Indian Institute of Technology Kanpur

National Programme on Technology Enhanced Learning (NPTEL)

Course Title
Basics of Finite Element Analysis

Lecture – 15
Weak Formulation

by
Prof. Nachiketa Tiwari
Dept. of Mechanical Engineering
IIT Kanpur

Hello, welcome to basics of finite element analysis, today is the third lecture for the current
week, and yesterday we had discussed weak formulation and we will continue that discussion
today as well.

(Refer Slide Time: 00:31)

So in the last class we had developed a weak formulation for a particular differential equation,
and that weak formulation was 0 equals 0 to L dw over dx a(x) cj, so this is the weak formulation
okay. Now I wanted to make a couple of very important points in context of this weak
formulation. First is that as we have several times earlier discussed, that the requirement because
of this weak formulation for differentiability on the, on the dependent variable u, it gets reduced,
if it is a fourth order differential equation we do it two times I need only a function which is
differentiable two times. If it is second order then I need only at, a linear function which will
meet the requirements. So that is one thing, the second thing is that,

(Refer Slide Time: 02:16)

This and this is also we had discussed in the last class was that, this boundary term which comes
out through the process of integration by parts it actually helps us, it actually helps us satisfy the
natural boundary conditions identically. So in their strong form we would have to choose these
shape functions in such a way that they meet natural as well as essential boundary conditions. If I
use the weak far, then I have more flexibility in choosing the shape function and the only thing I
have to worry about is that the shapes and functions should satisfy the essential boundary
condition.
(Refer Slide Time: 03:00)

The third thing is, it also helps us classify what kind of boundary conditions are involved. So we
will discuss this in detail, so let us look at this, so you have in the boundary condition there is
this thing which is the shape function which is in green, and then there is something in red okay.
So there are two parts, so the boundary condition it involves weight function and that is in green,
and then it involves dependent variable in red.
(Refer Slide Time: 04:08)

Actually at this stage I will replace this by u, du over dx okay which is mathematically the same
okay. So it is dependent on yeah it is, it is a function of a dependent variable, dependent variable
function. It is not dependent variable itself, but it is a function of dependent variable. So this is
like the co-efficient of weight function, there is w and we can treat a(x) du over dx as the co-
efficient of weight function.

Co-efficient of weight function, now in second order equation there is weight function, in a
fourth order equation we will have, we will have to integrate two times so first we will get a
weight function and second time when we integrate we will get a derivative of the weight
function okay. So if there are,
(Refer Slide Time: 05:26)

Weight functions in the boundary terms or there are derivatives of the weight function then they
are coefficients, co-efficient of weight function and its derivatives okay, they are known as
secondary variables (SV) secondary variables. And if I, so in this case it is a(x) du over dx okay
that is my secondary variable. And if I specify this, then I say that I have specified the natural
boundary condition okay. So this is the logical way to say which one is natural boundary
condition and which one is the essential boundary condition.
(Refer Slide Time: 06:33)

So this is one, the second thing is if we express the dependent variable in same form as weight
function and its derivatives then it is called as primary variable. And if I specify it I say that I
have specified the essential boundary condition okay. So to make things clear in this boundary
term the weight function appears as w, in the boundary term weight function appears as w, if I
express the dependent variable in this equation the dependent variable is u. We had discussed
mentioned this right in the beginning. So if I,
(Refer Slide Time: 07:52)

Replace w by u in this boundary term, then u is my primary variable, and if I specify it I say that
I have specified the essential boundary condition. In case of a beam, so this is the equation for a
bar right, this is, this equation represents a bar or rope intension or electro static problem and all
that, but suppose there was a beam.
(Refer Slide Time: 08:21)

It will be a fourth order differential equation. And there when we integrate it two times.
(Refer Slide Time: 08:25)

We will get these things, there will be a weight function and there will be its co-efficient. This is
what we will get in the first integration, and when we integrate it second time we will get this
and, so this will be the first boundary term and this will be the boundary term after second
integration okay. So in this case and this happens because the governing equation for a beam
Euler Bernoulli beam is a fourth order differential equation right.

