Finite Element Analysis

Prof. Dr. B. N. Rao

Department of Civil Engineering
Indian Institute of Technology, Madras

Lecture - 24

So in todays class, we will look at quadrilateral elements; and we will first look at
derivation of shape functions or both rectangular elements and square elements with the
different number of nodes; and then we look at similar to 1-D elements. We will look at
isoparametric mapping concepts for quadrilateral element and also its limitation, and also
we will discuss about numerical integration and for two-dimensional elements, and also
derivation of element equations for two-dimensional boundary value problem using
quadrilateral element with different number of nodes.

(Refer Slide Time: 01:04)

So, now higher order elements and elements with curved boundaries are effective, when
good approximate solutions are required with relatively few elements. So, that is the
basic idea behind these quadrilateral elements. Theoretically shape functions for any of
these elements can be developed by starting from a polynomial of appropriate degree,
and then expressing the coefficients in the polynomial in terms of nodal parameters or
the nodal values, similar to the way we did for one-dimensional elements both 2 node
elements and 3 node elements and also similar kind of approach we also adopted for
deriving shape functions for 3 node triangular elements which are linear.

This was the procedure used in the development of linear triangular element. This
approach however, becomes tedious and impractical for higher order elements that is
starting from a polynomial of appropriate degree and then trying to find the coefficients
of this polynomial by substituting the nodal values or nodal parameters and
corresponding nodal coordinates and solving these coefficients and substituting back
these coefficients in to the polynomial and grouping terms containing common nodal
parameters.

So, that is how we derived for that is how we derived shape functions for 1-D elements
and also linear triangular element, but that approach becomes tedious or impractical for
higher order elements that it as the number of nodes for a particular element increases we
need to choose a polynomial having as many number of coefficients as the number of
nodes for that particular elements. So, solving these coefficients and substituting back
and grouping terms having the nodal parameters common nodal parameters becomes
tedious for higher order elements.

(Refer Slide Time: 03:41)

Fortunately, for second order problems simple formulas exist that give shape functions
directly for rectangular and triangular elements. So, we will be discussing some of these
approaches how to get these simple formulas in this class. So, this lecture presents shape
function formulas for higher order rectangular and triangular elements and these
formulas together with isoparametric mapping concept play a fundamental role in
development of elements for practical applications, because these higher order elements
are element with curved boundaries are really required for solving some of the practical
problems.

The concept of isoparametric mapping was introduced earlier for one-dimensional

problem as a way to map actual element to simpler parent element, basically this was
done for one-dimensional elements to integrate some of the matrices and vectors that we
get.

(Refer Slide Time: 05:15)

Shape functions this is how we did; shape functions were written for the parent element.
Integrations and differentiations were performed over parent element. So, in this lecture,
the concept of isoparametric mapping will be extended to two-dimensional problems.
Using this concept, it is possible to develop quadrilateral elements and element with
curved boundaries.
(Refer Slide Time: 06:01)

So, now let us start with derivation of shape functions for rectangular elements. For
rectangular elements, the shape functions are based either on Lagrange interpolation
formula or they are written directly from experience. The elements which can be the
shape functions of which can be obtained using Lagrange interpolation formula are
classified as Lagrange elements, and the other elements they are classified as serendipity
elements.

So, now shape functions based on Lagrange interpolation formula. The shape functions
for rectangular and square elements are products of Lagrange interpolation shape
functions in x and y directions as illustrated in the following examples. So, basically we
need to write Lagrange interpolation shape function in x direction, Lagrange
interpolation shape function in the y direction, multiply these two, then we get the shape
function for the particular rectangular or square element. This is how the procedure goes
based on Lagrange interpolation formula.
(Refer Slide Time: 07:31)

So, now let us take a 4 node rectangular element, A 4 node rectangular element is shown
in the figure. The coordinates of node 1 are denoted by x 1, y 1 and of node 2 are
denoted by x 2, y 2, similarly node 3, node 4 etcetera and also note that for this particular
element that is shown in the figure x coordinate of node 2 is same as x coordinate of
node 3. Similarly, y coordinate of node 4 is same as y coordinate of node 3. Similarly, x
coordinate of node 4 is same as x coordinate of node 1 and y coordinate of node 2 is
same as y coordinate of node 1. So nodes can be denoted using the coordinates or for
simplicity nodal coordinates are identified by the node numbers.
(Refer Slide Time: 08:47)

