half and theta is equal to pi by, pi by two. This is the third problem. Find the
stretch tensor; find the stretch tensor or in other words, sometimes U is called as stretch
tensor, U and R; U, which represents the pure stretch which is called the stretch tensor.
Find U and R. In other words, substituting these things, you can first write down what is
F.

(Refer Slide Time: 7:18)

What is F? Cos pi by 2 minus, so, minus half, first term; the second term is a cos theta
again that goes to zero, so, minus 1 and then 1, again C goes to zero and half. So, the first
step is to find out C. Look at the steps, what I am going to do. C is F transpose F.
Determine F transpose F. So, the steps we are going to follow is that we will find out C,
solve the Eigen value problem, determine the Eigen values of this and as well as Eigen
vectors, write it. Write the Eigen values of the U. What is the Eigen value of U? What is
the relationship between that and Eigen value of C transpose C? Root of the values and
then determine the Eigen vectors, transform U to the original coordinates and then
determine R, because you know F and then, you know U.

So, now determine F transpose F. If you have forgotten something, please refer to your
notes and tell me what is F transpose F? 5 by 4 1 1 and 5 by 4. The next step is to find out
lambdas for C. Please determine lambda, write down. It is very simple, straight forward,
because it is just 2 by 2. Write down the characteristic equation and then find out what
lambda is. For the lambdas determine the Eigen vectors as well; determine the Eigen
vectors as well. Very good; so, lambda1 is equal to 0.25 and lambda2 is equal to 2.25.
Find out the corresponding Eigen vectors for lambda1 and lambda2. This lambda1

lambda2 are the Eigen values for this C. We will take the square root for U, but
meanwhile, let me determine the Eigen vectors for these two quantities.

(Refer Slide Time: 11:08)

That is in other words, v of 1 and v of 2 corresponding to the Eigen values lambda1 and
lambda2. 1 by, you have to normalize it, 1 by root 2, correct, 1 minus 1 1 by root 2 1 1;
very good, it is excellent. Look at these steps. You know this is how you calculate
stretches and as well as, because these steps or these kind of calculations may become
very important when you do finite element analysis, when you implement finite element
analysis later. In other words, how do I write the C matrix. So, C matrix is in terms of,
this is spectral decomposition; in terms of these lambdas are written as 0.25 or 1 by 4. If
you want, you can write it as 1 by 4 and someone said 9 by 4 and the corresponding, what
do I mean by basis in this case?

In spectral decomposition remember that we had written for spectral decomposition,

lambda 1 v 1 v 1 plus lambda 2 v 2 v 2. So, the basis here are the Eigen vectors, correct.
So, those are the Eigen, Eigen I mean, these are the basis. Essentially, what is that we
have done? You know, just recapitulate what we know. Essentially what we have done is
to get a basis, new basis, in such a fashion that you get diagonal C matrix. This is exactly

what you do in, say for example, when you do vibration problems; you find out the Eigen
values and the Eigen vectors and then change over the basis. Your original basis is now
replaced by the Eigen vectors. What exactly you do? Decouple all those equations, the
governing differential equations, you decouple them.

Basically, Eigen vectors are used to change the basis, so that it becomes the matrix,
ensuing matrix; in other words, ensuing equation in the other case, problems, you
know, vibration problems rather, would be much easier to solve. Now, how do I
determine U in this basis? Simple; square root of these base here; 2 0 0 3 by 2, but my job
is not yet over. What is that I should do? What is that I should do? Now, this is in, this -
what we have got here - is in this basis. Now, I have to get it back to my original basis.
How do I do that?

Look at this and see, you have to do a transformation. Please look at that. Let us see how
much you remember from what we did earlier and then write it down. In other words, the
basis actually can be written as 1 by root 2 1 minus 1 1 and 1. This, multiplied by U,
multiplied by transpose of this, would give me the U value. That is the transformation
that I am going to use, U or the value of U in my old coordinate system. Let us see how
we do that.

