Example 1

𝑝 = 5𝑘𝑁/𝑚&𝐸 = 1000𝑘𝑁/𝑚2 &𝜈 = 0.3&ℎ = 0.2𝑚

Question: Compute unknown displacements (use 2 integration points)

Solution

Step 1: Consider symmetry of geometry and Boundary conditions

𝑢11 = 𝑢42
𝑢21 = 𝑢32
𝑢51 = 𝑢52

Step 2: Find nodal coordinates

𝑇
𝑋1 = [ 1.0 3.5 0.0 0.0 1.5 ]
𝑇
𝑋2 = [ 0.0 0.0 3.5 1.0 1.5 ]

Step 3: Find shape functions

𝑁1 = 1 · (1 − 𝜉1 ) · (1 − 𝜉2 )
4
𝑁2 = − 1 · (1 + 𝜉1 ) · (1 − 𝜉2 )𝜉2
4
3 1
𝑁 = · (1 + 𝜉1 ) · (1 + 𝜉2 )𝜉2
4
𝑁4 = 1 · (1 − 𝜉1 ) · (1 + 𝜉2 )
4
𝑁 = 12 · (1 + 𝜉2 ) · (1 − 𝜉2 ) · (1 + 𝜉1 )
5

Step 4: Find local gradient of shape functions

1
𝑁1, 1 = − 1 · (1 − 𝜉2 ) 𝑁1, 2 = − 1 · (1 − 𝜉1 )
4 4
𝑁2, 1 = − 1 · (1 − 𝜉2 )𝜉2 𝑁2, 2 = − 1 · (1 + 𝜉1 ) · (− 2𝜉2 + 1)
4 4
3 1
𝑁, 1 = · (1 + 𝜉2 )𝜉2 3 1
𝑁, 2 = · (1 + 𝜉1 ) · (2𝜉2 + 1)
4 4
𝑁4, 1 = − 1 · (1 + 𝜉2 ) 𝑁4, 2 = 1 · (1 − 𝜉1 )
4 4
5 1
𝑁, 1 = 2 · (1 + 𝜉2 ) · (1 − 𝜉2 )
5
𝑁, 2 = − (1 + 𝜉1 )𝜉2

Step 5: Find Jacobian

∂𝑋1 𝑛𝑜𝑑𝑒 ∂𝑁𝑖 𝑖 ∂𝑋2 𝑛𝑜𝑑𝑒 ∂𝑁𝑖 𝑖

= ∑ 𝑋 & = ∑ 𝑋
∂𝜉1 𝑖=1
∂𝜉1 1 ∂𝜉1 𝑖=1
∂𝜉1 2
∂𝑋1 𝑛𝑜𝑑𝑒 ∂𝑁𝑖 𝑖 ∂𝑋2 𝑛𝑜𝑐𝑘 ∂𝑁𝑖 𝑖
∑ 𝑋 ∑ 𝑋
∂𝜉2 = 𝑖 = 1 ∂𝜉2 1 & ∂𝜉2 = 𝑖 = 1 ∂𝜉2 2

Step 6: Find stiffness matrix using numerical integration

We should integrate the stiffness matrix using 2 integration points. However, due to
the symmetry of DOFs with respect to positions of integration points, we can perform
integration using reduced B- operator at only one integration point. The result should
be multiplied by 2.

𝜉1 = 0
𝜉2 = − 1 = − 0.57735
√3

⎛ ⎞ ⎡ −0.39 0.23 −0.06 −0.11 0.33 ⎤

𝑁'⎜0, − 1 ⎟ = ⎢ ⎥
⎝ √3⎠ ⎣ − 0.25 − 0.59 − 0.04 0.25 0.58 ⎦

⎛ ⎞ ⎡ ⎤ ⎡ ⎤
𝐽⎜0, − 1 ⎟ = ⎢ 0.9025 0.1808 ⎥ ⇒ 𝐽−1 = ⎢ 0.8798 −0.1622 ⎥&|𝐽| = 1.1146
⎝ √3⎠ ⎣ − 1.2693 0.9807 ⎦ ⎣ 1.1388 0.8097 ⎦

