Professional Documents
Culture Documents
1 Introduction
Where are we headed? We will be examining straight-line elements (i.e., 1-D) that contain an
unknown variable.
Ex: Heat Transfer from a Fin
We will represent the spatial variation of the unknown variable (e.g., temperature) using the fol-
lowing functions:
• Linear Functions (e.g., T = Ax + B)
• Quadratic Functions (e.g., T = Ax2 + Bx + C)
• Cubic Functions (e.g., T = Ax3 + Bx2 + Cx + D)
We will also examine:
1. Coordinate systems
2. Special Elements called Isoparametric
3. Numerical Integration
2 Linear Elements
Returning to the heat transfer fin, Let’s approximate the temperature distribution with a combi-
nation of linear functions.
Note:
• More elements...more accuracy (and more time)
• Elements do not need to be the same length so make elements shorter on the left side (large
gradient)
1
2.1 The Element
Approximate temperature within an element as
a linear function
T (e) = c1 + c2 X
B.C.’s:
T = Ti at X = Xi
T = Tj at X = Xj
Substituting
Ti Xj − Tj Xi Tj − Ti
T (e) = + X
Xj − Xi Xj − Xi
As with Axial Elements, T (e) is the temperature in the element at any location X.
Rearranging: ( ) ( )
(e) Xj − X X − Xi
T = Ti + Tj
Xj − Xi Xj − Xi
If shape functions are:
Xj − X Xj − X X − Xi X − Xi
Si = = and Sj = =
Xj − Xi ℓ Xj − Xi ℓ
Then: T (e) = Si Ti + Sj Tj
In matrix form, { }
(e)
[ ] Ti
T = Si Sj
Tj
Notice how this whole derivation is analogous to the axial elements derivation.
Ψ(e) = c1 + c2 X
Xj − X X − Xi
Si = Sj =
ℓ ℓ
{ }
[ ] Ψi
Ψ(e) = Si Sj
Ψj
2
3 Properties of Shape Functions
These properties seem trivial now, but will be very useful later :
1. Each shape functions has a value of 1 at its corresponding node. (i.e., Si = 1 at X = Xi and
Sj = 1 at X = Xj )
4 Quadratic Elements
• The accuracy of our finite element model can be increased via two different methods:
– Increasing the number of linear elements
– Using a higher order approximating function.
• Returning to the heat transfer fin, let’s approximate the temperature distribution with a
combination of quadratic elements.
Ex:
Ti = c1 + c2 Xi + c3 Xi2
Tj = c1 + c2 Xj + c3 Xj2
Tk = c1 + c2 Xk + c3 Xk2
Solving for c1 , c2 , and c3 simultaneously and then substituting back in, we get:
[ ] [ ] [ ]
2 2 −4
T = 2 (X − Xj )(X − Xk ) Ti + 2 (X − Xi )(X − Xk ) Tj + 2 (X − Xi )(X − Xj ) Tk
(e)
ℓ ℓ ℓ
3
If shape functions are:
2
Si = (X − Xj )(X − Xk )
ℓ2
2
Sj = 2 (X − Xi )(X − Xk )
ℓ
−4
Sk = 2 (X − Xi )(X − Xj )
ℓ
Then: T (e) = Si Ti + Sj Tj + Sk Tk
In matrix form:
[ ] Ti
T (e) = Si Sj Sk Tj
Tk
Ψ(e) = c1 + c2 X + c3 X 2
In matrix form:
[ ] Ψi
Ψ(e) = Si Sj Sk Ψj
Ψk
Notice the results of the shape functions at each node:
1. Si = 1 at X = Xi , Sj = 1 at X = Xj , Sk = 1 at X = Xk
3. Si + Sj + Sk = 1
5 Cubic Elements
• We can achieve even more accuracy using a higher order approximation function.
• Let’s approximate temperature distribution in the heat transfer fin using a combination of
cubic elements.
• One-dimensional cubic elements require four nodes to define the element. (Need 4 B.C.’s for
the 4 unknown coefficients.)
