Hello again, welcome to basics of finite element

analysis part 1, this is the first week and

today is the second lecture of this course.
In the last lecture we had covered a very
broad-based overview of the course and today
we will further elaborate on the concept of
finite element analysis by a very simple like,
you know numerical experiment or an example.
So what we are going to do today is, calculate
the area under a curve.
And somehow apply what we learn through this
example to finite element analysis method.
So suppose I have a function, so this is my
x-axis that is my y-axis and this is my y
equals some function of x, ok. And in this
case f(x) is equal to constant some C. So
the value of this f(x) is C as X is varying.
And what I am interested in is to compute
this area under the function, okay. So because
f(x) is equal to C we know that if I integrate
f(x) with respect to DX.
And if I evaluate the values of that integral
between the limits x1 and x2 I will be able
to compute this area. So that is what we will
do we will do so what it comes down to is
area. So this area is equal to integral from
x1 to x2 f(x) dx and that because f(x) is
C, so it is c times dx, x1 to x2 and this
is equal to C times x2-x1, okay. So this x2-
x1 is the width of this rectangle.
And C is the height of this rectangle and
this is a very trivial example but this is
how I wanted to begin with. So what I have
explained is that if you have a simple function
you can integrate it in a closed form way
and you can calculate this area, okay. And
your calculation will be accurate and the
error will be 0. So this is an approximate
this is a exact solution when we are trying
to calculate the area.
Now let us look at another example. Oh by
the way before I go to the second example,
a couple of things. So here the region of
our interest in which under which we wanted
to calculate area was what x1 to x2.
So this x1 to x2 is the domain. In this case
domain is one dimensional in nature because
it involves only variation of X. There could
be situations where domain could be two dimensional
in nature. Suppose I want to calculate the
volume, which is covered by let us say my
hand and this baseline surface. So there the
domain could be that is a two-dimensional
surface. And so, so there the Domain is a
2D thing, here domain is 1D.
Similarly I could have a three dimensional
domain, so that is there. So now moving on
I am going back, I have another function f(x).
So this is my y-axis, x-axis, but it is very
complicated, okay. It is very complicated
and I have to find the area, between x1 and
let us say X2. So this is the area I am interested
in. Now if this function was easy to integrate
I would have an expression.
So this is my function f(x), if this function
was easy to integrate, then I would just straight
away integrate it and I will do a integration
under limits x1 to x2 and I will get a exact
solution. But here first thing is I do not
know what is the, expression the for this
function and the second thing is even if I
knew it was it may have been away it may be
a very complicated function. So I may not
be able to integrate it.
So then I use a numerical approach, okay.
So what I do is, I break this area into small,
I break it into small slices. And this is
exactly what we do when we are trying to integrate
differential equations also. It is a more
complicated process but the philosophy is
same. So I break this overall area which we
are trying to calculate into small slices,
okay. And let us say this slice is S1, this
slice is s2, this slice is s3 and so on and
so forth, okay.
So my total area, here also our domain is
X1 to X2, it is a 1 dimensional domain. Total
area will be area of (s1) plus area of (s2)
plus area of (sn). So suppose I break it up
into n slices, then the total area will be
the area of each individual slices. So now
if I have a way to figure out how, what is
the area of each slice, then I can calculate
this. So what is the area of a slice?
Now I can calculate the area of a slice in
several ways. So approach one, approach one
here what I do is.
I assume that the height of this, so let us
put some a, b, c, d. So what we are trying
to figure out is area of ABCD. And our first
approach is that the height of this ABCD is
constant, okay.
So in that case height of ABCD is constant.
Suppose I assume, if that is the case and
it is, it is ab. So let us say this value
of ab is y1. Then area of ABCD is equal to
what? Y1 times X. So let us say this distance
is ? X then Y 1 times ?X, okay. If the height
of this ABCD is constant height means, the
distance in the y-direction, okay. Distance
in the y-direction is constant then it will
be y1 times ?X.
