LECTURE 4
1
THE FINITE ELEMENT METHOD
or NODAL APPROXIMATION METHOD:
The basic concept behind the Finite
element method is “going from part to whole”
Name “FINITE ELEMENT” coined by
Clough
Fitting of a number of piecewise continuous
polynomials to approximate the variation of
the field variable over the entire domain 2
STEPS INVOLVED IN THE FINITE ELEMENT
METHOD:
Discretisation of the structure
Selection of suitable displacement model
Derivation of elemental matrices and load
vectors
Assembly of elemental equations to obtain
overall stiffness matrix
3
STEPS INVOLVED IN THE FINITE ELEMENT
METHOD:…contd
Imposition of boundary conditions
Solutions for the unknown nodal
displacements
Computation of elemental strains and
stresses
4
1
L1 = 10 cm 1
L2 = 10cm 2
E= 2x107N/cm2 2
BC:
3
U1 = 0
A 1 = [Link]
Pl = 10kN
A2=
[Link] 10 kN
5
u1
N 1u1 + N 2u2
u2
N 1u2 + N 2u3
u3
6
u(x) = a1 + a2x
u(x) = N1 u1 + N2u2
Here Ni s are called Shape functions or
Interpolation functions
Shape functions are used to interpolate
the field variable over the element in
terms
N (x)of nodal
= 1 - x/values
of the
N (0field
) = 1. variable
N () = 0
1 1 1
N1 (x) = x/ N 2 (0) = 0. N1 () = 1
N1 + N2 = 1 7
It can be verified that
N i (x j ) = 0 i j
= 1 i = j
= ij
(Kronecker Delta Function)
N1 u2
N2
. u1
1 2 1 2
8
To provide for the possibility of a constant
or uniform field when u is constant at all
points in the domain
We have u(x) = c = n N (x) u = c n N (x)
i
j=1
i
j=1
i
u1 = u2 = ……. = un = c
N 1 c N 2 c c
n
or N i (x) = 1
j=1
The above properties are very important
properties of shape functions.
9
• In FEA, we use the nodal approximation to
specify the unknown function in terms of its
values at selected ‘nodal points’, through a
Nodal Approximation
10
Now let us consider the numerical example
of the tapered beam whose area of cross
section varies uniformly from A1 to A2 at the
free end and subjected to its own self
weight and a point load at the end.
11
Example
A(x) = A1 – (A1- A2) x/l
ie.A(x) = 80 – (80-20)x/300
= (80 – 0.2x)
Specific weight = 0.075 N/cm3
Young's Modulus E = 2 x 107 N/cm2
12
The governing equation is
d du
[EA(x) ] + A(x) = 0 in 0 < x < L
dx dx
With [Link] i) u(0) = 0
and du
ii)At x=l [EA(x) ] P
dx
13
Weak form is given by
l l
du dw
0
EA( x)
dx dx
dx = A(x) w(x) dx P(l )w(l ) - P(0) w(0)
0
Substituting in the weak form
u(x) = N1u1 +N2u2
And w(x) as N1 first and then N2 we get a
system of two equations in two unknowns
namely u1 and u2
14
l
d(N1u1 N 2 u 2 ) dN1
0
EA( x)
dx dx
dx =
A(x) N 1 dx P(l )w(l ) - P(0) w(0)
0
----1
l
d(N1u1 N 2 u 2 ) dN 2
0
EA( x)
dx dx
dx =
A(x) N
0
2 dx P(l )w(l ) - P(0) w(0)
----2
15
l l
d(N1 ) dN1 d( N 2 ) dN1
0
EA( x)
dx dx
dx u1
0
EA( x)
dx dx
dx u 2 =
A(x) N
0
1 dx P(l )w(l ) - P(0) w(0)
l l
d(N1 ) dN 2 d( N 2 ) dN 2
0
EA( x)
dx dx
dx u1
0
EA( x)
dx dx
dx u 2 =
A(x) N
0
2 dx P(l )w(l ) - P(0) w(0)
16
17
These 2 equations can be written in matrix form as
K 11 K 12 u r1
1
K 21 K 22 u 2 r2
[K ]e
u = r
e e
Where
l
dN i dN j
K e
ij = EA(x) dx
0
dx dx
l
r e
j = A( x) N j dx
0
18
We know that the shape functions for a
two noded element are given by
x x
N1 = 1 - N2 =
l l
dN 1 1 dN 2 1
= - = l
dx l dx
19
l
dN 1 dN 1
K11 = EA(x) dx
0
dx dx
l
= A1 - A 2 1 2
0
E A1 -
l
x dx
l
= E A1 A2 E ( A1 A2 )
( + )
l 2 2 2l
20
l
dN 1 dN 2
K12 = EA(x) dx
0
dx dx
l
A1 - A 2 1 1
dx
=
0
E A1 -
l
x
l l
E A1 A2 E ( A1 A2 )
= ( + )
l 2 2 2l
K12 = K 21
21
l
K 22 = dN 2 dN 2
0
EA(x)
dx dx
dx
l
A1 - A 2 1
2
=
0
E A1 -
l
l
dx
E A1 A2 E ( A1 A2 )
= ( + )
l 2 2 2l
Therefore the element stiffness matrix will be
e E A1 + A 2 1 -1
[K ] =
l 2 -1
1
22
Similarly the element nodal load vector will be
l
r1 A( x) N 1 dx
0
l
(A1 - A 2 ) x
=
0
[ A
1
-
l
(1 -
l
) ] dx
= A1 A2
l l
3 6
l
r2 A( x) N 2 dx
0
l
(A 1 - A 2 ) x
=
0
[ A
1
-
l
( ) ] dx
l
A1 A2
= 6 l 3 l
23
Therefore the assembled load vector will be
r
e =
Case - I: Discretize the Tapered Bar into 3
elements.
