# MAE4700/5700 Finite Element Analysis for Mechanical and Aerospace Design

Cornell University, Fall 2009 Nicholas Zabaras Materials Process Design and Control Laboratory Sibley School of Mechanical and Aerospace Engineering 101 Rhodes Hall Cornell University Ithaca, NY 14853-3801

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 1

Truss analysis
• The truss element forms the basic element of such structures. • It can only take forces along its span (no moments). • Many engineering structures consist of straight members connected at their ends by bolts, pins or welding.

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 2

Truss analysis • Internal forces in a truss element act along F1'( e ) = F2'(ye ) = 0 y the member • However, displacements at the nodes can have both components (x’- and y’-directions, in local coordinates). • This is due to rotation

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 3

Truss analysis • To analyze a truss element in the global coordinates x and y, you need to account for both components of displacement: (e) (e) (e) (e) u1x , u1 y , u2 x , u2 y • Also note that the cross section of truss elements can vary as shown.
CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 4

Stiffness of a truss element
F1( e )

⎧ F1( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u1( e ) ⎫ ⎪ ⎪ ⎪ ⎪ ⎨ (e) ⎬ = ⎢ (e) (e) ⎥ ⎨ (e) ⎬ k ⎦ ⎪u2 ⎪ ⎪ F2 ⎪ ⎣ − k ⎩ ⎭ ⎩ ⎭

• The internal force p e in the F2( e ) truss is given (see free body diagram) as: p e = F2( e ) = − F1( e ) = Aeσ e • Assuming elastic deformations: p e = F2( e ) = − F1( e ) = Ae E eε e • The (small) strain is given as:
Ae E e e F2( e ) = − F1( e ) = e (u2 − u1e ) = Finally: L e = k e (u2 − u1e )
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 5

e u2 − u1e εe = Le

Ae E e ke = e L

CORNELL
U N I V E R S I T Y

Truss element stiffness in local coordinates • We can re-write the element stiffness equations as:
⎧ F1'( e ) ⎫ x ⎡1 ⎪ '( e ) ⎪ ⎢0 ⎪ F1 y ⎪ = k (e) ⎢ ⎨ '( e ) ⎬ ⎢ −1 ⎪ F2 x ⎪ ⎢ ⎪ F '( e ) ⎪ ⎣0 ⎩ 2y ⎭
{ F '( e ) }

⎧ F1'( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u1'(xe ) ⎫ ⎪ x ⎪ ⎪ ⎪ = ⎢ (e) ⎨ '( e ) ⎬ ( e ) ⎥ ⎨ '( e ) ⎬ k ⎦ ⎪u2 x ⎪ ⎪ F2 x ⎪ ⎣ − k ⎩ ⎭ ⎩ ⎭
Notice (as it should be) that:

0 −1 0 0 0 1 0
[ K '( e ) ]

0

⎧u1'(xe ) ⎫ 0⎤ ⎥ ⎪u '( e ) ⎪ 0 ⎪ 1y ⎪ ⎥⎨ '( e ⎬ 0 ⎥ ⎪u2 x ) ⎪ ⎥ ⎪ '( e ) ⎪ 0⎦ u ⎩ 2y ⎭
{d '( e ) }

{F '( e ) } = [ K '( e ) ]{d '( e ) }

F1'( e ) = F2'(ye ) = 0 y

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 6

Element stiffness in global coordinates
• We need to be able to transform displacements from the x’ and y’ axes to displacements along the x and y axes. We start with the reverse:
The angle φ is measured anti-clockwise from x to x’
e

⎧u1'(xe ) ⎫ ⎡cos φ e ⎪ '( e ) ⎪ ⎢ e ⎪u1 y ⎪ ⎢ − sin φ ⎨ '( e ) ⎬ = ⎢ ⎪u2 x ⎪ ⎢0 ⎪u '( e ) ⎪ ⎢0 ⎩ 2y ⎭ ⎣
{ d '( e ) }

sin φ e cos φ e 0 0

0 0 cos φ e − sin φ e

⎤ ⎧u1x ⎫ ⎥ ⎪ (e) ⎪ 0 ⎥ ⎪u1 y ⎪ e ⎥ ⎨ (e) ⎬ sin φ ⎪u2 x ⎪ ⎥ e ⎪ (e) ⎪ cos φ ⎥ ⎩u2 y ⎭ ⎦ 0
(e) {d ( e ) }

T (e)

Transformation matrix T(e)

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 7

Coordinate transformation
{d 'e } = [T e ]{d e }
⎧u1'(xe ) ⎫ ⎡cos φ e ⎪ '( e ) ⎪ ⎢ e ⎪u1 y ⎪ ⎢ − sin φ ⎨ '( e ) ⎬ = ⎢ ⎪u2 x ⎪ ⎢0 ⎪u '( e ) ⎪ ⎢0 ⎩ 2y ⎭ ⎣
{d '( e ) }

⎤ ⎧u1x ⎫ ⎥ ⎪ (e) ⎪ e cos φ 0 0 ⎥ ⎪u1 y ⎪ e e ⎥ ⎨ (e) ⎬ 0 cos φ sin φ ⎪u2 x ⎪ ⎥ e e ⎪ (e) ⎪ 0 − sin φ cos φ ⎥ ⎩u2 y ⎭ ⎦ [T ( e ) ] {d ( e ) } Note that : sin φ e 0 0
(e)

[T e ]T [T e ] = [ I ]
{d e } = [T e ]T {d 'e }

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 8

Coordinate transformation Verify that:
⎡cos φ e ⎢ sin φ e ⎢ ⎢0 ⎢ ⎢0 ⎣ ⎡1 ⎢0 ⎢ ⎢0 ⎢ ⎣0 − sin φ e cos φ e 0 0 0 0 cos φ e sin φ e
Te
T

[T e ]T [T e ] = [ I ]4 x 4
sin φ e cos φ e 0 0 0 0 cos φ e − sin φ e
Te

⎤ ⎡cos φ e ⎥⎢ − sin φ e 0 ⎥⎢ − sin φ e ⎥ ⎢0 ⎥⎢ e cos φ ⎥ ⎢0 ⎦⎣ 0

⎤ ⎥ 0 ⎥ = e ⎥ sin φ ⎥ e cos φ ⎥ ⎦ 0

0 0 0⎤ 1 0 0⎥ ⎥ 0 1 0⎥ ⎥ 0 0 1⎦
I 4×4
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 9

