Scribd
Upload
8 views7 pages

Lecture 3

The document discusses finite-difference methods (FDM) for solving the heat equation, presenting a weighted average approximation that encompasses explicit, implicit, and Crank-Nicolson schemes. It outlines stability conditions for the schemes and provides derivations for boundary conditions using central difference approximations. The document concludes with expressions for updating temperature values at the boundaries and within the domain using the derived schemes.

Uploaded by

Bipin Attri
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
8 views7 pages

Lecture 3

The document discusses finite-difference methods (FDM) for solving the heat equation, presenting a weighted average approximation that encompasses explicit, implicit, and Crank-Nicolson schemes. It outlines stability conditions for the schemes and provides derivations for boundary conditions using central difference approximations. The document concludes with expressions for updating temperature values at the boundaries and within the domain using the derived schemes.

Uploaded by

Bipin Attri
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

FDM for The Heat Equation Contd..

Rajen Kumar Sinha


Department of Mathematics
IIT Guwahati

Rajen Kumar Sinha MA573


Weighted Average Approximation: A more general finite-difference
approximation to the heat equation Ut = Uxx is given by

ui,j+1 − ui,j 1
= θ(ui−1,j+1 − 2ui,j+1 + ui+1,j+1 )
k h2

+(1 − θ)(ui−1,j − 2ui,j + ui+1,j ) , (1)

where 0 ≤ θ ≤ 1. Observe that

θ=0 =⇒ the explicit scheme (Schmidt’s scheme)


θ=1 =⇒ the Euler’s implicit scheme
1
θ= =⇒ the Crank-Nicolson scheme
2
FACT:
• The scheme (1) is unconditionally stable for 21 ≤ θ ≤ 1.
• The scheme (1) is conditionally stable for 0 ≤ θ < 12 . The condition
for stability is
k 1
r= 2 ≤ .
h 2(1 − 2θ)

Rajen Kumar Sinha MA573


Define the central difference operators δx and δt as:
δx φi,j = φi+ 12 , j − φi− 12 , j
δt φi,j = φi, j+ 21 − φi, j− 12
Using this, the explicit scheme can be written as
1 1
δt ui,j+ 12 = 2 δx2 ui,j , where
k h

δt ui,j+ 12 = ui,j+1 − ui,j


δx2 ui,j = δx (δx ui,j ) = δx (ui+ 12 , j − ui− 12 , j )
= ui+1,j − 2ui,j + ui−1,j
The Euler’s implicit scheme:
1 1
δt ui,j− 12 = 2 δx2 ui,j .
k h
The Crank-Nicolson scheme:
 
1 1
δt ui,j+ 12 = 2 δx2 ui,j+1 + δx2 ui,j
k 2h

Rajen Kumar Sinha MA573


The weighted average scheme:
 
1 1 2 2
δt ui,j+ 2 = 2 θ δx ui,j+1 + (1 − θ) δx ui,j
1
k h

Derivative Boundary Conditions:

∂U
= C (U − V ) at x = 0, t > 0
∂x
∂U
= −C (U − V ) at x = 1, t > 0
∂x
V − the temperature of the
surrounding and it is assumed to
be constant.

Rajen Kumar Sinha MA573


Using forward difference approximation, we have at x = 0
u1,j − u0,j
= C (u0,j − V )
h
=⇒ u1,j = u0,j + Ch(u0,j − V )
=⇒ u1,j = (1 + Ch)u0,j − ChV

The truncation error (T.E.)=O(h), which leads to a loss of accuracy in h.


To obtain a better approximation ( of O(h2 )), use central difference
scheme to have
u1,j − u−1,j
= C (u0,j − V )
2h
=⇒ u−1,j = u1,j − 2hC (u0,j − V ). (2)

Recall the explicit scheme

ui,j+1 = rui−1,j + (1 − 2r )ui,j + rui+1,j . (3)

For i = 0,

u0,j+1 = ru−1,j + (1 − 2r )u0,j + ru1,j , j = 0, 1, 2, . . . . (4)

Rajen Kumar Sinha MA573


Now, substituting the value of u−1,j from (2) in (4), it follows that

u0,j+1 = r {u1,j − 2hC (u0,j − V )} + (1 − 2r )u0,j + ru1,j


= 2ru1,j + {(1 − 2r ) − 2rhC }u0,j + 2rhCV
= {1 − 2r (1 + hC )}u0,j + 2ru1,j + 2rhCV , j = 0, 1, 2, . . .

At the other end x = 1, That is, with i = N, we have


 
∂U
= C (V − U)N,j
∂x N,J

Use central difference approximation to obtain


uN+1,j − uN−1,j
= C (V − uN,j )
2h
=⇒ uN+1,j = uN−1,j + 2hC (V − uN,j )

Rajen Kumar Sinha MA573


From the explicit scheme (3), we have for i = N,

uN,j+1 = ruN+1,j + (1 − 2r )uN,j + ruN−1,j


= r {uN−1,j + 2hC (V − uN,j )} + (1 − 2r )uN,j + ruN−1,j
= 2ruN−1,j + {(1 − 2r ) − 2rhC }uN,j + 2rhCV
= 2ruN−1,j + {1 − 2r (1 + hc)}uN,j + 2rhCV , j = 0, 1, 2, . . .

For i = 1, . . . N − 1,

ui,j+1 = rui−1,j + (1 − 2r )ui,j + rui+1,j , j = 0, 1, . . . .

*** Ends ***

Rajen Kumar Sinha MA573

You might also like