Because learning changes everything.

Chapter 5
Numerical Methods
in Heat Conduction

© McGraw Hill LLC. All rights reserved. No reproduction or distribution without the prior written consent of McGraw Hill LLC.
 Objectives

• Understand the limitations of analytical solutions of

conduction problems, and the need for computation-
intensive numerical methods.
• Express derivates as differences, and obtain finite
difference formulations.
• Solve steady one- or two-dimensional conduction
problems numerically using the finite difference method.
• Solve transient one- or two-dimensional conduction
problems using the finite difference method.

©McGraw Hill LLC 2

 5.1: Why Numerical Methods? 1

In Chapter 2, we solved various

heat conduction problems in
various geometries in a systematic
but highly mathematical manner by,
1. deriving the governing differential
equation by performing an energy
balance on a differential volume
Solution: element,
2. expressing the boundary conditions
T ( r ) = T1 + ( r02 − r 2 )
e
6k in the proper mathematical form,
dT 4 r 3e and,
Q ( r ) = − kA =
dr 3 3. solving the differential equation and
Figure 5.1: The analytical solution of a applying the boundary conditions to
problem requires solving the governing determine the integration
differential equation and applying the constants.
boundary conditions.
Access the text alternative for slide images.

©McGraw Hill LLC 3

 5.1: Why Numerical Methods? 2

1. Limitations. Analytical solution methods are limited to

highly simplified problems in simple
geometries.
The geometry must be such that its entire
surface can be described mathematically
in a coordinate system by setting the
variables equal to constants.
That is, it must fit into a coordinate system
perfectly with nothing sticking out or in.
Even in simple geometries, heat transfer
problems cannot be solved analytically if
the thermal conditions are not sufficiently
simple.
Figure 5.2: Analytical solution methods Analytical solutions are limited to problems
are limited to simplified problems in that are simple or can be simplified with
simple geometries. reasonable approximations.

©McGraw Hill LLC 4

 5.1: Why Numerical Methods? 3

2. Better Modeling.
When attempting to get an analytical solution to a
physical problem, there is always the tendency to
oversimplify the problem to make the mathematical
model sufficiently simple to warrant an analytical
solution.
Therefore, it is common practice to ignore any effects
that cause mathematical complications such as
nonlinearities in the differential equation or the
boundary conditions (nonlinearities such as
temperature dependence of thermal conductivity and
the radiation boundary conditions).
A mathematical model intended for a numerical Figure 5.3: The approximate
solution is likely to represent the actual problem better. numerical solution of a real-world
The numerical solution of engineering problems has problem may be more accurate
now become the norm rather than the exception even than the exact (analytical)
when analytical solutions are available. solution of an oversimplified
model of that problem.
Access the text alternative for slide images.

©McGraw Hill LLC 5

 5.1: Why Numerical Methods? 4

3. Flexibility.
Engineering problems often require extensive parametric studies
to understand the influence of some variables on the solution in
order to choose the right set of variables and to answer some
“what-if” questions.
This is an iterative process that is extremely tedious and time-
consuming if done by hand.
Computers and numerical methods are ideally suited for such
calculations, and a wide range of related problems can be solved
by minor modifications in the code or input variables.
Today it is almost unthinkable to perform any significant
optimization studies in engineering without the power and flexibility
of computers and numerical methods.

©McGraw Hill LLC 6

 5.1: Why Numerical Methods? 5

4. Complications.

Some problems can be solved analytically, but the

solution procedure is so complex and the resulting
solution expressions so complicated that it is not
worth all that effort.
With the exception of steady one-dimensional or
transient lumped system problems, all heat
conduction problems result in partial differential
equations.
Solving such equations usually requires mathematical Analytical solution:
sophistication beyond that acquired at the T ( r , z ) − T  J 0 ( n r ) sinh  ( L − z )
=
n =1 n J1 ( n r0 ) sinh ( n L )
undergraduate level, such as orthogonality, T0 − T
eigenvalues, Fourier and Laplace transforms, Bessel
where n 's are roots of J 0 ( n r0 ) = 0
and Legendre functions, and infinite series.
In such cases, the evaluation of the solution, which Figure 5.4: Some analytical
often involves double or triple summations of infinite solutions are very complex and
series at a specified point, is a challenge in itself. difficult to use.

©McGraw Hill LLC 7

 5.1: Why Numerical Methods? 6

5. Human Nature. Analytical solutions are necessary because

insight to the physical phenomena and
engineering wisdom is gained primarily
through analysis.
The “feel” that engineers develop during
the analysis of simple but fundamental
problems serves as an invaluable tool
when interpreting a huge pile of results
obtained from a computer when solving a
complex problem.
A simple analysis by hand for a limiting
Figure 5.5: The ready availability of
high-powered computers with
case can be used to check if the results are
sophisticated software packages in the proper range.
has made numerical solution the In this chapter, you will learn how to
norm rather than the exception.
formulate and solve heat transfer problems
numerically using one or more approaches.

©McGraw Hill LLC 8

 5.2: Finite Difference Formulation of
Differential Equations 1

The numerical methods for solving differential Table 5.1: Year-end balance of a
equations are based on replacing the differential $100 account earning interest at an
equations by algebraic equations. annual rate of 18 percent for various
In the case of the popular finite difference method, compounding periods.
this is done by replacing the derivatives by Compounding Number of Year-end
differences. period periods, n balance
Below we demonstrate this with both first- and 1 year 1 $118.00
second-order derivatives.
6 months 2 118.81
An Example: 1 month 12 119.56
A = A0 (1 + i ) = ( $100 )(1 + 0.09 ) = $118.81
n 2
1 week 52 119.68
dA 1 day 365 119.72
= iA
dt 1 hour 8760 119.72
A = A0 exp ( it ) 1 minute 525600 119.72

A = ( $100 ) exp ( 0.18  1) = $119.72 1 second 31536000 119.72

Reasonably accurate results can be obtained Instantaneous  119.72

by replacing differential quantities by
sufficiently small differences.
©McGraw Hill LLC 9
 5.2: Finite Difference Formulation of
Differential Equations 2

df ( x ) f f ( x + x ) − f ( x )
= lim = lim
dx x → 0 x x → 0 x

df ( x ) f ( x + x ) − f ( x ) finite difference form of


dx x the first derivative

df ( x ) 1 2 d 2 f ( x )
f ( x + x ) = f ( x ) + x + x + ...
dx 2 dx 2

Taylor series expansion of the function f

Figure 5.6: The derivative of a about the point x,
function at a point represents the
The smaller the x, the smaller the error, and
slope of the function at that point.
thus the more accurate the approximation.

