Lecture 5: 3-D Rotation Matrices.
§3.7 Transformation Matrix and
Stiffness Matrix in Three-
Dimensional Space. G
G z
The displacement vector d is a real
d
ẑ
vector entity.
y
It is independent of the frame used to
define it. o
G G G G x̂
d = d x i + d y j + d z k = dˆx iˆ + dˆ y ˆj + dˆz kˆ x
We are interested in calculating what ŷ
the global coordinate representation
is based on elemental coordinates,
and vice versa.
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Solve for the global displacements one at a time:
⎧ dˆx ⎫
G G G ˆ G ˆ G ˆ G G G ⎪⎪ ˆ ⎪⎪
d ⋅ i = d x = (i ⋅ i )d x + ( j ⋅ i )d y + (k ⋅ i )d z = ⎡⎣(i ⋅ i ) ( j ⋅ i ) ( k ⋅ i ) ⎤⎦ ⎨ d y ⎬
ˆ ˆ ˆ ˆ ˆ ˆ
⎪ˆ ⎪
⎪⎩ d z ⎭⎪
⎧ dˆx ⎫
G G G ˆ G ˆ G ˆ G G G ⎪⎪ ˆ ⎪⎪
d ⋅ j = d y = (i ⋅ j )d x + ( j ⋅ j )d y + (k ⋅ j )d z = ⎣(i ⋅ j ) ( j ⋅ j ) (k ⋅ j ) ⎤⎦ ⎨d y ⎬
ˆ ˆ ˆ ⎡ ˆ ˆ ˆ
⎪ˆ ⎪
⎪⎩ d z ⎭⎪
⎧ dˆx ⎫
G G G G G G G G ⎪⎪ ⎪⎪
d ⋅ k = d z = (iˆ ⋅ k )d x + ( ˆj ⋅ k )d y + (k ⋅ k )d z = ⎡⎣ (iˆ ⋅ k ) ( ˆj ⋅ k ) (k ⋅ k ) ⎤⎦ ⎨ dˆ y ⎬
ˆ ˆ ˆ ˆ ˆ
⎪ˆ ⎪
⎪⎩ d z ⎭⎪
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Or, we could collect the three scalar equations into a matrix
form:
G G ˆ ⋅ iG ) ⎤ ⎧ dˆ ⎫
⎧d x ⎫ ⎢ ⎡ (iˆ ⋅ i ) ( ˆ
j ⋅ i ) ( k
G ⎪ ˆ ⎪⎪
⎥ ⎪
x
⎪ ⎪ G G
⎨d y ⎬ = ⎢ (i ⋅ j ) ( j ⋅ j ) (k ⋅ j ) ⎥ ⎨d y ⎬
ˆ ˆ ˆ
⎪d ⎪ ⎢ ˆ G G G ⎥⎪ ⎪
⎩ z ⎭ ⎢⎣ (i ⋅ k ) ( j ⋅ k ) (k ⋅ k ) ⎥⎦ ⎩⎪ dˆz ⎭⎪
ˆ ˆ
R3×3
Looking at the columns of the rotation matrix R it is apparent
that it must be orthonormal.
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
The entries of the rotation matrix are referred to as the
direction cosines.
G G
( )
G G
iˆ ⋅ i = iˆ i cos θ x = cos θ x ( )
G G
ˆj ⋅ i = ˆj i cosψ x = cosψ x ( )
k ⋅ i = k i cos φx = cos φx
ˆ ˆ
G G G G G G
( )
ˆi ⋅ j = iˆ j cos θ y = cos θ y ( )
ˆj ⋅ j = ˆj j cosψ y = cosψ y ( )
kˆ ⋅ j = kˆ j cos φ y = cos φ y
G G G G G G
( )
iˆ ⋅ k = iˆ k cos θ z = cos θ z ( )
ˆj ⋅ k = ˆj k cosψ z = cosψ z
( )
k ⋅ k = k k cos φz = cos φz
ˆ ˆ
Although it appears that there are nine values that make
up the rotation matrix only three are independent.