So in this case my second primary variables will be, there is w so corresponding to that there
will be u, and there is wʹ so corresponding to that will be uʹ, where u is the deflection of the beam
okay, u is the deflection of the beam.
(Refer Slide Time: 09:45)

And the secondary variables will be ei, so excuse me this should be, this should have been not w
is should have been u, because u is the unknown we are trying to solve.
(Refer Slide Time: 10:05)

So it is (EI uʺ)ʹ and this is what it is the sheer force, it is the sheer force in a beam this is the
formula for the sheer force right. And EI uʺ this is my sheer force and this is EI uʺ is moment
bending moment okay. So my essential boundary conditions are u and uʹ, my natural boundary
conditions are sheer force and bending moment. So if I have a beam, I have to specify two
conditions at both the boundaries, at either I specify so total number of boundary conditions I
have to specify are four and two on each end.

So either I specify u and boundary can a bending moment or uʹ also have combination of these
two. So at a pin end my displacement is zero right. And corresponding to the second thing my
bending moment is zero, and at the fixed end at the rigidly fixed end can till leave you know. My
uʹ which is deflection is zero and the slope is zero okay. So all these boundary condition terms
they naturally come out when we do the weak formulation, and you can classify them and
separate them, and you then by looking at the physics of the problem you can figure out which
conditions are known, and which are unknown okay. So you can have situations where all the
essential boundary conditions are known, for instance you can have a beam like this, in this case
uʹ and u are known at both the ends, so all EBCs are known or you can have another situation so
it is a beam but I am bending it by applying a moment at both the ends.
(Refer Slide Time: 12:47)

So in this case, bending moment is known at both ends and because I am not applying any
vertical sheer force, so sheer force is also known. So in this case, both at both ends I know the
natural boundary conditions, and in the third case which may be something like this, I may have
a mixture of some essential boundary conditions are known and some natural boundary
conditions are known okay.

So this is the way we can specify and classify different boundary conditions, and the next point
in context of weak formulation we are going to discuss is, that how do we make the choice of
weight functions, because still so far what we have been saying is that we, if there are n
unknowns corresponding to n cj’s right c1, c2, c3 then we will choose n different weight functions
and because of that we will get n different equations and we will solve those to find the values of
c’s. So now we will discuss that atleast in context of weak formulation how do we go around
choosing appropriate weight functions.
(Refer Slide Time: 14:00)

So choice of weight functions okay, so in general we choose w such that it vanishes at the
boundary if the primary variable is known at the boundary okay.
(Refer Slide Time: 14:26)

What does that mean, suppose I have a bar and I am pulling it and I am also applying some
traction on this, x at x=0, u=0 right. Then when I am making a choice for weight function, I have
to ensure that w is zero at x=0 okay. What could be an example of such a weight function, one
example could be x, another example could be x2, another example could be sinΠx and so on and
so forth okay.

So this is one, another case so, so here I am applying a force in this case, this is same case but
here instead of, instead of applying a force I am displacing it let us say by five millimeters. So u
at L is equal to 0.005, in this case what is our primary variable, primary variable is u right, we
are discussed what is the primary variable and secondary variable. So primary variable is u and u
is known at, at x=0 and x=L.

So if u is known at both x=0 and x=L, then I have to choose w which vanishes at x is equal to at
both ends okay. It has to be zero at both ends, w should be such that it is zero at both ends. It
could be any function but it has to meet because not because u is known even though u is non
zero at x=L, my choice of w has to be such that it is should be zero. In the first case u was not
known, I was applying a force but I do not know how much the bar is going to stretch.
Here I am actually pulling it by five millimeters so x is known this u is known at x=L, and also u
is known at x=0, so, so in this case we will look at few boundary conditions so it could be one
sinΠx over L, another could be sin2Πx over L, sin3Πx over L, this could be one set of functions
right, another function could be x times x-L, another function could be x times (x-L)2 and so on
and so forth.

So this is how we choose the functions, now these functions will not work so let us say this is
case A and this is case B. So these functions will not work for case A why, because if we choose
these functions then we have overly excessively restricted our choice of functions okay. Here the
only restriction is at x=0, u0. So I cannot have the weight function behave in such a way that the
weight function becomes zero at the other end of the rod okay.

So the next thing is that why do we make such choices, why do we make such choices okay? So
the first reason is, that we have to choose these weight functions in such a way that they
represent variations in the primary variable. They represent variations in the primary variations,
so.

(Refer Slide Time: 19:37)

Choose so in a general sense choose w such that w is a variation of u okay. So variation of u and
if you remember our discussion of variations the variation of u suppose you have a string and it
is, and it is suppose u is zero at this end and u is zero at this end, then the variation of u could be
any function as long as, as long no not small because the smallness is dictated by that amplitude
alpha if you remember.

