0% found this document useful (0 votes)
28 views9 pages

Week 3 (II) - Weighted Residual Method

The document discusses the weighted residual method for finite element analysis. It describes using weighting functions to minimize the residuals between an approximate solution and the governing equations. The method is applied to a 1D heat conduction problem, expressing the temperature distribution within each element in terms of nodal shape functions. This allows setting up a system of equations to solve for the unknown nodal temperatures.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
28 views9 pages

Week 3 (II) - Weighted Residual Method

The document discusses the weighted residual method for finite element analysis. It describes using weighting functions to minimize the residuals between an approximate solution and the governing equations. The method is applied to a 1D heat conduction problem, expressing the temperature distribution within each element in terms of nodal shape functions. This allows setting up a system of equations to solve for the unknown nodal temperatures.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

29-01-2021

Module 2
Fundamentals of finite element (FE) method

Weighted residue technique

Weighted residual method


Suppose the governing equation and boundary conditions are given as
Governing Eq:
𝐴 𝑇 =0 (1)
Boundary condition:
𝐵 𝑇 =0

If an arbitrary temperature distribution is assumed in the domain and it is substituted in


Eq. 1, these equations will not be satisfied. The obtained results will be as follows:

𝐴 𝑇′ = 𝑅 ≠ 0 (2)
𝐵 𝑇′ = 𝑅 ≠ 0

These obtained quantities are referred to as residues.

1
29-01-2021

Weighted residual method

1 𝐴 𝑇′
3 n
1 2
4
2
𝑤

Fig. 1. Arbitrary points within domain A 1 2 B x


Fig. 2. Residue and weighting function
 The residue R1 will different at different points shown in Fig. 1, whereas the
residue R2 will vary in the boundary points.
 For 𝑇′ to be the solution of the governing equation, these residues should be zero.
3

Weighted residual method


 Considering a 1-D domain extending from point A to B as shown in Fig. 2.
 If the distribution of the residue for some arbitrary temperature distribution 𝑇′ is
given by A(𝑇 ), then we consider another function w.
 The function w is chosen such that when A 𝑇 and w is multiplied at every point
and added, we obtain the quantity,

𝑤 𝐴 𝑇 𝑑𝑥

where, w is an arbitrary function. It is also called weighting function.


 A 𝑇 will not be zero at all the points for a distribution of 𝑇 which is not the
solution of the problem.
Therefore, the possibility of the above integral being zero is minimal.

2
29-01-2021

Weighted residual method


If we consider several arbitrary distributions of w (e.g. w1, w2…) and obtain a
distribution of T which makes this integral zero for all values of w, then the only
possibility in which this can happen is when T is the desired solution of the
problem.

Thus the condition for a temperature distribution T to be the solution of a problem


stated in the form of expression (1) is given by,
(3)
   
𝑤 𝐴 𝑇 𝑑𝑣 + 𝑤 𝐵 𝑇 𝑑𝑆 = 0

where, w1 and w2 comprise many sets of arbitrary weighting functions.

Weighted residual method


Based on the selection of weight function the weighted residual method can be
divided into five categories:

• Collocation method – Weighting function is chosen from the family of


Dirac 𝛿 functions in the domain.

• Sub-domain method – A modification of the collocation method.

• Least Squares method – Summation of all the squared residuals is


minimized.

• Galerkin method – It will be discussed further in the course.

• Method of moments – The weight function are chosen from the family of
polynomials.
6

3
29-01-2021

Weighted residual method


Condition of a function to be used as a weighting function:
i. All functions should be totally independent.
ii. The functions should be simple in order to avoid computational difficulties.
N

1 2 3 4 5 6 7 8 n
Fig. 3 Variation of nodal shape function

Fig. 3 shows a 1-D element with n nodes within, 1-2, 2-3, 3-4… are each separate
elements.
N1 , N2,…Nn are nodal shape functions.
7

Weighted residual method


Nodal shape function has relevance at a particular node and adjoining elements.

The value of nodal shape function is 1 at a particular node and its value is zero at
every other node. This feature can be seen in Fig. 3.

If these shape functions are used as the weighting function then this approach is
called the Galerkin approach.

e.g., w1=Ni; and w2= -Ni; where (i = 1, 2, 3…., n). Thus, the weighted residual
statement given in Eq. 3 becomes,
   
𝑁 𝐴 𝑇 𝑑𝑣 − 𝑁 𝐵 𝑇 𝑑𝑆 = 0 (4)

4
29-01-2021

Heat conduction analysis


The steady state heat conduction equation is given by:

𝜕 𝑇 𝜕 𝑇 𝜕 𝑇
𝑘 + + + 𝑄̇ = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧

where, 𝑄̇ is the rate of heat generation in unit volume.


The boundary conditions specified at the surface is expressed as:

𝜕𝑇
𝑘
+ 𝛼𝑇 + 𝑞 = 0 𝑎𝑡 𝑆
𝜕𝑛
where, S1 represent the portions of the boundary on which boundary conditions are
specified.

