Sie sind auf Seite 1von 7

Assignment 3

Q1. For a linear element


φ ( x) = α1 + α 2 x

g
and express it matrix notation and in terms of
φ ( x) = N1 ( x)Φ1 + N 2 ( x)Φ 2

un
Find α1, α2 and N1, N2 in terms of Φ1, Φ2.

The pressure distribution of fluid flowing in a pipe is approximate by a linear element. If the pressure at location
2.5 cm and 7.5 cm from the origin in the pipe are 100 Pa and 55 Pa respectively, determine the pressure distribution

Ke
and its gradient in the element and also find the values at 3.5 cm from the origin.

Answer: φ
Consider a one-dimensional linear element with a length L and

ok
two nodes, one at each end shown as Figure. The nodes are
denoted by 1 and 2 and the nodes values by Φ1, Φ2. The origin φ = α 1+α 2x
of the coordinate system is to the left of node 1.
Kw
Assume a parameter φ varies linearly between the nodes, and
Φ2
the equation for φ is given by
Φ1
φ ( x) = α1 + α 2 x

The coefficient α1 and α2 can be determined by using the nodal


x
conditions 1 2
O

x1 L
φ = Φ1 at x = x1
x2
φ = Φ2 at x = x2
H

to develop the pair of equations The one-dimensional linear element


Φ1 = α1 + α 2 x1
Φ 2 = α 1 + α 2 x2
by

Solving for α1 and α2, we have


Φ1 x2 − Φ 2 x1 Φ1 x2 − Φ 2 x1 x2 x
α1 = = = Φ1 − 1 Φ 2
x2 − x1 L L L

Φ 2 − Φ1 Φ 2 − Φ1 1 1
ne

α2 = = = − Φ1 + Φ 2
x2 − x1 L L L

Substituting the above two equations into the equation for φ, we have
Do

x x   1 1  x x x x 
φ ( x) = α1 + α 2 x =  2 Φ1 − 1 Φ 2  +  − Φ1 + Φ 2  x =  2 −  Φ1 +  − 1  Φ 2
 L L   L L   L L L L
 x −x  x − x1 
= 2  Φ1 +   Φ 2 = N1 ( x)Φ1 + N 2 ( x)Φ 2
 L   L 

x2 − x x − x1
where N1 ( x) = and N 2 ( x) =
L L

1 of 7
The equation for φ can be written in matrix form as
Φ 
φ = [ N1 ( x) N 2 ( x)]  1 
Φ 2 

g
By differentiating N1(x) and N2(x) w.r.t. x, we have
d ( N1 ( x) ) d  x2 − x  −1

un
= =
dx dx  L  L

d ( N 2 ( x) ) d  x − x1  1
= =
dx dx  L  L

Ke
The gradient of the element can be found by

dφ d Φ − Φ1
= (α1 + α 2 x ) = α 2 = 2
dx dx L

ok
Now given Φ1 = 100Pa at x1 = 2.5cm and Φ2 = 55Pa at x1 = 7.5cm,
Φ1 x2 − Φ 2 x1 100Pa × 7.5cm − 55Pa × 2.5cm
α1 = = = 122.5Pa
x2 − x1 7.5cm − 2.5cm
Kw
Φ 2 − Φ1 55Pa − 100Pa
α2 = = = −9Pa/cm
x2 − x1 7.5cm − 2.5cm

The pressure distribution of the element is


φ ( x) = 122.5 − 9 x Pa

The gradient of the element is


O


= α 2 = −9Pa/cm
dx
H

When x = 3.5cm, the approximated pressure is


φ (3.5) = 122.5 − 9 × 3.5 Pa = 91 Pa
by

Q2. The nodal temperature of a linear triangular element at (10, 10), (30, 20) and (20, 40) are 120ºC, 90ºC and 80ºC
respectively. Determine the temperature distribution and its gradient in the element and also find the values at the
ne

(20, 20).