Is this clear? So basically what I am saying
is this, that this complicated shape ABCD
is not that but I can approximate it as something
like this, right? So effectively what I have
calculated is not ABCD but the area a, b,
e, d, you understand, okay. And in this case
what is the error? The error will be b, e
and c area of b, e and c that will be the
error, okay.
This is my first approach, then I will say
okay maybe I can have a better approach, first
two. So second approach is, that I do not
treat ABCD as a rectangle but I treat it as
a trapezoid. So if I treat it as a trapezoid,
then I draw a straight line between B and
C, right? And then my in this case suppose
the height at point c is y2, then area of
ABCD is y1 plus y2/ 2 times ?x, okay.
So this is the second approach, I can have
a third approach. I will say the third approach
could be that I will say, okay. So first approach
is that I treat this complicated shape as
just a simple rectangle area, error is large
and it equals the area of b, e and c. Second
approach is I treat it as a trapezoid, in
that case the area is the error is this region
between the true two green lines, you know.
I do not want to shade it right now because
I will draw another curve there. And I can
have a third approach and I will say, okay.
In the third approach this transition of height
between B and C, see right now, in approach
to the transition of height between B and
C is linear. I will say okay it is not linear
but it is some quadratic in nature.
So my line will be like this, okay. And I
say that this blue line is a quadratic curve,
okay.
So now if I know that this is a quadratic
curve, then I know the equation for the quadratic
curve, right? And I can again calculate. So
basically if at the moment I say it is a quadratic
curve it is not a straight. See in first case
it was a horizontal line, in second case it
was a straight inclined line, in third case
it is a quadratic curve. So in approach three,
area of ABCD is area under quadratic curve.
So using principles of geometry I can develop
the equation for this quadratic curve easily,
right? And then I can interact, integrate
that quadratic equation between limits x1
and x1+?x and I can get that area.
So what happened to my error? It went down
further. Then I will say that okay.
Approach four could be that it is not a quadratic
curve; it is a cubic club curve. So I can
make this transition between B and C more
and more accurate or closer to reality or
closer to reality. If I keep on increasing
the number of polynomials which capture the
transition from B to C more and more accurately,
okay.
So, so this is what happens in finite element
analysis also. So we can take either of those
estimates, the first estimate is approach
one where the height is constant, second approach
is where height is linearly increasing, third
is it is quadratically increasing, fourth
could be it is cubically increasing. And in
each step of sophistication, I am getting
more and more accurate what, solution, more
and more accurate solution for the area under
area for this rectangle ABCD.
And similarly, I can get I can calculate areas
for the second slice. Again I can have different
approaches. Approach ABCD right?
Then I can have for the third slice. So whatever
approach I think is most accurate, I can pick
up that approach and I can then compute area
of each of these slices, add up all the areas
of these slices and get the final area under
the curve, okay.
So in this context I would like to make a
couple of points. First, as ?X goes to 0,
error goes to 0 also. It does not matter whether
it is a constant curve or it is a incline
curve or it is a, you know this is one thing.
So, so this is one which means that if the
size of ? X is dx, which is infinitesimally
is small, infinitesimally small, then I will
have zero error, agree?
This is one thing but what does that mean?
If I have infinitesimally small size of the
slice, it will mean that I will have to do
these computations infinite number of times,
okay. So but in finite element method the
size of the slice and now we will use this
term called element.
So here what we were doing is, that we have
broken this line which is for f(x) into slices
I can call it or I can call them as elements,
okay I can call them as elements. So if the
size of the element goes to 0, I have to do
that computation infinite number of times,
which is physically not possible.
So what I do is that I reduce the size to
a significantly small number but it is still
not infinite, infinitesimally small. So I
still have not infinite number of elements
but finite number of elements, okay. That
is where the term finite element comes into
picture, that the size is not extremely small
and vanishingly small but it is finite in
size and also the number of elements is finite.