As long as it does not violate the end conditions right, wherever the variable is known at that
place variation has to be zero. At all other places it can be non zero, so if I choose these functions
for case A I will not truly represent the variation in u because I will be specifying that variation
at the other end of the rod is also zero. So once again I will like to restate that we choose w in
such a way that they represent variations in the dependent variable.

Wherever the dependent variable is known at those places the value of w has to be zero,
wherever it is not known the w need not be zero, w need not be zero.

(Refer Slide Time: 21:29)

These variational so these, so w represents variations or δu and in mechanics it is also known as
virtual displacement, and maybe in the next lecture we will see why we are doing all this. Why
we are doing all this? So if I do this then my governing equation.

(Refer Slide Time: 22:10)

So my original governing equation was this weak form, this is the original weak form of the, for
the system right, and my boundary condition was u at zero is equal to some constant u0 and a du
over dx at x is equal to L is zero okay. So what I can do is I get the same term here, [-w auʹ]
where x = L, and +[w auʹ] x=0 okay, so what I have done is I have just expanded this term in the
parentheses.

Now I know that du dx at x equals L is zero I, I am now going to enforce this boundary
condition, so this term becomes zero okay, so this, this whole term goes away right or I can yeah,
so anyway so this is what I am left with this is the thing. Now if I replace w by variation then I
get, this is the equation I get where w is represented by a variation in w, this form where I am
replacing w by variations is known as variational form, it is variational form.
And a lot of times this weak form and a lot of times this weak form and variational forms are lot
of times they are interchangeably used, they are used interchangeably, other point that if.

(Refer Slide Time: 25:27)

The differential equation is linear and of even order then we will see that weak form is
symmetric and bilinear in u and w, so if you look at our original equation.
(Refer Slide Time: 26:09)
(Refer Slide Time: 26:09)

This equation is linear in u, we do not have a u2 term right, and what is the order of this
differential equation, this is a second order differential equation right. Iuʹ and then it is going to
be a differentiated one more time, so it is a second order differential equation.
(Refer Slide Time: 26:26)

So it is of even order and also it is a linear equation, and for such equations when we construct
the weak form, there is another form weak form or this is another, this is weak form then you see
that you have this term wʹ a(x) uʹ, or you have this thing. So this term is we have discussed this
earlier, it is symmetric; if I replace w by u and u by w I get the same thing right. So this weak
form is symmetric and bilinear okay, it is symmetric and bilinear.

The third thing is, so we have said that differential equation is linear and of even order, weak
form is symmetric and bilinear. And the next point to make is that because w represents variation
in u, it represents variation in u its nature should be similar as that of u, it should belong to the
same space as u, it should belong to the same space as u okay. And what that means is that when
we are choosing different options for picking up w’s, we do not have to see when we were, we
had made, we have made some choices for picking a fees, hej’s right, which represent u’s, we
can use same functions to represent w’s also.

So that makes our life easier, makes our life easier. So this concludes our discussion on this
particular topic atleast in context of weak formulation, now in the next class what we will discuss
is three other concepts, linear, bilinear and quadratic functionals, and then we will show that
when we are multiplying the error with the, with the weight function and we are integrating it
over the domain, and then when we are creating a weak form of this, we are actually minimizing
some overall function.

And in context of solid mechanics essentially what we are doing is that we are minimizing the
total potential energy of the system, which and whenever in a solid mechanics we minimize
some total potential energy that is the state, that is the condition for its equilibrium. So that is
what we will discuss in the next class. Thank you.

Acknowledgement
Ministry of Human Resource & Development

Prof. Satyaki Roy

Co-ordinator, NPTEL IIT Kanpur

NPTEL Team
Sanjay Pal
Ashish Singh
Badal Pradhan
Tapobrata Das
Ram Chandra
Dilip Tripathi
Manoj Shrivastava
Padam Shukla
Sanjay Mishra
Shubham Rawat
Shikha Gupta
K. K. Mishra
Aradhana Singh
Sweta
Ashutosh Gairola
Dilip Katiyar
Sharwan
Hari Ram
Bhadra Rao
Puneet Kumar Bajpai
Lalty Dutta
Ajay Kanaujia
Shivendra Kumar Tiwari

an IIT Kanpur Production

©copyright reserved