FINITE ELEMENTS FOR HEAT TRANSFER

PROBLEMS

1
 HEAT CONDUCTION ANALYSIS
• Analogy between Stress and Heat Conduction Analysis

Structural problem Heat transfer problem

Displacement Temperature (scalar)
Stress/strain Heat flux (vector)
Displacement B.C. Fixed temperature B.C.
Surface traction force Heat flux B.C.
Body force Internal heat generation
Young’s modulus Thermal conductivity

– In finite element viewpoint, two problems are identical if a proper

interpretation is given.
• More Complex Problems
– Coupled structural-thermal problems (thermal strain).
– Radiation problem
2
 THERMAL PROBLEM
• Goals:

[K T ]{T } = {Q}
Thermal load

Nodal temperature

Conductivity matrix

– Solve for temperature distribution for a given thermal load.

• Boundary Conditions
– Essential BC: Specified temperature
– Natural BC: Specified heat flux

3
 STEADY-STATE HEAT TRANSFER PROBLEM
• Fourier Heat Conduction Equation:
– Heat flow from high temperature to low temperature

dT
qx = −kA
dx
Thermal conductivity (W/m/C )
Heat flux (Watts)

• Examples of 1D heat conduction problems

Thigh

Thigh Tlow
qx qx

Tlow
4
 GOVERNING DIFFERENTIAL EQUATION
• Conservation of Energy
– Energy In + Energy Generated = Energy Out + Energy Increase

Ein + Egenerated = Eout + U

• Two modes of heat transfer through the boundary

– Prescribed surface heat flow Qs per unit area
(
– Convective heat transfer Qh = h T  − T)
– h: convection coefficient (W/m2/C )

¥
T
Qs

dqx
qx Qg qx + Dx
dx
A

dx
5
 GOVERNING DIFFERENTIAL EQUATION cont.
• Conservation of Energy at Steady State
– No change in internal energy (U = 0)
 
( ) dq
qx + QsPx + h T  − T Px + Qg Ax =  qx + x x 
 dx 
Ein Egen
Eout
– P: perimeter of the cross-section
dqx
dx
( )
= Qg A + hP T  − T + QsP, 0 x L

• Apply Fourier Law

d  dT 

dx 
kA 
dx 
( )
+ Qg A + hP T  − T + QsP = 0, 0 x L

– Rate of change of heat flux is equal to the sum of heat generated and
heat transferred

6
 GOVERNING DIFFERENTIAL EQUATION cont.
• Boundary conditions
– Temperature at the boundary is prescribed (essential BC)
– Heat flux is prescribed (natural BC)
– Example: essential BC at x = 0, and natural BC at x = L:
T(0) = T0

 dT
kA dx = qL
 x =L

7
 DIRECT METHOD
• Follow the same procedure with 1D bar element
– No need to use differential equation
• Element conduction equation
– Heat can enter the system only through the nodes
– Qi: heat enters at node i (Watts)
– Divide the solid into a number of elements
– Each element has two nodes and two DOFs (Ti and Tj)
– For each element, heat entering the element is positive
1 2 N

Q1 Q2 Q3 QN
i j
e
qi(e ) q(je )
Ti L(e) Tj
xi
xj
8
 ELEMENT EQUATION
• Fourier law of heat conduction

q (e)
= −kA
dT
= −kA
( Tj − Ti )
i
dx L(e)
• From the conservation of energy for the element
(Tj − Ti )
q(e)
i +q(e)
j =0 (e)
q
j = +kA
L(e)
• Combine the two equation
q(e)
i 
 kA  1 −1  Ti 
 (e)  = (e)  −1 1  T  Element conductance matrix
q j  L   j

– Similar to 1D bar element (k = E, T = u, q = f)

9
 ASSEMBLY
• Assembly using heat conservation at nodes
– Remember that heat flow into the element is positive
– Equilibrium of heat flow:
 T1   Q1 
 T  Q 
   