CORNELL
U N I V E R S I T Y

Stiffness of a truss element • Similarly for the forces:
{F 'e } = [T e ]{F e }
⎧ F1'( e ) ⎫ ⎡cos φ e x ⎪ '( e ) ⎪ ⎢ e ⎪ F1 y ⎪ ⎢ − sin φ ⎨ '( e ) ⎬ = ⎢ ⎪ F2 x ⎪ ⎢0 ⎪ F '( e ) ⎪ ⎢0 ⎩ 2y ⎭ ⎣
{ F 'e }

sin φ e cos φ e 0 0

0 0 cos φ e − sin φ e

⎤ ⎧ F1x ⎫ ⎥ ⎪ (e) ⎪ 0 ⎥ ⎪ F1 y ⎪ e ⎥ ⎨ (e) ⎬ sin φ ⎪ F2 x ⎪ ⎥ e ⎪ (e) ⎪ cos φ ⎥ ⎩ F2 y ⎭ ⎦ 0
(e) {F e }

[T ( e ) ]

{F e } = [T e ]T {F 'e }

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 10

Stiffness of a truss element • Using and the transformation equations, {F '(e) } = [T (e) ]{F (e) }
{d '( e ) } = [T ( e ) ]{d ( e ) } {F '( e ) } = [ K '( e ) ]{d '( e ) }

{F '( e ) } = [ K '( e ) ]{d '( e ) } ⇒

we can write the stiffness in the x,y system as follows:

[T ( e ) ]{ F ( e ) } = [ K '( e ) ][T ( e ) ]{d ( e ) } ⇒
(e)

{F } = [T
(e)

(e) T

] [ K '( e ) ][T ( e ) ]{d ( e ) } ⇒
[ K (e) ]

[ K ] = [T ] [ K ' ][T ]
(e) T (e) (e)

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 11

Truss element stiffness

[ K ] = [T ] [ K ' ][T ]
(e) (e) T (e) (e)

⎡1 ⎢0 [ K '( e ) ] = k ( e ) ⎢ ⎢ −1 ⎢ ⎣0

0 − 1 0⎤ 0 0 0⎥ ⎥, 0 1 0⎥ ⎥ 0 0 0⎦

⎡cos φ e ⎢ − sin φ e [T ( e ) ] = ⎢ ⎢0 ⎢ ⎢0 ⎣

sin φ e cos φ e 0 0

0 0 cos φ e − sin φ e

⎤ ⎥ 0 ⎥ sin φ e ⎥ ⎥ cos φ e ⎥ ⎦ 0

⎡cos 2 φ e sin φ e cos φ e − cos 2 φ e − sin φ e cos φ e ⎤ ⎢ e e 2 e e e 2 e ⎥ − sin φ cos φ − sin φ ⎥ sin φ (e) ( e ) ⎢sin φ cos φ [K ] = k ⎢ − cos 2 φ e − sin φ e cos φ e cos 2 φ e sin φ e cos φ e ⎥ ⎢ ⎥ e e 2 e e e 2 e ⎢ − sin φ cos φ − sin φ sin φ cos φ sin φ ⎥ ⎣ ⎦

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 12

Truss element stiffness
(e) e e 2 e e e ⎧ F1(xe ) ⎫ ⎡cos 2 φ e ⎤ ⎧u1x ⎫ sin φ cos φ − cos φ − sin φ cos φ ⎪ (e) ⎪ ⎪ (e) ⎪ ⎢ e e 2 e e e 2 e ⎥ F1 y ⎪ sin φ − sin φ cos φ − sin φ ⎥ ⎪u1x ⎪ ⎪ ( e ) ⎢sin φ cos φ ⎨ (e) ⎬ = k ⎢ 2 e e e 2 e e e ⎥ ⎨ (e) ⎬ cos φ sin φ cos φ ⎪u2 x ⎪ − cos φ − sin φ cos φ ⎪ F2 x ⎪ ⎢ ⎥ ( ⎪ e e 2 e e e 2 e ⎪ F (e) ⎪ ⎢ − sin φ cos φ − sin φ sin φ cos φ sin φ ⎥ ⎪u2ey) ⎭ ⎣ ⎦⎩ ⎩ 2y ⎭

• •

Note the 2x2 symmetric submatrix structure This implies that you can reverse the numbering of nodes (1 and 2) without any changes in the element stiffness.

(e) ⎧ F2(xe ) ⎫ ⎡cos 2 φ e sin φ e cos φ e − cos 2 φ e − sin φ e cos φ e ⎤ ⎧u2 x ⎫ ⎪ (e) ⎪ ⎪ (e) ⎪ ⎢ e e 2 e e e 2 e ⎥ sin φ − sin φ cos φ − sin φ ⎥ ⎪u2 x ⎪ ⎪ F2 y ⎪ ( e ) ⎢sin φ cos φ ⎨ (e) ⎬ = k ⎢ 2 e e e 2 e e e ⎥ ⎨ (e) ⎬ cos φ sin φ cos φ ⎪u1x ⎪ − cos φ − sin φ cos φ ⎪ F1x ⎪ ⎢ ⎥ e ⎪ e e 2 e e e 2 e ⎪ F (e) ⎪ ⎢ − sin φ cos φ − sin φ sin φ cos φ sin φ ⎥ ⎪u1( y ) ⎭ ⎣ ⎦⎩ ⎩ 1y ⎭

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 13

Assembly process • The assembly process is identical to the one discussed for `spring structures’ and it will not be repeated here in its general form (no need to show at this point complicated looking matrix operations). • We will however provide soon a simple example demonstrating this assembly process.

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 14

Generalizing the application of essential BCs • You already have seen through an example how essential boundary conditions are applied to the global system of eqs: [ K ]{d } = {F } • In essence, we partition the stiffness matrix in a way that separates known from unknown degrees of freedom as follows:
⎡ K E K EF ⎤ ⎧d E ⎫ ⎧ f E ⎫ ⎢ T ⎥⎨ ⎬ = ⎨ ⎬ ⎣ K EF K F ⎦ ⎩d F ⎭ ⎩ f F ⎭
d E : Known displacements
d F : Unknown displacements

fF :

Applied (known) forces

f E : Unknown reaction forces corresponding to nodes/directions with prescribed displacement
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 15

CORNELL
U N I V E R S I T Y

Generalizing the application of essential BCs
⎡ K E K EF ⎤ ⎧d E ⎫ ⎧ f E ⎫ ⎢ T ⎥⎨ ⎬ = ⎨ ⎬⇒ ⎣ K EF K F ⎦ ⎩ d F ⎭ ⎩ f F ⎭

K E d E + K EF d F = f E
T K EF d E + K F d F = f F

The unknown displacements are obtained from the 2nd eq. as:
T − T K EF d E + K F d F = f F ⇒ d F = K F 1 ( f F − K EF d E )

With known d F , we can return to the 1st eq. to compute the reaction forces:

f E = K E d E + K EF d F Note that the matrix K F is symmetric and positive definite, so a solution for d F always exists!

