Analysis of 1-D Problems

1 Where are we headed?

We are going to analyze 1-D problems in...
1. Heat transfer problems

2. Fluid mechanics problems

3. Using ANSYS

2 Heat Transfer with 1-D Linear Elements

Ex: Heat transfer through a fin surrounded by a fluid

Heat source: base

Heat transfer mode:
1. Conduction in the x-direction through fin
2. Convection to the surrounding fluid
through the fin surfaces

The Governing Differential Equation is:

d2 T
kA − hpT + hpTf = 0
dx2
where:
k ≡ Thermal Conductivity
A ≡ Cross-sectional area in direction of conduction
h ≡ Convective heat transfer coefficient
p ≡ Fin perimeter
Tf ≡ Temperature of surrounding fluid
Tb ≡ Temperature of base(i.e. source)

We are using linear elements so we need two boundary conditions.

1. T = Tb at x = 0

2. A defined tip condition (one of the follwing)

• Long fin
T = Tf at x = L
• Adiabatic tip
dT
−kA = 0 at x = L
dx
• Convection
dT
−kA = hA(TL − Tf ) at x = L
dx
The governing differential equation represents an energy balance of the fin. We will approximate
the solution with linear elements of the form:

T (e) = Si Ti + Sj Tj

where:
Xj − X X − Xi
Si = , Sj =
ℓ ℓ
In matrix form: { }
(e)
[ ] Ti
T = Si Sj
Tj
1
The general form of our differential equation is:

d2 T
C1 + C2 T + C3 = 0
dx2
where:
C1 = kA
C2 = −hp
C3 = hpTf
Let’s go one step further and make our differential equation very general and applicable to other
problems . . . set T = ψ. Now we need to solve this thing using . . . The Galerkin Formulation!!!
Remember the basic idea with weighted residuals is to assume an approximate solution, substi-
tute into the governing D.E., and solve. The result is residual or error. We then work to set that
error or residual equal to zero in different manners.
With Galerkin’s formulation, we multiply the residual by a weighting function, average it, and
set it equal to zero. The residual equation is:
∫ L
Ri = Φi (D.E.)dx = 0
0

The weighting function must be of the same form as the approximate solution. Since we are using
ψ (e) = Si ψi + Sj ψj as our approximate solution, we will use shape functions as the weighting func-
tion.

Consider three nodes in a row with elements between

The residual equation for node j then becomes:

(e) (e+1)
Rj = Rj + Rj
∫ Xj [ ](e) ∫ Xk [ ](e+1)
d2 ψ (e+1) d2 ψ
Rj = Sj (e) C1 + C 2 ψ + C 3 dX + S j C 1 + C 2 ψ + C 3 dX = 0
Xi dX 2 Xj dX 2
We will have a residual equation for each node in the finite element model. For example, if:

Then:     Where:
R1 
  
 0

(1)
R2 = R2 + R2
(2)

   
R2  0
   (2) (3)
R3 = R3 + R3
R3 = 0 (3) (4)

   
R4  0
 
R4 = R4 + R4
 
   
   ..
R5 0 .

Let’s examine the Nodal Residual Equations for nodes 1, 2, 3, 4 . . . You can draw in the arrows!
∫ X2 [ ](1)
(1) d2 ψ
R1 = S1 C1 + C2 ψ + C3 dX = 0
X1 dX 2
2
 (1) (2)
R2 = R2 + R2
∫ X2 [ ](1) ∫ X3 [ ](2)
(1) d2 ψ (2) d2 ψ
R2 = S2 C1 + C2 ψ + C3 dX + S2 C1 + C2 ψ + C3 dX = 0
X1 dX 2 X2 dX 2
(2) (3)
R3 = R3 + R3
∫ X3 [ ](2) ∫ X4 [ ](3)
(2) d2 ψ (3) d2 ψ
R3 = S3 C1 + C2 ψ + C3 dX + S3 C1 + C2 ψ + C3 dX = 0
X2 dX 2 X3 dX 2
(3) (4)
R4 = R4 + R4
∫ X4 [ ](3) ∫ X5 [ ](4)
(3) d2 ψ (4) d2 ψ
R4 = S4 C1 + C2 ψ + C3 dX + S4 C1 + C2 ψ + C3 dX = 0
X3 dX 2 X4 dX 2

..
.
In general for element (e) with nodes i and j
∫ Xj [ ](e)
(e) (e) d2 ψ
Ri = Si C1 + C2 ψ + C3 dX (1)
Xi dX 2
∫ Xk [ ](e)
(e) (e) d2 ψ
Rj = Sj C1 + C2 ψ + C3 dX (2)
Xj dX 2
Now, lets manipulate the 2nd-order term from the governing equation into a 1st-order term. Take
dψ
the derivative of (Si dX ) using the chain rule.
( )
d dψ d2 ψ dSi dψ
Si = Si +
dX dX dX 2 dX dX