Ni
Qi =  qi (e)
[K T ]  2  =  2 
e =1 (NN)    
TN  QN 
– Same assembly procedure with 1D bar elements
• Applying BC
– Striking-the-rows works, but not striking-the-columns because
prescribed temperatures are not usually zero
Q2

1 3
Element 1 q2(1) 2 q2(2) Element 2
10
 EXAMPLE
• Calculate nodal temperatures of four elements
– A = 1m2, L = 1m, k = 10W/m/C

Q4 = –200W
T1 T2 T3 T4 T5
200 C x
1 2 3 4

Q1 Q2 = 500W Q3 = 0 Q5 = 0

• Element conduction equation

 1 −1  T1  q1(1)   1 −1 T2  q(2) 
  =  (1)    =  (2) 
2
10   10  
 −1 1  T2  q2   −1 1  T3  q3 

 1 −1 T3  q(3)   1 −1 T4  q(4) 

  =  (3)    =  (4) 
3 4
10   10  
 −1 1  T4  q4   −1 1  T5  q5 

11
 EXAMPLE cont.
• Assembly
 Q1   q1(1)   1 −1 0 0 0   T1 
Q   (1) (2)   −1 2 −1 0 0  T 
q
 2   2 + q2     2 

Q3  = q3 + q3  = 10  0 −1 2 −1 0  T3 
(2) (3)

Q  q(3) + q(4)    
 0 0 −1 2 −1 T4 
 4   4 (4) 4 
Q5   q5   0 0 0 −1 1 T5 

• Boundary conditions (T1 = 200 oC, Q1 is unknown)

1 −1 0 00  200   Q1 
 −1 2 −1 0 0   T2   500 
    
10  0 −1 2 −1 0   T3  =  0 
 
0 0 −1 2 −1  T4  −200 
   
 0 0 0 −1 1   T5   0 

12
 EXAMPLE cont.
• Boundary conditions
– Strike the first row
200 
 −1 2 −1 0 0     500 
   T2   
0 −1 2 − 1 0    0 
10   T =
 0 0 −1 2 −1  3  −200 
   T4   
 0 0 0 −1 1     0 
 T5 
– Instead of striking the first column, multiply the first column with
T1 = 200 oC and move to RHS

 2 −1 0 0  T2   500  2000 

 −1 2 −1 0  T   0   0 
10   3
 =
  
+


 0 −1 2 −1 T4  −200   0 
   
 0 0 −1 1  T5   0   0 
– Now, the global matrix is positive-definite and can be solved for nodal
temperatures
13
 EXAMPLE cont.
• Nodal temperatures

{T }T = {200 230 210 190 190 }°C

• How much heat input is required to maintain T1 = 200oC?

– Use the deleted first row with known nodal temperatures

Q1 = 10T1 − 10T2 + 0T3 + 0T4 + 0T5 = −300 W

• Other example 100W 3

4
2

1 2 4 5
1 3 5
50 C x

200W Q=0 Q=0

14
 GALERKIN METHOD FOR HEAT CONDUCTION
• Direct method is limited for nodal heat input
• Need more advanced method for heat generation and
convection heat transfer
• Galerkin method in Chapter 3 can be used for this purpose
• Consider element (e) (e )
i
e
j
qi q(je )
• Interpolation T L(e) T
xi i j
T(x) = TN
i i (x) + TN
j j (x) xj
 x−x  x−x
Ni (x) =  1 − (e) i  , N j (x) = (e) i
 L  L
 Ti 
T(x) = N {T } = Ni (x) N j (x)   Temperature varies linearly
Tj 
• Heat flux
dT  1 1 
= − {T } = B  {T } Heat flow is constant
dx  L(e) L(e) 
15
 GALERKIN METHOD cont.
• Differential equation with heat generation
d  dT 
 kA  + Qg A = 0, 0 x L
dx  dx 
• Substitute approximate solution
d  dT 
 kA  + AQg = R(x) Residual
dx  dx 
• Integrate the residual with Ni(x) as a weight
xj
 d  dT  
x  dx  kA dx  + AQg Ni (x)dx = 0
i  
• Integrate by parts
xj xj xj
dT dT dNi
kA Ni (x) −  kA dx = −  AQ gNi (x)dx
dx x xi
dx dx xi
i

