Transient Conduction:

Finite-Difference Equations
and
Solutions
Chapter 5
Section 5.9
Finite-Difference Method

The Finite-Difference Method

An approximate method for determining temperatures at discrete (nodal) points
of the physical system and at discrete times during the transient process.

Procedure:
Represent the physical system by a nodal network, with an m, n notation used
to designate the location of discrete points in the network, and discretize the
problem in time by designating a time increment t and expressing the time
as t = pt, where p assumes integer values, (p = 0, 1, 2,).

Use the energy balance method to obtain a finite-difference equation for

each node of unknown temperature.

Solve the resulting set of equations for the nodal temperatures at

t = t, 2t, 3t, , until steady-state is reached.

What is represented by the temperature, Tmp, n ?

Storage Term

Energy Balance and Finite-Difference

Approximation for the Storage Term
For any nodal region, the energy balance is
Ein E g E st (5.76)

where, according to convention, all heat flow is assumed to be into the region.

Discretization of temperature variation with time:

T Tmp,n1 Tmp, n
(5.69)
t m, n t

Finite-difference form of the storage term:

Tmp,n1 Tmp, n
E st m, n c
t
Existence of two options for the time at which all other terms in the energy
balance are evaluated: p or p+1.
Explicit Method

The Explicit Method of Solution

All other terms in the energy balance are evaluated at the preceding time
corresponding to p. Equation (5.69) is then termed a forward-difference
approximation.

Example: Two-dimensional conduction

for an interior node with x=y.

Tmp,n1 Fo Tmp1, n Tmp1, n Tmp, n 1 Tmp, n 1 1 4Fo Tmp, n (5.71)

t
Fo finite-difference form of Fourier number
x 2

Unknown nodal temperatures at the new time, t = (p+1)t, are determined

exclusively by known nodal temperatures at the preceding time, t = pt, hence
the term explicit solution.
Explicit Method (cont.)

How is solution accuracy affected by the choice of x and t?

Do other factors influence the choice of t?
What is the nature of an unstable solution?
Stability criterion: Determined by requiring the coefficient for the node of interest
at the previous time to be greater than or equal to zero.

For a finite-difference equation of the form,

Tmp,n1 .............................. ATmp, n
A0
Hence, for the two-dimensional interior node:
1 4Fo 0
Fo 1
4
x 2
t
4

Table 5.2 finite-difference equations for other common nodal regions.

Implicit Method
The Implicit Method of Solution
All other terms in the energy balance are evaluated at the new time corresponding
to p+1. Equation (5.69) is then termed a backward-difference approximation.

Example: Two-dimensional conduction for

an interior node with x=y.

1 4Fo Tmp,n1 Fo Tmp1,1n Tmp1,1n Tmp,n11 Tmp,n11 Tmp, n (5.87)

System of N finite-difference equations for N unknown nodal temperatures

may be solved by matrix inversion or Gauss-Seidel iteration.
Solution is unconditionally stable.

Table 5.2 finite-difference equations for other common nodal regions.

Marching Solution

Marching Solution
Transient temperature distribution is determined by a marching solution,
beginning with known initial conditions.

p t T1 T2 T3.. TN
Known
0 0 T1,i T2,i T3,i. TN,i

1 t -- -- -- --
2 2t -- -- -- --
3 3t -- -- -- --
. .
. .
. .
. .
. .
. .
Steady-state -- -- -- -- . --
Problem: Finite-Difference Equation

Problem 5.94: Derivation of explicit form of finite-difference equation

for a nodal point in a thin, electrically conducting rod
confined by a vacuum enclosure.

KNOWN: Thin rod of diameter D, initially in equilibrium with its surroundings, T sur,
suddenly passes a current I; rod is in vacuum enclosure and has prescribed electrical
resistivity, e, and other thermophysical properties.
FIND: Transient, finite-difference equation for node m.

