ME5130 - FEM
HW 1
RAHUL ALAM
ME15BTECH11030
1. Given differential equation is
d2 u
dx2
+ u = 1 , boundary conditions : u(0)= 1, u(1)=0
Ans:
u = up + uh . up = particular solution, uh = homogeneous solution
First we will find uh .
d2 u
For that we have to consider, dx2
+u = 0
So, the homogeneous solution is, uh = Acos(x) + Bsin(x)
The RHS of the equation is constant. So, for particular solution, we can up = k.
So, up ll = 0 and from the given differential we get k = 1.
So, the analytical solution is,
u = Acos(x) + Bsin(x) + 1
Now from Boundary conditions,
At x = 0 ,
1 = A+0+1 ⇒A = 0
At x = 1 ,
−1
0 = Bsin(1) + 1 ⇒ B = sin(1)
So final analytical solution is ,
sin(x)
u = 1 − sin(1)
2. In the differential equation in above question, let approx solution,
u = a0 + a1 x + a2 x2 ,
Ans.
u(0) = 1 = a0
u(1) = 1 + a1 + a2 = 0 ⇒ a1 = − a2 − 1
So, u = 1 − (a2 + 1)x + a2 x2 ⇒ u = (1 − x) + a2 (x2 − x)
Here we have only 1 D.O.F --> a2
d2 u
R(u) ≡ dx2
+ u -1
2
d
R(u) ≡ − (a2 + 1)x + a2 x2 ) + (1 − (a2 + 1)x + a2 x2 ) − 1
dx2
(1
⇒ R(u) = − x + a2 (x2 − x + 2)
Now we apply MWR (Method of weighted residue) to find the approximate solutions.
We will use the below 5 methods to find approximate solutions. Also we will find the RMS
√
1
error for each approximate function using , RMS error = ∫(u − u) 2 dx
0
a) Collocation method:
W i = δ(x − xi ) and ∫ W i R(xi )dx = 0
And we have R(xi ) = 0 at one point
We choose xi = 0.5
⇒ R(0.5) = 0
⇒ R(0.5) = − 0.5 + a2 (0.52 − 0.5 + 2) = 0
2
⇒ a2 = 7
2 2
⇒ u = (1 − x) + 7 (x − x) and RMS error = 0.00606
ii) Sub-domain method:
1
Wi = 1 and ∫ W i R(xi )dx = 0
0
1
⇒ ∫(− x + a2 (x2 − x + 2))dx = 0
0
3
⇒ a2 = 11
3 2
⇒ u = (1 − x) + 11 (x − x) and RMS error = 0.00598
iii) Least square method:
1
∂R
W = ∂a2
2
= x − x + 2 and ∫ W Rdx = 0
0
1
⇒ ∫(x2 − x + 2)(− x + a2 (x2 − x + 2))dx = 0
0
165
⇒ a2 = 606
165 2
⇒ u = (1 − x) + 606 (x − x) and RMS error = 0.00599
iv) Galerkin Method:
1
W = ∂u
∂a2 and ∫ W Rdx = 0
0
1
W = x2 − x ⇒ ∫(x2 − x)(− x + a2 (x2 − x + 2)) = 0 ⇒ a2 = 5
18
0
5 2
⇒ u = (1 − x) + 18 (x − x) and RMS error = 0.0059
v) Method of moments:
1
i
W = x , i = DOF and ∫ W Rdx = 0
0
1
1
W = x ⇒ ∫ x(− x + a2 (x2 − x + 2)dx = 0 ⇒ a2 = 4
11
0
4 2
⇒ u = (1 − x) + 11 (x − x) and RMS error = 0.0167
Below is the code for plotting analytical vs approximate solution for particular graph.
Here is the code for comparing all graphs in a plot.