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 16

A truss example
1,2
A1 ≡ A = 10−2

3,4
A2 = 2 A

5,6 2 7,8 3

Note that point D is free to move in the x direction
A3 = A

1

E = 107Pa

• Construct the global stiffness matrix and load vector • Partition the matrices and solve for the unknown displacements at point B, and displacement in x direction at point D. • Find the stresses in the three bars • Find the reactions at C, D and F

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 17

A truss example: Element 1

φ (1) = 1350

1

2

7

8

⎡1/ 2 − 1/ 2 − 1/ 2 1/ 2 ⎤ ⎢ −1/ 2 1/ 2 1/ 2 − 1/ 2 ⎥ EA ⎢ (1) ⎥ [K ] = 2 ⎢ −1/ 2 1/ 2 1/ 2 − 1/ 2 ⎥ ⎢ ⎥ − 1/ 2 1/ 2 ⎦ ⎣1/ 2 − 1/ 2

1 2 7 8

Note: Recall that you can number the corresponding global nodes in the sequence 7 8 1 2 without any changes in [ K (1) ] .

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 18

A truss example: Element 2
3,4
A2 = 2 A

φ = 900

3

4

7

8

0 0 0 ⎤ 3 ⎡0 ⎢ ⎥ 0 − 2 2⎥ 4 EA ⎢0 2 2 (2) [K ] = 0 0 0 ⎥ 7 2 ⎢0 ⎢ ⎥ ⎢0 − 2 2 0 2 2⎥ 8 ⎣ ⎦

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 19

A truss example: Element 3

A3 = A

φ = 450

5

6

7

8

1/ 2 ⎡1/ 2 ⎢1/ 2 1/ 2 EA ⎢ (3) [K ] = 2 ⎢ −1/ 2 − 1/ 2 ⎢ ⎣ −1/ 2 − 1/ 2

− 1/ 2 − 1/ 2 1/ 2 1/ 2

− 1/ 2 ⎤ − 1/ 2 ⎥ ⎥ 1/ 2 ⎥ ⎥ 1/ 2 ⎦

5 6 7 8

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 20

A truss example: Assembly (element 1)
1 2 7 8

⎡1/ 2 −1/ 2 −1/ 2 1/ 2 ⎤ ⎢ ⎥ EA ⎢−1/ 2 1/ 2 1/ 2 −1/ 2 ⎥ (1) [K ] = 2 ⎢−1/ 2 1/ 2 1/ 2 −1/ 2 ⎥ ⎢ ⎥ ⎣1/ 2 −1/ 2 −1/ 2 1/ 2⎦

1 2 7 8

1

2

3
0 0 0 0 0 0 0 0

4
0 0 0 0 0 0 0 0

5
0 0 0 0 0 0 0 0

6
0 0 0 0 0 0 0 0

7
− 1/ 2 1/ 2 0 0 0 0 1/ 2 − 1/ 2

8
− 1/ 2 ⎤ − 1/ 2 ⎥ ⎥ ⎥ 0 ⎥ 0 ⎥ 0 ⎥ ⎥ 0 ⎥ − 1 / 2⎥ ⎥ 1/ 2 ⎥ ⎦

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢0 [K ] = 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢1/ 2 − 1/ 2 ⎣

1 2 3 4 5 6 7 8

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 21

A truss example: Assembly (element 2)
3 4 7 8

⎡0 ⎢ EA ⎢ 0 [ K (2) ] = 2 ⎢0 ⎢ ⎢0 ⎣

0 2 2 0 −2 2

⎤ 3 ⎥ 0 − 2 2⎥ 4 0 0 ⎥ 7 ⎥ 0 2 2⎥ 8 ⎦ 0 0

1

2

3

4

5

6

7

8

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢0 [K ] = 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢1/ 2 − 1/ 2 ⎣

0 0 0 0 0 0 0

0 0 0 2 2 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

−1 / 2 1/ 2

0 −2 2

⎤ ⎥ ⎥ ⎥ 0 0 ⎥ −2 2 0 ⎥ ⎥ 0 0 ⎥ 0 0 ⎥ 1/ 2 + 0 − 1/ 2 + 0 ⎥ ⎥ − 1/ 2 1/ 2 + 2 2 ⎥ ⎦

− 1/ 2 −1 / 2

1 2 3 4 5 6 7 8

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 22

A truss example: Assembly (element 3)
5 6 7 8

1/ 2 ⎡1/ 2 ⎢ 1/ 2 EA ⎢1/ 2 (3) [K ] = 2 ⎢ −1/ 2 − 1/ 2 ⎢ ⎣ −1/ 2 − 1/ 2

− 1/ 2 − 1/ 2 1/ 2 1/ 2

− 1/ 2 ⎤ − 1/ 2 ⎥ ⎥ 1/ 2 ⎥ ⎥ 1/ 2 ⎦

5 6 7 8

1

2

3
0 0 0 0 0 0 0

4
0 0 0 2 2 0 0 0

5
0 0 0 0 1/ 2 1/ 2 − 1/ 2 0 0 0

6
0

7
− 1/ 2 1/ 2 0 − 1/ 2 − 1/ 2 1/ 2 + 0 + 1/ 2 − 1/ 2 + 1/ 2 − 1/ 2 − 1/ 2

8
⎤1 ⎥2 ⎥ ⎥3 0 ⎥ −2 2 ⎥4 ⎥5 − 1/ 2 ⎥ − 1/ 2 ⎥6 − 1/ 2 + 0 + 1/ 2 ⎥ 7 ⎥ 1/ 2 + 2 2 + 1/ 2 ⎥ 8 ⎦

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢0 [K ] = 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢1/ 2 − 1/ 2 ⎣

0 1/ 2 1/ 2 − 1/ 2 − 1/ 2

0 − 2 2 − 1/ 2

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 23

A truss example: Assembly

1

2

3

4

5

6

7

8

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢ 0 [K ] = 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢1/ 2 − 1/ 2 ⎣

0 0 0 0 0 0 0

0 0 0 2 2 0 0 0

0 0 0 0 1/ 2 1/ 2

0 0 0 0 1/ 2 1/ 2

− 1/ 2 1/ 2 0 0 − 1/ 2 − 1/ 2 1 0

− 1/ 2 − 1/ 2

0 − 2 2 − 1/ 2 − 1/ 2

⎤ ⎥ ⎥ ⎥ 0 ⎥ −2 2 ⎥ ⎥ − 1/ 2 ⎥ − 1/ 2 ⎥ ⎥ 0 ⎥ 1+ 2 2⎥ ⎦

− 1/ 2 − 1/ 2

1 2 3 4 5 6 7 8

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 24

A truss example: Partition and BCs

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢0 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢ ⎣1/ 2 − 1/ 2

