INTRODUCTION TO FINITE ELEMENT ANALYSIS

FUNDAMENTALS

1. INTRODUCTION

In their search for appropriate numerical methodology, so as to solve complex problems of

continuum mechanics, Prof. Turner and his associates (1956) from USA, proposed a
computer based finite element method of analysis. Employing their method to the
problems of structural analysis of space vehicles, they demonstrated that the method could
accommodate such aspects as arbitrary geometrical characteristics, non-homogeneous and
anisotropic mechanical constitution and non-linear material behavior of the continuum
into the analysis. The power arising from such capabilities led to the revolution in, the
fields of engineering and technology. in this last four decades the method which is
popularly known as FEM has become the only reliable tool for rational design of
engineering entities.

Finite element in concept is an extremely simple method. It comprises application of three

solution steps. By the way of illustration let us consider a problem of a two dimensional
continuum as shown in FIG 1.1. For the sake of generality let us assume that the
continuum is composed of two zones of material and in each of them there is a cavity or
an opening. Let us assume that the continuum is subjected to in plane forces and its outer
boundary is completely restrained. We are required to determine the deformation
characteristics of the continuum, The finite element solution for the problem proceeds in
the following manner.

Material Zone
B
Material Zone
A

Opening
Or
cavity

Opening
Or
cavity

Restrained
Boundary

Figure 1.1 : TWO DIMENSIONAL CONTINUUM

Step 1

31
The domain to be analysed is transformed into a finite element idealization scheme, as
indicated in FIG 1.2. In this connection the following points are noteworthy.

a) Triangular and quadrilateral plane elements are employed. These are finite
elements.

b) The size and orientation of the finite elements are selected in such a manner that an
element is exclusively composed of only one material. Thus as may be seen from
the figure, at the nodes over the interface between the two material zones, receive
elements of different materials.

NODES

NODES FOR
MATERIALS INTERFACE

FIGURE 1.2 : FINITE ELEMENT IDEALISATION FOR THE

CONTINUUM SHOWN IN FIGURE 1.1

c) The assembly of the finite elements while being interconnected through their
corners or nodes could negotiate sufficiently complex geometrical configurations
of the domain boundary, interface, periphery of the openings etc.

From all these details, we could conclude that though the domain being investigated has
complex geometry and non-homogeneous mechanical constitution, it could be transformed
into relative simple, non-trivial assembly of finite elements inter-connected~ through their
nodes.

Step 2

In the second step the equations governing the equilibrium of the idealized system are
formulated and solved. For the illustrative domain this aspect has following features.

a) The applied loads are transformed into the equivalent loads applied at the
nodes of the idealized system.
b) The idealized system would suffer nodal displacements due to the applied
nodal loads.

32
 c) A linear set of simultaneous equations involving the nodal loads and nodal
displacements is formulated, which ensures the structural nodal equilibrium for
the idealized system.
d) The set of simultaneous equations is solved subject to the boundary conditions,
represented by the zero prescribed displacements for the nodes lying over the
restrained boundary.

In this manner the solution details comprise the displacements suffered by the free nodes
and the reactions induced at the restrained nodes.

Step 3

The solution details are interpreted towards the desired goal. For example in the above
stated problem the goal could be the structural design. Consequently, the nodal response
derived above is utilised towards the evaluation of the stresses and strains induced over
the domain.

Above details pertained to the presentation of the basic philosophy of a finite element
analysis. This philosophy has wide scale applications in the civil engineering and
mechanical engineering practice. The areas of applications thus include, structural
mechanics, soil mechanics, fluid mechanics etc. For each of them the types of analysis
belong to following categories,

a) Linear static analysis

b) Linear dynamic analysis
c) Linear field problems
d) Von-linear static analysis
e) Non-linear dynamic analysis
f) Non-linear field problems

In this session we are concerned with only the linear static, dynamic and field analysis.

2. THEORY OF ELASTICITY

Problems of linear deformations demand considerations to theory of elasticity. The details

relevant to the scope of our study are presented herein.

2.1. Definitions and Notations

We shall consider a case of a three dimensional elastic body. To represent the deformations
we shall employ the Cartesian reference axes (x, y, z) as shown in FIG 2.1.

Let us consider a point A (x, y, z) over the domain of the elastic body undergoing the
linear deformations. Due to the deformation, the point A would be displaced to a position
A ( x+u, y+v, z+w) wherein, u, v and w represent the displacement components in x, y
and z directions respectively as shown in FIG 2.1.

33
 Z B

dz

dx
Y dy
A

FIG 2.1. CONCEPT OF DISPLACEMENTS DUE TO

DEFORMATION

Z
A

u
Y v A

FIG 2.2. CONSIDERATIONS TO AN INFINITISIMAL BLOCK

Differential displacements between two points spaced at infinitisimal distance viz; A and
B as shown in FIG 2.2, induce strains at the point A. The strain at a point could be
represented through six components. viz; three normal components and three shear
components, as shown in FIG 2.3a. We shall denote them through the following notations:

x -Normal strain in x-direction

y -Normal strain in y-direction
z -Normal strain in z-direction
xy -Shear strain over (x, y) plane
xz -Shear strain over (y, z) plane
zx -Shear strain over (z, x) plane

34
Strains would generate stresses, wherein, the stresses at a point could also be represented
through six components, viz. three normal components and three shear components as
shown in FIG 2.3 b. We shall denote them through the following notations.

(a) Strains
z (b) Stresses z

zx yz zx yz
yz
z
yz
zx
zx
y x

xy y xy y
x xy x xy

FIGUR E 2.3 : THR EE DIMENSION AL STR AINS

AND STRESSES

x-Normal stress in x-direction

y-Normal stress in y-direction
z- Normal stress in z-direction
xy-Shear stress over (x,y) plane
yx-Shear stress over (y,z) plane
zy-Shear stress over (Z,X) plane

2.2 Strain vector

Let us consider deformation of a segment ABCD of the size (dx X dy) over (x, y) plane, as
shown in FIG 2.4 Let the point A suffer displacements u and v in x and y directions
respectively and thereby get displaced to the position, A. The remaining corners B, C and
D, would get displaced to the position B, C and D respectively as shown in FIG 2.5.
Assuming the, displacements to be of a small order, we could write.