SCHEMATIC:
Problem: Finite-Difference Equation

ASSUMPTIONS: (1) One-dimensional, transient conduction in rod, (2) Surroundings are

much larger than rod, (3) Constant properties.

ANALYSIS: Applying conservation of energy to a nodal region of volume Ac x,

where A c D2 / 4,

Ein Eout Eg Est

Hence, with Eg I2 R e , where R e e x/Ac , and use of the forward-difference

representation for the time derivative,
p+1 p
2 Tm Tm
qa q b q rad I R e cV
t

4

p p p p
Tm-1 Tm Tm+1 Tm p 4 4 2 e x p+1
Tm Tm
p
kAc kAc Dx Tm Tsur I cAc x .
x x A c t

p+1
Dividing each term by cAc x/t and solving for Tm ,
p+1
Tm
k

c x 2 t p
Tm-1 p
Tm+1
2
k

t
c x 2
p
1 Tm

Ac c
Tm Tsur
P t p 4 4 I2 e t
.
Ac2 c
Problem: Finite-Difference Equation

2
or, with Fo = t/x ,

p 4
I2 e x 2
2
p Px
Tm Fo Tm-1 Tm+1 1 2 Fo Tm
p+1 p p 4
Fo Tm Tsur Fo.
kAc kAc2

p
Basing the stability criterion on the coefficient of the Tm term, it would follow that
Fo .

p 4
However, stability is also affected by the nonlinear term, Tm , and smaller values of Fo may be
needed to insure its existence.
Problem: Cold Plate

Problem 5.127: Use of implicit finite-difference

method with a time interval of
t = 0.1s to determine transient
response of a water-cooled cold
plate attached to IBM multi-chip
thermal conduction module.

Features:
Cold plate is at a uniform temperature,
Ti=15C, when a uniform heat flux
of qo 10 5 W/m 2 is applied to its base
due to activation of chips.
During the transient process, heat
transfer into the cold plate q in increases
its thermal energy while providing for
heat transfer by convection to the
water qconv . Steady state is reached
when q conv q in.
Problem: Cold Plate (cont.)

ASSUMPTIONS: (1) Steady-sate conditions, (2) Two-dimensional conduction, (3) Constant

properties.
Problem: Cold Plate (cont.)

ANALYSIS:
Nodes 1 and 5:
2t 2t p+1 2t p+1 2t p+1
1 T1 T2 T6 T1p
x 2 y2 x 2 y2

2t 2t p+1 2t p+1 2t p+1
1 T5 T4 T10 T5p
x 2 y2 x 2 y2

Nodes 2, 3, 4:
2t 2t p+1 t p+1 t p+1 2t p+1 p
1 Tm,n Tm-1,n Tm+1,n Tm,n-1 Tm,n
x 2 y2 x 2 x 2 y2

Nodes 6 and 14:

2t 2t 2ht p+1 2t p+1 2t p+1 2ht
1 T6 T1 T7 T +T6p
x 2 y2 ky y2 x 2 ky

2t 2t 2ht p+1 2t p+1 2t p+1 2ht p
1 T14 T15 T19 T +T14
x 2 y2 ky x 2 y2 ky

Problem: Cold Plate (cont.)

Nodes 7 and 15:

2t 2t 2ht p+1 2t p+1 t p+1 t p+1 2ht p
1 T7 T2 T6 T8 T +T7
x
2
y 2 ky y 2 x 2 kx 2 ky

2t 2t 2ht p+1 t p+1 t p+1 2t p+1 2ht p

1 T15 T14 T16 T20 T +T15
x
2
y 2 ky x 2 x 2 y2 ky

Nodes 8 and 16:

2t 2t 2 ht 2 ht p+1 4 t p+1 2 t p+1
1 T8 T3 T7
x 2 y 2 3 k x 3 ky 3 y 2 3 x 2

4 t p+1 2 t p+1 2 ht 1 1
T9 T11 T T8p
3 x 2 3 y 2 3 k x y

2t 2t 2 ht 2 ht p+1 2 t p+1 2 t p+1

1 T16 T11 T15
x 2 y 2 3 k x 3 ky 3 y 2 3 x 2

4 t p+1 4 t p+1 2 ht 1 1 p
T17 T21 T T16
3 x 2 3 y 2 3 k x y
Problem: Cold Plate (cont.)