∂𝑁 = 𝐽−1 𝑁' = ⎡⎢ −0.3064 0.2877 −0.0474 0.1335 0.1996 ⎤⎥

∂𝑥 ⎣ − 0.6515 − 0.1769 − 0.1008 0.08210 0.8471 ⎦

⎡ 0 ⎤
⎛ ⎞ ⎢ −0.3064 0.2877 0 0 0.1996
⎥
𝐵⎜0, − 1 ⎟=⎢ 0 0 − 0.1008 0.08210 0 0.8471 ⎥
⎝ √ ⎠ ⎢
3
⎣ −0.6515 −0.1769 −0.0474 0.1335 0.8471 0.1996 ⎥⎦

⎡ −0.3064 0.2877 0.1996 ⎤

⎢ ⎥
𝐵𝑟𝑒𝑑
= ⎢ 0.08210 − 0.1008 0.8471 ⎥
⎢⎣ −0.7850 −0.2243 1.0467 ⎥⎦

2
 ⎡𝑙 𝑣 𝑜 ⎤ ⎡ ⎤
⎢ ⎥ ⎢ 1 0.3 0 ⎥
𝐶 = 𝐸 2 · ⎢ 𝑣 1 0 ⎥ = 1000 · ⎢ 0.3 1 0 ⎥
1 − 𝑣 ⎢ 0 0 1 − 𝑣 ⎥ 0.91 ⎢ 0 0 0.35 ⎥
⎢⎣ 2 ⎥⎦ ⎢⎣ ⎦⎥

1 2 𝑇
𝐾𝑟𝑒𝑑 𝑟𝑒𝑑 𝑟𝑒𝑑
𝑖𝑝1 = 𝑊1 · 𝑊1 · (𝐵𝑖𝑝1 ) 𝐶𝐵𝑖𝑝1 · |𝐽𝑖𝑝1 | · ℎ

𝑊11 = 1 → 𝑖𝑛𝜉1 𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛

𝑊21 = 2 → 𝑖𝑛𝜉2 𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛

⎡ 147.5814 −9.0377 −172.5377 ⎤

⎢ ⎥
𝑘𝑟𝑒𝑑 = ⎢ − 9.0377 45.6335 − 21.0879 ⎥ ⇒ 𝐾𝑟𝑒𝑑 = 2𝑘𝑟𝑒𝑑
⎢⎣ −172.5377 −21.0879 608.6748 ⎥⎦
𝑖𝑝1 𝑖𝑝1

Step 7: Find Load Vector

i. local load vector

𝑑𝛤 = √(dx 1 )2 + (dx 2 )2
∂𝑥 ∂𝑥
dx 1 = ∂𝜉 1 𝑑𝜉1 + ∂𝜉 1 𝑑𝜉2 = (0.5𝜉2 − 1.75)𝑑𝜉2
1 2
∂𝑥 dx
dx 2 = ∂𝜉 2 𝑑𝜉1 + 2 𝑑𝜉2 = (0.5𝜉2 + 1.75)𝑑𝜉2
1 ∂𝜉2

⎡ ⎤ ⎡ 1 · (1 − 𝜉 )𝜉 ⎤
⎢ 2 ⎥ ⎢ 2 ⎥
1 ⎢ ⎥ 1⎢ ⎥
2 2
𝑁
𝑟𝑙𝑜𝑐𝑎𝑙 = ∫ ⎢ 𝑁5 ⎥𝑃𝑑𝛤 = ∫ ⎢ 1 · (1 + 𝜉 ) · (1 − 𝜉 ) ⎥𝑃𝑑𝛤
−1 ⎢ ⎥ −1⎢ ⎥
2 2 2
⎢ 𝑁3 ⎥ ⎢ 1 · (1 + 𝜉 )𝜉 ⎥
⎢⎣ ⎥⎦ ⎢⎣ 2 2 2 ⎥⎦

Using 2 integration points

⎡ ⎤ ⎡ 𝛿𝑢 ⎤
⎢ 4.1805 ⎥ ⎢ 2 ⎥
𝜉12 = − 1 &𝜉22 = 1 ⇒ 𝑟𝑙𝑜𝑐𝑎𝑙 = ⎢ 16.7222 ⎥ → ⎢ 𝛿𝑢5 ⎥
√3 √3 ⎢ 4.1805 ⎥ ⎢ 𝛿𝑢 ⎥
⎣ ⎦ ⎣ 3 ⎦