Answer:
Do

It is given that
T1 = 120ºC at x1 = 10 and y1 = 10
T2 = 90ºC at x2 = 30 and y2 = 20
T3 = 80ºC at x3 = 20 and y3 = 40

2 of 7
The temperature distribution equation of linear triangular element can be approximated by
T ( x, y ) = α1 + α 2 x + α 3 y
where

g
1 x1 y1
1 1
A = 1 x2 y2 = [ x1 y2 + x2 y3 + x3 y1 − x1 y3 − x2 y1 − x3 y2 ]

un
2 2
1 x3 y3
1 10 10
1 1 500
= 1 30 20 = [10 × 20 + 30 × 40 + 20 × 10 − 10 × 40 − 30 × 10 − 20 × 20] = = 250
2 2 2
1 20 40

Ke
a1 = x2 y3 − x3 y2 = 30 × 40 − 20 × 20 = 1200 − 400 = 800
a2 = x3 y1 − x1 y3 = 20 × 10 − 10 × 40 = 200 − 400 = −200
a3 = x1 y2 − x2 y1 = 10 × 20 − 30 × 10 = 200 − 300 = −100
b1 = y2 − y3 = 20 − 40 = −20 ; b2 = y3 − y1 = 40 − 10 = 30 ; b3 = y1 − y2 = 10 − 20 = −10

ok
c1 = x3 − x2 = 20 − 30 = −10 ; c2 = x1 − x3 = 10 − 20 = −10 ; c3 = x2 − x1 = 30 − 10 = 20

1 1
α1 = [ a1T1 + a2T2 + a3T3 ] = 800 × 120 + ( −200 ) × 90 + ( −100 ) × 80  = 140
2A 2 × 250 
Kw
1 1
α2 = [b1T1 + b2T2 + b3T3 ] = ( −20 ) × 120 + 30 × 90 + ( −10 ) × 80  = −1
2A 2 × 250 

1 1
α3 = [c1T1 + c2T2 + c3T3 ] = ( −10 ) × 120 + ( −10 ) × 90 + 20 × 80  = −1
2A 2 × 250 
O

Hence, the temperature distribution is


T ( x, y ) = α1 + α 2 x + α 3 y = 140 − x − y
H

The gradients of the element are


∂T ∂T
= α 2 = −1 and = α 3 = −1
∂x ∂y
by

When x = 20 and y = 20, the approximated temperature is


T (20, 20) = 140 − 20 − 20 = 100 (°C)

Alternative Method:
ne

It is given that
T1 = 120ºC at x1 = 10 and y1 = 10
T2 = 90ºC at x2 = 30 and y2 = 20
Do

T3 = 80ºC at x3 = 20 and y3 = 40

The temperature distribution equation of linear triangular element can be approximated by


T ( x, y ) = N1 ( x, y )T1 + N 2 ( x, y )T2 + N 3 ( x, y )T3
where

3 of 7
1 x1 y1
1 1
A = 1 x2 y2 = [ x1 y2 + x2 y3 + x3 y1 − x1 y3 − x2 y1 − x3 y2 ]
2 2
1 x3 y3

g
1 10 10
1 1 500
= 1 30 20 = [10 × 20 + 30 × 40 + 20 × 10 − 10 × 40 − 30 × 10 − 20 × 20] = = 250
2 2 2

un
1 20 40

a1 = x2 y3 − x3 y2 = 30 × 40 − 20 × 20 = 1200 − 400 = 800


a2 = x3 y1 − x1 y3 = 20 × 10 − 10 × 40 = 200 − 400 = −200
a3 = x1 y2 − x2 y1 = 10 × 20 − 30 × 10 = 200 − 300 = −100

Ke
b1 = y2 − y3 = 20 − 40 = −20 ; b2 = y3 − y1 = 40 − 10 = 30 ; b3 = y1 − y2 = 10 − 20 = −10
c1 = x3 − x2 = 20 − 30 = −10 ; c2 = x1 − x3 = 10 − 20 = −10 ; c3 = x2 − x1 = 30 − 10 = 20

1 1
N1 ( x, y ) = [ a1 + b1 x + c1 y ] = [800 − 20 x − 10 y ] = 1.6 − 0.04 x − 0.02 y
2A 2 × 250

ok
1 1
N 2 ( x, y ) = [ a2 + b2 x + c2 y ] = [ −200 + 30 x − 10 y ] = −0.4 + 0.06 x − 0.02 y
2A 2 × 250