0 0 0 0 0 0 0

0 0 0 2 2 0 0 0

0 0 0 0 1/ 2 1/ 2

0 0 0 0 1/ 2 1/ 2

− 1/ 2 1/ 2 0 0 − 1/ 2 − 1/ 2 1 0

− 1/ 2 − 1/ 2

0 − 2 2 − 1/ 2 − 1/ 2

⎤ ⎧d1 = 0 ⎫ ⎧r1 ⎫ ⎥⎪ ⎪ d 2 = 0 ⎪ ⎪r2 ⎥⎪ ⎪ ⎪⎪ ⎥ ⎪d3 = 0 ⎪ ⎪r3 0 ⎪ ⎥⎪ ⎪ ⎪⎪ r4 − 2 2 ⎥ ⎪d 4 = 0⎪ ⎪ ⎪ ⎨ ⎬⎨ ⎬ ⎥ d − 1/ 2 ⎪ 5 ⎪ ⎪0 ⎪ ⎥ ⎪ − 1/ 2 ⎥ ⎪d 6 = 0 ⎪ ⎪r6 ⎪⎪ 3 ⎪ ⎥⎪ 0 ⎪ ⎪10 Nt ⎪ ⎥ ⎪d7 ⎪ ⎪0 ⎪ ⎥⎪ 1 + 2 2 ⎦ ⎩ d8 ⎭⎩ ⎭

− 1/ 2 − 1/ 2

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 25

A truss example: Partition and BCs
dE = 0
fE

[KE ]

[ K EF ]

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢0 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢ ⎣ −1/ 2 − 1/ 2

⎤ ⎧d1 = 0 ⎫ ⎧r1 ⎫ ⎥⎪ ⎪ 1/ 2 d 2 = 0 ⎪ ⎪r2 ⎥⎪ ⎪ ⎪⎪ ⎥ ⎪d3 = 0 ⎪ ⎪r3 0 0 0 0 0 0 ⎪ ⎥⎪ ⎪ ⎪⎪ r4 −2 2 0 2 2 0 0 0 ⎪ ⎥ ⎪d 4 = 0 ⎪ ⎪ ⎨ ⎬⎨ ⎬ ⎥ d =0 r 0 0 1/ 2 1/ 2 − 1/ 2 − 1/ 2 ⎪⎪ 6 ⎪ ⎥⎪ 6 ⎪ ⎪0 ⎪ − 1/ 2 ⎥ ⎪d5 0 0 1/ 2 1/ 2 − 1/ 2 ⎪⎪ 3 ⎪ ⎥⎪ − 1/ 2 − 1/ 2 0 0 1 0 ⎪ ⎪10 Nt ⎪ ⎥ ⎪d7 ⎪ ⎪0 ⎪ 0 − 2 2 − 1/ 2 − 1/ 2 0 1 + 2 2 ⎥ ⎪ d8 ⎩ ⎭⎩ ⎭ ⎦ 0 0 0 0 0 0 0 0
[K T ] EF

− 1/ 2

− 1/ 2 − 1/ 2

[KF ]

dF

fF

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 26

A truss example: Partition and BCs

⎡1/ 2 − 1/ 2 EA ⎢ ⎢ −1/ 2 1 2⎢ ⎣ −1/ 2 0
KF

⎫ − 1/ 2 ⎤ ⎧d5 ⎫ ⎧0 ⎥⎪ ⎪ ⎪ ⎪ 0 ⎥ ⎨d 7 ⎬ = ⎨103 Nt ⎬ ⎪ ⎥⎪ ⎪ ⎪ 1 + 2 2 ⎦ ⎩ d 8 ⎭ ⎩0 ⎭
dF fF

⎧d5 ⎫ ⎧0.038284m ⎫ ⎪ ⎪ ⎪ ⎪ ⎨d 7 ⎬ = ⎨0.033284m ⎬ ⎪d ⎪ ⎪0.005m ⎪ ⎭ ⎩ 8⎭ ⎩

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 27

A truss example: Reaction calculation
f E = K E d E + K EF d F
[KE ]

dE = 0
fE

[ K EF ]

⎡1/ 2 − 1/ 2 ⎢ −1/ 2 1/ 2 ⎢ ⎢0 0 ⎢ 0 EA ⎢0 0 2 ⎢0 ⎢ 0 ⎢0 ⎢ −1/ 2 1/ 2 ⎢ ⎢ ⎣1/ 2 − 1/ 2

0 0 0 0 0 0 0

0 0 0 2 2 0 0 0

0 0 0 0

0 0 0 0

− 1/ 2 1/ 2 0 0 − 1/ 2 − 1/ 2 1 0
[KF ]

1/ 2 1/ 2 1/ 2 1/ 2 − 1/ 2 − 1/ 2

0 − 2 2 − 1/ 2 − 1/ 2
[K T ] EF

⎤ ⎧d1 = 0 ⎫ ⎧r1 ⎫ ⎥⎪ ⎪ d 2 = 0 ⎪ ⎪r2 ⎥⎪ ⎪ ⎪⎪ ⎥ ⎪d3 = 0 ⎪ ⎪r3 0 ⎪ ⎥⎪ ⎪ ⎪⎪ r4 − 2 2 ⎥ ⎪d 4 = 0⎪ ⎪ ⎪ ⎨ ⎬⎨ ⎬ ⎥ d =0 r − 1/ 2 ⎪⎪ 6 ⎪ ⎥⎪ 6 ⎪ ⎪0 ⎪ − 1/ 2 ⎥ ⎪d5 ⎪⎪ 3 ⎪ ⎥⎪ 0 ⎪ ⎪10 Nt ⎪ ⎥ ⎪d7 ⎪ ⎪ ⎪0 ⎪ 1 + 2 2 ⎥ ⎩ d8 ⎭⎩ ⎭ ⎦
dF fF

− 1/ 2 − 1/ 2

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 28

A truss example: Reaction calculation

⎡0 − 1/ 2 1/ 2 ⎤ ⎧r1 ⎫ ⎢ ⎥ ⎪r ⎪ 0 1/ 2 − 1/ 2 ⎥ ⎧d5 ⎫ ⎪ 2 ⎪ EA ⎢ ⎪ ⎪ ⎪ ⎪ ⎢0 0 0 ⎥ ⎨d 7 ⎬ ⇒ r3 ⎬ = ⎨ ⎥⎪ ⎪ 2⎢ ⎪r ⎪ 0 − 2 2 ⎥ ⎩ d8 ⎭ ⎢0 ⎪ 4⎪ ⎢ ⎥ ⎪r6 ⎪ 1/ 2 − 1/ 2 − 1/ 2 ⎦ d F ⎩ ⎭ ⎣
fE K EF