Change in length (AB) (u + u/x * dx) u u

x = ------------------------------ = ---------------------------- = ---
length (AB) dx x

y = Change in length (AC) = (v + v/y * dy) v v

------------------------------ --------------------------- = ----
length (AC) dy y

v . dx
x v
1 = tan 1 = ------------ = -----
dx x

35
 u . dy
y u
2 = tan 2 = ------------ = -----
dy y

Hence,
u v
xy = 1 + 2 = ---- + ---
y x

In the similar manner, we could also derive the expressions for the strains over (y, z) and
(z, x) planes. Denoting the six strain components through strain vector (), we have

x u/x
y v/y
z = w/z
[]=
xy u/y + v/x
yz v/z + w/y
zx w/x + u/z

u .dy
y
y

2 D
v+ v .dy C
y C D
B v .dx
A 1
x
v
A B

x
u
u+ u .dx
x

FIGURE 2.4 : STRAIN DISPLACEMENT RELATIONSHIP

2.3 Stress Vector

Applying Hookes law

x = x/E - y/E - z/E

36
y = - x/E + y/E - z/E
z = - x/E - y/E + z/E
xy = xy/G
yz = yz/G
zx = zx/G

wherein

E is the modulus of elasticity of the material

is the poissons ratio of the material
G is the shear modulus of the material such that
G = E/2 (1 + )

We shall denote the stress components through a stress vector [ ], as defined below ;

x
y
[] =
z
xy
yz
zx
Then,

[] = [A] []

Wherein,
1/E - /E - /E
- /E 1/E - /E
[A]= - /E - /E 1/E
1/G
1/G
1/G

In view of the above details, we could also write

[] = [A] 1 [E] = [C] [E]

Denoting

E(1-)
C1 = ---------------------
(1+) (1-2)

E
C2 = ---------------------- ( 2.4 )
(1+) (1-2)

37
 C3 = G = E
--------------
2(1+)

[C] the elastic coefficient matrix turns out to be,

C1 C2 C2
C2 C1 C2
(C ) = C2 C2 C1 (2.5)
C3
C3
C3

2.4 Plane strain Deformations

More often then not, we encounter a situation, wherein out of three dimensions of the
elastic body, one of the dimension is very large compared to the other dimensions. In such
cases, we could simplify the situation, by treating it as a two dimensional plane strain
continuum. For example, consider an earth dam, whose height and width are measurable
in meters, whereas length runs into kilometers. As the length of the dam is too large
compared to its sectional dimensions, the deformations in the longitudinal direction are too
insignificant to need considerations. We could therefore assume that the displacements and
strains are confined to only transverse vertical planes. The problem of this kind is referred
to as plane strain deformation problem,

Assuming that the length of the dam is oriented in Z direction, we should note that W = 0,
and (u,v) are functions of only (X,Y). Consequently, the strain components (z, yz, zx) are
zero. In other words, the strain components in only (x,y) plane, viz. (x, y, xy) exist. In
view of this, we consider a section of the dam, with a thickness of 1.0m in the longitudinal
direction, and refer it to (x,y) axes as shown in FIG. 2.5. We could note the following
features of such plane strain deformations.

38
 y

1m

v x
u

FIGURE 2.5 : STATE OF PLANE STRAINS

a) The displacements have components u and v in x and y directions respectively.
b) The strain vector [] and the stress vector [] has components as shown in FIG 2.6
and these vectors are as defined in Eq 2.6.

x u/x x (2.6)
() = y = v/y ; () = y
xy u/y + v/x xy

c) To derive the relationship between [] and [E],

we shall substitute z = zx = yz = 0 into Eq. 2.3 consequently in the relationship [] = [c]
[E], [C] has the structure as defined in Eq 2.7.

C1 C2 0
[C] = C2 C1 0
0 0 C3
y
y
xy
xy

x x
x x

y
xy
xy
y
x y

39
 (a) Strains (b) Stresses

(d) Even though the normal strain x=0, the normal stress z exists; because for z = 0,

z = - x - y + z = 0
E E E

Or z = (x + y )

However, the shear stress components yz = zx = 0. This means z the normal stress in Z
direction is one of the principal stresses.

(e) In case of the three dimensional deformations, there exists three principal stresses.
1 the major principal stress, 2 the intermediate principal stress and 3, the
minor principal stress. In view of the plane strain deformations, 2 turns out to be
the normal stress 2.

Employing the concept of Mohr's stress circle, we could prove that (1, 3) and the
angle made by the major principal plane with the x axis are as defined in Eq. 2.8.
1 = (x + y + R)

2 = (x + y - R)

= tan-1 2 xy
y - x

R = (y x)2 + 4xy2

3. LINEAR PLANE STRAIN FINITE ELEMENT ANALYSIS - STATIC ACTION

3. 1 INTRODUCTION

Finite element analysis of plane strain deformations involves three basic steps;

a) Finite element idealization of the system being analyzed.

b) Formulation and solution of equations governing equilibrium of the idealized

system.

c) Evaluation of structural response comprising the variation of displacements,

strains, stresses etc. over the idealized system.

We shall discuss herein the methodology pertaining to these basic steps of analysis. For
the purpose of illustration we shall consider the problem of plane strain deformation in a
triangular earth embankment resting over hard strata, due to the self weight of the
embankment material. The details pertaining to sectional dimensions and material
properties are presented in FIG 3. 1.