So, now let us see how to derive shape functions for this particular element. If T is the
field variable and T 1, T 2, T 3, T 4 etcetera are the nodal variables, then the trial
solutions in terms of shape functions is expressed as; first let us see only along 1 2 you
can see from the figure y is going to be constant, y is equal to y 1; therefore shape
functions must be function of x only. So, now we are going to write shape functions
along line 1 2. So, that is denoted with T 1 that is field variable variation along line 1 2 is
denoted with T 1, and it is going to be function of x alone n 1, n 2 are going to be
Lagrange interpolation functions T 1 T 2 are the field variable values at node 1 and node
2 and from the knowledge of one-dimensional elements we already know how to get n 1
and n 2.
(Refer Slide Time: 10:21)

Using one-dimensional Lagrange interpolation formula, we know n 1 is equal to this n 2

is equal to the value or the quantity that is given there. So, we know how the field
variable is varying along 1-2. Now, let us look at alongside 4-3 and from the figure it can
be easily noticed that y value along 4-3 is equal to y 3 is equal to y 4 and field variable
along 4-3 is denoted with T, roman letter II and that can be written in terms of shape
functions of node 4 and node 3 in the manner that is shown there, that is T 2 is equal to n
4 times T 4 plus n 3 times T 3, which can be written in matrix and vector form in the
manner that is shown.
(Refer Slide Time: 11:58)

Again, from one-dimensional Lagrange interpolation formula T n n 4 and n 3 can be

written like this. So, we have seen how the field variable T is varying alongside 1-2 and
also alongside 4-3. Now, let us look at take one of the sides, which is along y direction in
y direction, so along 1-2, we know T 1 and along 4-3, we know T 2 from the previous
equations. So, once we know the value of field variable alongside 1-2 and 4-3 in the y
direction variation of T in the y direction that is alongside 1-4 or 2-3 can be written as T
is equal to n 1 T 1 plus n 4 T 2, which can be written in matrix and vector form in the
way that is shown there.
(Refer Slide Time: 13:20)

And now substituting T 1 T 2 and n 1 n 4 can be obtained by writing one-dimensional

Lagrange interpolation in y direction, so that is how n 1 and n 4 are obtained.
Expressions for side 1-2 and 4-3 can be written in matrix form, that is T I T II are written
together in a matrix and vector form. So, substituting T I T II vector in to the previous
equation we get this one. So, carrying out multiplications of n 1 n 4 vector with the
matrix containing n 1 n 2 0 0 0 0 n 3 n 4.

(Refer Slide Time: 14:21)

We get this, which can be compactly written like this. So capital N 1 is defined as small
n 1 as a function of x times small n 1 as a function of y, similarly capital n 2, which is
shape function corresponding to node 2 is equal to small n 1 as a function of y times
small n 2 as a function of x. Similarly, shape function of node 3, which is denoted with
capital N 3 it is equal to small n 4 as a function of y times small n 3 as a function of x.
Similarly, shape function at node 4, which is denoted with capital N 4 is equal to small n
4 as a function of x times small n 4 as a function of y. So we can write what is capital N
1, capital N 2, capital N 3.

(Refer Slide Time: 15:31)

And substituting what is small n 1 as a function of x small n 1 as a function of y we get

N 1 like this, which is basically derived based on Lagrange interpolation formula in x
direction multiplied by Lagrange interpolation formula in y direction at node 1.
Similarly, shape function of node 2, shape function of node 3 and shape function of node
4. It can be easily observed that all of these shape functions, let us say N 1, N 1 is going
to be equal to 1 at x is equal at x is equal to x 1 and y is equal to y 1 and it is going to be
equal to zero at all other locations. Similarly, N 2 is going to be 1 at x is equal to x 2 y is
equal to y 2 and it is going to be zero at other nodal position. Similarly, N 3 and N 4 are
going to be N 3 is going to be equal to 1 at node 3 and it is going to be equal to zero at
rest of the nodes. Similarly, N 4 is going to be 1 at node 4 and it is going to be equal to
zero at rest of the nodes.
(Refer Slide Time: 17:11)