(Refer Slide Time: 15:44)

In other words, 1 by root 2 1 minus 1 1 1, half cipher cipher 3 by 2 into, I will write it
here; 1 by root 2 1 1 minus 1 1 would be the transformation matrix. What is the result of
that? 1 by 2 1 by 2, so, you can take it as half. So, the answer is, if I am correct, 2 by 3 2
minus 1 minus 1 2. Just check this up, see whether you get this. Please check that.
Essentially, the steps are very clear, just find out the Eigen value, go to the spectral
decomposition mode, standard step; determine the Eigen values. This is a very practical
way of decomposing F into the stretch term, U and the rotation term, R. Hope this is
correct. Please check this up. You know, I just did that fast, I hope it is okay.

Now, how do you determine R? Yes, that is all. So, invert it, F U inverse, calculate F U
inverse and calculate R; please invert that and find out what R is.

(Refer Slide Time: 17:55)

R is equal to F U inverse. F you already know; invert U to get U inverse and then, please
attempt the problems, all the problems; as we go along please attempt it.

(Refer Slide Time: 18:44)

I think, I think this is U inverse, actually. Sorry, this is U inverse; U inverse is this and U
is actually half of, yeah, U is half of 2 1 1 2, because 1 by 1 by root 2 1 by root 2 is half

10

of it, then we multiply this 2 1 1 2. This is U and actually this is U inverse. We have
already inverted it, so, that is U inverse into 2 by 3.

(Refer Slide Time: 19:17)

Obviously, we will not get 2 by 3 here, because 1 by root 2 1 by root 2 is here. So, we
will get only half. So, U inverse is equal to 2 by 3 2 minus 1 minus 1 2; so, 2 minus 1
minus 1 2. So, R is equal to 0 minus 1 1 0.

(Refer Slide Time: 19:57)

11

Note that, note that this is U inverse, not U. This U is here, just you have to multiply
those terms there; 1 by root 2, 1 by root 2 goes out to half. So, other multiplication gives
you this. Inverse of that is given by ., take the determinant, then cofactors; write down
the cofactors, it is just 2 by 2 matrix. So, R gives this. You initially note that R R
transpose is equal to I. Any question on this problem?

Now, take down the next problem. The uniaxial, we are talking about, now we are going
to talk about uniaxial stress and the associated strain. How this condition can be
expressed in terms of the first Piola-Kirchhoff stress, the second Piola-Kirchhoff stress
and of course, the Cauchy stress and what are the strains? Different strains are capital E,
small e and so on.

(Refer Slide Time: 21:15)

Consider a bar with length l, b and a. Let us say that its original dimensions are l0, b0 and
a0 and its final dimensions are say, given by a b and l. Let us say that a uniaxial stress
state is what is applied to this bar. Determine the nominal stress, second P-K, first for this
bar.

12

(Refer Slide Time: 22:20)

Take the directions to be say, this is the X direction or capital X or capital X1. If you
want, write it as capital X1 and then that to be capital X2 or in other words, X2 is in the or
you can X2 and X3; to be careful that is X3. Right handed coordinates should be
maintained and so, X1 and X2 and X3. Our first job here, what is that we should do? Of
course, sigma you know. So, the Cauchy stress is nothing but say, sigma11 or sigma X 0
0, 0 0 0, 0 0 0; that is fine. But, in order that we convert this into the first Piola-Kirchhoff
stress and second Piola-Kirchhoff stress, what is that we need, the first thing? F, very
good; so, we need F.

13

(Refer Slide Time: 23:37)

Now, I will give you two minutes. Let us see how many of you come up with what is F
for this? No, it is not l minus l, not. Please, first, your first step should be to write down
say, what is small x1 co-ordinate, small x1 in terms of capital X1 and then small y1 and
small z1. How do you write this term? Choose, I mean, I have just given x1 x2 x3. Choose
an origin and try to write down x1 y1 and z1. Yes, so, first x1, write it, write down. So, l by
l0, same what you said, l by l0 into say, X1; not x1, y1; y1, say, x2. x2 is a; so, a by a0 into
capital X2 and X3 is equal to b by b0 into X3, very simple. When x3 is equal to b, capital
X3 is equal to b0. In other words, this is some sort of a, what is that we have followed
here; linear interpolation, one end to the other end; that is all.