40
- Modulus of elasticity = 1000.0 T/m2

- Poissions Ratio = 0.4

D- Density = 2.0 T/m3

2 2

1 1 20 m

80 m

FIGURE 3.1 : ILLUSTRATIVE PROBLEM DATA DETAILS

3.2 FINITE ELEMENT IDEALIZATION

A finite element analysis begins with a proposal of finite element idealization scheme for
the problem under investigation. Chief objective behind such an exercise is to transform
a complex continuum into an assembly of relatively simple elements called finite
elements. Concept of finite element idealization is however not new. It happens to be
logical extension of the idealization technique employed in conventional structural
analysis. For example a plane frame is analyzed by representing it through the assembly of
line segments as shown in FIG 3.2. The line segments are analogous to finite elements and
their assemblage is similar to finite element idealization.

In the problem of plane frame the choice of line segments as elements and their
assemblage to represent the frame is obvious, the later automatically simulating the joints
between the frame members. In case of a continuum such as the one represented by the
embankment section, however, the development of finite element idealization is not that
obvious. In theory any arbitrary division of the continuum should provide a basis for the
development of its finite element idealization.

------ CENTRE LINE NODES OR JUNCTIONS LINE

SEGMENTS

41
 a) PLANE FRAME b) IDEALISED FRAME

Figure 3.2 Concept Of Finite Element Idealization

This implies availability of indefinite number of alternative finite element idealization

schemes. For the problem of the embankment, the choice is however restricted to only
triangular and quadrilateral forms. In FIG 3.3, typical details are presented,

(a) ELEMENTS CHARACTERISED BY

TRIANGULAR FORMS

42
 (b) ELEMENTS CHARACTERISED
BY QUADRILATERAL FORMS

From amongst these element forms the three noded triangular element is most simple. In
addition the experience till date has demonstrated its universality. In any case, the three
noded triangular element form is most convenient as a tool for demonstrating the finite
element methodology. In view of this, we shall idealize, the triangular embankment into
an assembly of four triangular elements, interconnected through six nodes as shown in
FIG 3.4.

3.3. Essence Of Method

The plane strain finite element analysis of the! idealized embankment has following
features.

3.3.1 Definitions and Notations

We shall refer the idealized system to (x,y) cartesian axes as shown over FIG 3.4

At any point (x,y) over the system there exists two degrees of freedom. This is represented
by displacements u and v in x and y directions respectively.

A load acting at a point (x,y) could be represented through its components Fx and Fy in
x and y directions respectively.

43
 y

4 Fy 10 m

Fx
4 5

y
1 2 10 m
3
u
1 2 3 x
40 m 40 m

0 Nodes 1 to 6 . Elements (1) to (4)

FIG 3.4 : FINITE ELEMENT IDEALISATION SCHEME

FOR ILLUSTRATIVE EMBANKMENT

These details are illustrated over Fig. 3.4

3.3.2 Element characteristics

We shall denote an element through its nodes numbered (i, ii, iii) in an anticlock-wise
fashion as shown in FIG 3.5 correspondingly the element nodal displacements and the
element nodal forces are as indicated over the figure. We shall denote them.

Vii Fyii
Viii Fyiii
uii Fxii
ii
uiii ii iii Fxiii
iii Fyi
Vi
ui Fxi
i
i

(a) Element Nodal (b) Element Nodal

Displacements forces

Figure 3.5 Element Characteristics

through the element nodal displacement vector [Se] and element nodal force vector [Fe]
as defined in Eq.3.1

44
 ui Fxi
vi Fyi
uii Fxii
[e] = vii [Fe] = Fyii
3.1
uiii Fxiii
viii Fyiii

The element equilibrium is defined by Eq 3.2.

wherein (Ke) is element stiffness matrix.

We should note that the element has (3 x 2 = 6) 6 nodal degrees of freedom.

Consequently, (Ke) is a (6 x 6) matrix. Further, in view of the reciprocal load deflection
behavior (Ke) would be symmetrical matrix.

3.3.3. Charactristics of the idealized system

Structural nodal displacements and structural nodal forces are as shown in FiG.3.6.

v6

6
u
6

v4 v5

u
4 u4 5
5

v1 v2 v3

1 u3
u1 2 u2 3

(a) Structural Nodal Displacements

45
 Fy6

6
Fx6
Fv5
Fy4

4
Fx4 5
Fx5

Fy1 Fy2
Fy3

1 Fx3
Fx1 2 Fx2 3

(a) Structural Nodal Forces

FIGURE 3.6 CHARACTERISTICS OF THE IDEALIZED SYSTEM

Denoting the nodal displacements and the nodal forces for the idealized system through
the vectors (S) and (F) respectively, these vectors are as defined in Eq.3.3

46
 u1 Fx1
v1 Fy1
u2 Fx2
v2 Fy2
u3 Fx3
v3 Fy3
[S] = u4 ; [F] = Fx4
v4 Fy4 (3.3)
u5 Fx5
v5 Fy5
u6 Fx6
v6 Fy6

Equilibrium of the idealized system is defined by Eq.3.4

[K] [S] = [F] (3.4)

wherein [K] is structural stiffness matrix. We should note that the idealized system has
(6x2=12) 12 nodal degrees of freedom. Consequently, [K] is a (12 x 12) matrix. Further
like [Ke] [K] is also a symmetrical matrix. [K] is derived by super positioning of [Ke] of
the elements meeting at the nodes, whereas [F] is derived by super positioning of [Fe] of
the elements meeting at the nodes of the idealized system,

3.3.4. Solution of Equations

Eq. 3.4. represent a set of 12 simultaneous equations, with 12 unknown nodal parameters,
The unknown parameters are as follows:

a) As the embankment rests over unyielding strata, the base-nodes 1,2 and 3 lying over the
strata are completely restrained. Consequently, u1 = v1 = u2 = v2 = u3 = v3 = 0.0. These
nodes would however be subjected unknown reactive force components (
equation ) due to the nodal restraints. In this the suffix R denotes reaction,

b) Nodes 4,5 and 6 are free to displace due to the self weight of the embankment. Thus the
six displacement components (u4, v4, u5, v5, u6, v6) are unknown. On solving Eq.3.4 the
six unknown displacement components and the six unknown reactive components are
derived.