⎧r1 ⎫ ⎧ RFx ⎫ ⎧−1000 Nt ⎫ ⎪r ⎪ ⎪ R ⎪ ⎪ ⎪ ⎪ Fy ⎪ ⎪1000Nt ⎪ 2⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ r3 ⎬ ≡ ⎨ RCx ⎬ = ⎨0 ⎨ ⎬ ⎪ r ⎪ ⎪ R ⎪ ⎪ −1000 Nt ⎪ ⎪ 4 ⎪ ⎪ Cy ⎪ ⎪ ⎪ ⎪ r6 ⎪ ⎪ RDy ⎪ ⎪0 ⎪ ⎭ ⎩ ⎭ ⎩ ⎭ ⎩

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 29

A truss example: Compute the stresses
' ' u2ex − u1' e u2ex − u1' e x x εe = ⇒ σ e = Ee Le Le

However:
⎧u1'(xe ) ⎫ ⎡cos φ e ⎪ '( e ) ⎪ ⎢ e ⎪u1 y ⎪ ⎢ − sin φ ⎨ '( e ) ⎬ = ⎢ ⎪u2 x ⎪ ⎢0 ⎪u '( e ) ⎪ ⎢0 ⎩ 2y ⎭ ⎣ sin φ e cos φ e 0 0 0 0 cos φ e − sin φ e ⎤ ⎧u1x ⎫ ⎥ ⎪ (e) ⎪ 0 ⎥ ⎪u1 y ⎪ e ⎥ ⎨ (e) ⎬ sin φ ⎪u2 x ⎪ ⎥ ⎪ ( ⎪ cos φ e ⎥ ⎩u2ey) ⎭ ⎦ 0
(e)

Combining the 2 Eqs gives:
Ee σ = e ⎡ − cos φ e L ⎣
e

− sin φ

e

cos φ

e

sin φ ⎤ {d } ⎦
e e

{d e }

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 30

A truss example: Compute the stresses
Ee σ = e ⎡ − cos φ e L ⎣
e

− sin φ e

cos φ e

sin φ e ⎤ {d e } ⎦

• Applying this to each element, we have:

⎧0 ⎫ ⎪ ⎪ 2 2 2 ⎤ ⎪0 E ⎡ 2 ⎪ (1) − − σ = ⎬ = −141.421kPa ⎢ ⎥⎨ 2 2 2 ⎦ ⎪0.038284m ⎪ 2⎣ 2 ⎪0.005m ⎪ ⎩ ⎭ ⎧0 ⎫ ⎪0 ⎪ ⎪ ⎪ (2) σ = E [ 0 − 1 0 1] ⎨ ⎬ = 50kPa 0.033284m ⎪ ⎪ ⎪0.005m ⎪ ⎩ ⎭ ⎧0.038284m ⎫ ⎪ ⎪ E ⎡ 2 2 2 2 ⎤ ⎪0 ⎪ (3) − − σ = ⎨ ⎬ = 0kPa ⎢ ⎥ 2 2 2 ⎦ ⎪0.033284m ⎪ 2⎣ 2 ⎪0.005m ⎪ ⎩ ⎭
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 31

CORNELL
U N I V E R S I T Y

Three-dimensional (space) truss structures

• The local stiffness eqs are exactly as before, i.e. ⎧ F1'(xe) ⎫ ⎡ k (e) − k (e) ⎤ ⎧u1'(xe) ⎫ e ⎪ ⎪ ⎪ ⎪
⎨ '( e ) ⎬ = ⎢ ( e ) ⎪ ⎪ ⎩ F2 x ⎭ ⎣ −k k
(e)

⎥ ⎨ '( e ) ⎬ , ⎦ ⎪u2 x ⎪ ⎩ ⎭

k

(e)

A Ee =

Le

• However, we now have to include displacements and forces in the x, y and z axes.
CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 32

Three-dimensional (space) truss structures
⎧ F1'( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u1'(xe ) ⎫ ⎪ x ⎪ ⎪ ⎪ = ⎢ (e) ⎨ '( e ) ⎬ ( e ) ⎥ ⎨ '( e ) ⎬ k ⎦ ⎪u2 x ⎪ ⎪ F2 x ⎪ ⎣ −k ⎩ ⎭ ⎩ ⎭

• A unit vector along the direction x’ of a 3D truss element has the components (direction cosines of the axes between x’ and x,y,z, respectively):
l se = x e − x1e 2 Le , me = s
e y 2 − y1e

Le

, ne = s

z e − z1e 2 Le

e e e Le = ( x 2 − x1e ) 2 + ( y 2 − y1e ) 2 + ( z 2 − z1e ) 2

where ( x e , y e , z e ) and ( x2e , y2e , z e ) are the nodal coordinates in the (x,y,z) system.
1 1 1 2

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 33

Three-dimensional (space) truss structures
⎧ F1'( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u1'(xe ) ⎫ ⎪ x ⎪ ⎪ ⎪ ⎨ '( e ) ⎬ = ⎢ ( e ) ( e ) ⎥ ⎨ '( e ) ⎬ k ⎦ ⎪u2 x ⎪ ⎪ F2 x ⎪ ⎣ −k ⎩ ⎭ ⎩ ⎭

• The displacement transformation then takes the form:

e e ⎧u1'(xe ) ⎫ ⎡l s m s ⎪ ⎪ ⎨ '( e ) ⎬ = ⎢ ⎪u2 x ⎪ ⎢0 0 ⎩ ⎭ ⎣

ne s 0

0 l se
[T e ]

0 me s

• Similar transformation is applied

e ⎧u1(x ) ⎫ ⎪ (e) ⎪ ⎪u1 y ⎪ ⎪ ⎪ 0 ⎤ ⎪u1(ze ) ⎪ ⎥ ⎨ ( e ) ⎬ ≡ [T e ]{d e } n e ⎥ ⎪u2 x ⎪ s ⎦ ⎪u ( e ) ⎪ ⎪ 2y ⎪ ⎪u ( e ) ⎪ ⎩ 2z ⎭ ⎧ F1'( e ) ⎫ ⎪ x ⎪ e e e {d } ⎨ '( e ) ⎬ = [T ]{F } for the forces: ⎪ F2 x ⎪ ⎩ ⎭

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 34

Three-dimensional (space) truss structures
⎧ F1'( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u1'(xe ) ⎫ ⎪ x ⎪ ⎪ ⎪ = ⎢ (e) ⎨ '( e ) ⎬ ( e ) ⎥ ⎨ '( e ) ⎬ k ⎦ ⎪u2 x ⎪ ⎪ F2 x ⎪ ⎣ −k ⎩ ⎭ ⎩ ⎭