Rearrange algebraically ( )
d2 ψ d dψ dSi dψ
Si 2
= Si − (3)
dX dX dX dX dX
Xj −X Xj −X
Substitute Equation 3 into Equation 1 and solve . . . knowing Si = ( ℓ ) and ψ = ( ℓ ψi +
X−Xi
ℓ ψj ):
∫ [ ( )] ∫ Xj ( ) ∫ Xj ∫ Xj
(e)
Xj
d dψ −dSi dψ
Ri = C1 Si dX + C1 dX + Si (C2 ψ)dX + Si C3 dX
Xi dX dX Xi dX dX Xi Xi

z [ }| {
z }| { ∫ X −X ( )]
(e) dψ dψ Xj −d( j ℓ ) d Xj − X X − Xi
Ri = C1 Si |X=Xj −C1 Si |X=Xi + C1 ψi + ψj dX
dX dX Xi dX dX ℓ ℓ
z }| { z}|{
C2 ℓ C2 ℓ ℓ
+ ψi + ψj + C3
3 6 2
Knowing Si = 1 at X = Xi and Si = 0 at X = Xj

(e) dψ C1 C2 ℓ C2 ℓ ℓ
Ri = −C1 |X=Xi − (ψi − ψj ) + ψi + ψj + C3
dX ℓ 3 6 2

3
Substituting Equation 3 into Equation 2 and solving gives a similar result:

(e) dψ C1 C2 ℓ C2 ℓ ℓ
R j = C1 |X=Xj − (−ψi + ψj ) + ψi + ψj + C3
dX ℓ 6 3 2
Wow!! Remember ψ is the temperature for our problem.

Assemble into Matrix form

{ } { dψ } [ ]{ } [ ]{ } { }
Ri −C1 dX |X=Xi C1 1 −1 ψi C2 ℓ 2 1 ψi C3 ℓ 1
= − + +
Rj dψ
C1 dX |X=Xj ℓ −1 1 ψj 6 1 2 ψj 2 1

Setting the residual equations to zero gives:

{ } [ ]{ } [ ]{ } { }
dψ
C1 dX |X=Xi C1 1 −1 ψi −C2 ℓ 2 1 ψi C3 ℓ 1
+ + =
dψ
−C1 dX |X=Xj ℓ −1 1 ψj 6 1 2 ψj 2 1

2.1 All Elements

Notice the form! { }
dψ
C1 dX |X=Xi { } {ψ }
(e) (e) i (e)
dψ + [K]C1 + [K]C2 = {F}C3
−C1 dX |X=Xj ψj

Substituting for the constants, we get

{ dψ } { }
C1 dX |X=Xi dT
kA dX |X=Xi
=
dψ
−C1 dX |X=Xj −kA dX
dT
|X=Xj
[ ] [ ]
(e) C1 1 −1 kA 1 −1
[K]C1 = = ←
ℓ −1 1 ℓ −1 1
[ ] [ ]
(e) −C2 ℓ 2 1 hpℓ 2 1
[K]C2 = = ←
6 1 2 6 1 2
{ } { }
(e) C3 ℓ 1 hpℓTf 1
{F}C3 = =
2 1 2 1
We can make an additional substitution to the first term when we assume the 3rd B.C. from page
280,
dT
−kA = hA(TL − Tf ) at X = L
dX
Therefore: { } { } [ ]{ } { }
dψ
C1 dX |X=Xi 0 0 0 Ti 0
= = −
dψ
−C1 dX |X=Xj hA(Tj − Tf ) 0 hA Tj hATf

There are two parts to this result: an element “stiffness” contribution and a forcing function
contribution. [ ]
(e) 0 0
[K]B.C. =
0 hA
{ }
(e) 0
{F}B.C. =
hATf
Now combine the whole thing!!
[K](e) {T} = {F}(e)
4
2.2 Tip Element ONLY
(e) (e) (e)
For only the tip element, [K](e) = [K]C1 + [K]C2 + [K]B.C.
{ [ ] [ ] [ ]}
(e) kA 1 −1 hpℓ 2 1 0 0
[K] = + +
ℓ −1 1 6 1 2 0 hA
(e) (e)
and {F}(e) = {F}C3 + {F}B.C.
{ } { }
hpℓTf 1 0
{F} (e)
= +
2 1 hATf

Wow!! This should all look familiar! Now assemble global stiffness matrix and load matrix for
the whole finite element model. Solve same as we did the mechanics solution: [K]{T} = {F}. See
two examples in textbook.