i. Transformation

𝛼1 = 180 − 28.74 = 151.26 → 𝛿𝑢2

𝛼2 = 180 − 45 = 135 → 𝛿𝑢5
𝛼3 = 90 + 28.74 = 118, 74 → 𝛿𝑢3

3
 ⎡ cos𝛼1 sin 𝛼1 0 0 0 0 ⎤
⎢ ⎥
⎢ − sin 𝛼1 cos𝛼1 0 0 0 0 ⎥
⎢ ⎥
𝑇=⎢ ⎥
0 0 cos𝛼2 sin 𝛼2 0 0
⎢ 0 0 −sin 𝑎2 cos𝛼2 0 0 ⎥
⎢ ⎥
⎢ 0 0 0 0 cos𝑎3 sin 𝛼3 ⎥
⎢⎣ 0 0 0 0 −sin 𝛼3 cos𝛼3 ⎥⎦

⎡ ⎤ ⎡ 𝛿𝑢21 ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎢ 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ 4.1805 ⎥ ⎢ 𝛿𝑢22 ⎥ ⎢ −2.0101 ⎥ ⎢ 𝛿𝑢21 ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 𝛿𝑢 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 0 ⎥ ⎢ 𝛿𝑢11 ⎥
= 𝑇𝑇 ⎢ ⎥ → ⎢ 51 ⎥ ⇒ 𝑟 = ⎢ − 11.8243 ⎥ → ⎢ 𝛿𝑢51 ⎥ ⇒ 𝑅𝑟𝑒𝑑 = ⎢ − 5.6757 ⎥ → ⎢ 𝛿𝑢 ⎥
0
𝑟𝑔𝑙𝑜𝑏𝑎𝑙
⎢ 16.7222 ⎥ ⎢ 𝛿𝑢52 ⎥ ⎢ −11.8243 ⎥ ⎢ 𝛿𝑢52 ⎥ ⎢ ⎥ ⎢ 21 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ −3.6655 ⎥ ⎢ ⎥ ⎢ −23.6987 ⎥ ⎢ 𝛿𝑢51 ⎥
⎢ 0 ⎥ ⎢ 𝛿𝑢31 ⎥ ⎢ ⎥ ⎢ 𝛿𝑢32 ⎥ ⎢ ⎥ ⎢ ⎥
⎢⎣ 4.1805 ⎥⎦ ⎢⎣ 𝛿𝑢32 ⎥⎦ ⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦

Step 8: Compute the unknown displacements

⎡ 𝑢11 ⎤ ⎡ −0.0474 ⎤
⎢𝑢 ⎥ −1 ⎢ ⎥
⎢ 21 ⎥ = (𝐾 ) 𝑅 = ⎢ − 0.0882 ⎥
𝑟𝑒𝑑 𝑟𝑒𝑑

⎢⎣ 𝑢51 ⎥⎦ ⎢⎣ −0.0359 ⎥⎦

4
Example 2

i. Find physical coordinate of a point with isoparametric coordinates

(-0.2, 0.7, 0.3)

i. Evaluate the Jacobian at node 8

Solution

𝑁1 = 1 · (1 − 𝜉1 ) · (1 − 𝜉2 ) · (1 − 𝜉3 )
8
𝑁2 = 1 · (1 + 𝜉1 ) · (1 − 𝜉2 ) · (1 − 𝜉3 )
8
𝑁 = 1 · (1 + 𝜉1 ) · (1 + 𝜉2 ) · (1 − 𝜉3 )
3
8
𝑁4 = 1 · (1 − 𝜉1 ) · (1 + 𝜉2 ) · (1 − 𝜉3 )
8

𝑁5 = 1 · (1 − 𝜉1 ) · (1 − 𝜉2 ) · (1 + 𝜉3 )
8
𝑁6 = 1 · (1 + 𝜉1 ) · (1 − 𝜉2 ) · (1 + 𝜉3 )
8
𝑁 = 1 · (1 + 𝜉1 ) · (1 + 𝜉2 ) · (1 + 𝜉3 )
7
8
𝑁8 = 1 · (1 − 𝜉1 ) · (1 + 𝜉2 ) · (1 + 𝜉3 )
8

𝑇
𝑋1 = [ 0 1 1 0 0 1 1 0 ]
𝑇
𝑋2 = [ 0 0 1 1 0 0 1 1 ]
𝑇
𝑋3 = [ 0 0 0 0 1 1.2 1.2 1 ]