Heat conduction analysis


Considering A(T) as the governing equation and B(T) as the boundary equation in
Eq. 4, and applying the concept of weighted residual method to the 3-D steady state
heat transfer problem, we have:
   
𝜕 𝑇 𝜕 𝑇 𝜕 𝑇 𝜕𝑇
𝑁 𝑘 + + + 𝑄̇ 𝑑𝑉 − 𝑁 𝑘 + 𝛼𝑇 + 𝑞 𝑑𝑆 = 0 (5)
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑛
or,
         
𝜕𝑇
𝑘𝑁 𝛻 𝑇𝑑𝑉 + 𝑁 𝑄̇ 𝑑𝑉 − 𝑘𝑁 𝑑𝑆 − 𝑁 𝛼𝑇𝑑𝑆 − 𝑁 𝑞𝑑𝑆 = 0 (6)
𝜕𝑛

First form of Green’s theorem,

𝑢𝛻 𝜙 + 𝛻𝑢. 𝛻𝜙 𝑑𝑉 = 𝑢𝛻𝜙 . 𝑑𝑆⃗


10

5
29-01-2021

Heat conduction analysis


Using the first form of Green’s theorem to rewrite Eq. 6, we have,
 
𝜕𝑁 𝜕𝑇 𝜕𝑁 𝜕𝑇 𝜕𝑁 𝜕𝑇
− 𝑘 + + 𝑑𝑉 +
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧
         
𝜕𝑇 𝜕𝑇
𝑘𝑁 𝑑𝑆 + 𝑁 𝑄̇ 𝑑𝑉 − 𝑘𝑁 𝑑𝑆 − 𝑁 𝛼𝑇𝑑𝑆 − 𝑁 𝑞𝑑𝑆 = 0 (7)
𝜕𝑛 𝜕𝑛
or,
       
𝜕𝑁 𝜕𝑇 𝜕𝑁 𝜕𝑇 𝜕𝑁 𝜕𝑇
𝑘 + + 𝑑𝑉 − 𝑁 𝑄̇ 𝑑𝑉 + 𝑁 𝛼𝑇𝑑𝑆 + 𝑁 𝑞𝑑𝑆 = 0 (8)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

On discretizing the domain into four-noded tetrahedron elements, the temperature


within the element can be expressed in terms of nodal temperature as,

𝑇= 𝑁𝑇 = 𝑁 𝑇 (9)

11

Heat conduction analysis


The volume integrals and the surface integrals in Eq. 8 are expressed as the sum over
all the elements.

If there exits n nodes within the volume there will be n unknown temperatures Ti in
Eq. 8.

To obtain the Ti values, n number of equations are required that can be solved
simultaneously.

 The equations can be obtained by considering n separate values of weight functions,


Ni, each n value giving a separate equation.

Thus, if the number of nodes is n, then the number of Ni values is also n.


12

6
29-01-2021

Heat conduction analysis


Therefore Eq. 8 can be expressed as:

     
𝜕𝑁 𝜕𝑇 𝜕𝑁 𝜕𝑇 𝜕𝑁 𝜕𝑇
𝑘 + + 𝑑𝑉 − 𝑁 𝑄̇ 𝑑𝑉 + 𝑁 𝛼𝑇𝑑𝑆 + 𝑁 𝑞𝑑𝑆 = 0
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧
(10)

where, we have m elements, of which r elements have boundary of domain as their face.

Considering a tetrahedron mesh that has only 4 nodes, only 4 non-zero values of Ni will
exist for an element.
Thus an element will contribute terms only in four equation of a set of n equations
expressed as Eq. 10.
𝟒

𝑻= 𝑵𝒊 𝑻 𝒊 = 𝑵 𝑻 𝒆
𝒊 𝟏
13

Heat conduction analysis


If four nodes associated with an element ‘e’ are designated as 1, 2, 3, 4 for simplicity, the
corresponding term-wise contribution from this element towards Eq. 10 will be:
(i) For first term:
 
𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁]
𝑘 𝑇 + 𝑇 + 𝑇 𝑑𝑉
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

 
𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁]
𝑘 𝑇 + 𝑇 + 𝑇 𝑑𝑉
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

 
𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁]
𝑘 𝑇 + 𝑇 + 𝑇 𝑑𝑉
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

 
𝜕𝑁 𝜕[𝑁] 𝜕𝑁 𝜕[𝑁] 𝜕4 𝜕[𝑁]
𝑘 𝑇 + 𝑇 + 𝑇 𝑑𝑉 (11)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

14

7
29-01-2021

Heat conduction analysis


• The terms of 𝑇 are nodal temperatures that do not vary within the element,
• Therefore, this vector is taken out of the integral sign and on re-arranging the
remaining terms we obtain the elemental contribution in matrix form as:

ℎ ℎ ℎ ℎ 𝑇
ℎ ℎ ℎ ℎ 𝑇 (12)
= ℎ {𝑇 }
ℎ ℎ ℎ ℎ 𝑇
ℎ ℎ ℎ ℎ 𝑇
where,
 
𝜕𝑁 𝜕𝑁
ℎ = 𝑘 𝑑𝑉 i = 1, 2, 3, 4
𝜕𝑥 𝜕𝑥
j = 1, 2, 3, 4
m = 1, 2, 3

15

Heat conduction analysis


(ii) For 2nd term
The element contribution towards the
 
second term is made of four terms given as:
𝑓 = − 𝑄̇ 𝑁 𝑑𝑉 𝑜𝑟 𝑣𝑒𝑐𝑡𝑜𝑟 {𝑓 } i=1, 2, 3, 4

(iii) For 3rd and 4th terms


The contribution towards the 3rd and the 4th terms can similarly be obtained as:
ℎ 𝑇 and 𝑓
where,
   
ℎ = ∫ 𝛼𝑁 𝑁 𝑑𝑆 𝑓 = ∫ 𝑞𝑁 𝑑𝑆 and i, j = 1, 2, 3, 4

The total contribution from an element ‘e’ is thus obtained as:


ℎ 𝑇 + ℎ 𝑇 + 𝑓 + 𝑓

16

8
29-01-2021

Thank you
for your kind attention

17

You might also like