So, the shape functions for rectangular elements are product of Lagrange interpolations
in two coordinate directions. So, that is how we derived and this N 1 and note that this is
equal to 1 at node 1 and zero at other nodes and it is linear function of x along 1-2 side 1-
2 and linear function of y alongside 1-4 and 0 alongside 2-3 and 3-4, because node 1 is
not part of side 2-3 and 3-4, so it is going to be shape function of node 1 is going to be
zero alongside 2-3 and 3-4. So, these properties not only node shape function of node 1,
but other shape functions of other nodes also satisfy these properties. So, shape function
of a node is going to be zero alongside to which it is not going to be part and shape
function of a particular node is going to be equal to 1 at it is own location and it is going
to be equal to 0, at all other locations, at all other nodal locations.
(Refer Slide Time: 18:46)

So, other shape functions have similar behavior, because of these characteristics the i eth
shape function is considered associated with node i of the element.

(Refer Slide Time: 19:04)

And the shape functions that we derived based on Lagrange interpolation formula the
same shape functions can also be derived from starting with a polynomial and that
polynomial is given here. So, starting with this polynomial we can derive same shape
functions, similar to the procedure that we adopted for triangular element. And if you see
this polynomial note that because of presence of term x y, the x and y derivatives of T
are not constant and if you recall shape function or the derivatives of shape functions of a
linear triangular element of constant and also derivatives of T field variable T are also
constant for linear triangular element, because if you recall the polynomial that we used
for deriving shape functions for linear triangular element do not contain this x y term.

But now for this 4 node quadrilateral element one way of deriving shape function is
starting with a Lagrange interpolation formula or the other way is by starting from a
polynomial like this and if you start with polynomial like this you can see there is a
presence of this x y term. Basically, please note that this 1 x y; x y all these terms are
coming from Pascals triangle, so because of the presence of this x y term in this
expression the derivatives of T with respect x and y are not constant, which was the case
for triangular element linear triangular element. So, therefore this element generally
gives better results than a triangular element. So, we have seen how to derive shape
functions for a 4 node rectangular element using Lagrange interpolation formula.

(Refer Slide Time: 21:48)

So, we can solve an example like this for shape functions a 4 node rectangular element is
shown all the coordinates of all the nodes are also shown x y coordinate system x y axis
are also indicated clearly in the figure and these are the shape function expressions that
we derived, so now if somebody is interested in writing shape functions for each of these
nodes N 1 to N 4, simply we need to plug in the corresponding coordinate values into
this expressions for N 1, N 2, N 3, and N 4. So x 1 is equal to 0, y 1 is equal to 0, x 2 is
equal to 3, y 2 is equal to 0, x 3 is equal to 3, y 3 is equal to 2, x 4 is equal to 0, y 4 is
equal to 2 substituting these quantities in to N 1, N 2, N 3, and N 4 expressions, we can
get the shape functions.

(Refer Slide Time: 22:59)

Substituting the numerical values of nodal coordinates in to the above shape function
formulas, the explicit expressions for shape function for this rectangular element are as
follows. So, this is N 1 after simplification N 2, N 3 and N 4. To visualize how the shape
functions varies as a function of x and y over the domain of that particular element we
can actually plot N 1, N 2, N 3 and N 4 as a function of x and y with x varying from 0 to
3 and y varying from zero to 2.
(Refer Slide Time: 23:49)

So, similar kinds of plots are shown here for N 1 and N 2, three-dimensional plots of N 1
N 2 are shown and these plots can be obtained using any of the commercial software like
MATLAB or Mathematica by just giving the expression for shape function and also the
range over which plot is required that is x going from 0 to 3 y going from 0 to 2. So, this
is how we can derive shape functions for 4 node rectangular element using Lagrange
interpolation formula.

(Refer Slide Time: 24:33)

So, now let us take a 6 node rectangular element like this and here also shape functions
can be written in the manner are following the procedure that we adopted for 4 node
rectangular element writing shape function expressions along x direction and shape
function expressions along y direction multiplying both we get shape functions of each
of the nodes. So, the coordinates of node 1 are x 1 y 1, and those of node 2 are x 2 y 2,
similarly other nodes, and you can see here in this 6 node rectangular element, which is
shown node 2 and node 5 are interior to side 1 2 and side 6 4 and these nodes 2 and 5 are
located arbitrarily on the side 1 3 and 6 4.