©McGraw Hill LLC 10

 5.2: Finite Difference Formulation of
Differential Equations 3

Consider steady one-dimensional heat d 2T e

conduction in a plane wall of thickness L 2
+ =0
dx k
with heat generation.
dT Tm − Tm −1 dT Tm +1 − Tm
 and 
dx m−
1 x dx m+
1 x
2 2

dT dT
− Tm +1 − Tm Tm − Tm −1
dx dx −
T − 2Tm + Tm +1
2 1 1
m+ m−
d T
 2 2
=  x x = m −1
dx 2 m
x x x 2

Tm −1 − 2Tm + Tm +1 em
+ =0 m = 1, 2,3,..., M − 1
x 2 k
Finite difference representation of the second
derivative at a general internal node m. Figure 5.7: Schematic of the
nodes and the nodal temperatures
1
Tm = (Tm−1 + Tm+1 ) no heat generation. used in the development of the
2 finite difference formulation of
heat transfer in a plane wall.

Access the text alternative for slide images.

©McGraw Hill LLC 11

 5.2: Finite Difference Formulation of
Differential Equations 4

Figure 5.8: The differential equation is valid at

every point of a medium, whereas the finite
difference equation is valid at discrete points
(the nodes) only.

Access the text alternative for slide images.

©McGraw Hill LLC 12

 5.2: Finite Difference Formulation of
Differential Equations 5

Figure 5.9: Finite difference mesh

for two-dimensional conduction in
rectangular coordinates.

Finite difference formulation for steady two-dimensional heat conduction in a region with
heat generation and constant thermal conductivity in rectangular coordinates.

Tm +1, n − 2Tm , n + Tm −1, n Tm , n +1 − 2Tm , n + Tm , n −1 em , n

+ + =0
x 2
y 2
k

Access the text alternative for slide images.

©McGraw Hill LLC 13

 5.3: One-dimensional Steady Heat
Conduction 1

We develop the finite Figure 5.10: The nodal

difference formulation of heat points and volume
conduction in a plane wall elements for the finite
difference formulation of
using the energy balance
one-dimensional
approach. conduction in a plane
The energy balance method is wall.
based on subdividing the
medium into a sufficient
number of volume elements
and then applying an energy
balance on each element.
( Rate of heat conduction at the left surface) + ( Rate of heat conduction at the right surface) +
( Rate of heat generation inside the element ) = ( Rate of change of the energy content of the element )

Eelement T
Qcon, left + Qcond,right + Egen, element = =0 Eelement = 0 Qcond = kA
t L
Egen, element = emVelement = em Ax

©McGraw Hill LLC 14

 5.3: One-dimensional Steady Heat
Conduction 2

Tm −1 − Tm Tm +1 − Tm
Qcond, left = kA Qcond, right = kA
x x
Tm −1 − Tm T −T
kA + kA m +1 m + em Ax = 0
x x
Tm −1 − 2Tm + Tm +1 em m = 1, 2,3,..., M − 1
+ =0
x 2 k
This equation is applicable to each of the M −
1 interior nodes, and its application gives M −
1 equations for the determination of
temperatures at M + 1 nodes.
The two additional equations needed to solve
for the M + 1 unknown nodal temperatures are
obtained by applying the energy balance on Figure 5.11: In finite difference
the two elements at the boundaries (unless, of formulation, the temperature is assumed
course, the boundary temperatures are to vary linearly between the nodes.
specified).

Access the text alternative for slide images.

©McGraw Hill LLC 15

 5.3: One-dimensional Steady Heat
Conduction 3

Figure 5.12: The assumed direction of

heat transfer at surfaces of a volume
element has no effect on the finite
difference formulation.
Access the text alternative for slide images.

©McGraw Hill LLC 16

 5.3: One-dimensional Steady Heat
Conduction 4

Boundary Conditions.
Boundary conditions most commonly
encountered in practice are the specified
temperature, specified heat flux, convection,
and radiation boundary conditions, and here
we develop the finite difference formulations
for them for the case of steady one-
dimensional heat conduction in a plane wall of
thickness L as an example.
The node number at the left surface at x = 0 is 0,
and at the right surface at x = L it is M. Note
that the width of the volume element for either
boundary node is x / 2.
Specified temperature boundary condition, Figure 5.13: Finite difference
formulation of specified
T ( 0 ) = T0 = Specified value temperature boundary conditions
T ( L ) = TM = Specified value on both surfaces of a plane wall.

©McGraw Hill LLC 17

 5.3: One-dimensional Steady Heat
Conduction 5

When specified heat flux, convection, radiation, or

combined convection and radiation conditions are
specified at a boundary, the finite difference equation for
the node at that boundary is obtained by writing an
energy balance on the volume element at that boundary.


All sides
Q + Egen, element = 0

T1 − T0
Qleft surface + kA + e0 ( Ax / 2 ) = 0
x

The finite difference form of various boundary

conditions at the left boundary:
1. Specified Heat Flux Boundary Condition. Figure 5.14: Schematic for the
finite difference formulation of
T1 − T0
+ e0 ( Ax / 2 ) = 0
the left boundary node of a
q0 A + kA
x plane wall.
Special case: Insulated Boundary ( q0 = 0 ) .
T1 − T0
kA + e0 ( Ax / 2 ) = 0
x Access the text alternative for slide images.

