Finite Element Analysis

Prof. Dr. B. N. Rao

Department of Civil Engineering
Indian Institute of Technology, Madras

Lecture - 36

In last class, we have derived element equations for two d elasticity problems starting
with governing potential energy functional. And using shape function corresponding to 3
node triangular element. And also the solution procedure is illustrated through an
example, and at the end we found that, before that we have taken a cantilever plate, and
using two elements, two triangular elements with 3 nodes each. We descritized, that
cantilever plate and the solution that we got, when we compare it with exact solution.

We can actually compare the stresses along the common side for the two elements, that
is interface between the two elements. For exact solution stresses on two sides should be
equal and opposite of each other; and the solution that we obtained using two linear
triangular elements, for this cantilever plate problem. We observed that there is large
discontinuity in stresses across element boundaries, so this solution is not good this is
because, we used only two elements, the mesh is very coarse and obviously, we cannot
expect very good results, but as we increase the number of elements, its discontinuity in
stresses should reduce.

So, now in today’s class, we will look at 4 node quadrilateral element that is derivation
of finite element equations using 4 node quadrilateral elements for two d elasticity
problem. And subsequently illustration to the same cantilever plate problem now is using
only one 4 node quadrilateral element. Later introduce, let class we will be looking at 8
node isoparametric element as well.
(Refer Slide Time: 02:38)

So, now, let us get started a 4 node quadrilateral element is shown on the left hand side,
and for isometric mapping purpose, will be using the element that is shown, the parent
element that is shown on the right hand side. The element equations can easily be
derived using isoparametric mapping concept, introduced earlier, that we are familiar
with.

The trial solutions are written in isoparametric mapping, the trial solutions are written in
terms of shape function for parent element. So, we need to know what are the shape
functions for the 4 nodes of this quadrilateral element in the parent element domain.
(Refer Slide Time: 03:35)

Once we have the shape functions of all 4 nodes of the parent element, we can write this
equation which interpolates displacement, using the nodal values at the 4 nodes and the
interpolation functions or the shape functions. Here, N 1, N 2, N 3 are nothing but, shape
functions corresponding to the 4 node parent element.

And once, we know these shape function expressions we can as well write expressions
for isoparametric mapping, that is how x y are related to s and t, how the axial element
coordinate are related to the parent element coordinates, this is isoparametric mapping,
written in matrix and vector form.
(Refer Slide Time: 04:52)

And once we have this information, we can easily find what is Jacobian matrix, and
determinant of Jacobian matrix compresses of all partial derivative of x with respect s, x
with respect to t, y with respect s and y with respect to t. So, once we have these two
equations, we can easily find those derivatives, so Jacobian matrix is defined like this,
and from here we can find determinant of Jacobian.

There is nothing new in all this process, it is a same as for 3 node triangular element
except that the number of nodes now became 4 instead of 3, so the matrix sizes becomes
bigger. So, once we have x and y related to the nodal values like this, nodal coordinate
values we can easily find the partial derivatives.
(Refer Slide Time: 06:01)

Using these relations, which can be simplified further as shown there, similarly
derivative of x with respect to t.

(Refer Slide Time: 06:20)

And we can also find the derivatives of y with respect to s and t, using y coordinates of
node instead of x, in the previous two equations. And now, we also require knowing how
the shape function derivatives in the axial element coordinate system or related on the
parent element coordinate system that is given by this one. Derivatives with respect to x
and y can be computed as follows, which is given there for which we require to calculate
determinant of Jacobian.

(Refer Slide Time: 07:10)

So, once we have all this information, we can easily write strain-displacement
relationship, strains can be expressed in terms of nodal displacements; we are familiar
with this definition of strain. So, now, substitute or this vector consisting of partial
derivative of displacements can be rearranged, as shown in the later part of the equation,
and partial derivatives of displacements with respect x and y can easily be calculated
using this relation.
(Refer Slide Time: 07:48)

This is derivatives of displacement component in the x direction similarly, derivatives of

displacement component in the y direction. And we can put these two equations together,
and get this one and here in the last equation, we have derivatives of displacement
components both in x and y directions with respect to x and y.