Now, write down F. Now you have crossed a major thing. Now write down F. That is
straight forward dow x1 by dow capital X1 that is l by l0, no other terms there, so, all these
terms becomes 0; 0 a by a0 0, 0 0 b by b0. Now, what is the determinant or J determinant
of F or J, what is it? Yes, so, it is very simple. You know dv by d capital V; dv d small v
is abc, you may not even go through this, so, say a b l, rather a b l by a0 b0 l0; that is the
determinant of J. Now, we have to calculate F inverse. It is a diagonal matrix.

14

(Refer Slide Time: 27:19)

So, F inverse l0 by l 0 0, 0 0 0 0; very good. Knowing sigma, knowing J, what is that I

want to calculate? P, ., nominal, what is the formula? J, no, no, that is for from P. So, J
F inverse sigma; yeah, J sigma F inverse transpose, calculate that. Actually, we had sigma
is equal to J P F transpose, J inverse P F transpose. So, this will become J sigma is equal
to P F transpose. So, J sigma F inverse transpose equal to P.

Pardon? What is sigma? Sigma is sigma. What is that you get for P? a b, let me write it
here.

15

(Refer Slide Time: 29:36)

So, P is equal to ab by a0 b0 sigma11, rest of it are zero. Look at this. What this indicates?
ab into sigma11, ab is what? The current area, so Cauchy stresses for the current area. So,
that area multiplied by sigma11, what you get? The force that is acting, that divided by the
original area. That is why this is called as a nominal stress. Obviously, I mean you could
have guessed it, but nevertheless it is very important that you go through the whole steps,
because that will give you the whole thing. This is the force. The force divided by the
original area would give you P.

Now, calculate S, the second Piola-Kirchhoff stress. Second Piola-Kirchhoff stress, we

already saw that it is the pull back that we have to apply on the Cauchy stress. Rework
that here.

16

(Refer Slide Time: 31:18)

What is the relationship between sigma and S? F inverse sigma F transpose is equal to S.
Sigma transpose is equal to sigma, anyway it is the same. So, let us take the S11 term, just
work it out and tell me what the first term is, the S11 term. Yeah, a b sigma11 by a0 b0 into
l0 by l. Everyone got this answer, into l0 by l? Yes, I think that is right. Look at that. Here
for P, you are able to give a physical meaning for P. But for second Piola-Kirchhoff
stress, we are not in a position to give any physical meaning for it. It is just a pullback
operation and that is how you should view it. You should not, you know look at Piola-
Kirchhoff stress as you know, something which we can physically relate to.

Yeah, this poses problems. This poses problems, especially when you give results and
when you go to plasticity, when you deal with second Piola-Kirchhoff stress, since it does
not have a physical meaning and many of the quantities that we deal with ultimately,
especially when with respect to yielding or failure and so on, they have a physical
meaning. So, when you deal with this Piola-Kirchhoff stress, second Piola-Kirchhoff
stress, you have a problem. Agreed that? So, we will overcome that by different things
and in fact, you would notice that none of the softwares, none of the softwares give
results in terms of second P-K. Though internally many of them work with second P-K,
they will not give you results on second P-K, basically because it does not have the

17

physical meaning. But, that should not worry you when we apply these things, because
you should view it as a mathematical operation of pullback. So, that is the reason, why
you will not, apart from theory, you will not see these thing in results. The results are still
given in terms of the Cauchy stress. All the results that are still given, in most of the
softwares, are in terms of the Cauchy stress.