©McGraw Hill LLC 18

 5.3: One-dimensional Steady Heat
Conduction 6

2. Convection Boundary Condition.

T1 − T0
hA (T − T0 ) + kA + e0 ( Ax / 2 ) = 0
x
3. Radiation Boundary condition.
T1 − T0
 A (Tsurr
4
− T04 ) + kA + e0 ( Ax / 2 ) = 0
x
4. Combined Convection and Radiation Boundary
Condition.
T1 − T0
hA (T − T0 ) +  A (Tsurr
4
− T04 ) + kA + e0 ( Ax / 2 ) = 0 Figure 5.14: Schematic for
x the finite difference
or formulation of the left
boundary node of a plane
T1 − T0
hcombined A (T − T0 ) + kA
+ e0 ( Ax / 2 ) = 0 wall.
x
5. Combined Convection, Radiation, and Heat Flux Boundary Condition.
T1 − T0
q0 A + hA (T − T0 ) +  A (Tsurr
4
− T04 ) + kA + e0 ( Ax / 2 ) = 0
x
Access the text alternative for slide images.

©McGraw Hill LLC 19

 5.3: One-dimensional Steady Heat
Conduction 7

Figure 5.15: Schematic for the finite

difference formulation of combined
convection and radiation on the left
boundary of a plane wall.

T1 − T0 x
hA (T − T0 ) +  A (Tsurr
4
− T04 ) + kA + e0 A =0
x 2
Access the text alternative for slide images.

©McGraw Hill LLC 20

 5.3: One-dimensional Steady Heat
Conduction 8

6. Interface Boundary Condition Two different solid media A

and B are assumed to be in perfect contact, and thus at the
same temperature at the interface at node m.
Tm −1 − Tm T −T
kA A + k B A m +1 m + eA, m ( Ax / 2 ) + eB , m ( Ax / 2 ) = 0
x x

Figure 5.16: Schematic for the finite difference

formulation of the interface boundary condition for two
media A and B that are in perfect thermal contact.

Tm −1 − Tm T −T x x
kA A + k B A m +1 m + eA,m A + eB ,m A =0
x x 2 2

Access the text alternative for slide images.

©McGraw Hill LLC 21

 5.3: One-dimensional Steady Heat
Conduction 9

Treating Insulated Boundary Nodes as Interior Nodes: The Mirror Image Concept.

The mirror image approach can also be used

for problems that possess thermal symmetry
by replacing the plane of symmetry by a
mirror.
Alternately, we can replace the plane of
symmetry by insulation and consider only
half of the medium in the solution.
The solution in the other half of the medium
is simply the mirror image of the solution
obtained.

Tm +1 − 2Tm + Tm −1 em
+ =0
x 2
k
Figure 5.17: A node on an insulated
boundary can be treated as an interior node T1 − 2T0 + T1 e0
→ + =0
by replacing the insulation with a mirror. x 2 k
Access the text alternative for slide images.

©McGraw Hill LLC 22

 5.3: One-dimensional Steady Heat
Conduction 10

Example:
L 0.04m
x = = = 0.02 m
M −1 3 −1

Node 1:
T0 − 2T1 + T2 e1 0 − 2T1 + T2 e1
+ = 0 → + =0
x 2 k x 2 k
e1x 2
→ 2T1 − T2 = Figure 5.18: Schematic for Example 5.1.
k
Node 2:
T = 30C h = 45 W / m 2 .K
T1 − T2
hA (T − T2 ) + kA + e2 ( Ax / 2 ) = 0 2T1 − T2 = 71.43 (in C)
x
 hx  hx e2 x 2 T1 − 1.032T2 = 36.68 (in C)
T1 − 1 +  T2 = − T −
 k  k 2k

Access the text alternative for slide images.

©McGraw Hill LLC 23

 5.3: One-dimensional Steady Heat
Conduction 11

Finite difference solution:

T2 = 136.1C
Figure 5.18: Schematic for Example 5.1.
Exact solution:
Exact solution: T2 = 136.0C

0.5ehL2 / k + eL + T h ex 2 Figure 5.19: Despite being approximate

T ( x) = x− in nature, highly accurate results can be
hL + k 2k
obtained by numerical methods.

Access the text alternative for slide images.

©McGraw Hill LLC 24

 5.3: One-dimensional Steady Heat
Conduction 12

The finite difference formulation of steady

heat conduction problems usually results
in a system of N algebraic equations in N
unknown nodal temperatures that need to
be solved simultaneously.
There are numerous systematic
approaches available in the literature, and
they are broadly classified as direct and
iterative methods.
The direct methods are based on a fixed
number of well-defined steps that result in
the solution in a systematic manner.
The iterative methods are based on an
initial guess for the solution that is refined Figure 5.22: Two general categories
by iteration until a specified convergence of solution methods for solving
criterion is satisfied. systems of algebraic equations.

Access the text alternative for slide images.

©McGraw Hill LLC 25

 5.3: One-dimensional Steady Heat
Conduction 13

One of the simplest iterative methods is the Gauss-Seidel iteration.

Table 5.2: Application of the Gauss-Seidel iterative method to the finite difference equations
of Example 5.2.

Finite difference equations Nodel Nodel Nodel Nodel Nodel

° °
in explicit form, Iteration ° °
Temperature, C Temperature, C Temperature, C Temperature, C Temperature, ° C
T1 T2 T3 T4 T5
T 1 = 0.4371 T 2
+ 112.4137. Initial Guess 195.0 195.0 195.0 195.0 195.0

T 2 = 0.5826 T 1 + 0.4161. 1 197.6 196.3 195.5 194.7 193.4

T 3 + 0.0348 T 3 = 0.6238. 198.2 196.9 194.5 193.2
2 195.8
T 2 + 0.3743 T 4 + 0.0521.
3 198.5 197.2 195.9 194.5 193.2
T 4 = 0.7470 T 3 + 0.2490
T 5 + 0.1041. 4 198.6 197.3 195.9 194.5 193.2

T 5 = 0.9921 T 4 + 0.2073. 5 198.7 197.3 195.9 194.5 193.2

6 198.7 197.3 195.9 194.5 193.2

7 198.7 197.3 195.9 194.5 193.2

©McGraw Hill LLC 26

 5.4: Two-dimensional Steady Heat
Conduction 1

Sometimes we need to consider

heat transfer in other directions as
well when the variation of
temperature in other directions is
significant.
We consider the numerical
formulation and solution of two-
dimensional steady heat
conduction in rectangular
coordinates using the finite
difference method.