(Refer Slide Time: 08:37)

So, now once we have that information strains, we know definition of strain and we just
found what are the derivatives of displacement components with respect x and y, so we
can plug in that and rearrange the equations and finally we get this. Where A is a matrix
which is defined like this, please note that this kind of definition of A and other
intermediate variables is choice of the teacher. Sometimes, if you refer some other books
you may see different definition, are all together they may avoid this kind of intermediate
definition or intermediate variable definition. Now, derivatives of trial solution with
respect to s and t are easy to compute, here if you see A matrix you have derivatives of
trial solution with respect to s and t, the vector beside A matrix it consists of derivatives
of trial solution with respect to s and t, which can be easily calculated.

Because, trial solution is function of s and t, because the shape functions or functions of s
and t. And this can be done in this manner that is calculation of derivatives of trial
solution with respect s and t can be obtained through this equation, which can be
compactly written like this.

(Refer Slide Time: 10:50)

So, finally, strain-displacement matrix can be written like this, so the key thing is we
need to get the nodal coordinate information, and then we can easily follow these steps.
And finally get the required quantities to calculate strains, once we knew the nodal
values.
(Refer Slide Time: 11:27)

So, now let us look at element stiffness matrix, element stiffness matrix is defined like
this, please note here for 3 node linear triangular element B matrix is a constant whereas,
for 4 node quadrilateral element B matrix is not a constant, so we need to perform
numerical integration. So, for that purpose we need to change the limits of integration
from minus 1 to 1, and the details of how we do that we already discussed several times
in the earlier lectures.

And here in this equation h is the element thickness; C is the appropriate constitutive
matrix depending on plane stress or plane strain. The stiffness matrix is obtained using
Gaussian quadrature can be evaluated using Gaussian quadrature, which is a numerical
integration procedure.
(Refer Slide Time: 12:37)

2 by 2 integration is generally considered adequate for this element, and so element

stiffness matrix can be approximated by taking two integration points along s direction
and two integration points along along t direction; and weight at each integration point is
weight in the s direction times weight in the t direction. And evaluate all the quantities in
the integrand, at these integration points and sum up and multiply the integrand value at
each integration point with the corresponding weight, and sum up the contribution from
all the integration points.

So, that is basically what numerical integration schemes will do, and that is here
expressed in the form of an equation. Where s and t are the locations of Gauss point and
w’s are the corresponding weights, m and n are the number of integration points in s and
t directions respectively. These integration points in s and t direction need not be same,
which we already looked at; it depends on the order of the integrand in the s direction,
and order of integrand in the t direction, so this is how element stiffness matrix can be
evaluated.
(Refer Slide Time: 14:10)

So, now, let us look at how to assemble or how to obtain equivalent nodal forces, and for
illustration purpose, we consider uniformly distributed pressure along element side. If T
x, T y are the components of applied surface pressure in x and surface pressure is same
as surface traction are the components of applied surface pressure in the x and y
directions, the equivalent nodal load vector is given by this one.

So, here we need to substitute N, N is nothing but, a shape function matrix of all the
shape functions along that particular side, along which the pressure is applied.
Integrations can be performed in a closed form, if the specified surface tractions T x, T y
are simple functions of x and y. This also we discussed in the last class, if T x, T y are
simple functions we can easily perform the integrations, we can perform integrations in a
closed form, but if T x, T y are complicated functions; then we need to adopt numerical
integration. And the simplest case that you can have is T x, T y being constant or
uniform.
(Refer Slide Time: 15:48)

Simplest case is when T x, T y is specified as constant along one or more sides of an

element. And also please note that, we are dealing with 4 node elements, so the trial
solution is linear along each side and depends upon the two end nodes, so we can easily
write the shape function of the nodes using Lagrange interpolation formula, which we
discussed earlier. Here, both actual element and the parent element are shown and also
the traction is applied on side 2-3.