⎡ ⎤ r1 ⋅ r1 = 1 r2 ⋅ r2 = 1
R = ⎢⎢ r1 r2 r3 ⎥⎥ r1 ⋅ r2 = 0 r2 ⋅ r3 = 0
⎢⎣ ⎥⎦ 3×3 r1 ⋅ r3 = 0 r3 ⋅ r3 = 1
iˆ ĵ k̂
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
The course textbook uses the
variable ‘θ’ to refer to the
direction angles for the element
x axis.
Note of the three direction
cosines: cosθx, cos θy and
cosθz, only two are
independent.
To specify the orientation of
the elemental frame, we would
need one more direction cosine
of either of the element y or z
axes.
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
We need to go between element coordinates and global
coordinates for any vector.
⎧ xˆ ⎫ ⎧ x ⎫ ⎡ cθ x cθ y cθ z ⎤ ⎧ x ⎫
⎪ ⎪ ⎪ ⎪ ⎢ ⎥⎪ ⎪
⎨ yˆ ⎬ = R ⎨ y ⎬ = ⎢cψ x cψ y cψ z ⎥ ⎨ y ⎬
T
⎪ zˆ ⎪ ⎪ ⎪ cφz ⎥⎦ ⎩⎪ z ⎭⎪
⎩ ⎭ ⎩ z ⎭ ⎢⎣ cφx cφ y
Applying the transformation above to the element node force and
displacement vectors…
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
⎧ ⎧ f1x ⎫⎫ ⎡1 0 0 −1 0 0 ⎤ ⎧ ⎧ d1x ⎫⎫
⎪ ⎪ f ⎪⎪ ⎢0 ⎪ ⎪ d ⎪⎪
⎪ ⎪ ⎪⎪1 y ⎢ 0 0 0 0 0 ⎥⎥ ⎪ ⎪ 1 y ⎪⎪
⎪⎪ ⎡ R 0 ⎤ ⎪⎪ f1z ⎪⎪ ⎪⎪ AE ⎢ 0 0 0 0 0 0 ⎥ ⎪⎪ ⎡ RT 0 ⎤ ⎪⎪ d1z ⎪⎪⎪⎪
T
⎨⎢ T ⎥⎨ ⎬⎬ = ⎢ ⎥ ⎨⎢ ⎥ ⎨ ⎬⎬
⎪⎣ 0 R ⎦ ⎪ 2 x ⎪⎪
f L ⎢ −1 0 0 1 0 0 ⎥ ⎪ ⎣ 0 RT ⎦ ⎪ d 2 x ⎪ ⎪
⎪ ⎪ f 2 y ⎪⎪ ⎢0 0 0 0 0 0⎥ ⎪ ⎪d 2 y ⎪⎪
⎪ ⎪ ⎪⎪ ⎢ ⎥⎪ ⎪ ⎪⎪
⎪⎩ T ⎪⎩ f 2 z ⎪⎭⎪⎭ ⎣0 0 0 0 0 0 ⎦ ⎪⎩ ⎪⎩ d 2 z ⎪⎭⎪⎭
T
f d
fˆ d̂
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
The revised element stiffness matrix only depends on the
direction cosines of the x̂ axis.
Tf = kTd
ˆ
f = ⎡⎣T T kT
ˆ ⎤d
⎦
k Logan’s notation.
⎡ Cx2 Cx C y Cx Cz −Cx2 −Cx C y −Cx Cz ⎤
⎢ ⎥
⎢ C y2 C y Cz −Cx C y −C y2 −C y Cz ⎥ C x = cos θ x
AE ⎢ Cz2 −Cx Cz −C y C z −Cz2 ⎥ C y = cos θ y
k= ⎢ ⎥
L ⎢ Cx2 Cx C y Cx Cz ⎥
⎢ Cz = cos θ z
C y2 C y Cz ⎥
⎢ ⎥
⎢⎣SYM Cz ⎥⎦
2
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Logan notes how the 6x6 element stiffness matrix for the spatial
truss element can be broken down into a tiling of a 3x3
component.
AE ⎡ λ −λ ⎤
k=
L ⎢⎣ −λ λ ⎥⎦
⎡ Cx2 CxC y CxCz ⎤
⎢ ⎥
λ = ⎢C y C x C y2 C y Cz ⎥
⎢ Cz Cx Cz C y Cz2 ⎥⎦
⎣
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
It only remains to calculate the x̂ direction cosines.