Figure 5.23: The nodal network for the finite

difference formulation of two dimensional
conduction in rectangular coordinates.

Access the text alternative for slide images.

©McGraw Hill LLC 27

 5.4: Two-dimensional Steady Heat
Conduction 2

( Rate of heat condution at the left, top,right,and bottom surfaces ) +

( Rate of heat generation inside the element ) =
( Rate of change of the energy content of the element )
Eelement
Qcond, left + Qcond, top + Qcond, right + Qcond, bottom + Egen, element = =0
t
Ax = y  1 = y Ay = x  1 = x
Tm −1, n − Tm , n Tm , n +1 − Tm , n Tm +1, n − Tm , n
k y + k x + k y
x y x
Tm ,n −1 − Tm ,n
+ k x + em ,n xy = 0
y
Tm −1, n − 2Tm , n + Tm +1, n Tm , n −1 − 2Tm , n + Tm , n +1 em , n
+ + =0
x 2 y 2 k
For square mesh:
em ,n l 2
Tm −1,n + Tm +1,n + Tm ,n +1 + Tm ,n −1 − 4Tm ,n + =0 Figure 5.24: The volume element of a
k general interior node (m, n) for two-
enode l 2 dimensional conduction in rectangular
Tleft + Ttop + Tright + Tbottom − 4Tnode + =0
k coordinates.
Tnode = (Tleft + Ttop + Tright + Tbottom ) / 4
Access the text alternative for slide images.
no heat generation.
©McGraw Hill LLC 28
 5.4: Two-dimensional Steady Heat
Conduction 3

Boundary Nodes.
The region is partitioned between the
nodes by forming volume elements
around the nodes, and an energy balance
is written for each boundary node.
An energy balance on a volume element
is,


All sides
Q + eVelement = 0
Qleft + Qtop + Qright + Qbottom + e2V2 = 0

We assume, for convenience in Figure 5.25: The finite difference

formulation, all heat transfer to be into formulation of a boundary node is
the volume element from all surfaces obtained by writing an energy
except for specified heat flux, whose balance on its volume element.
direction is already specified.
Access the text alternative for slide images.

©McGraw Hill LLC 29

 5.4: Two-dimensional Steady Heat
Conduction 4

Example:

Figure 5.27: Schematics for energy balances

on the volume elements of nodes 1 and 2.
x y T2 − T1 x T4 − T1 x y
Figure 5.26: Schematic for Node 1 0+h (T − T1 ) + k +k + e1 =0
2 2 x 2 y 2 2
Example 5.3 and the nodal
x = y = 1 −  2 + hl  T + T + T = − hl T − e1l
2

network (the boundaries of   1 2 4 

 k  k 2k
volume elements of the nodes Node 2
are indicated by dashed lines).
y T3 − T2 T −T y T1 − T2 y
hx (T − T2 ) + k + k x 5 2 + k + e2 x =0
2 x y 2 x 2
 2hl  2hl e2l 2
T1 −  4 +  T2 + T3 + 2T5 = − T −
 k  k k

Access the text alternative for slide images.

©McGraw Hill LLC 30

 5.4: Two-dimensional Steady Heat
Conduction 5

Node 3
 x y  x T6 − T3 y T2 − T3 x y
h +  (T − T3 ) + k +k + e3 =0
 2 2  2 y 2 x 2 2
 2hl  2
2hl el
T2 −  2 +  T3 + T6 = − T − 3
 k  k 2k
e4l 2
Node 4 T5 + T1 + T5 + T10 − 4T4 + =0
k
Figure 5.28: Schematics for
e4l 2
energy balances on the volume T10 = 90 C, T1 − 4T4 + 2T5 = −90 −
k
elements of nodes 3 and 4.
e5l 2
Node 5 T4 + T2 + T6 + T11 − 4T5 + =0
k
e5l 2
T11 = 90 C, T2 + T4 − 4T5 + T6 = −90 −
k
 2hl  2
2hl 3e l
Node 6 T12 = 90 C, T3 + 2T5 −  6 +  T6 + T7 = −180 − T − 6
 k  k 2k

 x y  y T7 − T6 T −T
Figure 5.29: Schematics for h +  (T − T6 ) + k + K x 12 6
 2 2  2 x y
energy balances on the volume T5 − T6 x T3 − T6 3xy
elements of nodes 5 and 6. + k y +k + e6 =0
x 2 y 4
Access the text alternative for slide images.

©McGraw Hill LLC 31

 5.4: Two-dimensional Steady Heat
Conduction 6

Nodes 7, 8
y T8 − T7 T −T
hx (T − T7 ) + k + k x 13 7
2 x y
y T6 − T7 y T13 = 90 C,
+k + e7 x =0
2 x 2
 2hl  2
2hl el
T6 −  4 +  T7 + T8 = −180 − T − 7
 k  k k

 2hl  2
2hl el
Figure 5.30: Schematics for T7 −  4 +  T8 + T9 = −180 − T − 8
 k  k k
energy balances on the volume
−2.064T1 + T2 + T4 = −11.2
elements of nodes 7 and 9.
T1 − 4.128T2 + T3 + 2T5 = −22.4
Node 9 T2 − 2.128T3 + T6 = −12.8
x y x T15 − T9 y T8 − T9 x y T1 − 4T4 + 2T5 = −109.2
h (T − T9 ) + qR + k +k + e9 =0
2 2 2 y 2 x 2 2 T2 + T4 − 4T5 + T6 = −109.2
x = y = 1 T15 = 90 C, T3 + 2T5 − 6.128T6 + T7 = −212.0

 hl  qRl hl e9l 2 T6 − 4.128T7 + T8 = −202.4

T8 −  2 +  T9 = −90 − − T −
 k  k k 2k T7 − 4.128T8 + T9 = −202.4
T8 − 2.064T9 = −105.2
Access the text alternative for slide images.