Node 11:
2t 2t 2ht p+1 t p+1 t p+1 t p+1 2ht p
1 2 2 T11 2 T8 2 2 T12 2 T16 T +T11
x y kx y x y kx

Nodes 9, 12, 17, 20, 21, 22:

2t 2t p+1 t p+1 p+1 t p+1 p+1 p
1 Tm,n Tm,n+1 Tm,n-1 Tm-1,n Tm+1,n Tm,n
x 2 y2 y2 x 2

Nodes 10, 13, 18, 23:

2t 2t p+1 t p+1 p+1 2t p+1 p
1 Tm,n Tm,n+1 Tm,n-1 Tm-1,n Tm,n
x 2 y2 y2 x 2

Node 19:

2t 2t p+1 t p+1 p+1 2t p+1 p
1 T19 T14 T24 T20 T19
x 2 y2 y2 x 2

Nodes 24, 28:

2t 2t p+1 2t p+1 2t p+1 2qot p
1 T24 T19 T25 +T24
x 2 2
y y 2 x 2 ky

2t 2t p+1 2t p+1 2t p+1 2qot p
1 T28 T23 T27 +T28
x 2 2
y y 2 x 2 ky

Problem: Cold Plate (cont.)

Nodes 25, 26, 27:

2t 2t p+1 2t p+1 t p+1 p+1 2q t p+1
1 Tm,n Tm,n+1 Tm-1,n Tm+1,n o +Tm,n
x 2 2
y y 2 x 2 k y

The convection heat rate per unit length is

q conv h x/2 T6 T x T7 T x y T8 T / 2 y T11 T x
y T16 T / 2 x T15 T x/2 T14 T q out.

The heat input per unit length is

qin qo 4x

On a percentage basis, the ratio of convection to heat in is

n qconv / qin 100.
Problem: Cold Plate (cont.)

Results of the calculations (in C) are as follows:

Time: 5.00 s; n = 60.57% Time: 20.00 s; n = 98.16%
19.612 19.712 19.974 20.206 20.292 23.574 23.712 24.073 24.409 24.535
19.446 19.597 20.105 20.490 20.609 23.226 23.430 24.110 24.682 24.861
21.370 21.647 21.730 25.502 25.970 26.115
24.217 24.074 23.558 23.494 23.483 28.682 28.543 28.042 28.094 28.122
25.658 25.608 25.485 25.417 25.396 30.483 30.438 30.330 30.291 30.282
27.581 27.554 27.493 27.446 27.429 32.525 32.502 32.452 32.419 32.409

Time: 10.00 s; n = 85.80% Time: 23.00 s; n = 99.00%

22.269 22.394 22.723 23.025 23.137 23.663 23.802 24.165 24.503 24.630
21.981 22.167 22.791 23.302 23.461 23.311 23.516 24.200 24.776 24.957
24.143 24.548 24.673 25.595 26.067 26.214
27.216 27.075 26.569 26.583 26.598 28.782 28.644 28.143 28.198 28.226
28.898 28.851 28.738 28.690 28.677 30.591 30.546 30.438 30.400 30.392
30.901 30.877 30.823 30.786 30.773 32.636 32.613 32.563 32.531 32.520
Time: 15.00 s; n = 94.89%
23.228 23.363 23.716 24.042 24.165
22.896 23.096 23.761 24.317 24.491
25.142 25.594 25.733
28.294 28.155 27.652 27.694 27.719
30.063 30.018 29.908 29.867 29.857
32.095 32.072 32.021 31.987 31.976
Temperatures at t = 23 s are everywhere within 0.13C of the final steady-state values.