(Refer Slide Time: 26:00)

So, now let us write the shape functions for all the nodes of this six node rectangular
element. Following same reasoning as for four node element it is obvious that shape
functions have quadratic variation in x direction and linear variation in y direction. If you
see this six node rectangular element alongside 1 3 we have three nodes alongside 1 6 we
have only two nodes, so two nodes in y direction gives us linear variation in y direction
three nodes in x direction gives quadratic variation in x direction.

So, following the procedure that we adopted for rectangular a four node rectangular
element, we can derive in a similar manner shape functions for all six nodes of this
particular element. Here N 1 is shown; N 1 the shape function of node one is Lagrange
interpolation formula in x direction, which is going to be quadratic, because there are
three nodes in x direction times Lagrange interpolation in y direction, which is going to
be linear, because there are two nodes in y direction. The product of those two gives us
shape function for node 1.

(Refer Slide Time: 24:40)

Similarly, node 2, node 3, node 4, node 5 and node 6 and once we have all the shape
functions the trial solution can be written like this T is equal to N 1 times T 1 plus, N 2
times T 2 plus, N 3 times T 3 plus, N 4 times T 4 plus, N 5 times T 5 plus, N 6 times T 6,
which can be written in a matrix and vector form the way that is shown.

(Refer Slide Time: 28:19)

So, now let us take a numerical example with all the coordinate values given. So, here in
the figure a 6 node rectangular element is shown, x y coordinates of all the nodes can
also be easily obtained using the information that is given in figure that is x 1 is equal to
0, y 1 is equal to 0, x 2 is equal to 2, y 2 is equal to 0, x 3 is equal to 3, y 3 is equal to
zero and x 4 is going to be 3 and y 4 is going to be 2, x 5 is going to be 2, y 5 is going to
be 2, x 6 is going to be 0, y 6 is going to be 2, so with all this information what we can
do is we can plug in these coordinates of these nodes in to the expressions that we have
for shape functions N 1 to N 6 and we can get the shape function values of all the nodes
and also we can write the trial solution.

(Refer Slide Time: 29:50)

Substituting the numerical values of nodal coordinates into the shape function formulas
explicit expressions for shaped functions for 6 node rectangular element are given here N
1, N 2, N 3, N 4 and N 5 and to visualize how the shape functions looks or how they vary
along x and y directions, we can plot three-dimensional plots of N 1 N 2 are shown in
figure below.
(Refer Slide Time: 30:26)

And similarly shape function of other nodes can be plotted. So this is a 6 node
rectangular element based on Lagrange interpolation formula.

(Refer Slide Time: 30:51)

So, now let us look at another Lagrange element, which is 9 node rectangular element. A
9 node rectangular element is shown here also x y axis are shown in the figure and nodes
2, 4, 6, 8 can be located at any place on respective sides and node 9 is located inside the
element and coordinates of node 1 are x 1 y 1 and similarly for the other nodes and here
you can see shape functions varies quadratically both in x direction y direction, because
we have three nodes along x direction and three nodes along y direction.

(Refer Slide Time: 32:06)

Here shape functions vary quadratically in both directions. So, writing Lagrange
interpolation formula in x direction multiply with Lagrange interpolation formula in y
direction we can write shape function expressions for all the nine nodes; N 1 Lagrange
interpolation x direction times Lagrange interpolation y direction.

(Refer Slide Time: 32:49)

Similarly, N 2 and the rest and the properties that we have seen for 4 node quadrilateral
element the shape functions for these nine node rectangular elements also satisfies. So if
you see this node 1 this is the expression for shape function of node 1 and it can be easily
verified that N 1 is going to be equal to 0 at all other nodes except node 1 where it is
equal to 1 and also it can be verified that N 1 is going to be 0 alongside 3-4-5 and
alongside 7-6-5 and alongside 1-2-3 this expression is quadratic function in x and
alongside 1-8-7 is quadratic function of y.