3.3.5. Structural Response

With the nodal displacements ( ) being completely known, we could derive detailed
structural response of the idealized system. This may include evaluation of the stresses
and strains induced in the embankment section.

3.4. STRUCTURAL CHARACTERISTICS OF A PLANE STRAIN

TRIANGULAR ELEMENT

Eq.3.4 defining the equilibrium of the idealized system ,could be derived through the
super positioning of Eq.3.2,defining equilibrium of1the elements meeting at the nodes of

47
the system. It therefore follows that in the formulation of Eq.3.4, we require the
coefficients of [Ke] and [Fe] of the elements present in the idealized system. In this
section, we shall derive them;

3.4.1 - Interpolation Functions or Shape Functions

Displacements (u, v) suffered by a point (x, y) over the element (FIG 3.7.a), could be
expressed as the functions of the nodal displacements as defined in Eq.3.5
P iii
u Niui Niiuii Niiiuiii N u
P i
p p

P iii
V NiVi NiiVii NiiiViii N V
P i
p p

wherein; (Ni, Nii, Niii) are called the interpolation functions or shape functions.
y

iii
v

u
(x,y)

FIG. 3.7 DISPLACEMENTS @ (x, y,)

We shall derive the expressions for (Ni, Nii, Niii) by assuming the variation of (u, v)
over the element domain i.e. assume the displacement function. The choice of
displacement function, however, cannot be arbitrary. We should assume only such a
displacement function, that satisfies the inter element compatibility between the
neighboring elements of the idealization scheme. By this we mean that the element
displacement field should be such that due to deformation the neighboring elements
neither overlap, nor develop gaps at their common edges as indicated over FIGS 3.7 (b)
and 3.7 (c)
OVERLAP
GAP

FIGURE 3.7(b) OVERLAP FIGURE 3.7 (c) GAP

48
The inter element compatibility is easily satisfied by assuming linear variation of (u, v) as
defined in Eq.3.6.

u = a1 + a2x + a3y
v = a4 + a5x + a6y

Wherein (a1 , a2, a6) are arbitrary constants.

The displacement functions defined by Eq.3.6 signify that the element will deform as
shown in FIG 3.7 (d). This means the element maintains its triangular shape even after
deformation; consequently, the inter-element compatibility between the neighboring
elements gets automatically satisfied.

UNDEFORMED
iii
DEFORMED iii

ii

i
ii

FIG. 3.7 Deformation Characteristics Of 3 Noded Triangular Element

Let us consider the first of Eqs. 3.6 and substitute in it the conditions at the element
nodes. This leads to Eq. 3.7

[ue] = [A] [a]

wherein, the matrices [ue], [A] and [a] are as shown below :-

ui 1 xi yi a1
[ue] uii ; [A] 1 xii yii ; [a] = a2 (3.8)
uiii 1 xiii yiii a3

From Eq. 3.7 it follows that ; [a] = [A] 1[Ue] (3.9)

Wherein (A) 1 is as shown in Eq. (3.10)

49
 ai aii aiii
[A]-1 = bi bii biii (3.10)
ci cii ciii

So that;

= Area of the element (i, ii, iii)

ai = (xii, yiii, -yii, xiii)
bi = (yii - yiii), etc.
ci = - (xii xiii);etc.

Denoting;

[P] = 1 X y

We can write

u = [P] [a]

From Eq. 3.9 and Eq. 3.12

u = [P] [A] 1 [ue] = [L] [ue]

Wherein;

[P] [A] 1 = [L] = Li Lii Liii

Performing the matrix multiplication, we shall find that

(ai + xbi + yci)

Li = -------------------
2

(aii + xbii + ycii)

Lii = ---------------------
2

(aiii + xbiii + yciii)

Liii = ----------------------
2
comparing with Eq.3.13 with the first of Eq.3.5, we find that; N i = Li, Nii = Lii, Niii = Liii.
The interpolation functions or shape functions Np are as defined in Eq.

(ap + xbp + ycp)

Np = ----------------------
2

For Np we shall arrive at the same result by considering, second of (3.6)

50
(Li, Lij Liii) presented in Eq. (3.15) are the well known area co-ordinates and as such they
have a physical interpolation For example at a point A = (x, y) shown in Fig 3.8. We can
prove that;

Li = Area (A-ii-iii) /

Lii = Area (A-iii-i) / (3.17)

Liii = Area (A-i-ii) /

iii

ii

FIG.3.8 CONCEPT OF THE AREA COORDINATES

Obviously ; Li + Lii + Liii = 1

In view of the fact that Nps and Lps are identical, it also follows that;
P iii
Ni Nii Niii Np 1 (3.19)
p 1

Even otherwise the property of interpolation functions as represented by Eq. 3,19 has
obvious validity. These functions are in fact the nodal weightage factors, whose sum has
got to be equal to 1, to permit appropriate inter-polations.

Further, we can also prove that

p iii p iii
x L x
p i
p p N x
p i
p p

p iii p iii
y Lpyp
p i
N y
p i
p p

(3.20)

From Eqs. 3.5 and 3.20, it is clear that the co-ordinates of a point, and the displacements at
that point are related to their nod counterparts, through the same set of interpolation (or
shape) functions. The elements with characteristics are called "Isoperimetric Elements.
The three noded triangular element considered herein is a linear isoperimetric triangular

51
element. The other element forms, with nodes over the element sides or even internal
nodes as shown in FIG. 3.3 are also isoperimetric elements. The order of these elements
however increase with the increase in the number of nodes. For example, the six noded
triangular element and eight noded quadrilateral element are quadratic' isoperimetric
elements i.e. they permit assumption of quadratic variations for (x, y, u, v) over the
element domain.