Yeah, first Piola-Kirchhoff stress, so, you can develop a relationship between P and S. I
will leave it to you, it is very straight forward to develop; please develop a relationship
between P and S. But, keep this in mind and we will encounter second Piola-Kirchhoff
stress, quite often in the implementation levels, but you as a, if you just use it, you will
not encounter second P-K or second Piola-Kirchhoff stress. But, there is today a
tendency, especially in biomechanics, to actually plot graphs in terms of second P-K
itself. There are, I mean, I am not sure whether how much that is the correct way to
approach things, because this does not have a physical meaning, but still what people do
is to calculate sigma, calculate F and plot the graph; second P-K versus e. There has been
recent literatures, which show experimentalist plotting second Piola-Kirchhoff stress.
But, personally I feel that that does not take you anywhere, because you do not get a
physical picture often.

Let us look at the next problem. I will just give the problem to you. This will be our
fourth problem. The deformation of a body is described.

18

(Refer Slide Time: 35:57)

The deformation of the body is described by x1 is equal to minus 6X2, x2 is equal to half
of X1 and x3 is equal to 1 by 3 capital X3. If sigma is given by 0 0 0, 0 50 0, 0 0 0, sigma
is given by this, determine Cauchy stress vector t, Cauchy stress vector t and the first
Piola-Kirchhoff stress vector. Note the difference between stress tensor and stress vector.
We have emphasized that in our first course, we know that stress can be looked at as a
vector and as a tensor.

(Refer Slide Time: 37:06)

19

So, sigma n or sigma transpose n is equal to a vector, that is a tensor and calculate the
first P-K stress vector taking n to be e2, taking n to be e2, the unit vector along the second
direction and then calculate what the stress vector is for both Cauchy stress and the Piola-
Kirchhoff stress. Our first job, of course, is to calculate from this F. Of course, what you
get here, p, what you get here? Same P N; note the difference between small n and capital
N. Small n is the normal at the point of interest along a surface ds in the current
coordinates and capital N is the same normal in the reference configuration, in the
reference coordinates, coordinate; so, note the difference between the two.

Your first job now is to find out F dow x1 by dow capital X1 is 0 minus 6 0, half 0 0, 0 0 1
by 3. So, I need F inverse transpose.

(Refer Slide Time: 39:48)

Calculate F inverse, calculate F inverse; 0 2 0, minus 1 by 6 0 0, 0 0 3; that is F inverse.

Yeah, F inverse transpose, this is F inverse transpose. Yeah, F inverse that is F inverse.
So, you have to calculate F inverse transpose and then calculate P. Of course, determinant
J, I think, happens to be 1. Calculate the determinant of F that is determinant of F. This is
equal to J is equal to 1. So, calculate P now. P happens to be 0 100 0, 0 0 0, 0 0 0. Now, I
have to calculate capital N, the normal. How do I calculate? Nansons formula.

20

(Refer Slide Time: 41:32)

Just calculate what capital N is. So, NdS is equal to J inverse F transpose n ds. What
happens to this? Half e1, yeah, ds. Check that up; is that right? Yes, very good. So, half e1
ds. What is N? N is actually the, yeah, the so, what is N? It is the unit vector along e1,
which is equal to e1 itself. So obviously, so, ds is half of ds; that is it. Because, this is a
unit vector, e1 is a unit vector, so, I can rearrange this as e1 half ds. That is the magnitude.
This is a unit vector, this is a unit vector; this is the magnitude. So, ds is equal to half ds.
That is straight forward. So, look at that N, which was e1 has become now small n which
is e2. Yes, so, find out the traction or the stress vector sigma n.

Very good; so, that happens to be 0 50 0. Now, what happens to p? Yeah, 0 100 0; very
good, 0 100 0. Obviously, we need not even calculate that, because I see that, I told you
already, I gave the clue that ds will be half ds; obviously that is 0 100 0. So, that
completes few problems that we were able to do. You now know, how application is
done, it is very straight forward. Then, any question? With that, we will close the topic on
stress. We will move over to few more fundamentals that we have to learn before we get
back to the non-linear finite element analysis.

21

Yes, the softwares have these things as routines - calculation of U, calculation of R and
so on. We will see more about that when we come to the formulations of elements in
terms of these quantities later. Yes, of course, there are shape functions that are defined.
We will see about them during the formulation. We will close now, see you.

22