©McGraw Hill LLC 32

 5.4: Two-dimensional Steady Heat
Conduction 7

Irregular Boundaries. Many geometries encountered in practice

such as turbine blades or engine blocks do
not have simple shapes, and it is difficult to
fill such geometries having irregular
boundaries with simple volume elements.
A practical way of dealing with such
geometries is to replace the irregular
geometry by a series of simple volume
elements.
This simple approach is often satisfactory for
practical purposes, especially when the
Figure 5.31: Approximating an nodes are closely spaced near the boundary.
irregular boundary with a More sophisticated approaches are available
rectangular mesh.
for handling irregular boundaries, and they
are commonly incorporated into the
commercial software packages.

©McGraw Hill LLC 33

 5.5: Transient Heat Conduction 1

The finite difference solution of transient

problems requires discretization in time in
addition to discretization in space.
This is done by selecting a suitable time step
t and solving for the unknown nodal
nodal temperatures repeatedly for each t
until the solution at the desired time is obtained.
In transient problems, the superscript i is used
as the index or counter of time steps, with i = 0 Figure 5.37: Finite difference
corresponding to the specified initial condition. formulation of time-dependent
problems involves discrete
( Heat transferred into the volume element from all of its surfaces during t ) +
points in time as well as
( Heat generated within the volumeelement duringt ) =
space.
(The change in theenergy content of the volume element duringt )

t  
All sides
Q + t  Egen,element = Eelement Eelement = mc p T = Velement c p T
Eelement T Tmi +1 − Tmi

All sides
Q + Egen, element =
t
= Velement c p
t

All sides
Q + Egen,element = Velement c p
t
Access the text alternative for slide images.

©McGraw Hill LLC 34

 5.5: Transient Heat Conduction 2

Figure 5.38: The change in the energy content of

the volume element of a node during a time interval
t.

Tmi +1 − Tmi

All sides
Q + Egen, element = Velement c p
t

©McGraw Hill LLC 35

 5.5: Transient Heat Conduction 3

Explicit method: If temperatures at the previous time step i is used.

Implicit method: If temperatures at the new time step i + 1 is used.

Figure 5.39: The formulation of explicit and

implicit methods differs at the time step
(previous or new) at which the heat transfer and
heat generation terms are expressed.

Tmi +1 − Tmi
Explicit method: 
All sides
Q +Ei i
gen, element = Velement c p
t

Tmi +1 − Tmi
Implicit method: 
All sides
Q i +1
+E i +1
gen, element = Velement c p
t

©McGraw Hill LLC 36

 5.5: Transient Heat Conduction 4

Transient Heat Conduction in a Plane Wall.

Tm −1 − Tm Tm +1 − Tm Tmi +1 − Tmi
kA + kA + em Ax = pAxc p
x x t
em x 2 x 2 i +1
Tm −1 − 2Tm + Tm +1 +
k
=
t
( Tm − Tmi )  = k / pc p

t
= mesh Fourier number
x 2
em x 2 Tmi +1 − Tmi
Tm −1 − 2Tm + Tm +1 + =
k 
emi x 2 Tmi +1 − Tmi
Tm −1 − 2Tm + Tm +1 +
i i i
= ( explicit )
k 
emi x 2
Ti +1
m =  (T i
m −1 +Ti
m +1 ) + (1 − 2 )T +  k
i
m

Figure 5.40: The nodal points and i +1 i +1 emi +1x 2 Tmi +1 − Tmi
i +1
volume elements for the transient finite T − 2T + T +
m −1 m +1 = (implicit )
m
k 
difference formulation of one-dimensional
emi +1x 2
conduction in a plane wall.  Tm −1 − (1 + 2 ) Tm +  Tm +1 + 
i +1 i +1 i +1
+ Tmi = 0
k
Access the text alternative for slide images.

©McGraw Hill LLC 37

 5.5: Transient Heat Conduction 5

T1i − T0i x x T0i +1 − T0i

hA (T − T ) + kAi
+ e0 A
i
= pA c p
x t
0
2 2

 hx  i hx e0i x 2

T 0
i +1
= 1 − 2 − 2  T0 + 2 T1 + 2
i
T + 
 k  k k

No heat generation and  = 0.5.

(Tmi +1 = (Tmi −1 + Tmi +1 ) / 2

The temperature of an interior node at

the new time step is simply the average
of the temperatures of its neighboring Figure 5.41: Schematic for the explicit
nodes at the previous time step. finite difference formulation of the
convection condition at the left boundary
of a plane wall.

Access the text alternative for slide images.

©McGraw Hill LLC 38

 5.5: Transient Heat Conduction 6

Stability Criterion for Explicit Method: Limitation on t .

The explicit method is easy to use, but it suffers from an
undesirable feature that severely restricts its utility: the
explicit method is not unconditionally stable, and the largest
permissible value of the time step t is limited by the
stability criterion.
If the time step t is not sufficiently small, the solutions
obtained by the explicit method may oscillate wildly and
diverge from the actual solution.
To avoid such divergent oscillations in nodal temperatures,
the value of t must be maintained below a certain upper
limit established by the stability criterion.
t 1 (interior nodes, one-dimensional heat
= 
x 2 2 transfer in rectangular coordinates). Figure 5.42: The stability
( 0.01 m ) criterion of the explicit method
2
1 x 2
Example t  = = 111 s = 1.85 min
2   −6
2
 requires all primary coefficients
2  0.45  10 m  s

  to be positive or zero.
hx 1
1 − 2 − 2 ~ 0 or  
k 2 (1 + hx / k )
Access the text alternative for slide images.