For each side the coordinate’s capital S for actual element and the corresponding small s
for the parent element are shown for the shape functions, for each side can be expressed
as follows, alongside 1-2. If you take parent element along side 1-2, shape functions of
node nodes 3 and 4 are going to be 0 alongside 1-2.
(Refer Slide Time: 17:08)

So, the shape function vector, shape function matrix is going to be like this, because
shape functions of nodes 3 and 4 alongside 1-2 are going to be 0.

(Refer Slide Time: 17:30)

And similarly, side 2-3, shape functions of nodes 1 and 4 are going to be 0 alongside 2-3.
Anyway we discussed all these several times in the earlier classes, but for completeness
all those are repeated here, for side 3-4, shape functions of nodes 1, 2 are going to be 0.
And the shape function matrix for evaluating the equivalent nodal vector.
(Refer Slide Time: 18:00)

The shape function matrix that is required, if the traction is applied along side 1-2 looks
like this, similarly 4-1, side 4-1. So, once we obtain this shape function matrix, we can
easily evaluate equivalent nodal load vector; for illustration consider the case, when
uniform pressure is applied along side 2-3.

(Refer Slide Time: 18:38)

So, to proceed further we need to write, what is isoparametric mapping for side 2-3 and
only non-zero shape functions are going to be shape function corresponding to node 2
and 3. Other shape functions are going to be 0 and shape functions of node 2 and 3 for
side 2-3 can easily be obtained using Lagrange interpolation formula. And finally, we
can write the relationship between x coordinate or local coordinate along that side 2-3 for
the actual element, the relationship between the local coordinate system for a side 2-3 of
actual element, and local coordinate system of side 2-3 of the parent element. How they
are related, we can write through this isoparametric mapping. So, this is how x and y are
related to S the local coordinate system of the parent element.

So, once we have this information of x and y, we can easily take partial derivative of x
with respect s partial derivative of y with respect to s, this is partial derivative of x with
respect s, which can be rearranged as the weight is shown there d x is equal to x 3 minus
x 2 divided by 2 times d s.

(Refer Slide Time: 20:26)

Similarly, partial derivative of y with respect to s, and once we have this we can easily
write Jacobian transformation of local coordinate system, along side 2-3 of the actual
element. And local coordinate system along side 2-3 of the parent element, so the
Jacobian transformation from capital S to small s is defined as, this one and we just
calculated what is partial or derivative of x with respect to s, derivative of y with respect
to s.
(Refer Slide Time: 21:21)

So, using that information and with reference to this figure, which map side 2-3 of the
actual element to side 2-3 of the parent element. Differential element in the actual
element capital D capital S is from the geometry, we can easily see that, it is going to be
square root of d x square plus d y square. And substituting what is d x, d y we get this
and here, there is a small type typographical error, it should be d s instead of that d t is
typed.

So, finally, we get d capital S is equal to L 2 3 over 2 times d small s, where L 2 3 is

nothing but, length of side 2-3. So, the Jacobian is d capital S over d small s is nothing
but, length of side 2-3 over 2 half the length of that side, Jacobian for each side is equal
to half the length of that side.
(Refer Slide Time: 22:49)

So, we can easily now evaluate numerically by changing limits of integration, once we
got this Jacobian, once we obtained the relationship between d capital S and d small s we
can change the limits of integration; and adopt numerical integration if required, if the
tractions applied are constant, then we do not need to adopt numerical integration, we
can do integrations in a closed form. So, since here we assume T x, T y are uniform or in
other words they are constant.

(Refer Slide Time: 23:29)

So, we can easily obtain this Q vector in a closed form, so carrying out integrations we
get Q in this manner. Similar to linear triangular element this equation says total pressure
along the side is divided equally among 2 nodes along that side.

(Refer Slide Time: 24:01)

If there are known forces distributed over an element, than the work done by these body
forces can be evaluated easily using two-dimensional integration used for stiffness
matrix evaluation. All quantities needed for potential energy functional have now been
expressed in terms of nodal unknowns, so finally, from potential energy functional, we
get k d equal to Q t.