G x −x
Cx = i ⋅ i = cos θ x = 2 1
ˆ
L
G y − y1
C y = i ⋅ j = cos θ y = 2
ˆ
L
G z −z
C z = iˆ ⋅ k = cos θ z = 2 1
L
Only the node locations (in the nominal configuration)
are required.
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Sample problems: E = 30 × 106 psi
A = 6 in 2
Example 3.7 pg.# 89.
P. 3.43 pg.# 132.
Determine element
lengths.
Calculate the direction
cosines of the x̂ axis for
each element.
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Element 1.
L(1) = 362 + 02 + 1442 = 148.4 in
x −x 36 y −y 0.0 z − z 144.0
C x(1) = 4 (1) 1 = = 0.2426 C y(1) = 4 (1) 1 = = 0.0 C z(1) = 4 (1) 1 = = 0.9704
L 148.6 L 148.6 L 148.6
Element 2.
L(2) = 1442 + 722 + 1442 = 216.0 in
x −x 144 y −y −72 z −z 144
C x(2) = 4 (2) 2 = = 0.6667 C y(2) = 4 (2) 2 = = −0.3333 C z(2) = 4 (2) 2 = = 0.6667
L 216.0 L 216.0 L 216.0
Element 3.
L(3) = 1442 + 722 + 1442 = 216.0 in
x −x 144 y −y −72 z −z 144
C x(3) = 4 (3) 3 = = 0.6667 C y(3) = 4 (3) 3 = = −0.3333 C z(3) = 4 (3) 3 = = 0.6667
L 216.0 L 216.0 L 216.0
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Prior to assembly note the structure of the 3-D truss stiffness
matrix.
⎡ Cx2 Cx C y Cx Cz −Cx2 −Cx C y −Cx Cz ⎤
⎢ ⎥ We need only evaluate this
⎢ C y2 C y Cz −Cx C y −C y2 −C y Cz ⎥ 3x3 “building block”, λ.
AE ⎢ Cz2 −Cx Cz −C y C z −Cz2 ⎥
k= ⎢ ⎥
L ⎢ Cx2 Cx C y Cx Cz ⎥
⎢ C y2 C y Cz ⎥
⎢ ⎥
⎢⎣SYM Cz2 ⎥⎦
Exploit the homogeneous boundary conditions.
d1x = d1 y = d1z = d 2 x = d 2 y = d 2 z = d3 x = d3 y = d3 z = 0.0
Combining steps 5 (assembly) and 6 (solution)…
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
⎧ F1x ⎫ ⎡− − − − − − − − − − − − ⎤ ⎧ d1x ⎫
⎪F ⎪ ⎢− − − − − − − − − − − − ⎥⎥ ⎪⎪ d1 y ⎪⎪
⎪ ⎪1 y ⎢
⎪ F1z ⎪ ⎢− − − − − − − − − − − − ⎥ ⎪ d1z ⎪
⎪ ⎪ ⎢ ⎥⎪ ⎪
F
⎪ 2x ⎪ ⎢− − − − − − − − − − − − ⎥ ⎪d2 x ⎪
⎪ F2 y ⎪ ⎢− − − − − − − − − − − − ⎥ ⎪d 2 y ⎪
⎪ ⎪ ⎢ ⎥⎪ ⎪
⎪ F2 z ⎪ AE ⎢ − − − − − − − − − − − − ⎥ ⎪d2 z ⎪
⎨ ⎬ = (1) ⎢ ⎨ ⎬
⎪ F3 x ⎪ L ⎢ − − − − − − − − − − − − ⎥ ⎪ d3 x ⎪
⎥
⎪ F3 y ⎪ ⎢− − − − − − − − − − − − ⎥ ⎪ d3 y ⎪
⎪ ⎪ ⎢− − − − − − ⎥ ⎪ ⎪
⎪ 3z ⎪
F − − − − − − ⎪ d3 z ⎪
⎢ ⎥
⎪F ⎪ ⎢− − − − − − − − − − − −⎥ ⎪d4 x ⎪
⎪ 4x ⎪ ⎢ ⎥⎪ ⎪
⎪ F4 y ⎪ ⎢− − − − − − − − − − − − ⎥ ⎪d 4 y ⎪
⎪F ⎪ ⎢⎣ − − − − − − − − − − − − ⎥⎦ ⎩⎪ d 4 z ⎭⎪
⎩ 4z ⎭
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
The 3x3 block that must be assembled is the bottom right corner
of each of the element stiffness matrices.