• Similarly to the derivation for 2D trusses, the stiffness matrix in global coordinates is then:

[ K ] ≡ [T ]
6 x6 6x2

e T

{K } [T ]
2x2 2 x6

e

e

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 35

Stiffness of a space element
⎡l se 2 mel se nel se − l se 2 s s ⎢ ee me 2 me n e − mel se ⎢ ms l s s s s s ⎢ e e nel e me ne n e 2 − n el se E A ⎢ s s s s s s (e) [K ] = e ⎢ e2 L ⎢ −l l se 2 − me l se − nel se s s s ⎢ ee −m s l s − me 2 − me ne 2 me l se s s s s ⎢ ⎢ −nel e − me ne − ne 2 nel se ⎣ s s s s s s
x e − x1e 2 Le
e y 2 − y1e

− mel se s − me 2 s − me ne s s mel se s me 2 s me ne s s
z e − z1e 2 Le

− n el se ⎤ s ⎥ e e − ms ns ⎥ ⎥ e2 − ns ⎥ ⎥ ne l se ⎥ s ⎥ me ne ⎥ s s ne 2 ⎥ ⎦ s

where :

l se =

, me = s

Le

, ne = s

e e Le = ( x 2 − x1e ) 2 + ( y 2 − y1e ) 2 + ( z e − z1e ) 2 2

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 36

Computing the stresses in a space truss element
' u2ex − u1' e x εe = ⇒ e L 'e 'e E e 'e e e u2 x − u1 x σ =E = e (u2 x − u1' e ) x Le L

However:
e e ⎧u1'(xe ) ⎫ ⎡l s m s ⎪ ⎪ ⎨ '( e ) ⎬ = ⎢ ⎪u2 x ⎪ ⎣0 0 ⎩ ⎭ ⎢

ne s 0

0 l se

0 me s

Combining the 2 Eqs gives:
Ee σ = e ⎡ −l se L ⎣
e

− me s

− ne s

l se

me s

n e ⎤ {d e } s ⎦

e ⎧u1(x ) ⎫ ⎪ (e) ⎪ ⎪u1 y ⎪ 0 ⎤ ⎪u1(ze ) ⎪ ⎪ ⎪ ⎥ ⎨ ( e ) ⎬ ≡ [T e ]{d e } ne ⎦ ⎪u2 x ⎪ ⎥ s ⎪u ( e ) ⎪ ⎪ 2y ⎪ ⎪u ( e ) ⎪ ⎩ 2z ⎭

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 37

Accounting for thermal effects in truss analysis
• Consider a truss structure that is heated. We need to e e e account for thermal expansion effects. Note: ε = ε elastic + ε thermal total strain • Hooke’s law is now modified as follows (using the x’ coordinate system):
e σ e = E eε elastic e u2 − u1e e = E e ( ε e − ε thermal ) = E e ( − α e ΔT e ) Le total strain

F

(e) 2

= −F
e

(e) 1

e u2 − u1e = p =Aσ =A E ( − α e ΔT e ) = e L εe e e e e e
0

Ae E e = k (u − u ) − A E ε , k = e L
e 2 e 1 e e e 0 e

⎧ F1( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u1( e ) ⎫ ⎪ ⎪ ⎪ ⎪ e e e ⎧ −1⎫ ⎨ (e) ⎬ + A E ε 0 ⎨ ⎬ = ⎢ (e) ⎥ ⎨ (e) ⎬ k ( e ) ⎦ ⎩u2 ⎭ ⎪ ⎪ ⎪ ⎪ ⎩1 ⎭ ⎣ − k ⎩ F2 ⎭
Therma l n odal vector

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 38

Element equations with thermal effects
• Expanding these equations to include nodal displacements in the y’ axis gives:
⎧ F1'( e ) ⎫ x ⎧−1⎫ ⎡1 ⎪ '( e ) ⎪ ⎪0 ⎪ ⎢0 ⎪ F1 y ⎪ ⎪ e ⎪ + Ae E eε 0 ⎨ ⎬ = k ( e ) ⎢ ⎨ '( e ) ⎬ ⎢ −1 ⎪ F2 x ⎪ ⎪1 ⎪ ⎢ ⎪ F '( e ) ⎪ ⎪0 ⎪ ⎩ ⎭ ⎣0 ⎩ 2y ⎭
{F '
(e)

0 −1 0 0 0
[ K '( e ) ]

0 1 0

⎧u1'(xe ) ⎫ 0⎤ ⎥ ⎪u '( e ) ⎪ 0 ⎪ 1y ⎪ ⎥⎨ '( e ⎬ 0 ⎥ ⎪u2 x ) ⎪ ⎥ 0 ⎦ ⎪u '( e ) ⎪ ⎩ 2y ⎭
{ d '( e ) }

}

e { F 'thermal }

• We need to transform this to x and y displacements (our degrees of freedom for this element)
{d 'e } = [T e ]{d e } {F 'e } = [T e ]{F e }
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 39

CORNELL
U N I V E R S I T Y

Element equations with thermal effects
⎧ F1'( e ) ⎫ x ⎧−1⎫ ⎡1 ⎪ '( e ) ⎪ ⎪ ⎪ ⎢ ⎪ F1 y ⎪ e e e ⎪0 ⎪ ( e) ⎢0 ⎨ '( e ) ⎬ + A E ε 0 ⎨ ⎬ = k ⎢ −1 ⎪ F2 x ⎪ ⎪1 ⎪ ⎢ ⎪ F '( e ) ⎪ ⎪0 ⎪ ⎩ ⎭ ⎣0 ⎩ 2y ⎭
{ F '( e ) }
e { F 'thermal }

0 −1 0 0 0 1 0 0
[ K '( e ) ]

⎧u1'(xe ) ⎫ 0⎤ ⎥ ⎪u '( e ) ⎪ 0 ⎪ 1y ⎪ ⎥⎨ '( e ⎬ 0 ⎥ ⎪u2 x ) ⎪ ⎥ 0 ⎦ ⎪u '( e ) ⎪ ⎩ 2y ⎭
{d '( e ) }

{d 'e } = [T e ]{d e } {F 'e } = [T e ]{F e }

• We can transform these element equations as follows:
e [T e ]{F e } + {F 'thermal } = [ K '( e ) ][T e ]{d e }

• From these equations, we conclude that:
e {F e } + [T e ]T {F 'thermal } = [T e ]T [ K '( e ) ][T e ]{d e }
e { Fthermal }