The length of each element 'l ' = 100 cm.
24
25
1E A1 + A 2 1 -1 E 70 70
K -1
l1 2 1 100 70 70
2 E A2 + A3 1 -1 E 50 50
K -1
l2 2 1 100 50 50
3 E A3 + A4 1 -1 E 30 30
K -1
l3 2 1 100 30 30
26
The global stiffness matrix will become
[K1 ]
[K] = 2
[K ]
3
[K ]
70 - 70
E 70 70 + 50 - 50
100 - 50 50 + 30 - 30
- 30 30
70 - 70 0 0
E 70 120 - 50 0
100 0 - 50 80 - 30
0 0 - 30 30
27
r 1 l 2 A1 A2
r2 6 2 A2 A 1
220
r
1 6
x 100
200
6
160 100
r
2 6
x 100
140
r
3 6
x 100
80
6
6
28
Similarly the assembled global load vector
will become
| r 1 | P1
[R] = | r2 | + P2
| r 3 |
P 3
29
The global load vector is
220 R
6
[R] = 200 160 + O
6 6
x 100
140 100 O
6 6
80
P
6
220 R
360 O
= x 100
+
6 240 O
30
80 P
Now the total system of equation will be
70 - 70 u1 220 R
E -70 u O
120 - 50 2 x 100 360
100 =
50 80 - 30 u3 6 240 O
- 30 30 u 4 80 P
Now applying the Boundary conditions i.e. u 1 = 0 ..
Delete the first row and first column of elements and
the system of equation will reduce to
120 - 50 u2 360 O
= x 100
- 50 80 - 30 u3 240 O
6 80 P
30 30 u
4 31
The data are E = 2 x 107 N/cm2 = 0.075
N/cc and P = 1 x 105 N.
On solving the above equation we get
u4 = 0.035501997 cm
u3 = 0.018818567 cm
u2 = 0.008778557 cm
The deflection at mid section of the bar by
interpolation is
u2 + u3
Ux50 =
2 = 0.01379856 cm 32
Example 2 Let us consider the discretization
with 2 elements
h = 150 cm
The assembled stiffness matrix will be
E 65 - 65
[K] = 65 65 + 35
150
- 35 35
Similarly the assembled load vector will be
210 R
6
180 120
[R] = x 150 + + O
6 6
90
6 P 33
After applying the [Link] the global system of
equation will become
100 - 35 u 2 240 O
E 6
150 = x 150
80
-35 35 u P
3 6
On solving the above set of simultaneous
equations we get
u3 =0.033068406 cm (Tip displacement)
u2 =0.011607692 cm (Mid section
displacement)
34
e E A1 + A 2 1 -1
[K ]
l 2 -1
1
For a bar of constant cross section A1= A2
e EA 1 -1
[K ]
l
-1
1
r
e = Al 1
2 1
35
Example 3
36
37
38
39
40
41
WEAK FORM OF GOVERNING
EQUATION FOR THERMAL
PROBLEMS
42
43
where
k = Thermal conductivity coefficient
h = Thermal convection coefficient
A = Area of cross section subjected to
CONDUCTION
p= Perimeter is the area exposed to
CONVECTION
T = Atmospheric Temp. , T = Variable
Q = Heat Source
44
(q + dq) – q + hp dx(T - T )=0
by dx we get
dq + hp(T - T ) = 0
dx
d(-kA(x) dT ) + hp(T - T ) =0
dx dx
45
Boundary conditions:
i) At x= 0 T = To
ii) At the free end any one of the following
three possible boundary conditions could
be specified
1. If free end is insulated _ kA dT/dx = 0
2. If free end is open to atmosphere
_ kA dT/dx|=l = hA(T- T)
3. Specified temperature T(l) = Tl
46