©McGraw Hill LLC 39

 5.5: Transient Heat Conduction 7

Tmi +1 =  (Tmi −1 + Tmi +1 ) + (1 − 2 ) Tmi

The implicit method is
unconditionally stable, and thus
we can use any time step we
please with that method (of
course, the smaller the time
step, the better the accuracy of
the solution).
The disadvantage of the implicit
method is that it results in a set
of equations that must be solved
Figure 5.43: The violation of the simultaneously for each time
stability criterion in the explicit method step.
may result in the violation of the
second law of thermodynamics and Both methods are used in
thus divergence of solution. practice.

Access the text alternative for slide images.

©McGraw Hill LLC 40

 5.5: Transient Heat Conduction 8

Example:

Figure 5.44: Schematic

for Example 5.5.

L 0.04m
x = = = 0.02m
M −1 3 −1
Explicit finite difference formulation.
e1x 2
T1i +1 =  (T0 + T2i ) + (1 − 2 ) T1i +  Node 1
k
T1i − T2i x x T i +1 − T2i
hA (T − T2i ) + kA + e2 A =  A cp 2 Node 2 Figure 5.45: Schematic for the
x 2 2 t
explicit finite difference formulation
2hx e2 x 2 T2i +1 = T2i
k
(  2) ( 2 2) k = 
T − T i
+ 2 T i
− T i
+ of the convection condition at the
right boundary of a plane wall.
 hx  i  i hx e2 x 2 
T2i +1 = 1 − 2 − 2  2
T +   1
2T + 2 T + 
 k   k k 

Access the text alternative for slide images.

©McGraw Hill LLC 41

 5.5: Transient Heat Conduction 9

hx 1 x 2
1 − 2 − 2  0 →  → t   = t / x 2
k 2 (1 + hx / k ) 2a (1 + hx / k )

( 0.02m )
2

t  = 15.5 s
 2

2 12.5  10−6 m s  1 + (45W / m 2 .K ) ( 0.02m ) / 28 W / m.K 
 

=
t
=
(12.5 10 −6
m 2 / s ) (15s )
= 0.46875 (for t =15 s )
x 2 ( 0.02 m )
2

T1i +1 = 0.0625T1i + 0.46875T2i + 33.482

T2i +1 = 0.9375T1i + 0.032366T2i + 34.386

Implicit finite difference formulation.

e0 x 2
 T0 − (1 + 2 ) T 1
i +1
+ T 2
i +1
+ + T2i = 0 Node 1
k
2hx e2 x 2
k
(  2 ) ( 1 2 ) k + T2i = 0
T − T i +1
+ 2 T i +1
− T i +1
+ Node 2
 hx  i +1 e2 x 2
2 T 1
i +1
− 1 + 2 + 2  T2 + 2 + T2i = 0
 k  k
−1.9375T1i +1 + 0.46875T2i + 33.482 = 0

0.9375T1i +1 − 1.9676T2i +1 + 34.386 = 0

©McGraw Hill LLC 42
 5.5: Transient Heat Conduction 10

Node Node
Table 5.3: The variation of the Time step, i Time, s Temperature, Temperature,
°C T1i °C T2i
nodal temperatures in Example
5.5 with time obtained by the 0 0 200.0 200.0
explicit method.
1 15 139.7 228.4

2 30 149.3 172.8
3 45 123.8 179.9
4 60 125.6 156.3
5 75 114.6 157.1
6 90 114.3 146.9
7 105 109.5 146.3
8 120 108.9 141.8
9 135 106.7 141.1
10 150 106.3 139.0
20 300 103.8 136.1
30 450 103.7 136.0
40 600 103.7 136.0

©McGraw Hill LLC 43

 5.5: Transient Heat Conduction 11

Node Node
Table 5.4: The variation of the Time step, i Time, s Temperature, Temperature,
°C T1i °C T2i
nodal temperatures in example
5.5 with time obtained by the 0 0 200.0 200.0
implicit method.
1 15 168.8 199.6

2 30 150.5 190.6
3 45 138.6 180.4
4 60 130.3 171.2
5 75 124.1 163.6
6 90 119.5 157.6
7 105 115.9 152.8
8 120 113.2 149.0
9 135 111.0 146.1
10 150 109.4 143.9
20 300 104.2 136.7
30 450 103.8 136.1
40 600 103.8 136.1

©McGraw Hill LLC 44

 5.5: Transient Heat Conduction 12

Two-Dimensional Transient Heat Conduction.

Tm −1, n − Tm , n Tm , n +1 − Tm , n Tm +1, n − Tm , n
k y + k x + k y
x y x
Tm ,n −1 − Tm ,n Tmi +1 − Tmi
+ k x + em ,n xy = xy c p
y t
( x = y = 1)
em , nl 2 Tmi +1 − Tmi
Tm −1, n + Tm +1, n + Tm , n +1 + Tm , n −1 − 4Tm , n + =
k 
 = k / cp  = t / l 2
enodel 2 Tnode
i+1
− Tnode
i
Tleft + Ttop + Tright + Tbottom − 4Tnode + =
k 
Figure 5.49: The volume
i
enode l 2 Tnode
i+1
− Tnode
i
element of a general interior
Tleft + Ttop + Tright + Tbottom − 4Tnode +
i i i i i
=
k  node (m, n) for two-dimensional
i
l2
Tnode =  (Tleft + Ttop + Tright + Tbottom ) + (1 − 4 ) Tnode + 
i+1 i i i i i enode transient conduction in
k rectangular coordinates.

Access the text alternative for slide images.

©McGraw Hill LLC 45

 5.5: Transient Heat Conduction 13

i
l2
=  (T ) + (1 − 4 )T enode
T i+1
node
i
left +T +T
i
top
i
right +Ti
bottom
i
node + Explicit formulation.
k

= (Tleft )
1
In the case of no heat generation and  = i+1
Tnode i
+ Ttop
i
+ Tright
i
+ Tbottom
i
/4
4

Stability criterion
t 1 (interior nodes, two-dimensional
= 
l2 4 heat transfer in rectangular
coordinates).
1
Figure 5.50: In the case of no heat generation and  = ,
4
the temperature of an interior node at the new time
step is the average of the temperatures of its
neighboring nodes at the previous time step.

Access the text alternative for slide images.