[ K (e) ]
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 40

CORNELL
U N I V E R S I T Y

Element equations with thermal effects
e {F e } + [T e ]T {F 'thermal } = [T e ]T [ K '( e ) ][T e ]{d e }
e { Fthermal }

[ K (e) ]

Use:

⎡cos φ e ⎢ e ⎢ − sin φ (e) [T ] = ⎢ 0 ⎢ ⎢0 ⎣

sin φ e cos φ e 0 0

0 0 cos φ
e

− sin φ e

⎤ ⎥ 0 ⎥ sin φ e ⎥ ⎥ e cos φ ⎥ ⎦ 0

(for 2D trusses)

Finally we obtain:
(e) ⎧ F1(xe ) ⎫ ⎡cos 2 φ e ⎧ − cos φ e ⎫ sin φ e cos φ e − cos 2 φ e − sin φ e cos φ e ⎤ ⎧u1x ⎫ ⎪ (e) ⎪ ⎪ ⎪ ⎢ ⎥ ⎪ (e) ⎪ F1 y ⎪ sin φ e cos φ e sin 2 φ e − sin φ e ⎪ − sin φ e cos φ e − sin 2 φ e ⎥ ⎪u1 y ⎪ ⎪ e e e ⎪ (e) ⎢ ⎨ (e) ⎬ + A E ε 0 ⎨ ⎬=k ⎢ ⎨ ( ⎬ cos φ e ⎪ cos 2 φ e sin φ e cos φ e ⎥ ⎪u2ex) ⎪ F2 x ⎪ − cos 2 φ e − sin φ e cos φ e ⎪ ⎪ ⎢ ⎥ e e 2 e e e 2 e ⎪ F (e) ⎪ ⎪sin φ e ⎪ ( ⎢ − sin φ cos φ − sin φ sin φ cos φ sin φ ⎥ ⎪u2ey) ⎪ ⎩ ⎭ ⎣ ⎦⎩ ⎭ ⎩ 2y ⎭ {F ( e ) }
e { Fthermal }

[ K (e) ]

d (e)

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 41

Element equations with thermal effects
What do you need to do to account for thermal effects in truss analysis? For each truss element that is heated, simply add to the element force, the following extra term
⎧ − cos φ e ⎫ ⎪ ⎪ − sin φ e ⎪ e ⎪ e e e Ae E eε 0 ⎨ ⎬ , where ε 0 = α ΔT e ⎪cos φ ⎪ ⎪sin φ e ⎪ ⎩ ⎭
e { Fthermal }

You will need to define at which truss elements thermal effects take place and for each of them read the values α e and ΔT e .

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 42

Principle of minimum potential energy • An alternative equivalent approach to solving many structural problems is the principle of minimum potential energy. From all the possible compatible displacements of a structure, the one that minimizes the total potential energy is the exact solution.
Potential energy for given displacements = Strain energy for these given displacements Work done by external loads on these given displacements

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 43

Principle of minimum potential energy • Let us see this principle applied to the truss problems discussed earlier.
Assembly process

PE e , PE e = U e − W e min ∑
{d } e e

PE =

Ωe

1 e e σ ε 2
Elastic strain energy density ( work / volume )

'( e dV e − ( F1'( e ) u1'(xe ) + F2'(xe ) u2 x ) ) x External Work

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 44

Principle of minimum potential energy • Lets apply this principle to one truss element. We need to minimize with respect to the nodal displacements (local '( e u1'(xe ) , u2 x ) . Recall that: coordinates)
' e u2ex − u '1x e , σ = E eε e εe = Le

1 e e2 e '( e PE = ∫ E ε dV − F1'( e ) u1'(xe ) − F2'(xe ) u2 x ) = x 2 Ωe
e ' e 1 e u2ex − u '1x 2 '( e E ( ) dV e − F1'( e ) u1'(xe ) − F2'(xe ) u2 x ) x e ∫e 2 L Ae dx ' Ω

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 45

Principle of minimum potential energy

' e 1 u2ex − u '1x 2 '( e PE e = minE e Ae Le ( ) − F1'( e ) u1'(xe ) − F2'(xe ) u2 x ) x min 'e e 'e e 2 Le u1 x ,u '2 x u1 x ,u '2 x

• Take partial derivatives of PE wrt u1 x , u ' 2 x :
∂PE e E e Ae ' e = 0 ⇒ e (u1x − u 'e x ) − F1'( e ) = 0 ⎧ F '( e ) ⎫ ⎡ k ( e ) − k ( e ) ⎤ ⎧u '( e ) ⎫ x 2 e ⎪ 1x ⎪ ⎪ 1x ⎪ ∂u '1x L ⇒ ⎨ '( e ) ⎬ = ⎢ ( e ) ( e ) ⎥ ⎨ '( e ) ⎬ e e e k ⎦ ⎪u2 x ⎪ ∂PE E A 'e ⎪ F2 x ⎪ ⎣ −k ⎩ ⎭ ⎩ ⎭ e '( e ) = 0 ⇒ e (u2 x − u '1x ) − F2 x = 0 ∂u 'e x L 2
These are the same Eqs as those obtained with the direct method!

e

'e

e

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 46

Principle of minimum potential energy • In general (not just for mechanics problems!), the principle of minimum potential energy takes the following form:
⎛ ⎛1 ⎞ ⎞ PE e = min ⎜ ∑ ⎜ {d ( e ) }T [ K ( e ) ]{d ( e ) } ⎟ − {d }T {F } ⎟ min ∑ ⎠ {d } {d } ⎝ e ⎝ 2 ⎠ e

or after assembly:
⎛1 ⎞ = min ⎜ {d }T [ K ]{d } − {d }T {F } ⎟ mind }PE {d } ⎝ 2 ⎠ {

• Note that the mimimization gives us the familiar solution: [ K ]{d } = {F }
CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 47

Principle of minimum potential energy
• We will not use this method to repeat the truss calculations. • However, it will be our starting point for computing the stiffness of beam elements (lecture 4). • The method of minimum potential energy can be applied to many problems not related to mechanics – however there are many problems where this technique is not applicable. • After discussing beam bending problems, we will need to look for more powerful (`unfortunately’ also more mathematical) methods (weak (Galerkin) formulations).