4
Ex:
T (e) = c1 + c2 X + c3 X 2 + c4 X 3
Resulting Equation
T (e) = Si Ti + Sj Tj + Sk Tk + Sm Tm
Ψ(e) = c1 + c2 X + c3 X 2 + c4 X 3
1. Si = 1 at X = Xi , Sj = 1 at X = Xj ,
Sk = 1 at X = Xk , Sm = 1 at X = Xm
2. Si = 0 at X = Xj , Xk , & Xm , Sj = 0 at X = Xi , Xk , & Xm ,
Sk = 0 at X = Xi , Xj , & Xm , Sm = 0 at Xi , Xj & Xk
3. Si + Sj + Sk + Sm = 1
• Lagrange Method
– Keeps us from solving for c1 , c2 , c3 , c4 etc. by using simultaneous equations
– Instead, we directly write shape functions by multiplying several linear functions.
∏
N
X − XM omitting (X − XK ) (X − X1 )(X − X2 ) · · · (X − XN )
SK = =
XK − XM omitting (XK − XK ) (XK − X1 )(XK − X2 ) · · · (XK − XN )
M =1
for an (N - 1)-order polynomial. Therefore, for a 3rd -order polynomial, N = 4; for a 5t h-order
polynomial, N = 6.
Example
Use the Lagrange Polynomial formula to find the shape functions for the cubic element.
N − 1 = 3 ∴ N = 4, K = 1, 2, 3, 4
∏
N
X − XM omitting (X − XK )
SK =
XK − XM omitting (XK − XM )
M =1
For node i, K = 1:
(X − Xj )(X − Xk )(X − Xm )
Si =
( −ℓ −2ℓ
3 )( 3 )(−ℓ)
−9
Si = (X − Xj )(X − Xk )(X − Xm )
2ℓ3
For node k, K = 2:
(X − Xi )(X − Xm )(X − Xj )
Sk =
( 3ℓ )( −ℓ −2ℓ
3 )( 3 )
6
27
Sk = (X − Xi )(X − Xm )(X − Xj )
2ℓ3
For node m, K = 3:
(X − Xi )(X − Xk )(X − Xj )
Sm = ℓ −ℓ
( 2ℓ
3 )( 3 )( 3 )
−27
Sm = (X − Xi )(X − Xk )(X − Xj )
2ℓ3
For node j, K = 4:
X = Xi + x
Linear shape functions in Local Coordinates
Xj − X X − Xi
Si = Sj =
ℓ ℓ
Xj − (Xi + x) (Xi + x) − Xi
Si = Sj =
ℓ ℓ
x x
Si = 1 − Sj =
ℓ ℓ
Notice: 0 ≤ x ≤ ℓ
7
8 5.4 Natural Coordinates in 1-D
• Natural Coordinates are basically local coordinates in a dimensionless form.
• Natural Coordinates simplify numerical integration
• Notation is ξ, lower case Xi
• The objective is for ξ to be −1 at i and 1 at j.
2x ℓ
ξ= − 1 or x = (ξ + 1)
ℓ 2
at x = 0 → ξ = −1
at x = ℓ → ξ = 1
Let’s see what Linear Shape Functions look like in natural coordinate form:
x x
Si = 1 − Sj =
ℓ ℓ
( ) ( )
1 ℓ 1 ℓ
Si = 1 − (ξ + 1) Sj = (ξ + 1)
ℓ 2 ℓ 2
1 1
Si = 1 − (ξ + 1) Sj = (1 + ξ)
2 2
1 1
Si = 1 − ξ −
2 2
1
Si = (1 − ξ)
2
Notice shape function properties are retained for natural coordinates:
1. Si = 1 at X = Xi (ξ = −1)
Sj = 1 at X = Xj (ξ = 1)
2. Si = 0 at X = Xj (ξ = 1)
Sj = 0 at X = Xi (ξ = −1)
3. Si + Sj = 1