Analysis of Two-Dimensional Elements Heat Transfer Problems:

Stiffness Matrices, ANSYS

1 Governing Equation
Three Modes of Heat Transfer
1. Conduction
2. Convection
3. Radiation

1.1 Fourier’s Law for Conduction

Two-Dimensional Heat Transfer Rate is Defined by:
∂T
qX = −kA
∂X
∂T
qY = −kA
∂Y
where:
qX , qY ≡ conduction heat transfer rate in X-Y Directions
k ≡ thermal conductivity of medium
A ≡ cross-sectional area of medium
∂T
≡ temperature gradients
∂X
1.2 Conservation of Energy
For Two-Dimensional conduction with convective and radiation boundary conditions: (Steady
state)
∂2T ∂2T
kX + k Y + q̇ = 0
∂X 2 ∂Y 2
where:
q̇ ≡ heat generation (i.e. input) per unit volume

1.3 Various Boundary Conditions

1. Constant surface temperature due to fluid phase change such as condensation or evaporation

T (0, Y ) = T0

2. Heat loss or gain at surface may be neglected

∂T
| =0
∂X (X=0, Y)

3. Constant heat flux (i.e., heat transfer rate per area) on surface
∂T
−k |X=0 = q0
∂X

4. Heat loss or gain at surface by convection

∂T
−k |X=0, Y = h [T (0, Y ) − Tf ]
∂X

5. Heat loss or gain at surface by radiation

6. Both conditions 4 and 5 simultaneously

1
2 Formulation: Linear Rectangular Elements
Equation for nodal temperatures within a linear 2-D Rectangular element:
 

 Ti 

[ ]  
(e) Tj
T = Si Sj Sm Sn

T 
 m 
Tn

where the shape functions are: ( )( )

X Y
Si = 1 − 1−
ℓ w
( )
X Y
Sj = 1−
ℓ w
XY
Sm =
ℓw
( )
Y X
Sn = 1−
w ℓ

2.1 Galerkin Approach

With the conduction heat transfer equation: remember the idea is to reduce residual (i.e. error)
to zero. Galerkin uses an average with a weighting function of the same form as solution equation.
Therefore our weight functions are the shape functions.
∫ ( )
(e) ∂2T ∂2T
Ri = Si kX + kY + q̇ dA = 0
A ∂X 2 ∂Y 2
∫ ( )
(e) ∂2T ∂2T
Rj = Sj kX + kY + q̇ dA = 0
A ∂X 2 ∂Y 2
∫ ( )
(e) ∂2T ∂2T
Rm = Sm kX + kY + q̇ dA = 0
A ∂X 2 ∂Y 2
∫ ( )
(e) ∂2T ∂2T
Rn = Sn kX + kY + q̇ dA = 0
A ∂X 2 ∂Y 2
In matrix form ∫ ( )
T ∂2T ∂2T
[S] kX + kY + q̇ dA = 0
A ∂X 2 ∂Y 2
[ ]
where: [S] = Si Sj Sm Sn . Separating into three integrals
∫ ( ) ∫ ( ) ∫
∂2T ∂2T
[S]T kx dA + [S]T
k Y dA + [S]T q̇dA = 0
A ∂X 2 A ∂Y 2
A

Now, set:
C 1 = kX
C 2 = kY ψ=T
C3 = q̇
Therefore: ∫ ( ) ∫ ( ) ∫
T ∂2ψ T ∂2ψ
[S] C1 dA + [S] C2 dA + [S]T C3 dA = 0
A ∂X 2 A ∂Y 2 A
Now the solution will apply to other problems that have similar governing differential equations.
Since our integrals are over an area, we will need to perform a very complicated double integral.
∫ ℓ∫ w
dY dX
0 0
2
The resulting ”Stiffness” Matrix is:
   
2 −2 −1 1 2 1 −1 −2
C1 w 
−2 2 1 −1 
 + C2 ℓ  1 2 −2 −1

[K](e) =   
6ℓ −1 1 2 −2 6w −1 −2 2 1
1 −1 −2 2 −2 −1 1 2

where: ℓ, w ≡ Length of rectangular element sides.