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 48

Revisiting the 2-node truss element
• Up to now we used the direct method to express the nodal loads vs. nodal displacements for the 2-node truss element. • Let us linearly interpolate the displacement ux'(e) of any point x’ in the element in terms of the nodal displacements:
'( ux e)

x ' '( e ) x ' '( e ) x' x' = (1 − e )u1x + e u2 x = [1 − e , e ] L L L L
( N1 e ) ( N 2e )

⎧u1'(xe ) ⎫ = [ N ( e ) ] {d ( e ) } ⎨ '( e ) ⎬ ⎩u2 x ⎭ basis functions
matrix

' du xe • The strain in this 2-node element ε = can now dx ' e

element basis functions

Nodal displacements

be computed as follows:

u

'( e ) x

' du xe d [ N ( e ) ]{d ( e ) } dN ( e ) ]{d ( e ) } ≡ [ B ( e ) ]{d ( e ) } = [ N ]{d } ⇒ ε = = =[ dx ' dx ' dx ' (e) (e) e

'( e ) '( e ( dN1( e ) dN 2e ) 1 1 1 1 ⎧u1x ⎫ u2 x ) − u1'(xe ) (e) e [B ] = [ , ] = [− e , e ] ⇒ ε = [− e , e ] ⎨ '( e ) ⎬ = dx ' dx ' L L L L ⎩u2 x ⎭ Le Gradient of basis functions matrix

• This is exactly the same approximation we used before (constant strain element)

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 49

Revisiting the 2-node truss element
• Let us use these interpolation formulas u '( e ) = [ N ( e ) ]{d ( e ) },ε e = [ B ( e ) ]{d ( e ) } to express the potential energy in a format that will become very familiar as we proceed forward in this course.
PE e = 1 e e2 e '( e E ε dV − F1'( e ) u1'(xe ) − F2'(xe ) u2 x ) = x ∫e 2 Ω

1 (e) T e T e e {d } [ B ] E [ B ]{d (e) } dV e − F1'(xe)u1'(xe) − F2'(xe)u2'(xe) = ∫2 Ae dx ' Ωe
εe
T

εe

=

⎤ 1 (e) T ⎡ '( e d } ⎢ ∫ [ B e ]T E e [ B e ]dV e ⎥ {d ( e ) } − F1'( e ) u1'(xe ) − F2'(xe ) u2 x ) { x 2 e ⎣Ω ⎦
[Ke ] Element stiffness matrix

For now these calculations are identical to our earlier approach! Indeed:
⎡ 1⎤ ⎢ − Le ⎥ e 1 1 e Ae E e ⎡ 1 −1⎤ e e T e e e e T e e e [ K ] = ∫ [ B ] E [ B ]dV = ∫ [ B ] E [ B ] A dx ' = ∫ ⎢ ⎥ E [− e , e ] A dx ' = e ⎢ 1 ⎥ L L L ⎣ −1 1 ⎥ e ⎦ 0 0 ⎢ Ω e e ⎥ k ⎢ L ⎦ ⎣
Le Le

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 50

Truss analysis with displacement constraints
• Up to this point, we imposed essential boundary conditions in terms of prescribed x- or y- nodal displacements. How about if the support is inclined as in the figure below:
For this problem, the constraint is that there is no normal displacement at the support 1

• Here, we don’t know the displacements at node 1 but we know the relation between u1x and u1y. In general we write these constrains on our nodal degrees of freedom as: Cd=q.

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 51

Truss analysis with displacement constraints
• Note that at node 1 we don’t have essential boundary conditions – we have a displacement constraint.

• To solve this problem we use the principle of minimum potential energy with the constraint Cd=q:
Find d and λ such that

1 min L = {d }T [ K ]{d } − {d }T {F } + {λ} {d } 2 Lagrange
Potential energy of unconstra in ed system multiplier enforcing the constra int

T

([C ]{d } − {q})
Constra int

• Here, we enforce the constraint using Lagrange multipliers.

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 52

Truss analysis with displacement constraints
Find d and λ such that

1 min L = {d }T [ K ]{d } − {d }T {F } + {λ} {d } 2 Lagrange
Potential energy of unconstra in ed system multiplier enforcing the constra int

T

([C ]{d } − {q})
Constra int

λ is the Lagrange multiplier that enforces the constraint – it is nothing else but the reaction force at node 1 (normal to the support!)

• Minimization is now performed with respect to both d and λ.
⎛K ⎜ ⎝C CT ⎞ ⎛ d ⎞ ⎛ F ⎞ ⎟⎜ ⎟ = ⎜ ⎟ ⇒ 0 ⎠⎝ λ ⎠ ⎝ q ⎠
Apply essential boundary conditions and then solve for {dF} and λ (here, reaction force at node 1)
MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 53

CORNELL
U N I V E R S I T Y

Displacement constraints: Implementation
• How do you implement this in the MatLab libraries of HW1?
u x1 sin 30 + u y1 cos 30 = 0

• Introduce the constraints in the InputData.m and then modify the stiffness and load vectors in the NodalSoln.m. • Apply the essential boundary conditions first before you augment the reduced global equations (Kf) with the Lagrange multiplier.
C = zeros(1,neq-length(debc)); % Read information for constraints %The dimension of C is neq minus the % prescribed DOF via essential BCs % Here there is only one constraint

InputData.m
C(1) = sin(pi/6); C(2) = cos(pi/6); q = 0;

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 54

Displacement constraints: NodalSoln.m
function [d, rf, lambda] = NodalSoln(K, R, debc, ebcVals, C, q) % K=global stiffness, R=global force, debc=degrees of freedom with specified values, ebcVals=specified displacements dof = length(R); df = setdiff(1:dof, debc); % Extract the total degreess-of-freedom % Sets the difference between two sets of indices, i.e. the global degrees of freedom minus the % degrees of freedom with prescribed essential boundary conditions % Remove eqs. corresponding to prescribed displacements

Kf = K(df, df);

Rf = R(df) - K(df, debc)*ebcVals; % Modify the remaining load vector to account for the essential boundary conditions [m n] = size(C); Kf = [Kf C'; C zeros(m)]; Rf = [Rf;q]; dfVals = Kf\Rf; % Extract number of constraints % Augment global equations with the Lagrange multipliers

⎛K ⎜ ⎝C

CT 0

⎞⎛ d ⎞ ⎛ F ⎞ ⎟⎜ ⎟ = ⎜ ⎟ ⎠⎝ λ ⎠ ⎝ q ⎠

% Augment load vector % Solve the linear system of equations. Here for simplicity, we use Gauss elimination.

d = zeros(dof,1); % Restore the solution vector (i.e. include back the nodes with prescribed displacements). d(debc) = ebcVals; % Use the originally established ordering of the nodes. d(df) = dfVals(1:(length(dfVals)-m)); rf = K(debc,:)*d - R(debc); % Calculate the reaction vector at nodes with prescribed displacements lambda = dfVals((length(dfVals)-m+1):length(dfVals)); % Calculate Langrange multipliers (reactions at nodes with constraints)

CORNELL
U N I V E R S I T Y

MAE 4700 – FE Analysis for Mechanical & Aerospace Design N. Zabaras (9/7/2009) 55

Sign up to vote on this title