So, to illustrate all these details, let us take an example, example is same as what we used
for illustrating 3 node triangular elements, except that there we adopted two elements 2,
3 node triangular elements. Whereas, here we will be using only one 4 node quadrilateral
element for solving the same problem, cantilever plate problem.
(Refer Slide Time: 25:21)

Again all the quantities are given both in FPS units and SI units and since, the quantities
that are given in FPS units are round numbers, we will proceed with FPS units, but it is
not going to make much difference, because as long as we are consistent with the units
we get solution.

(Refer Slide Time: 25:40)

Element stiffness matrix can be written by evaluating following quantities at Gauss

points, so we require to evaluate, we are now adopting 4 node quadrilateral element. So,
we need to use 4 node element, 4 node parent element shape functions and try to find
what is partial derivative of x with respect s, partial derivative of x with respect to t,
partial derivative of y with respect s, partial derivate of y with respect to t, and after that
we can easily find determinant of J.

(Refer Slide Time: 26:27)

And once, we have determinant of J, we can even calculate what is that intermediate
matrix A required for getting strain displacement relation, and also one more
intermediate matrix is required; and once we have these two, we can get strain
displacement matrix. And constitutive matrix for plane stress, and once we evaluate all
these quantities at each integration point, and then depending on the number of
integration points.
(Refer Slide Time: 27:25)

Depending on the number of integration points, evaluate the integrand at each of the
integration point multiply with corresponding weight, and sum it up we are going to
entire approximate value of entire stiffness matrix.

(Refer Slide Time: 27:51)

So, to proceed further, we need to make a note of all the coordinates of all nodes, x
coordinates and y coordinates.
(Refer Slide Time: 28:00)

And once we make a note of it, they are ready to evaluate stiffness matrix using Gauss
integration, so here we are adopting 2 by 2 integration, so in total there are going to be 4
integration points, how to obtain coordinates and weights of the corresponding
integration points, we already discussed several times earlier. So, now let us take 1
integration point, and see the details and similar procedure can be repeated in the other
integration points. So, Gauss point that is selected is given the details are given
coordinates and weights.

(Refer Slide Time: 28:53)

So, evaluate all the required quantities at this integration point and finally, we can
evaluate at this integration point, what is strain displacement matrix, and once we have
that we can easily evaluate what is stiffness matrix at this integration point. Performing
similar calculations for other integration points, and adding the resulting matrices, we get
the following or we get the element stiffness matrix.

(Refer Slide Time: 29:17)

So, we obtained the element stiffness matrix, and there is not much inward as far as
assembly of equivalent load vector is concerned, because all the point loads are given, so
the all the loads that are applied are concentrated loads. So, only we need to just plug in
the corresponding value at the appropriate location in the load vector. And if you see this
problem, there are 4 nodes and nodes 1 and 4 are fixed, so all degrees of freedom
corresponding to these nodes are going to be 0.

So, in the final global equation system, we are going to anyway eliminate the rows and
columns corresponding to degrees of freedom which are 0. So, finally, we can directly
write the reduced equation system, which consists of degrees of freedom corresponding
to nodes 2 and 3, which are u 2, v 2, u 3, v 3 because, degrees of freedom corresponding
to nodes 1 and 4 are going to be 0.
(Refer Slide Time: 30:42)

Global equations and nodal solution, since there is only one element the global stiffness
matrix is same as element stiffness matrix, essential boundary conditions. These are the
essential boundary conditions; imposing these boundary conditions gives the following
reduced system of equations.

(Refer Slide Time: 31:01)

And solving this we get the displacements at nodes 2 and 3, once we obtained the
displacements at nodes 2 and 3, we can easily do post processing like calculation of
stresses.
(Refer Slide Time: 31:23)

So, calculation of stresses for quadrilateral elements, the strains and stresses vary linearly
over the element because, if you see the derivatives or if you see the B matrix, B matrix
is going to be linear over the element. So, for quadrilateral element strains and stresses
vary linearly over element; stresses at any point over element can be computed using
strain-displacement and stress-strain relations, as an illustrations stresses at the nodes and
element centroid are computed as follows.