2.2 Stiffness Matrices: Conduction

Substituting for C1 and C2 , the conductance matrix is:
   
2 −2 −1 1 2 1 −1 −2
kX w −2 2 1 −1 
 + kY ℓ  1 2 −2 −1

[K](e) =
6ℓ −1 1 2 −2 6w −1 −2 2 1
1 −1 −2 2 −2 −1 1 2

2.3 Stiffness Matrices: Convection

Any convective boundary conditions along different edges will contribute to the conductance matrix
with the following matrices:  
2 1 0 0
hℓij 
 1 2 0 0

[K](e) = 
6 0 0 0 0
0 0 0 0
 
0 0 0 0
hℓjm 
0 2 1 0

[K](e) = 
6 0 1 2 0
0 0 0 0
 
0 0 0 0
hℓmn 
 0 0 0 0

[K](e) =
6  0 0 2 1
0 0 1 2
 
2 0 0 1
hℓni 
0 0 0 0

[K](e) = 
6 0 0 0 0
1 0 0 2

2.4 Load Matices: Convection

The Load Matrices along the edges due to convection are:
 

 1
hT ℓ  
f ij 1
{F}(e) =
2  0
 
0
 

 0
(e) hTf ℓjm 1
{F} =
2 
 1
  
0
 

 0
hT ℓ   
f mn 0
{F}(e) =
2 
 1
  
1
3
  

1
hTf ℓni  
0
{F}(e) =
2 
0
 
1

2.5 Load Matrices: Heat Generation

Any heat generation within an element will contribute to the thermal-load matrix for the element
with:  

 1
q̇A  
1
{F}(e) =
4 
 1
 
1

2.6 Solve
[K] {T} = {F}

3 Formulation: Linear Triangular Elements

Equation for Nodal Temperatures within a linear 2-D Triangular element:
 
[ ]  Ti 
T (e) = Si Sj Sk Tj
 
Tk

where shape functions:

1
Si = (αi + βi X + δi Y )
2A
1
Sj = (αj + βj X + δj Y )
2A
1
Sk = (αk + βk X + δk Y )
2A
And:
2A = Xi (Yj − Yk ) + Xj (Yk − Yi ) + Xk (Yi − Yj )
αi = Xj Yk − Xk Yj βi = Yj − Yk δi = Xk − Xj
αj = Xk Yi − Xi Yk βj = Yk − Yi δj = Xi − Xk
αk = Xi Yj − Xj Yi βk = Yi − Yj δk = Xj − Xi

3.1 Galerkin Approach

Again, reduce Residual (i.e. error) to zero.
∫ ( )
T ∂2T ∂2T
R= [S] kX + kY + q̇ dA = 0
A ∂X 2 ∂Y 2

3.2 Stiffness Matrices: Conduction

 2   2 
βi βi βj βi βk δi δi δj δi δk
k X  k
[K](e) = βi βj βj2 βj βk  + Y  δi δj δj2 δj δk 
4A 4A
βi βk βj βk βk2 δi δk δj δk δk2
where: A ≡ Face area of element

4
3.3 Stiffness Matrices: Convection
For convective boundary conditions along different edges
 
2 1 0
hℓij 
[K](e) = 1 2 0
6
0 0 0
 
0 0 0
hℓjk 
[K](e) = 0 2 1
6
0 1 2
 
2 0 1
hℓ ki 
[K](e) = 0 0 0
6
1 0 2

3.4 Load Matrices: Convection

Thermal Load Matrices along the edges due to convection:
 
hTf ℓij  
1
(e)
{F} = 1
2  
0
 
hTf ℓjk  
0
(e)
{F} = 1
2  
1
 
hTf ℓki  
1
(e)
{F} = 0
2  
1

3.5 Load Matrices: Heat Generation

Heat Generation within an element will contribute to the thermal-load matrix:
 
q̇A  
1
(e)
{F} = 1
3  
1

Solution: [K] {T} = {F}

5
4 ANSYS Example: Steady-State Heat Conduction
Small chimney constructed from two different materials
• Inner layer: concrete (k = 0.07 Btu/hr * in * ◦ F)
• Outer layer: bricks (k = 0.04 Btu/hr * in * ◦ F)

• Inside surface: exposed to hot gas

– T = 140◦ F
– h = 0.037 Btu/hr * in2 * ◦ F
• Outside surface
– T = 10◦ F
– h = 0.012 Btu/hr * in2 * ◦ F
• Dimensions of the chimney are shown in
figure

Determine: state conditions; plot the heat fluxes

through each layer