(Refer Slide Time: 34:08)

It is 0 at all other nodes, except node 1 where it is equal to 1, zero along edges 3-4-5 and
7-6-5 and along edge 1 or side 1-2-3 it is going to be quadratic function of x and
alongside are edge 1-8-7 it is a quadratic function of y and not only for shape function of
node, one similar observations can be made for other shape functions. So, here when we
are deriving shape function expression for this 9 node Lagrange element, basically we
use Lagrange interpolation formula, instead of that we can also start with a polynomial
having 9 coefficients and we can derive same shape functions.
(Refer Slide Time: 35:20)

Same shape functions can also be derived from the following polynomial using
procedure employed for linear triangular element, so this is the element for which we
need to derive shape functions. There are 9 nodes, so we need to start with a polynomial
having 9 coefficients like this and we can adopt the procedure that we adopted for
deriving shape functions for linear triangular element; and once we do that, we get same
shape functions as we obtained using Lagrange interpolation formula. But only thing is
this procedure is going to be tedious, and also it is going to be cumbersome since we
need to solve for nine coefficients and we need to group terms containing same nodal
parameters or nodal values to get the shape function expressions. Now, we have the
shape function expressions, explicit expressions based on Lagrange interpolation formula
for this nine node element, we can write shape function for any element once we knew
the nodal coordinates.
(Refer Slide Time: 36:37)

So, now let us take an example here a 9 node element is shown x y axis are also shown;
and also from the information that is given, we can easily figure out what are the x y
coordinates of each of the nodes. So, once we have that information, we can plug this
information into the explicit expressions for shape functions that we obtained using
Lagrange interpolation formula to get the shape function expressions.

(Refer Slide Time: 37:10)

Substituting the numerical values of nodal coordinates in to the above shape function
formulas explicit expressions for N 1, N 2 and N 9 shape functions for this rectangular
element. Here even though N 1, N 2 and N 9 are shown we can easily write or we can
easily simplifying the substituting the nodal coordinates in to the previous explicit
formulas and we can get the node shape function expressions for other nodes as well and
to visualize how is shape function N 1 and N 2 varies we can even plot.

(Refer Slide Time: 37:53)

So, here three-dimensional plot of N 1 and N 2 are shown for this particular 9 node
element. So, we have seen 4 node rectangular element, 6 node rectangular elements and
9 node rectangular element and we have also seen how to derive shape functions of all
these elements using Lagrange interpolation formula. So, these are one the elements for
which we can adopt Lagrange interpolation formula to derive the shape functions, but
there are some other elements for which we need to adopt some other procedure, so those
set of elements are called serendipity elements.
(Refer Slide Time: 38:56)

So, now let us look at those elements serendipity shape functions for rectangular
elements. Following shape functions for rectangular elements have been developed
intuitively hence the name serendipity based on basic characteristics of shape functions
that is N 1 is equal to 1 at node 1 and 0 at other nodes or N i is equal to 1 at node i and 0
at other nodes. So, instead of using Lagrange interpolation or sometimes it is not possible
to use Lagrange interpolation formula to derive shape functions for certain rectangular
elements containing certain number of nodes, in that case we need to derive shape
functions intuitively without violating the conditions that shape function of node i is
equal to 1 at node i and equal to 0 at other nodes.
(Refer Slide Time: 40:03)

Using that basic characteristics of shape functions and intuitively, if you can derive the
shape functions and that is what serendipity element. Elements based on these shape
functions are very popular, their main advantage is that all the nodes are placed on
element sides and thus there are no interior nodes.

(Refer Slide Time: 40:31)

So, now let us look at 8 node serendipity element and it is a quadratic element, so it is
called 8 node quadratic serendipity element and x y axis are indicated in the figure. And
also based on the information that is given in the figure, we can easily figure out what
are the x y coordinates of each of the nodes. And if you compare this element with the 9
node element that we have just seen, you can notice that only the middle interior node,
which is 9th node is missing, so that is the only difference.

So, here before I show the shape function expressions for this element, let us see if you
want to derive shape function of node 1 and you can see from the figure node 1 is going
to be 0 along edge 3-6 and node 1 should also be zero along edge 7-6-5, in addition to
edge 3-4-5 and if you can include the equation of line 3-4-5 equation of line of edge 3-4-
5 and equation of line of edge 7-6-5 in to the shape function expression of node 1, so
then node 1 is going to be zero along edge 3-4-5 and it is going to be 0 along edge 7-6-5.