First to do that, we require vector of nodal displacement since, we calculated

displacements at nodes 2 and 3, and the other displacements are 0, we can easily write
the vector of nodal displacements. And now, we are interested in finding stresses at
nodes, so we need to know what is the corresponding s and t coordinates.
(Refer Slide Time: 32:37)

So, at node 1 s is equal to minus 1, t is equal to minus 1, so A matrix, G matrix once we

have these two matrices, we can easily get what is B matrix. And once we know B
matrix we can get this strain, because strain is related to nodal displacements y of this B
matrix.

(Refer Slide Time: 33:04)

Hence, once we have these strains, we can calculate stresses and then do rest of the post
processing similar to what we did in the last class, like calculating principle stresses and
other details.
(Refer Slide Time: 33:21)

Similar, calculations can be repeated at other nodes by taking appropriate s and t values,
this is at node 2, node 3 (Refer Slide Time: 33:38).

(Refer Slide Time: 33:46)

And that node 4, and please note that element centroid corresponds to s is equal to 0, t is
equal to 0.
(Refer Slide Time: 34:02)

So, if somebody is interested in finding the stresses and strains, and the corresponding
principle stresses and equivalent stress quantities are effective stress quantities at the
centroid, we need to calculate do the perform the calculations by taking s is equal to 0, t
is equal to 0. So, strain, stresses and all those details at this point are given.

(Refer Slide Time: 34:24)

So, the next element that we are going to look at is 8 node isoparametric element, using
isoparametric mapping higher order elements can be formulated in a manners, similar to
4 node quadrilateral element.
(Refer Slide Time: 34:48)

As an illustration an 8 node element based on serendipity shape functions is presented,

element can have curved boundaries, like this actual 8 node element is shown and also 8
node parent element is also shown. So, for by this time you are familiar for writing
isoparametric mapping relations, we require to know, what are the shape functions of all
the 8 nodes of 8 node parent elements.

So, first thing is we require to write or note down what are the shape functions of all the
8 nodes of 8 node serendipity element, which is shown serendipity element, parent
element which is shown on the right hand side, and the actual element in the figure is
shown on the left hand side.
(Refer Slide Time: 36:13)

This shape functions for the parent element are as follows, all the shape function
expressions we have seen earlier, and also we discussed how to derive this.

(Refer Slide Time: 36:29)

Using the parent element shape functions, the trial solution can be written as follows
displacement component in the x direction is given by, it can be obtained using all the 8
node shape functions, and the corresponding nodal values at the 8 nodes. So, suppose if
you want to displacement in the x direction, then it can be interpolated using
displacement component at all the 8 nodes in the x direction, and the corresponding
shape functions. Similarly, in the y direction, so this is how trial solution can be written
and isoparametric mapping, once we have these two relations, we can write what is
Jacobian matrix, determinant of Jacobian and also shape function derivatives an x with
respect to x and y; all this is again same as for 4 node quadrilateral element.

(Refer Slide Time: 37:55)

So, strains except the dimension of the matrices get increased, because we have 8 nodes,
so we have more degrees of freedom. And derivatives of trial solution with respect to s
and t.

(Refer Slide Time: 38:10)

Again the dimension of G matrix increases, because there are 8 nodes, so once we have
this we can write strain displacement matrix.

(Refer Slide Time: 38:21)

And also which can be written as epsilon is equal to B transpose d in compactly, so then
we can write the element stiffness matrix, again now B matrix becomes more complicate,
a B matrix is more complicated for 8 node case. So, the integrand becomes more
complicated, so better we adopt numerical integration, because it is not possible to obtain
or evaluate the integral in a closed form, where h is thickness element thickness, c is
appropriate constitutive matrix.
(Refer Slide Time: 39:09)

Stiffness matrix is evaluated using Gaussian quadrature 2 by 2 formulas is the lowest

order integration that can be used for this element; you can easily guess why it is the
lowest order of integration that can be used for this element, because we are 8 node
element is the quadratic element. So, if you go back and apply the formula 2 n minus 1
criterion for deciding the number of integration points that are required to exactly
evaluate an integral that we discussed earlier.