For example, consider the contribution of element 1.
0
⎧f (1)
⎫ ⎡ C
2
x Cx C y Cx Cz −C 2
x −Cx C y −Cx Cz ⎤ ⎧ d1x ⎫
⎢ ⎥
1x
⎪ ⎪
−C y Cz ⎥ ⎪⎪ d1 y ⎪⎪
0
⎪f
(1)
1y ⎪ ⎢ C 2
y C y Cz −Cx C y −C y2
0
⎪⎪ f (1) ⎪⎪ AE ⎢ Cz2 −Cx Cz −C y Cz −Cz2 ⎥ ⎪⎪ d1z ⎪⎪
⎨ ⎬ = (1) ⎢ ⎥⎨ ⎬
1z
⎪f
(1)
4x ⎪ L ⎢ Cx2 Cx C y Cx Cz ⎥ ⎪ d 4 x ⎪
⎪f (1) ⎪ ⎢ Cx C y C y2 C y Cz ⎥ ⎪d 4 y ⎪
⎪ 4y
⎪ ⎢ ⎥⎪ ⎪
⎩⎪ f
(1)
4z ⎭⎪ ⎢⎣SYM Cx Cz C y Cz Cz ⎥⎦ ⎪⎩ d 4 z ⎭⎪
2
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
Evaluating this 3x3 for all 3 elements:
⎡0.05885 0 0.2354 ⎤
AE ⎢ ⎥
element 1 → 0 0 0
148.4 ⎢ ⎥
⎢⎣ 0.2354 0 0.9417 ⎥⎦
⎡ 0.4425 -0.2222 0.4445 ⎤
AE ⎢ ⎥
element 2 → -0.2222 0.1111 -0.2222
216.0 ⎢ ⎥
⎢⎣ 0.4445 -0.2222 00.4445⎥⎦
⎡ 0.4425 0.2222 0.4445 ⎤
AE ⎢ ⎥
element 1 → 0.2222 0.1111 0.2222
216.0 ⎢ ⎥
⎢⎣ 0.4445 0.2222 00.4445⎥⎦
MECH 420: Finite Element Applications
Lecture 5: 3-D Rotation Matrices.
The reduced system becomes…
0
⎧ F4 x ⎫ ⎡0.05885+2(0.4425 ⋅ 0.687) 0 0.2354+2(0.4445 ⋅ 0.687) ⎤ ⎧ d 4 x ⎫
⎪ ⎪ 0 AE ⎢ ⎥ ⎪d ⎪
⎨ F4 y ⎬ = ⎢ 0 2(0.1111 ⋅ 0.687) 0 ⎥ ⎨ 4y ⎬
⎪ ⎪ 148.4
⎩ F4 z ⎭ ⎣⎢ 0.2354+2(0.4445 ⋅ 0.687) 0 0.9417+2(0.4445 ⋅ 0.687) ⎦⎥ ⎩⎪ d 4 z ⎭⎪
−4000
d4 y = 0
⎧ 0 ⎫ 6 ⋅ 30 × 106 ⎡0.666845 0.8461⎤ ⎧d 4 x ⎫
⎨ ⎬= ⎨ ⎬
⎩ −4000 ⎭ 148.4 ⎣⎢ 0.8461 1.5524 ⎦⎥ ⎩ d 4 z ⎭
⎡0.666845 0.8461⎤
⎢ ⎥
148.4 ⎣ 0.8461 1.5524 ⎦ ⎧ 0 ⎫ = ⎧d 4 x ⎫ = ⎧ 0.0087 ⎫
⎨ ⎬ ⎨ ⎬ ⎨ ⎬
6 ⋅ 30 × 106 ( 0.6668 ⋅ 1.5524-0.84612 ) ⎩−4000 ⎭ ⎩ d 4 z ⎭ ⎩−0.0069 ⎭
MECH 420: Finite Element Applications