1 1
N 3 ( x, y ) = [ a3 + b3 x + c3 y ] = [ −100 − 10 x + 20 y ] = −0.2 − 0.02 x + 0.04 y
Kw
2A 2 × 250

The corresponding partial derivatives of N1 ( x, y ) , N 2 ( x, y ) and N 3 ( x, y ) are

∂N1 ( x, y ) ∂  a1 + b1 x + c1 y  b1 ∂N1 ( x, y ) ∂  a1 + b1 x + c1 y  c1
=   = 2 A = −0.04 ; =   = 2 A = −0.02
∂x ∂x  2A  ∂y ∂y  2A 

∂N 2 ( x, y ) ∂  a2 + b2 x + c2 y  b2 ∂N 2 ( x, y ) ∂  a2 + b2 x + c2 y  c2
O

=   = 2 A = 0.06 ; =   = 2 A = −0.02
∂x ∂x  2A  ∂y ∂y  2A 

∂N 3 ( x, y ) ∂  a3 + b3 x + c3 y  b3 ∂N 3 ( x, y ) ∂  a3 + b3 x + c3 y  c3
=   = 2 A = −0.02 ; =   = 2 A = 0.04
H

∂x ∂x  2A  ∂y ∂y  2A 

Hence, the temperature distribution is


by

T ( x, y ) = N1 ( x, y )T1 + N 2 ( x, y )T2 + N 3 ( x, y )T3


= (1.6 − 0.04 x − 0.02 y ) × 120 + ( −0.4 + 0.06 x − 0.02 y ) × 90 + ( −0.2 − 0.02 x + 0.04 y ) × 80
= 140 − x − y

The gradients of the element are


ne

∂T ∂N ∂N 2 ∂N 1 1
= T1 1 + T2 + T3 3 = [b1T1 + b2T2 + b3T3 ] = ( −20 ) × 120 + 30 × 90 + ( −10 ) × 80  = −1
∂x ∂x ∂x ∂x 2A 2 × 250 

∂T ∂N ∂N 2 ∂N 1 1
= T1 1 + T2 + T3 3 = [c1T1 + c2T2 + c3T3 ] = ( −10 ) × 120 + ( −10 ) × 90 + 20 × 80  = −1
2 × 250 
Do

∂y ∂y ∂y ∂y 2A

When x = 20 and y = 20,


N1 (20, 20) = 1.6 − 0.04 × 20 − 0.02 × 20 = 0.4
N 2 (20, 20) = −0.4 + 0.06 × 20 − 0.02 × 20 = 0.4
N 3 (20, 20) = −0.2 − 0.02 × 20 + 0.04 × 20 = 0.2

4 of 7
The corresponding approximated temperature distribution at point (20, 20) is
T (20, 20) = N1 (20, 20)T1 + N 2 (20, 20)T2 + N 3 (20, 20)T3 = 0.4 × 120 + 0.4 × 90 + 0.2 × 80
= 100 (°C)

g
un
Q3. Describe the general procedures of Finite Element Method for solving the 2D Laplace equation.

Answer:

Ke
To solve the two-dimensional Laplace equation
∇ 2φ = 0
by Finite Element Method, we need to follow the following procedures:

(1) Discretization of the Flow Domain

ok
The actual continuum field φ is simplified as an assemblage of subdivisions called finite elements. The finite
elements are considered to be interconnected at specified joint called nodes.
Kw
(2) Interpolation Functions

The variation of the field variables inside a finite element is Φ (x, y)

approximated by a simple function called interpolation function Φ(x3, y3)


Ni which are defined in terms of variables at the nodes.
Φ (x1, y1) Φ(x, y)
O

Consider the field variable φ is a linear variation in a 2D simplex


Φ (x2, y2)
triangular element
H

φ ( x, y ) = α1 + α 2 x + α 3 y
with nodal conditions
y
φ = Φ1 at (x1, y1)
φ = Φ2
by

at (x2, y2)
(x1, y1)
φ = Φ3 at (x3, y3). (x3, y3)
(x, y)
Therefore, the variable φ(x,y) can be in term of Φ1, Φ2 and Φ3 by
x (x2, y2)
φ ( x, y ) = N1 ( x, y )Φ1 + N 2 ( x, y )Φ 2 + N 3 ( x, y )Φ 3
ne