You can easily check that, the lowest order integration that can be used for this element
is 2 by 2, for better accuracy 3 by 3 formulas, is 3 by 3 integration is preferred. So, the k
matrix stiffness matrix can be numerical evaluated like this, where rest of the details are
similar to 4 node quadrilateral element, so this is about element stiffness matrix.
(Refer Slide Time: 40:31)

So, what about equivalent load vector, if T x, T y are the components of applied surface
pressure in x and y directions, the equivalent nodal load vector is given by this one. So,
to evaluate equivalent nodal load vector again we require knowing, what is the shape
function matrix along that particular side, along which fractions are specified. Integral
must be evaluated separately for each side along which a pressure is specified, as an
example considers evaluation of Q T, when T is specified alongside 1-2-3.

(Refer Slide Time: 41:19)

So, T is specified along side 1-2-3, so we need to write shape function matrix along side
1-2-3, and then we need to write isoparametric equations and because, we need to know
what is the Jacobian, how differential element in the parent element along that particular
side is related to differential element, in the axial element along that particular side. So,
for that we require isoparametric mapping relation for this side, so that is given by the
shape functions, which are non-zero along that side 1-2-3.

And since, shape functions of other nodes alongside 1-2-3 except nodes 1, 2, 3 are 0, we
can easily obtain shape functions of nodes 1, 2, 3 along side 1-2-3 using Lagrange
interpolation formula. And then subsequently we can write the relationship between x
and local s coordinates system, and y and local s coordinate system.

(Refer Slide Time: 42:32)

And once we have these two relations, we can easily calculate what is partial derivative
of x with respect s, partial derivative of y with respect s, and this schematic shows the
differential element, how differential element in the axial element is related to
differential element in the parent element. So, differential arc length is given by in the
axial element d capital S is given by square root of d x square plus d y square, this is
similar to 4 node quadrilateral element except that now, we need to substitute d x, d y
corresponding to 8 node elements.
(Refer Slide Time: 43:17)

And dividing on both sides with d small s, we get this and now substituting what is d
derivative of x with respect s, derivative of y with respect s in this, we can easily find
what is the relationship between d capital S and d small s are the Jacobian.

(Refer Slide Time: 43:39)

Once we have that relation, we can change the limits of integration and evaluate the
integral using numerical integration, one-dimensional numerical integration; two point
integration is usually considered adequate for evaluating Q T.
(Refer Slide Time: 44:05)

Body force terms can be evaluated in a similar way, all quantities needed for potential
energy functional have now been expressed in terms of nodal unknowns, so finally, we
get k d equal to Q T. So, all these details will illustrate using example 8 node element.

(Refer Slide Time: 44:26)

Here, also we will be taking cantilever plate, all the geometrical details, material details
are given both in FPS units and SI units. And also pressure here, the difference between
the earlier examples, which we have seen, when we are looking at 3 node triangular
element and 4 node quadrilateral element is there we have concentrated forces, where it
is not required for us to evaluate the equivalent nodal load vector by doing integration,
we can directly obtain that from the values of the concentrated forces that are applied.
Whereas here, uniform are applied pressure is applied, so we need to perform
integration, and we will see or we will find an interesting observation when we do that.

So, now let us proceed this is the problem, so we need to find displacements and stresses
in this cantilever plate using 1, 8 node element and pressure is applied on side 5-6-7, and
its applied in the y direction and the value is given both in FPS units and SI units. This 8
node element, this entire cantilever plate is discretized using only one 8 node element
this, and 8 node element, actual element we are going to mapped on to a parent element
which is shown on the right hand side.

And we will be using 3 by 3 Gaussian integration, so total you can guess total 9
integration points are there and I guess, you know, how to obtain the corresponding
weights and coordinates for all this points, so now the calculations will be illustrated for
one integration point. So, similar process can be repeated for the other integration points,
and contribution from all integration points can be added up to get the final element
stiffness matrix.

(Refer Slide Time: 46:53)

So, now let us take 1 integration point, coordinates and weights are given.
(Refer Slide Time: 47:03)

And we require, at that point we require to find what is partial derivative of shape
function vector with respect to shape function derivative; shape function vector
derivative with respect to s and t those details are given here.