©McGraw Hill LLC 46

 5.5: Transient Heat Conduction 14

Tmi +1 − Tmi
Example 
All sides
Q + eVelement
i
= pVelement c p
t

Figure 5.52:
Schematics for energy
balances on the
volume elements of
nodes 1 and 2.

Node 1
x y T2i − T1i x T4i − T1i
h (T − T1 ) + k
i
+k
2 2 x 2 y
x y x y T1i +1 − T1i
+ e1 = cp Figure 5.51: Schematic and nodal
2 2 2 2 t network for Example 5.7.
e1l 2 T1i +1 − T1i
2hl
k
(T − T1 ) + 2 (T2 − T1 ) + 2 (T4 − T1 ) + k = 
i i i i i

 hl  i  i hl e1l 2 
T 1
i +1
= 1 − 4 − 2  T1 + 2  T2 + T4 + T +
i

 k   k 2k 
Access the text alternative for slide images.

©McGraw Hill LLC 47

 5.5: Transient Heat Conduction 15

Node 2
y T3i − T2i T5i − T2i
hx (T − T ) + k i
+ k x
2 x y
2

y T1i − T2i y y T2i +1 − T2i

+k + e2  x = x c p
2 x 2 2 t
 hl   2hl e l2 
Figure 5.52: Schematics for energy T2i +1 = 1 − 4 − 2  T2i +   T1i + T3i + 2T5i + T + 2 
 k   k k 
balances on the volume elements of
nodes 1 and 2. Node 3
 x y  x T6i − T3i
+  (T − T3 ) + k
i
h
 2 2  2 y
y T2i − T3i x y x y T3i +1 − T3i
+k + e3 = cp
2 x 2 2 2 2 t
 hl  i  i hl e2l 2 
T3
i +1
= 1 − 4 − 4  T3 + 2  T2 + T6 + 2 T +
i

 k   k 2k 
Figure 5.53: Schematics for energy
balances on the volume elements of
Node 4
 i e4l 2 
nodes 3 and 4. T4
i +1
= (1 − 4 ) T +   T3 + 2T5 + 90 +
4
i i

 k 
Access the text alternative for slide images.

©McGraw Hill LLC 48

 5.5: Transient Heat Conduction 16

Node 5

 i e5l 2 
T 5
i +1
= (1 − 4 ) T +   T2 + T4 + T6 + 90 +
5
i i i

 k 
Node 6
Figure 5.54: Schematics for energy
 x y  y T7i − T6i T12i − T6i T5i − T6i
(  6 )
balances on the volume elements of h + T − T i
+ k + k x + k y
nodes 5 and 6.  2 2  2 x x x

x T3i − T6i 3xy 3xy T6i +1 − T6i

+ + e6 = cp
2 y 4 4 t
 hl  i
T6i +1 = 1 − 4 − 4  T3
 3k 
 hl e6l 2 
Figure 5.55: Schematics for energy +  2T + 4T + 2T + 4  90 + 4 T + 3
3
i
5
i
7
i

3 k k 
balances on the volume elements of
nodes 7 and 9.
Access the text alternative for slide images.

©McGraw Hill LLC 49

 5.5: Transient Heat Conduction 17

Nodes 7, 8
 i
− i i
− i
hx (T − T7i ) + k
y T T T T
8 7
+ k x 13 7
2 x y

y T6i − T7i y y T7i +1 − T7i

+k + e7 x = x c p
2 x 2 2 t

 hl  i  i 2hl e2l 2 
T 7
i +1
= 1 − 4 − 2  T7 +  T6 + T8 + 2  90 +
i
T + 
 k   k k 

 hl  i  i 2hl e8l 2 
T 8
i +1
= 1 − 4 − 2  T8 +  T7 + T9 + 2  90 +
i
T + 
 k   k k 

©McGraw Hill LLC 50

 5.5: Transient Heat Conduction 18

Node 9
x y x T15i − T9i
h
2
(  9 ) R 2 2 y
T − T i
+ q + k

k y T8i − T9i x y x y T9i +1 − T9i

+ + e9 = cp
2 x 2 2 2 2 t
 hl   q l hl e l2 
T9i +1 = 1 − 4 − 2  T9i + 2  T8i 90 + R + T + 9 
 k   k k 2k 

T1i +1 = 0.0836T1i + 0.444 (T2i + T4i + 11.2 )

Figure 5.55: Schematics for
energy balances on the volume
elements of nodes 7 and 9. T2i +1 = 0.0836T2i + 0.222 (T1i + T3i + 2T5i + 22.4 )

1 − 4 − 4
hl
 0 → 
1 l2 T3i +1 = 0.0552T3i + 0.444 (T2i + T6i + 12.8 )
4 (1 + hl / k )
→ t 
4 (1 + hl / k )
 = t / 2

T4i +1 = 0.112T4i + 0.222 (T1i + 2T5i + 109.2 )

k

( 0.012m ) T5i +1 = 0.112T1i + 0.222 (T2i + T4i + T6i + 109.2 )

2

t  = 10.6s
4 ( 3.2  10 m / s ) 1 + (80W / m .K ) ( 0.012m ) / (15W / m.K ) 
6 2 2

T6i +1 = 0.0931T6i + 0.074 ( 2T3i + 4T5i + 2T7i + 424 )

( 3.2  10 −6
m 2 /s ) (10s ) T7i +1 = 0.0836T7i + 0.222 (T6i + T8i + 202.4 )
t
= = = 0.222
T8i +1 = 0.0836T8i + 0.222 (T7i + T9i + 202.4 )
( 0.012m )
2
l2
T9i +1 = 0.0836T9i + 0.444 (T8i + 105.2 )

©McGraw Hill LLC 51

 5.6: A Brief History of Computational Heat
Transfer 1

Heat transfer problems can be solved by one of three methods,

analytical, experimental, or computational.
Each of these methods represents a powerful tool to solve problems; and
often problems are solved using multiple methods; the approaches
complement each other rather than compete with each other.
Analytical methods (“theoretical methods”) are the approaches taught in
textbooks with equations based on simplified assumptions of real-world
situations. Analytical methods are very useful because they are relatively
easy to apply, quick to implement, and are inexpensive. Analytical
methods are limited to very simple geometries and boundary conditions.
Experimental methods are attractive because they can be very accurate
and can account for the actual physical processes which are occurring.
Experiments only provide information where measurements are made.