3.4.2 Element strain and Element Stresses

Basic features of the theory of elasticity pertaining to the plane strain deformations have
already been presented earlier. For the sake of ready reference, however, the relevant
details are repeated herein.

a) strains

In a plane strain field over the element domain, the strain at a point has, normal
components, x and y in x and y directions respectively, and shear strain xy over the xy
plane. The strain components are denoted through a vector (E] and the same are related to
the derivatives of the displacements (u, v). All these details are defined in Eq.3.21.

x u/x
[] = y v/y
xy u/y + v/x

From Eqs. 3.5 and 3.16, it follows that

p iii p iii
u 1 u 1
y

2
bpup :
p i y

2
c u
p i
p p

p iii p iii
u 1 v 1
y

2
b p vp :
p i y

2
c v
p i
p p

(3.22)

wherein (bp, cp)could defined as per Eq.3.11

From Eqs. 3.21 and 3,22, we have,

[] = [B] [e]

wherein [e] is the element nodal displacement vector and [b] is element strain
transformation matrix. Details of (B) are shown in Eq.3.4

bi 0 bii 0 biii 0
[B]=1 / (2*) 0 Ci C Cii 0 Ciii (3.24)

52
 Ci bi Cii bii Ciii biii

We should note that [B] the strain transformation matrix is composed of only constant
coefficients, therefore, the three noded triangular element is also referred to as a constant
strain triangular (CST) element.

(b) stresses

In a plane strain field over the element domain, the stress at a point has, normal
components, x and y in x and y direction respectively and shear stress xy over the XY
plane. The stress components are denoted through a vector [] and the same are related to
the strain vector [], through the elastic deformation coefficients. All these details are
presented below :-

[] = [C] [] = [C] [B] [ e] (3.25)

C1 C2 0
[C] = C2 C1 0
0 0 C3

Wherein ;
E (1-)
C1 = ---------------------------
(1+) ( 1-2)

E
C2 = ---------------------------
(1+) ( 1-2)

E
C3 = ---------------------------
2(1+)

In Eq, 3,25 as both [B] and [C] are composed of constant coefficients, the element may
also be called constant stress element.

3.4.3 Element Stiffness Matrix

Employing the principles of virtual work, we shall derive the element stiffness Matrix
[Ke] in the following manner.

Let the element be subjected to the virtual displacements [e]. The WE the external virtual
work done on the element is as defined in Eq.3.27.

53
WE e T [ Fe ] (3.27)

wherein [Fe] is the element nodal force vector already defined.

An elementary area (dx*dy) of the element would suffer the strain [] and stress []. The
internal work dwI stored within the elementary area is given by;

dwI = []T [] dx.dy.1 (3.28)

`1 in this equation is due to the 1m thickness assumed as the unit thickness of the dam
section.

Thus, WI total internal virtual work is give by;

WI dw []
T
I [ ]dxdy (3.29)
A A

wherin A denotes the integration over the element area, substituting for [] and [] from
Eq. 3.23 and 3.25 respectively in to Equation 3.29;

WI [e]T [ B]T [C ][ B][e]dxdy

A

As WE = WI, due to the condition of equilibrium, from Eqs. 3.27 and 3.30 it follows that,

T [C ][ B ][e]dxdy
[ Fe] [ B]
A

However, [Ke] = [] = [Fe], therefore;

T [C ][ B ]dxdy
[ Ke] [ B]
A

Eq. 3.31 represents a standard finite element statement defining [Ke] of an element. The
same equation with appropriate character of [B] and [C] are composed of constant
coefficients, hence,

[ Ke] [ B ]T [C ][ B ]

wherein, represents the area of the triangular element

3.4.4. Element Nodal Force Vector

In the finite element analysis, applied loads will have to be in the form of nodal forces.
The load due to self weight of the element which happens to be in form of distributed
body force, will therefore be required to be transformed into equivalent nodal load vector
[Fe].

Let us consider a general problem of body forces in which x and y are the intensities of
the body forces per unit volume, in x and y directions respectively, Let [F e] represent

54
valent element nodal forces, due to the distributed body forces integrated over volume of
the element. We subject the

y T/3
y y
ELEMENTARY x T/3
VOLUME (dx*dy*1)
y y T/3
x x T/3
y T/3

x T/3

(a) x (b) x
y

/3

/3

/3
(c) x

Figure 3.9

element nodes to virtual displacement [e], then the external work WE is given by Eq.3.33

WE [e] [ Fe] T
(3.33)

An elementary area (dx *dy *1) of the element is acted upon by body forces X and Y as
shown in Fig 3.9(a), such that;

x x dxdy
y y dxdy (3.34)

The internal work dwB over the elementary volume, due to these body forces is given in
Eq. 3.35

dwB x.u y.v ( x u y v ) dxdy (3.35)

wherein (u, v) are the displacements at the center of the elementary volume. From Eqs 3.5
and 3.35 we find that ;

p iii p iii
dwB [ x N p u p y N p u p ]dxdy (3.36)
p i p i

Eq. 3.36 is expressed in a more convenient form as shown below

55
dwB = [e]T[N] T[] (3.37)

wherein (N) and () are as defined in Eq. 3.38

Ni 0 Nii 0 Niii 0

[N] =
O Ni 0 Nii 0 Niii

x
[]= (3.38)
y

Now, WB the internal work done due to the body forces is given by Eq. 3.39

WB dwB [e]T [ N ]T [ ]dxdy (3.39)

A A

As WE = WB, from Eqs. 3.33 and 3.39 it follows that

[ Fe] [ N ] [ ]dxdy
T
(3.40)
A

with (x, y) varying over the element domain, Eq. 3.40 could be tackled through an
appropriate numerical integration scheme. We shall discuss this aspect in greater detail at
a later stage. Presently, we assume x and y to be constant. In that case, [F e] of Eq.3.40
is as defined in Eq. 3.41

x
y
[Fe] = x
3 y
x
y

The above result merely confirms to the laws of statics. This may be seen from the fact
that the total body forces XT and YT are XT = x and YT = y acting through the C.
G. of the element and by the laws of statistics it follows that this is equivalent to the
forces XT/3 and YT/3 acting at each node of the element as shown in Figure 3.9 (b)

In case of the body forces due to the self weight of the element , x = 0, and y =
wherein is the soil density. is assigned a -ve sign because the weight acts in negative
y direction. From Eq.3.41, it therefore follows that, in case of the self weight of the
element;

0
-
(3.42)

56
 0
[Fe] = 3 -
0
-

The above forces are shown over Fig 3.9 (c)

3.5 Illustrative Computations

We shall present herein complete set of details pertaining to the methodology for the finite
element solution technique. For this, the illustrative embankment of Fig 3.1 is idealized
into an assembly of four train angular elements, inter-connected through six nodes as
shown in Fig 3.10.

y
6

4 5

1 2 3

1 2 3 x
Fig 3.10 : Idealization Scheme for Illustrative Embankment

3.5.1 Description of the Idealisation Scheme

The scheme of idealization is described through nodal co-ordinates, element material and
nodal connectivity and the boundary conditions.