 Φ1 
= [ N1 N2 N 3 ] Φ 2 
 Φ 3 
= [ N ( x , y ) ] Φe
Do

with

1 x1 y1
1 1
A = 1 x2 y2 = [ x1 y2 + x2 y3 + x3 y1 − x1 y3 − x2 y1 − x3 y2 ]
2 2
1 x3 y3

5 of 7
a1 = x2 y3 − x3 y2 ; a2 = x3 y1 − x1 y3 ; a3 = x1 y2 − x2 y1
b1 = y2 − y3 ; b2 = y3 − y1 ; b3 = y1 − y2
c1 = x3 − x2 ; c2 = x1 − x3 ; c3 = x2 − x1

g
1 1 1
N1 ( x, y ) = [ a1 + b1 x + c1 y ] ; N 2 ( x, y ) = [ a2 + b2 x + c2 y ] ; N 3 ( x, y ) = [ a3 + b3 x + c3 y ]
2A 2A 2A

un
∂φ ∂N ∂N 2 ∂N 1
= Φ1 1 + Φ 2 + Φ3 3 = [b1Φ1 + b2Φ 2 + b3Φ 3 ]
∂x ∂x ∂x ∂x 2A

∂φ ∂N ∂N 2 ∂N 1
= Φ1 1 + Φ 2 + Φ3 3 = [ c1Φ1 + c2Φ 2 + c3Φ 3 ]
∂y ∂y ∂y ∂y 2A

Ke
(3) Formulation of Element Characteristic Matrices and Vectors by Weighted Residual Approach (Galerkin’s
method)

For a differential equation

ok
∇ 2φ = 0 .

By taking the weights using Ni


Kw
∫∫ N (∇ φ ) dA = 0
2 e
i
Ae

the element characteristic matrices can then be found from the Green-Gauss Theorem to reduce the integral to a
matrix.

Green-Gauss Theorem is the integration by part theorem which states that


O

∂φ ∂Ψ
∫∫ Ψ ∂x dxdy = − ∫ ∂x φ dxdy + v∫ Ψφ l dC
S S C
x
H

∂φ ∂Ψ
∫∫ Ψ ∂y dxdy = − ∫ ∂y φ dxdy + v∫ Ψφ l dC
S S C
y

where lx is the cosine of the angle between the normal n and the x direction
ly is the cosine of the angle between the normal n and the y direction
by

The integral can be integrated for nodes i = 1, 2, 3, … for a triangular element and then the results can be expressed
in an element matrix such that

 K e  Φe = P e
ne

where  K e  element characteristic matrix


Do

Φe vector of nodal φ
P e vector of nodal flows

6 of 7
(4) Assembly of Element Matrices and Vectors and Derivation of System Equations

The procedure of assembling the element matrices and vectors is based on the requirement of ‘compatibility’ at the
nodes. This means that at the nodes, the value of the unknown nodal variable is the same for all elements joining the

g
node.

un
Let E – total number of element
M – total nodal degree of freedom
Φ – vector of M nodal degree of freedom
[K ] – assembled system characteristic matrix of order M×M

Ke
Since the element  k e  and  P e  of order n×n and n×1 can be expanded to order M×M and M×1 respectively by

including zeros in the remaining locations. e.g.

 k11 k12 0 0

ok
k 0 0 
 k11 k12   21 k22
k expand to
 21 k22  0 0 0 0
 
0 0 0 0
Kw
E E
Hence [ K ] = ∑  k e  and [ P ] = ∑  P e 
e=1 e=1

[K ] and [ P ] can be generated by identifying the locations of the element  k e  and  P e  in [ K ] and [ P ]

respectively and by adding them to the existing values as e change from 1 to E.


O

(5) Incorporation of the Boundary Conditions and Solving the Final Matrix Equation
H

The overall equations


[ K ]Φ = [ P ]
cannot be solved since [ K ] is singular.
by

The boundary conditions must be incorporated to obtain the final solution. Methods like the iterative Gauss-Seidel,
TDMA, ADI or other matrix inversion method can be used to solve the assembled system of equations.
ne
Do

7 of 7