2.3 Calculating Heat Loss

Heat Loss (Q, or Q̇ from our heat transfer textbook) from fin due to convection on surfaces
∑
Qtotal = Q(e)
∫ Xj
(e)
Q = hp [(Si Ti + Sj Tj ) − Tf ] dX
Xi
∫ Xj
Q (e)
= hp (T − Tf ) dX
Xi
[( ) ]
Ti + Tj
(e)
Q = hpℓ − Tf
2

where: h ≡ Convective Heat Transfer Coefficient

p ≡ Fin Perimeter
ℓ ≡ Element Length

3 Conductance Problems with Convection Only at End

Ex: Conduction through wall and convection to outside

In this example, convection only occurs at Node 5

Therefore, for all elements except (4):

 

 kA [ ] 0
[ ] [ ]
1 −1 hpℓ
7 2 1 0 0 
[K](e) = +  + * 0

 ℓ −1 1
 6 1 2 0 
hA 
[ ]
(1) k1 A 1 −1
[K] =
ℓ1 −1 1
5
 [ ]
(2)k2 A 1 −1
[K] =
ℓ2 −1 1
[ ]
(3) k3 A 1 −1
[K] =
ℓ3 −1 1
[ ] [ ]
(4) k4 A 1 −1 0 0
[K] = +
ℓ4 −1 1 0 hA
Again, for all elements except (4):
0 } {
{ }

>
hpℓT
f 1 0
{F}(e) =  + *0

2 1 
hAT f 
{ }
0
{F} (1)
= {F} (2)
= {F} (3)
=
0
{ }
0
{F} (4)
=
hATf
Assemble Global Stiffness Matrix and Global Force Matrix
    

T1 
 
 Tsurface 

      

 (G)  T2   0 
K  T3 =
    0 
 T    0  
 4
    

T5 hATf
where we have included the B.C.: T1 = Tsurface

4 Examples of One-Dimensional Elements in ANSYS

• LINK31: Models radiation heat transfer between two points in space.
• LINK33: Uniaxial heat conduction element. Models transfer of heat between its two nodes
via conduction mode. DOF is temperature. Defined by cross-sectional area and thermal
conductivity.
• LINK34: Uniaxial heat convection element. Models transfer of heat between its two nodes
via convection mode. DOF is temperature. Defined by surface area and convection heat
transfer (film) coefficient.

ANSYS Example

6
5 Fluid Mechanics with 1-D Linear Elements

Ex: Laminar Flow in a channel

The governing differential equation is:
d2 u dp
µ − =0
dy 2 dx
where:
u ≡ fluid velocity
µ ≡ dynamic viscosity of the fluid
dp
dx ≡ pressure drop in direction of flow (x-direct)
T ≡ temperature on surfaces and in the fluid

Boundary Conditions (two for linear elements):

u = 0 at y = 0 u = 0 at y = h
Note that the governing equation is of the same general form as the equation we developed for Heat
Transfer:
d2 ψ
C1 2 + C2 ψ + C3 = 0
dx
where:
ψ=u
x=y
C1 = µ
C2 = 0
C3 = −dp
dx

From derivation of [K](e) in Heat Transfer

(e) (e)
[K](e) = [K]C1 + [K]C2

where: [ ] [ ]
(e) C1 1 −1 µ 1 −1
[K]C1 = =
ℓ −1 1 ℓ −1 1
[ ]
(e) −C2 ℓ 2 1
[K]C2 = =0
6 1 2
Therefore: [ ]
µ 1 −1
[K](e)
= ← Flow Resistance Matrix
ℓ −1 1
From derivation of {F}(e) in Heat Transfer
{ }
C3 ℓ 1
{F} (e)
=
2 1
{ }
− dp ℓ 1
{F} (e)
= dx ← Flow Forcing Matrix
2 1
dp
For fully-developed laminar flow, dx is a constant
As before, assemble global “stiffness” and “force” matrices. Then solve the Global Matrix Equation:

[K]{u} = {F}

7
6 Typical 1-D Fluid Mechanics Finite Element Model

6.1 Mass Flow Rate in the Channel

∑
ṁtotal = ṁ(e)

ṁtotal = ṁ(1) + ṁ(2) + ṁ(3) + . . .

∫ Yj
(e)
ṁ = ρW udY
Yi

where:
ρ ≡ Fluid density at each element
W ≡ Channel width (in z-direction)
u ≡ Element flow velocity
∫ Yj
(e)
ṁ = ρW (Si ui + Sj uj )dY
Yi
( )
(e) ui + uj
ṁ = ρW ℓ
2

7 Verify Results
Again, answers must be verified to some degree of certainty!! (See book for conservation of
energy in each element.)