(a) Nodal co-ordinates

NODE x-(m) Y-(m)

1 0.00 0.00
2 40.00 0.00
3 80.00 0.00
4 20.00 10.00

57
 5 60.00 10.00
6 40.00 20.00

(b) Element details

ELEMENT NO. NODES MATERIAL NO.

1 1 2 4 1
2 2 5 4 1
3 2 3 5 1
4 4 5 6 1

Note : Besides element nodal connectivity (i, ii, iii), it is necessary to qualify the material
number for the element. In the illustrative problem considered herein, all the
elements of the idealization are of the same material, denoted by material No.1

(c) Boundary Conditions

As the dam is resting over a hard strata, the base nodes (1,2,3) are completely restrained,
therefore

u1=0, v1=0, u2=0, v2=0, u3=0, v3=0

3.5.2. Element Characteristics

The element characteristics comprise [B], [C] [K e] and [Fe] of the elements. These are
derived through appropriate equations of section 3.4

(a) Strain Transformation matrix [B] (Eq.3.24)

-0.025 0 0.025 0 0 0
[B]1 =
0 0.05 0 -0.05 0 0.10

-0.05 -0.025 -0.05 0 0.10 0

0 0 0.025 0 -0.025 0
[B]2 =
0 -0.10 0 0.05 0 0.05

-0.10 0 0.05 0.025 0.05 -0.025

-0.025 0 0.025 0 0 0
[B]3 =
0 -0.05 0 -0.05 0 0.10

58
 -0.05 -0.025 -0.05 0.025 0.10 0

-0.025 0 0.025 0 0 0
[B]4 =
0 -0.05 0 -0.05 0 0.10

-0.05 -0.025 -0.05 0.025 0.10 0

Note : In the above tables, outer suffix to [B] denote element number

(b) Elastic Coefficient Matrix [c] (eqs. 3.25, 3.26]

As the material of all the elements is dentical, their [c] matrix are also identical. For all of
them E=1000 t/m2 , = 0.4. Hence,

E(1-) 1000 x (1 0.4)

C1= ------------- = ----------------------- = 2143 T/m2
(H ) (1-2) (1+0.4) (1-0.8)

E 0.4 x1000
C2= ------------- = ----------------------- = 1429 T/m2
(H ) (1-2) (1+0.4) (1-0.8)

E 1000
C3= ------------- = ----------------------- = 357 T/m2
2 (1-2) 2 (1+0.4)

consequently, for all the elements;

2143 1429 0
[C] = 1429 2143 0
0 0 357

(c) Element stiffness Matrix (Ke) (Eq.3.30)

[Ke]1 = [B]1T [C] [B]1

448 448 -88 268 -356 -716
448 1118 -268 1026 -180 -2144
-88 -268 448 -448 -356 716
268 1026 -448 1118 180 -2144
-356 -180 -356 180 716 0

59
 -716 -2144 716 -2144 0 4288

[Ke]2 = [B]2T [C] [B]2

716 0 -356 -180 -356 180
0 4288 -716 -2144 716 -2144
-356 -716 448 448 -88 268
-180 -2144 448 1118 -268 1026
-356 716 -88 -268 448 -448
180 -2144 268 1026 -448 1118

[Ke]3 = [B]3T [C] [B]3

448 448 -88 268 -356 -716
448 1118 -268 1026 -180 -2144
-88 -268 448 -448 -356 716
268 1026 -448 1118 180 -2144
-356 -180 -356 180 716 0
-716 -2144 716 -2144 0 4288

[Ke]4= [B]4T [C] [B]4

448 448 -88 268 -356 -716
448 1118 -268 1026 -180 -2144
-88 -268 448 -448 -356 716
268 1026 -448 1118 180 -2144
-356 -180 -356 180 716 0
-716 -2144 716 -2144 0 4288

Note In the above tables, the outer suffix to [B] and [Ke] denote the element No.

(d) Element Load Vector [Fe] (Eq.3.42)

As all the elements have identical material and the same element area, the coefficients of
[Fe] for all of them are same. Now, = 2 t/m3 and = 200 m2 , hence [Fe] for all of them
is as given below :

0
-400/3
0
[Fe] =
-400/3
0
-400/3

3.5.3 Characteristics of Idealized section

[K] and [F] of the idealized section are derived through super positioning of [K e] and [Fe]
of the elements meeting at the nodes of the idealized section. This involves placement of

60
co-efficients of [Ke] and [Fe] in [K] and [F] respectively, followed by algebric summation
of the co-efficients so placed.