The governing equation for heat transfer in
a one dimensional problem is given by
d dT
KA hp (T T ) 0
dx dx
The weak form can be obtained by
w( x) R( x)dx 0
For a bar of length ‘l’ with wall temperature ‘T’
the weak form of the governing equation
becomes 47
l
d dT
0
w( x) KA hp (T T) dx 0
dx dx
l l
d dT
0
w( x) KA dx w( x)hp (T T)dx 0
dx dx 0
l
Let d dT
I 1 w( x) KA dx
0
dx dx
and u w(x) du dw
d dT dT
dv KA dx v KA
dx dx dx
48
I 1 uv vdu
l l
dT dT dw
I 1 w( x) KA KA dx
dx 0 0 dx dx
Substituting the above term in equation 1,
we get
l l l
dT dT dw
w( x) KA KA dx w( x)hp(T T)dx 0
dx 0 0 dx dx 0
49
l l l l
dT dT dw
w( x) KA KA dx hpw( x)T ( x)dx hpw( x)Tdx 0
dx 0 0 dx dx 0 0
Boundary term B1(T,w) B2(T,w) l(w)
l l l
dT dw
0 KA dx dx dx 0 hpw( x)T ( x)dx 0 hpw( x)Tdx w( x)hA(TL T)
50
Substituting in the weak form
T(x) = N1T1 +N2T2
And w(x) as N1 first and then N2 we get a
system of two equations in two unknowns
namely T1 and T2 which can be written as
51
K 11 K 12
T K 11 K 12
T q1
1 1
K
K 21 K 22 T2 21 K 22 T2 q 2
cond conv
l
dN i dN j
Where K e
ijcond = kA(x) dx
0
dx dx
l
K ije conv = hp(x)
0
N i N j dx
l
hpT
e
q j = N j dx
0
52
Let the elements be of equal length l
The element matrices are
e KA 1 -1 hP l 2 1 0 0
[K ] = -1 + 1 +
l 1 6 2 0 hA
e hPl T 1 0
[f ] = +
2 1 hA T
53
54
Boundary conditions:
at x = 0,T(0) = T
dT
at x = L, KA l = hA (Tl - T )
dx
conduction = convection loss
For a typical linear element
NI = 1 - (x/l)
NJ = (x/l)
55
Let the elements be of equal length l = 2 cm
The element matrices are
e kA 1 -1 hp l 2 1 0 0
[K ] = -1 + 1 +
l 1 6 2 0 hA
e
hpl T 1 0
[q ] = +
2 1 hA T
56
The element matrices for ELEMENT (1),
(2) & (3) are
e 6 6 20
[K ] cond = ; {qe } =
6 6 20
e 0.667 0.333 20
[K ] conv = ; {qe } =
0.333 0.667 20
e 6.666 -5.667 20
[K ] therm = ; {qe } =
-5.667 6.666 20
57
The element matrix for ELEMENT (4) is
e 6 6
[K ] cond =
6 6
e 0.667 0.333 0 0
[K ] conv =
0.333 0.667 0 0.4
20 0
{q e } =
20 8
e 6.666 -5.667 20
[K ] therm = ; {q e } =
-5.667 7.066 2858
On assembly we get
6.667 - 5.667 0 0 0 T1 20
-5.667 13.33 - 5.667 0 0 T2 20 + 20
0 - 5.667 13.33 - 5.667 0 * T3 = 20 + 20
T4 20 + 20
0 0 - 5.667 13.33 - 5.667
0 0 0 - 5.667
7.066 T5 28
59
By applying Boundary condition at
at x = 0 T = T0 = 80º
13.33 - 5.667 0 0 T2 40 + 5.667 *80
-5.667 13.33 - 5.667 0 T3 40
* =
0 - 5.667 13.33 - 5.667 T4 40
T5 28
0 0 - 5.667 7.066
By solving we get
T2 = 53.95 0
C; T3 = 39.88 0 C;
T4 = 32.82 0
C; T5 = 30.29 0 C;
60
Boundary condition: Free end insulated
61
h = 10 W/cm2 oC
K = 70 W /cm oC
T0 = 140oC
T∞ = 40oC
ℓ = 5 cm
Radius r = 1 cm
Area A = π r2 = π cm2
Perimeter p = 2πr = 2 π
62