(Refer Slide Time: 47:16)

And once we have that, we can easily calculate, what are these quantities.
(Refer Slide Time: 47:28)

And once, we have this we can also calculate what is G, and once we have A and G we
can calculate what is B matrix, and plane stress conditions are assumed here.

(Refer Slide Time: 47:40)

So, using that information we can easily using the material property details, we can
easily write constitutive matrix for this particular problem. So, stiffness matrix at this
Gauss point can now be written by carrying out matrix multiplications; since, nodes 1, 7,
8 have no displacements, only the 10 by 10 matrix associated with nodes 2 through 6 is
written to save space. So, basically the nodes along which are fixed since both degrees of
freedom are constraint. So, finally, if you go to the final reduced equation system the
equations are the rows and columns corresponding to the degrees of freedom of which
are 0 are going to be eliminated, so they are eliminated at this stage itself to save this
space. So, the stiffness matrix contribution, the global reduced equation system from
integration point can be obtained by carrying out matrix multiplications.

(Refer Slide Time: 49:12)

And performing similar calculations for other 8 Gauss points, and adding the resulting
matrices gives final element stiffness matrix.

(Refer Slide Time: 49:17)

And now, since pressure lode is applied along side 5-6-7, let us discuss how to assemble
the equivalent load vector for this applied pressure, so the value of applied pressure is
given here, T x is equal to 0, T y is equal to minus 10,000 psi. And also because, this
pressure load is applied on side 5-6-7, it will result in nodal loads in y direction only at
nodes 5-6-7.

A two point integration is used for evaluating the line integral and also you can see,
when we are trying to write the shape function matrix along side 5-6-7, the shape
functions of all other 5 nodes are going to be 0 along side 5-6-7. So, to simplify writing
only non-zero entries are written.

(Refer Slide Time: 50:43)

So, at integrate two point integration is used to evaluate this line integral to get the
equivalent nodal load vector. So, the first integration point shape function vector, which
comprises of shape function value of node 5, 6, 7 and the derivative of that and Jacobian.
(Refer Slide Time: 51:19)

And finally, evaluating equivalent nodal load vector at integration point one we get this.

(Refer Slide Time: 51:37)

Performing similar calculations at the other integration points and adding the results
together we get, and here the entire vector is written, entire load vector is written.
(Refer Slide Time: 52:01)

And if you look at this one, it is interesting to note that from tributary area concept,
lower at the node based on area tributary to that node; one would expect load at node 6 to
be twice that at nodes 5 and 7.

(Refer Slide Time: 52:25)

But, that is not happening here; however, equivalent nodal load obtained from the work
expression is in fact, smaller at middle node than at the end nodes. The simple tributary
area concept is valid only for linear elements, for higher order elements the distribution
must be determined by integrating appropriate work expressions, may not be intuitively
obvious, so keep this in mind.

(Refer Slide Time: 53:00)

So, we need to perform, so the basic thing is simply by using tributary area concept, if
we do the equivalent nodal load vector is going to be different from that we get by
performing integration of the work expression; complete equations after applying
boundary conditions are given here, complete reduced equation system.

(Refer Slide Time: 53:43)

So, solving this we can find solution at all the 5 nodes, and once we have the
displacements we can calculate stresses, similar to what we did for 3 node linear
triangular element, and 4 node quadrilateral elements.

(Refer Slide Time: 54:00)

Stresses and strains at any point can be calculated by evaluating B matrix at that point by
evaluating B matrix at the centroid, which corresponds to s is equal to 0, t is equal to 0,
this is only for illustration.

(Refer Slide Time: 54:22)

So, this in essence is what plane stress, plane strain problems are or 2D elasticity
problems using finite element method, so we discussed about the finite element
formulation for two d elasticity problems. And afterwards we looked at finite element
equations for triangular element, and 4 node quadrilateral element, and 8 node
isoparametric element; so in next class, we will look at the case of axisymmetric
problems.