16
 GALERKIN METHOD cont.
• Substitute interpolation relation
xj x
 dNi dNj  dNi j

x kA  Ti dx + Tj dx  dx dx = x AQgNi (x)dx − q(x j )Ni (x j ) + q(x i )Ni (x i )

i i

• Perform integration
xj
kA
L(e)
( Ti − Tj ) = Q (e)
i + q(e)
i Q(e)
i =  AQgNi (x)dx
xi

• Repeat with Nj(x) as a weight

xj

=  AQgNj (x)dx
kA
L(e)
( Tj − Ti ) = Q (e)
j + q(e)
j
Q(e)
j
xi

17
 GALERKIN METHOD cont.
• Combine the two equations
 1 −1  Ti  Qi + qi 
(e) (e)
kA [k (e) ]{ T } = {Q (e)
} + {q(e)
}
 −1 1  T  = Q(e) + q(e) 
T
L(e)    j   j j   Similar to 1D bar element

– {Q(e)}: thermal load corresponding to the heat source

– {q(e)}: vector of nodal heat flows across the cross-section
• Uniform heat source

 
xj
1
(e)
N (x) AQ L
{Q (e) } =  AQg  dx =
i g
 
xi Nj (x) 2 1

– Equally divided to the two nodes

• Temperature varies linearly in the element, and the heat
flux is constant

18
 EXAMPLE Insulated
No heat flow

• Heat chamber
Wall temperature = 200 C Wall

Uniform heat source inside 200 C

the wall Q = 400 W/m3.
Thermal conductivity of the x
wall is k = 25 W/mC.
1m
Use four elements through
the thickness (unit area) No heat flow
Boundary Condition:
x
T1 = 200, qx=1 = 0. 200 C

T1 T2 T3 T4 T5
1 2 3 4
19
 EXAMPLE cont.
• Element Matrix Equation
– All elements are identical

 1 −1  T1  50  q1(1) 

100 
−  T  = 50  +  (1) 
 1 1   2    q2 

– Assembly

 Q1   q1(1)  1 −1 0 0   T1   50 
0
Q   (1) (2)   −1 2 −1 0 0  T2  100 
q
 2   2 + q2      
(3)  
Q3  = q3 + q3  = 100  0 −1 2 −1 0  T3  − 100 
(2)

Q  q(3) + q(4)   
0 0 −1 2 −1 T4  100 
 4   4 (4) 4     
Q5   q5   0 0 0 −1 1  T5   50 

20
 EXAMPLE cont.
• Boundary Conditions
– At node 1, the temperature is given (T1 = 200). Thus, the heat flux at
node 1 (Q1) should be unknown.
– At node 5, the insulation condition required that the heat flux (Q5)
should be zero. Thus, the temperature at node 5 should be unknown.
– At nodes 2 – 4, the temperature is unknown (T2, T3, T4). Thus the heat
flux should be known.
1 2 3 4 5
Q1 Q5

1 −1 0 00  200  50 + Q1 
 −1 2 −1 0 0   T2   100 
    
100  0 −1 2 −1 0   T3  =  100 
 
0 0 −1 2 −1  T4   100 
   
 0 0 0 −1 1   T5   50 

21
 EXAMPLE cont.
• Imposing Boundary Conditions
– Remove first row because it contains unknown Q1.
– Cannot remove first column because T1 is not zero.
200 
 −1 2 −1 0 0    100 
T
 0 −1 2 −1 0   2  100 
100    T3  =  

 0 0 −1 2 −1    100 
   T4  
 0 0 0 −1 1   50 
 T5 
100( −1 200 + 2  T2 − 1 T3 ) = 100
– Instead, move the first column
100(2  T2 − 1 T3 ) = 100 + 20000
to the right.