(a) Element Equilibrium

Element equilibrium is defined by [K e] [e] = [Fe], in which the co-efficients of [Ke] and
[Fe] are associated with the element nodal degrees of freedom, represented by [e], we
should note that [e] of an element (i, ii, iii) is as defined in Eq.3.1. Consequently, the
information presented till now allow the formulation of equations of element equilibrium,
as per the details shown in Table 3.1

Table 3.1 : Element Equilibrium

Element (1) : [Ke]1[e]1 = [Fe]1

i ii iii [e]1 [Fe]1

1 2 4

448 448 -88 268 -356 -716 u1 Fx1 0

448 1118 -268 1026 -180 -2144 v1 Fy1 -400/3

-88 -268 448 -448 -356 716 u2 = Fx2 0
268 1026 -448 1118 180 -2144 v2 Fy2 -400/3
-356 -180 -356 180 716 0 u4 Fx4 0
716 -2144 716 -2144 0 4288 v4 Fy4 -400/3

[Ke]1

61
Element (2) : [Ke]2[e]2 = [Fe]2

i ii iii [e]2 [Fe]2

2 5 4

716 0 -356 -180 -356 180 u2 Fx2 0

0 4288 -716 -2144 716 -2144 v2 Fy2 -400/3
-356 -716 448 448 -88 268 u5 Fx5 0
-180 -2144 448 1118 -268 1026 v5 = Fy2 -400/3
-356 716 -88 -268 448 -448 u4 Fx4 0
180 -2144 268 1026 -448 1118 v4 Fy4 -400/3

[Ke]2

Element (3) : [Ke]3[e]3 = [Fe]3

i ii iii [e]3 [Fe]3

2 3 5

448 448 -88 268 -356 -716 u2 Fx2 0

448 1118 -268 1026 -180 -2144 v2 Fy2 -400/3

-88 -268 448 -448 -356 716 u3 = Fx3 0
268 1026 -448 1118 180 -2144 v3 Fy3 -400/3
-356 -180 -356 180 716 0 u5 Fx5 0
-716 -2144 716 -2144 0 4288 v5 Fy5 -400/3

[Ke]3

62
Element (4) : [Ke]4[e]4= [Fe]4

i ii iii [e]4 [Fe]4

4 5 6

448 448 -88 268 -356 -716 u4 Fx4 0

448 1118 -268 1026 -180 -2144 v4 Fy4 -400/3

-88 -268 448 -448 -356 716 u5 = Fx5 0
268 1026 -448 1118 180 -2144 v5 Fy5 -400/3
-356 -180 -356 180 716 0 u6 Fx6 0
-716 -2144 716 -2144 0 4288 v6 Fy6 -400/3

[Ke]4

(b) Placement and structural equilibrium

Structural equilibrium is defined by [K] [] = [F], wherein [] and [F] are as shown
below :

u1 Fx1
v1 Fy1
u2 Fx2
v2 Fy2
u3 Fx3
[]= V3 ; [F] = Fy3
u4 Fx4
v4 Fy4
u5 Fx5
v5 Fy5
u6 Fx6
v6 Fy6

The information in Table 3.1 therefore permit placement of [K e] and [Fe] into [K] and [F]
respectively. The details are presented in Table 3.2. The algebraic sum of the
coefficients in [K] and [F] of various tables of Table 3.2, defines [K] and [F] of the
structural equilibrium equations represented by [K] [] = [F]. The result pertaining to
this is given in Table 3.3

(c) Solution of Equations

With [K] and [F] derived, 12 simultaneous equations defining condition of equilibrium
and involving 12 unknowns are formulated. These are unknown nodal reactions at nodes
(1,2,3) and unknown nodal displacements at nodes (4,5,6). In the basic equation of
equilibrium [K] remains as formulated, but [] and [F] acquires character as shown below
:

63
Table 3.2 : Placement of [Ke]1 and [Fe]1 into [K] and [F] respectively

1 2 3 4 5 6 [] [F]

448 448 -88 268 -356 -716 u1 Fx1 0

448 1118 -268 1026 -180 -2144 v1 Fy1 -400/3
-88 -268 448 -448 -356 716 u2 Fx2 0
268 1026 -448 1118 180 -2144 v2 Fy2 -400/3
u3 = Fx3
V3 Fy3
-356 -180 -356 180 716 0 u4 Fx4 0
-716 -2144 716 -2144 0 4288 v4 Fy4 -400/3
u5 Fx5
v5 Fy5
u6 Fx6
v6 Fy6

[K]

64
Table 3.2: Placement of [Ke]2 and [Fe]2 into [K] and [F] respectively

1 2 3 4 5 6 [] [F]

u1 Fx1
v1 Fy1
716 0 -356 180 -356 -180 u2 Fx2 0
0 4288 716 -2144 -716 -2144 v2 Fy2 -400/3
u3 = Fx3
V3 Fy3
-356 716 448 -448 -88 -268 u4 Fx4 0
180 -2144 -448 448 268 -1026 v4 Fy4 -400/3
-356 -716 -88 448 448 448 u5 Fx5 0
-180 -2144 -268 1026 448 1118 v5 Fy5 -400/3
u6 Fx6
v6 Fy6

[K]

65
Table 3.2: Placement of [Ke]3 and [Fe]3 into [K] and [F] respectively

1 2 3 4 5 6 [] [F]

u1 Fx1
v1 Fy1
448 448 -88 268 -356 -716 u2 Fx2 0
448 1118 -268 1026 -180 -2144 v2 Fy2 -400/3
-88 -268 448 -448 -356 716 u3 = Fx3 0
268 1026 -448 1118 180 -2144 V3 Fy3 -400/3
u4 Fx4
v4 Fy4
-356 -180 -356 180 716 0 u5 Fx5 0
-716 -2144 716 -2144 0 4288 v5 Fy5 -400/3
u6 Fx6
v6 Fy6

[K]

66
Table 3.2: Placement of [Ke]4 and [Fe]4 into [K] and [F] respectively

1 2 3 4 5 6 [] [F]

u1 Fx1
v1 Fy1
u2 Fx2
v2 Fy2
u3 = Fx3
V3 Fy3
448 448 -88 268 -356 -716 u4 Fx4 0
448 1118 -268 1026 -180 -2144 v4 Fy4 -400/3
-88 -268 448 -448 -356 716 u5 Fx5 0
268 1026 -448 1118 180 -2144 v5 Fy5 -400/3
-356 -180 -356 180 716 0 u6 Fx6 0
-716 -2144 716 -2144 0 4288 v6 Fy6 -400/3

[K]