Similarly, node 1 has to be 0 along or at node 2 and 8. So, if you can come up or if you
can get the equation of line, which passes through node 2 and 8 that can be easily derived
based on the nodal coordinates of node 2 and node 8, we can easily write what is the
equation of straight line that passes through nodes 2 and 8. If you can include that
equation of that line also in to the shape function of node 1 then we get the shape
function including all the equations of sides 3-4-5 and 7-6-5 and also equation of line
passing through node 2 and 8. If you include all these into the shape function expression
for node 1 and normalize it we are going to get finally the shape function expression
explicit expression for this 8 node quadratic serendipity element for node 1.

Similarly, we can derive for node 2, node 3, node 4, node 5, node 6, node 7 and node 8.
So, based on that, we can easily write the shape functions for this element. Note that
nodes are at the corners and at the mid sides and the origin of the coordinate system is at
the element centroid for this element.
(Refer Slide Time: 44:18)

So, based on the procedure that I mentioned, when we are writing shape function for
node 1 include equation of line passing through sides 3-4-5 and also include equation of
line passing through 7-6-5 and equation of line passing through 2 and 8 and normalize it
then we are going to get shape function expression of node 1. Here it is written in terms
of s and t, where s and t are defined; s is equal to 2 x over a, t is equal to 2 over 2 y over
b.

(Refer Slide Time: 45:10)

So, this is how we can write shape functions for rest of the nodes. So, adopting the
explanation or the procedure that I mentioned, one can easily write shape functions for
rest of the nodes explicit expressions for all the nodes. All the 8 nodes are given here and
it can be easily verified that each of these nodes is equal to 1 at its own position and it is
equal to 0 at the other nodal locations.

(Refer Slide Time: 45:48)

It can be easily verified that the shape functions have the desired properties that is N i is
equal to 1 at node i and N i is equal to zero at other nodes. So, here we used some kind of
reasoning or we have developed whatever expressions that I have shown we have
developed for this element intuitively by making sure that it is shape function at a
particular node is equal to zero at other nodal location, we have derived the nodal shape
functions explicit expressions of nodal shape functions intuitively; instead of that we can
also start with a polynomial having 8 number of coefficients.
(Refer Slide Time: 46:57)

The shape functions can be derived from the following polynomial using method that we
adopted for linear triangular elements. So, since there are 8 nodes, we need to start with a
polynomial having 8 coefficients. So, this is the polynomial with which we can start and
adopt the procedure that we did for linear triangular elements and finally, we can get
same shape functions as we have seen and the three-dimensional surface plots of these
shape functions are similar to those of biquadratic lagrangian shape functions.

(Refer Slide Time: 47:59)

So, now we have derived shape functions for 4 node element, 6 node element, 9 node
element and 8 node element, so we can write shape functions for all elements with nodes
having nodes from 4 to 9, so here we will write a general set of shape functions to define
shape functions for any rectangular elements with nodes range in from 4 four 9. The
complete set of shape functions will be given in a table and but the ordering of nodes or
the ordering of element node numbers is very important.

So, the expressions that I am going to show you are valid only for this node numbering.
With this particular node numbering scheme shape functions for higher order elements
are constructed by adding terms in to shape function for lower order elements. Here there
is a typing mistake in the title it should be shape functions for 4 to 9 node rectangular
element, so complete set of shape functions for any element with nodes from 4 to 9 are
given in table below.

(Refer Slide Time: 49:39)

And the following notation is used for the expressions that are given in the table s is
defined like the way it is done earlier s is equal to 2 x over a t is equal 2 y over b and f 1,
f 2, f 3, f 4, f 5, f 6, f 7, f 8 and f 9 are defined like this. The shape functions expressions
for all nodes of an element having 4 to 9 nodes is expressed in terms of these fs, so this
definition of f 1 to f 9 is very important to read the table.
(Refer Slide Time: 50:31)

So, in the table, expressions for shape functions of all nodes for 4 to 9 node rectangular
elements are given, 4 node element, 5 node element, 6 node.

(Refer Slide Time: 50:54)

7 node, 8 node.
(Refer Slide Time: 50:58)

And 9 node element and please note that elements with number of nodes between 4 and
8 are known as transition elements. These elements are useful when transition from
quadratic to linear element is desired. So using this table, we can derive shape functions
for any nodded elements starting from 4 node to 9 node element and we will continue in
the next class.