©McGraw Hill LLC 52

 5.6: A Brief History of Computational Heat
Transfer 2

Numerical simulation involves creating a virtual version of a real-world

problem on a computer. The computer is used to calculate temperatures,
rates of heat transfer, flow patterns, pressures, and so on. The output of
numerical simulation is a continuous field of the solved variable
(temperature, heat transfer, pressure, velocity). An advantage of
numerical simulation is that changes to a simulation are trivially easy.

John P Abraham John P Abraham

Figure 5.57: Interactions between the arbitrary Figure 5.58: Illustration of cells that
object and the surrounding environment. define the object.
Access the text alternative for slide images.
©McGraw Hill LLC 53
 5.6: A Brief History of Computational Heat
Transfer 3

John P Abraham John P Abraham

Figure 5.59: A cell and its Figure 5.60: The four neighboring cells
neighbors. exchanging information with the central
cell.

Access the text alternative for slide images.

©McGraw Hill LLC 54
 5.6: A Brief History of Computational Heat
Transfer 4

Actual images of problems solved

with numerical simulation.

John P Abraham

Figure 5.61: Increasing zoomed-in

views of a hot rectangular block
(white square) that has flow
passing over it.

Access the text alternative for slide images.

©McGraw Hill LLC 55
 5.6: A Brief History of Computational Heat
Transfer 5

Actual images of problems

solved with numerical
simulation.

Figure 5.62: Temperature

patterns downstream of a
hot block with fluid flow
across the block.
John P Abraham

©McGraw Hill LLC 56

 5.6: A Brief History of Computational Heat
Transfer 6

Different Types of Numerical Simulation.

Two important developments have led to widespread implementation of numerical
simulations: Computer power and algorithm development.
Almost all numerical simulation algorithms use a grid of cells.
The most common and elementary approch to solving problems is the finite
difference/control volume approach. This approach treats each cell as a control
volume and conservation laws are applied directly to the cells.
The second most common method is the Finite Element Method (FEM). The FEM
uses a process called variational calculus and and the Galerkin method to
generate the matrix. A big advantage of the FEM is thet the elements can easily
be adapted to have non-uniform shapes. The drawback is that the mathematics is
more complex than for the control volume approach.
Often the control volume approach is preferred if a user is writing the code and if
the geometry is more or less rectangular. FEM is preferred for complex
geometries and commercially available codes.

©McGraw Hill LLC 57

 5.6: A Brief History of Computational
Heat Transfer 7

John P Abraham

Figure 5.63: Nodal locations for control volume

approach and finite element method.

Access the text alternative for slide images.

©McGraw Hill LLC 58

 5.6: A Brief History of Computational Heat
Transfer 8

Stability and Iteration of Numerical Simulations.

One of the first demonstrations of numerical simulation was
by Lewis Fry Richardson in 1910. The work focused on
iterative calculations of the Laplace equation.

Relaxation Methods and Time-Stepping Approaches.

The issue of stability was a major challenge to the originators
of numerical methods.
Relaxation method was developed, and still used today.
Relaxation refers to slowing the rate of change of outputs
from one iterations to next.

©McGraw Hill LLC 59

 5.6: A Brief History of Computational Heat
Transfer 9

Fourier Stability Method: A method for improving stability of unsteady

problems.
Numerical solutions of unsteady problems can be formulated as explicit
or implicit.
Explicit Method: It is simple to solve but not very stable. The time steps
have to be sufficiently small.
Implicit Method: It does not suffer from stability problems. The solutions
depend on both prior and new values.
Ti new = f ( T old values ) Explicit equation
Ti new = f ( T old and T new values ) Implicit equation

f ( T old values ) + f ( T new values )

Ti new
= Crank Nicolson Implicit equation
2

©McGraw Hill LLC 60

 5.6: A Brief History of Computational Heat
Transfer 10

Complex Geometry, Adaptive, and Multigrid Methods.

It is challenging to fit regularly shaped elements into a complex region.
One option is to use a stair-stepping approach.
An alternative approach is the body-fitted mesh. It uses a non-rectangular
elements to follow curved boundaries of the region.

John P Abraham

John P Abraham
Figure 5.65: Illustration of a body-
Figure 5.64: Illustration of a stair-step mesh. fitted mesh.

Access the text alternative for slide images.

©McGraw Hill LLC 61

 5.6: A Brief History of Computational Heat
Transfer 11

Multigrid methods solve the

problem using multiple meshes
ranging from coarse to fine.

Figure 5.66: Multigrid solutions –

speeds the transfer of information
from boundaries to interior of
domain.

John P Abraham

Access the text alternative for slide images.

©McGraw Hill LLC 62

 Summary 1

Why numerical methods?

Finite difference formulation of differential equations.
One-dimensional steady heat conduction:
• Boundary conditions.
• Treating Insulated Boundary Nodes as Interior Nodes: The Mirror Image Concept.
Two-dimensional steady heat conduction:
• Boundary Nodes.
• Irregular Boundaries.
Transient heat conduction:
• Transient Heat Conduction in a Plane Wall.
• Stability Criterion for Explicit Method: Limitation on t.
• Two-Dimensional Transient Heat Conduction.

©McGraw Hill LLC 63

 Summary 2

A Brief History of Computational Heat Transfer.

• Different Types of Numerical Simulation.
• Stability and Iteration of Numerical Simulations.
• Relaxation Methods and Time-Stepping Approaches.
• Complex Geometry, Adaptive, and Multigrid Methods.

©McGraw Hill LLC 64

 Because learning changes everything. ®

www.mheducation.com

© McGraw Hill LLC. All rights reserved. No reproduction or distribution without the prior written consent of McGraw Hill LLC.