 2 −1 0 0  T2  100  20000  20100 

 −1 2 −1 0  T  100   0   100 
100    3  =   
+
 
 
=
 


 0 −1 2 −1 T4  100   0   100 
   
 0 0 −1 1  T5   50   0   50 
22
 EXAMPLE cont.
• Solution
T1 = 200 C, T2 = 203.5 C, T3 = 206 C, T4 = 207.5 C, T5 = 208 C
208

207
FEM
206
Exact
205

204
T

203

202

201

200
0 0.2 0.4 x 0.6 0.8 1

• Discussion
– In order to maintain 200 degree at node 1, we need to remove heat

Q1 + 50 = 100T1 − 100T2 = −350

 Q1 = −400 W

23
 CONVECTION BC
• Convection Boundary Condition
– Happens when a structure is surrounded by fluid
qh Wall
– Does not exist in structural problems
– BC includes unknown temperature (mixed BC) T T

qh = hS(T  − T)
Fluid Temperature
Convection Coefficient
– Heat flow is not prescribed. Rather, it is a function of temperature on
the boundary, which is unknown
• 1D Finite Element
– When both Nodes 1 and 2 are convection boundary

 q1 = hAT1 − hAT1
  T1 T2
q2 = hAT2 − hAT2
T1 T2
24
 EXAMPLE (CONVECTION ON THE BOUNDARY)
• Element T1 T2 T3

equation T 1¥ T 3¥
h1 1 2 h3

kA  1 −1  T1  q1  kA  1 −1 T2  q2 

(1) (2)

    =  (1)      =  (2) 
L  −1 1  T2  q2  L  −1 1  T3  q3 

• Balance of heat flow


q
– Node 1: 1
(1)
= h1A(T1 − T1 )

– Node 2: q2 + q2 = 0
(1) (2)


– Node 3: q3 = h3 A(T3 − T3 )
(2)

• Global matrix equation

 1 −1 0   T1   h1A(T1 − T1 ) 


kA     
−1 2 −1 T2  =  0 
L  
 0 −1 1  T3  h3 A(T3 − T3 )
25
 EXAMPLE cont.
• Move unknown nodal temperatures to LHS
 kA kA 
L + h1A − 0 
L
  T1   h1AT1 


 − kA 2kA kA     
− T =
 2  0 
 L L L    
 kA kA   3  3 3 
T h AT
 0 − + h3 A 
 L L 
• The above matrix is P.D. because of additional positive terms
in diagonal
• How much heat flow through convection boundary?
– After solving for nodal temperature, use
q1(1) = h1A(T1 − T1 )

• This is convection at the end of an element

26
 EXAMPLE: FURNACE WALL
• Firebrick
Firebrick Insulating
k1=1.2W/m/oC brick
hi=12W/m2/oC Tf = 1,500 Ta = 20 C
• Insulating brick C
k2=0.2W/m/oC hi x
ho
ho=2.0W/m2/oC
0.25 m 0.12 m
16.8 −4.8 0   T1  18,000 
 −4.8 6.47 −1.67  T  =  0 
  2  
 0 −1.67 3.67  T3   40 
{T } T = {1,411 1,190 552}C Convection No heat flow
Convection
boundary boundary

1,500 C x 20 C

q(2)
3 = h0 (Ta − T3 ) = −1054 W/m 2

T1 T2 T3
Tf Ta
hi 1 2 ho
27
 CONVECTION ALONG A ROD

• Long rod is submerged into a fluid

• Convection occurs across the entire surface
• Governing differential equation
d  dT 

dx 
kA 
dx 
+ AQ g + hP T 
( )
− T = 0, 0  x  L P = 2(b + h)

Convection
Fluid T 

b
h
qi(e ) i
j q(je )

xi Convection
xj
28
 CONVECTION ALONG A ROD cont.
• DE with approximate temperature
d  dT 

dx 
kA
dx 

(
 + AQg + hP T − T = R(x) )
• Minimize the residual with interpolation function Ni(x)
xj
 d  dT  
x  dx  kA dx  + AQg + hP(T − T) Ni (x)dx = 0


i  

• Integration by parts
xj xj xj xj xj
dT dT dNi
kA Ni (x) −  kA dx −  hPTNidx = −  AQ gNi (x)dx −  hPT Nidx
dx x xi
dx dx xi xi xi
i