3. In the differential equation in above question, let approx solution,
u = b0 + b1 x + b2 x2 + b3 x3
Ans.
u(0) = b0 = 1
u(1) = 1 + b1 + b2 + b3 = 0 ⇒ b1 = − b2 − b3 − 1
⇒ u = (1 − x) + b2 (x2 − x) + b3 (x3 − x)
R(u) = d2 u
+ u − 1 ⇒ R(u) = − x + b2 (x2 − x + 2) + b3 (x3 + 5x)
dx2
Now we apply MWR (Method of weighted residue) to find the approximate solutions.
We will use the below 5 methods to find approximate solutions. Also we will find the RMS
√
1
error for each approximate function using , RMS error = ∫(u − u) 2 dx
0
a) Collocation method:
W i = δ(x − xi ) and ∫ W i R(xi )dx = 0
And we have R(xi ) = 0 at 2 point b/w 0 and 1. We choose xi = 0.25 and xi = 0.5 .
29 81 1
R(0.25) = 0 ⇒ 16 b2 + 64 b3 = 4
7 21 1
R(0.5) = 0 ⇒ b
4 2
+ b
8 3
= 2
Solving we get, b2 = 0.0092 and b3 = 0.1843
⇒ u = 1 − x + 0.0092(x2 − x ) + 0.1843(x3 − x)
and RMS error = 0.00147
ii) Sub-domain method:
0.5 1
Wi = 1 and ∫ W i R(xi )dx = ∫ W i R(xi )dx = 0
0 0.5
0.5
11 41 1
∫ [− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0 ⇒ b
12 2
+ b
64 3
= 8
0
1
11 135 3
∫ [− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0 ⇒ b
12 2
+ b
64 3
= 8
0.5
Solving we get, b2 = 0.0174 and b3 = 0.1702
⇒ u = 1 − x + 0.0174(x2 − x ) + 0.1702(x3 − x)
and RMS error = 0.00101
iii) Least square method:
1
W1 = ∂R
∂b2 , W2 = ∂R
∂b3 and ∫ W i Rdx = 0
0
1
W 1 = x2 − x + 2 ⇒ ∫(x2 − x + 2)[− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0
0
101 101 11
⇒ b
30 2
+ b
20 3
= 12
1
W 2 = x + 5x3
⇒ ∫(x3 + 5x)[− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0
0
101 220 28
⇒ b
20 2
+ b
21 3
= 15
Solving we get, b2 = 0.0181 and b3 =0.1695
⇒ u = 1 − x + 0.0181(x2 − x ) + 0.1695(x3 − x)
and RMS error = 0.00108
iv) Galerkin Method:
1
W1 = ∂u
∂b2 , W2 = ∂u
∂b3 and ∫ W i Rdx = 0
0
1
W 1 = x2 − x ⇒ ∫(x2 − x)[− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0
0
3 9 1
⇒ b
10 2
+ b
20 3
= 12
1
W2 = x − x 3
⇒ ∫(x3 − x)[− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0
0
9 76 2
⇒ b
20 2
+ b
105 3
= 15
Solving we get, b2 = 0.0216 and b3 =0.1707
⇒ u = 1 − x + 0.0216(x2 − x ) + 0.1707(x3 − x)
and RMS error = 0.00019
v) Method of moments:
1
1 2
i
W = x , i = DOF ⇒ W 1 = x , W 2 = x and ∫ W Rdx = 0
0
1
11 28 1
W1 = x ⇒ ∫ x[− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0 ⇒ b
12 2
+ b
15 3
= 3
0
1
37 17 1
W2 = x 2
⇒ ∫ x 2 [− x + b2 (x2 − x + 2) + b3 (x3 + 5x)]dx = 0 ⇒ b
60 2
+ b
12 3
= 4
0
Solving we get, b2 = 0.0376 and b3 =0.1600
⇒ u = 1 − x + 0.0376(x2 − x ) + 0.1600(x3 − x)
and RMS error = 0.00042
Here is the code for comparing all the approximate and analytical plots.