So this is 1, so as ?X approaches 0 my error
goes to 0, this is one thing. The second one
is, as our approach. See we in first approach
we assumed that height was constant, in second
approach we said height was linearly varying,
in third approach we said that height was
quadratically varying.
So as our approach becomes more sophisticated
and this is a general thing because more sophisticated,
error goes to 0, okay.
This transition from point B to C, right?
In first approach there was no transition.
It was constant, in second approach it was
linear transition, in third approach it was
quadratic transition.
And this function which helps in understanding
the transition, so here it is a constant function
in first approach, right? Constant form, constant
transit, constant function, here it is a linear
function in approach to, in third it is a
quadratic function. These functions which
explain how variables are changing over an
element, how variables are changing over an
element, not across elements over an element,
they are known as shape functions, okay.
Shape functions. Because they help us understand
the shape of the transitions from point A
to point B on an element.
Now these functions which are plotted in this
line are one-dimensional because they are
only functions of X. If you have a plate or
a shell it could be a two dimensional function.
If you have a solid object it could be a three
dimensional function, okay. So these functions
could be 1D, 2D or 3D. But they help us understand
that within an element how is u, suppose you
are trying to capture u, how is u changing
from one point to other? Is it linear transition,
is it a quadratic transition, is it a constant,
is it cubic, okay. So that is there.
So these are known as shape functions and
fourth is.
That these small slices which we called slices
initially.
This is what is element; this is what is an
element. These elements have at least in case
of a linear you know when the element it has
two end points, right? It has two end points
but they could have internal points also,
they could have internal points also.
Suppose so, so suppose you have a quadratic
function then to see, to see how the function
is changing over an element you have to develop
a quadratic expression, right?
To define a line you need two points, to define
a straight line you need two points. So this
linear function, for this you need the values
of this green line f(x).
Only at point B and C, right? But to define
a quadratic function which transits from b
to c, you need three values, you need three
values. So for a quadratic function which
is one dimensional in nature you have to get
value at one end point, you have to get value
at other end point, and you also have to get
a value at somewhere in between.
Because then because what is a quadratic function?
It is ax 2 + b x + c. So to find a, b and
c you need values of that function at three
points. Similarly for a line, linear line
it is y = ax + b, to calculate values of a
and b you need values at the end points.
But for quadratic you not only need the value
of the function at end point but also the
value at maybe midpoint or somewhere in between,
it is somewhere in between, okay.
So, so elements have nodes where we have to,
where we are interested in finding the values
at ends only for linear elements, and they
also have nodes at ends and in between for
higher order elements, okay. Is this clear?
Now how do we all, we use all of this study,
we will explain it in the next lecture. But
what I wanted to impress upon you is.
That the accuracy of our, when we are trying
to calculate the area it, we have seen in
this case it depends on two parameters. The
first parameter is how thin each slice is
or how small each element is, the more small
we make the better we get in terms of reducing
the error. The second way of reducing the
error is that not only we make the slice small,
but we also pick up a better transition between
two endpoints of the unknown variable, okay.
So, so a better, a higher order shape function,
another term for this shape function is interpolation
function.
So in this class I will use shape function
and interpolation function interchangeably.
They are known as interpolation function because
they help us interpolate the value of unknowns
between two nodes, okay. If I know the function,
using that function I can interpolate the
value of the unknown between two nodes using
interpolation methods.
So they are known as shape functions or interpolation
functions, some people also call them as basis
functions, basis functions. So these are,
so if we use higher order shape functions
that also helps us increase the accuracy,
if you reduce the size of the element we have
seen that also helps us reduce the error and
increase the accuracy. These are two ways
to increase accuracy of at least in context
of this example the area under a curve.
We will extend the same thing in several other,
we will continue you know subsequent lectures
to; we will use the same idea to when we are
trying to solve differential equations, thank
you.