29
 CONVECTION ALONG A ROD cont.
• Substitute interpolation scheme and rearrange
xj x
 dNi dNj  dNi j

x kA  Ti dx + Tj dx  dx dx + x hP(TN
i i + TN
j j )Ni dx
i i

xj

=  (AQg + hPT  )Ni dx − q(x j )Ni (x j ) + q(x i )Ni (x i )

xi

• Perform integration and simplify

(e)  Ti
Tj 
(e) ( i
T − Tj ) + hpL  +  = Q(e)
kA
i + q(e)
i
L  3 6 
xj

 g

Q(e)
i = (AQ + hPT )Ni (x)dx
xi

• Repeat the same procedure with interpolation function Nj(x)

30
 CONVECTION ALONG A ROD cont.
• Finite element equation with convection along the rod
 kA  1 −1 hPL(e) 2 1   Ti  Qi + qi 
(e) (e)

 (e)  −1 1  + 6  1 2  T  = Q(e) + q(e) 

L       j   j j  

T ] + [k h ]  T = {Q
[k (e) (e)
 (e)
} + {q(e) }

• Equivalent conductance matrix due to convection

hPL(e)  2 1
k  =
(e)
h  1 2
6  
• Thermal load vector
(e) 
 Q  AQ L(e)
+ hPL T 1
{Q } =   =
(e) i g

Q j  2 1

31
 EXAMPLE: HEAT FLOW IN A COOLING FIN
• k = 0.2 W/mm/C, h = 210−4 W/mm2/C
• Element conductance matrix
0.2  200  1 −1 2  10 −4  320  40 2 1
[k ] + [k ] =
(e) (e)
  +  1 2
−
T h
40  1 1  6  
• Thermal load vector
Convection
2  10  320  40  30 1
−4
T  = 30 C
{Q } =
(e)

2 1
160 mm
• Element 1 1.25 mm

330 C x Insulated
120 mm

T1 T2 T3 T4
1 2 3
32
 EXAMPLE: HEAT FLOW IN A COOLING FIN cont.
• Element conduction equation
 1.8533 −0.5733   T1  38.4  q1(1) 
 = + 
– Element 1  −0.5733 1.8533  T2  38.4  q(1)
2 

 1.8533 −0.5733  T2  38.4  q(2) 

 =  +  (2) 
2
– Element 2  
 −0.5733 1.8533  T3  38.4  q3 
 1.8533 −0.5733  T3  38.4  q(3) 
– Element 3    =  +  (3) 
3

 −0.5733 1.8533  T4  38.4  q4 

• Balance of heat flow
– Node 1 q1(1) = Q1

2 + q2 = 0
q(1) (2)
– Node 2

– Node 3 q(2)
3 + q3 = 0
(2)


– Node 4 4 = hA(T − T4 )
q(3)

33
 EXAMPLE: HEAT FLOW IN A COOLING FIN cont.
• Assembly
1.853 −.573 0 0   T1   38.4 + Q1 
 −.573 3.706 −.573 0  T2   76.8 

   =  
 0 −.573 3.706 −.573  T3   76.8 
   
 0 0 −.573 1.853  T4  38.4 + hA(T − T4 )


• Move T4 to LHS and apply known T1 = 330

1.853 −.573 0 0  330  38.4 + Q1 
 −.573 3.706 −.573 0   T2   76.8 
  = 
 0 −.573 3.706 −.573   T3   76.8 
 
 0 0 −.573 1.893  
 T4   39.6 
• Move the first column to RHS after multiplying with T1=330
3.706 −.573 0  T2  265.89 
 −.573 3.706 −.573  T  =  76.8 
  3  
 0 −.573 1.893  T4   39.6 

34
 EXAMPLE: HEAT FLOW IN A COOLING FIN cont.
• Solve for temperature
T1 = 330C, T2 = 77.57C, T3 = 37.72C, T4 = 32.34C

350
300
250
200
150
100
50
0
0 40 80 120

35