67
Table 3.3: Equations of Structural Equilibrium
[K] [] = [F]

448 448 -88 268 0 0 -356 -716 0 0 0 0 u1 0

448 1118 -268 1026 0 0 -180 -2144 0 0 0 0 v1 -400/3
-88 -268 1612 0 -88 268 -712 896 -712 -896 0 0 u2 0
268 1026 0 6524 -268 1026 896 -4288 -896 4288 0 0 v2 -400/3
0 0 -88 -268 448 -448 0 0 -356 716 0 0 u3 = 0
0 0 268 1026 -448 1118 0 0 -180 -2144 0 0 V3 -400/3
-356 -180 -712 896 0 0 1612 0 -176 0 -356 -716 u4 0
-716 -2144 896 -4288 0 0 0 6524 0 2052 -180 -2144 v4 -400/3
0 0 -712 -896 -356 -180 -176 0 1612 0 -356 716 u5 0
0 0 -896 4288 716 -2144 0 2052 0 6524 180 -2144 v5 -400/3
0 0 0 0 0 0 -356 -180 -356 180 716 0 u6 0
0 0 0 0 0 0 -716 -2144 716 -2144 0 4288 v6 -400/3

[K] [] [F]

68
 Node No. [] REMARK Node [] REMARK
No.

BOUNDARY CONDITIONS
0.0 0+ FX1R

UNKNOWN FORCES
1 0.0 1 -400/3 + Fy1R
0.0 0+ FX2R
2 0.0 2 -400/3 + Fy2R

0.0 UNKNOWN DISPLACEMENTS TO BE DETERMINED KNOWN DISPLACEMENTS 0+ FX3R

0.0 -400/3 + Fy3R

KNOWN FORCES TO BE DETERMINED

3 3

u4 0
4 v4 4 -400
u5 0
5 5
v5 -400
u6 0
v6 -400/3

6 6

Many methods are available for solution of simultaneous equations. Of these, the methods
based on Gausss elimination scheme are most efficient Frontal solution technique
incorporated in the standard finite element programme employs the Gausss elimination
technique. Its details are covered in the new chapter, wherein solution of the above set of
equations is carried out. For further presentation in this chapter the final results from the
solution details are produced herein.

69
(1) Unknown nodal displacements

u4 = -0.0504 m ; u5 = 0.0504m ; u6 = 0.0 m

v5 = -0.0781 m ; v5 = 0.0781m ; v6 = 0.1261 m

(2) Unknown nodal reactions

FX1R = 73.86 tons ; FX2R = 0.0 ; FX3R = 78.86 tons

Fy1R = 310.75 tons ; Fy2R = 979.5 tons ; Fy3R = 310.75 tons

In Fig 3.11, the nodal reactions and the deformed section due to the nodal displacements
are shown.

DEFORMED PROFILE

4 5

1 2 3
73.86 t 73.86 t

310.75 t 979.5 t 310.75 t

Figure 3.11 : Response of the idealized section

(d) Structural Response

We are now in position to derive the detailed structural response

(1) Element nodal displacement vectors [e]1 , [e]2, [e]3 and [e]4 are as shown in
table 3.4

Table 3.4 : Element nodal displacements

Node [e]1 Node [e]2 Node [e]3 Node [e]4

No. No. No. No.

70
 1 0.0 2 0.0 2 0.0 4 -0.0504
0.0 0.00 0.0781
2 0.0 5 0.0504 3 0.0 5 -0.0504
0.0 -0.0781 0.0 -0.0781
3 -0.0504 4 -0.0504 5 0.0504 6 0.0
-0.0781 -0.0781 -0.0781 -0.1261

Note Outer Suffix to [e] denotes the element no.

2) Calculation of element strains and element stresses require [B], [C] and [e] of the
elements. These are now fully available. Employing them, such the element strains
[]=[B] [e] and the element stresses [] = [C] [] = [C] [B] [e] are computed. The
results are presented in table 3.5. We should note that negative (x, y) denote the
compressive strains; negative (x, y) denote the compressive stresses; and the sign for
(xy) or (xy) indicate the directions of shear distributions.

Table 3.5 Element Strains and element stresses.

Element Strains Stresses t/m

No.
x y xy x y xy
1 0 -0.0078 -0.0050 -11.17 -16.76 -1.8
2 0.0025 -0.0078 0 -5.77 -13.15 0
3 0 -0.0078 0.0050 -11.17 -16.76 1.8
4 0.0025 -0.0048 0 -1.44 -6.67 0

The triangular element being constant strain and constant stress element, the strain and
stresses across the interface between the neighboring elements or the strains and stresses at
the nodes, arising from the elements should display discontinuities.

Consider for example the boundary (4.5). This is the common boundary between element
Nos.2 and 4, and from Table 3.5, we may easily note that inter-element discontinuities in
the strain and stress components prevail. Or at Node.2, the element 1,23 meet and each of
them would define different values of strains and stress components as may be seen from
Table 3.5. Such discontinuities are inconvenient for defining the strains and stress fields;
even otherwise it is desirable to smoothen these discontinuities for unique description of
the fields. It is therefore customary to represent the strains and stress fields through nodal
strains and nodal stresses, in which the same are derived through the method of direct
averaging. For example, at node 2, we assume that [] and [] are average of respective
[] and [] of the elements meeting at the node viz. elements 1,2 and 3. The nodal
parameters calculated in this manner are presented in table 3.6

Table 3.6 : Nodal Strains and nodal stresses.

71
Node Strains Stresses t/m
No.
x y xy x y xy
1 0 -0.0078 -0.0050 -11.17 -16.76 -1.8
2 0.000 -0.0078 0 -9.37 -15.56 0
3 0 -0.0078 0.005 -11.17 -16.76 1.8
4 0.0017 -0.0068 -0.002 -6.13 -12.19 -0.6
5 0.0017 -0.0068 0.002 -6.13 -12.19 0.6
6 0.0025 -0.0048 0 -1.44 -6.67 0
Note :

Negative (x, y) signify compressive strains

Negative (x, y) signify compressive stresses
The sign for xy or xy indicate the directions of shear distributions

72