𝑋1 = 𝑁2 · 1 + 𝑁3 · 1 + 𝑁6 · 1 + 𝑁8 · 1 = 1 · (𝜉1 + 1)
2
𝑋2 = 𝑁3 · 1 + 𝑁4 · 1 + 𝑁7 · 1 + 𝑁8 · 1 = 1 · (𝜉2 + 1)
2
𝑋3 = 𝑁 · 1 + 𝑁 · 1.2 + 𝑁 · 1.2 + 𝑁 · 1 = 1 · (𝜉3 + 1) · (𝜉1 + 11)
5 6 7 8
20

5
 ⎡ 1 × (1 − 0.2) ⎤ ⎡ ⎤
⎢ 2 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 0.4 ⎥
𝑋(−0.2, 0.7, 0.3) = ⎢ 1 × (1 + 0.7) ⎥ = ⎢ 0.85 ⎥
⎢ 2 ⎥ ⎢ ⎥
⎢ 1 × (1 + 0.3) × (11 − 0.2) ⎥ ⎢ 0.702 ⎥
⎢⎣ 20 ⎥⎦ ⎢⎣ ⎥⎦

⎡1 1 · (1 + 𝜉 ) ⎤ ⎡ ⎤
⎢2 0
20 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 0.5 0 0.1 ⎥
3

𝐽=⎢0 1 0 ⎥ ⇒ 𝐽(−1, 1, 1) = ⎢ 0 0.5 0 ⎥

⎢ 2 ⎥ ⎢ ⎥
⎢0 1 ⎥ ⎢ 0 0 0.5 ⎥
⎢⎣ 0 · 𝜉 + 11) ⎥⎦
20 ( 1 ⎢⎣ ⎥⎦

6
Example 3

i. Question: Integrate the linearly distributed load acting on the surface 1-2-3 to
obtain the consistent toad vector.

i. Question: Transform the local result into global coordinate system in node ③

Solution

Step 1: Shape functions

𝑁1 = 𝜉1
𝑁2 = 𝜉2
𝑁3 = 𝜉3
𝑁4 = 𝜉4 = 1 − 𝜉1 − 𝜉2 − 𝜉3

Step 2: Parametrize surface 1-2-3

𝜉4 = 0
𝑁1 = 𝜉1
𝑁2 = 𝜁2
𝑁3 = 1 − 𝜉1 − 𝜉2

Step 3: Parametrize load

𝑝(𝜉1 , 𝜉2 , 𝜉3 ) = 𝑁1 · 0 + 𝑁2 · 0 + 𝑁3 · (−5) = − 5 · (1 − 𝜉1 − 𝜉2 )
𝜉1
1 1 − 𝜉2
𝑟=∫ ∫ [ 𝜉2 ] · (− 5) · (1 − 𝜉1 − 𝜉2 ) · |𝐽|𝑑𝜉1 𝜉2
0 0
1 − 𝜉1 − 𝜉2
|𝐽| = 2𝐴

Step 4: Compute Area

7
 Define two vectors

𝑇
𝑋1 = [ 1 0 0 0 ]
𝑇
𝑋2 = [ 0 1.5 0 0 ]
𝑇
𝑋3 = [ 0 0 2 0 ]

−1
𝑡1 = 𝑋2 − 𝑋1 = [ 1.5 ]
0
−1
𝑡2 = 𝑋3 − 𝑋1 = [ 0 ]
2

Find an outer normal to the surface 1-2-3

|| 𝑒1 𝑒2 𝑒3 || ⎡ 3 ⎤
⎢ ⎥
𝑛 = 𝑡1 × 𝑡2 = || − 1 1.5 0 || = ⎢ 2 ⎥
0 2 || ⎢⎣ 1.5 ⎥⎦
| −1 |
||

𝐴 = 1 · ∥𝑛∥ = 1 · √32 + 22 + 1 · 5 = 1.9526

2
2 2
|𝐽| = 2𝐴 = 3.9051

Step 5: Numerical integration for load

⎛ ⎡1⎤ ⎡2⎤ ⎡1⎤ ⎞

⎜ ⎢6⎥ ⎢3⎥ ⎢6⎥ ⎟
⎜ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎟
𝑟 = 1 · ⎜− 1 · ⎢ 1 ⎥ · (1 − 1 − 1) · 5 · |𝐽| − 1 · ⎢ 1 ⎥ · (1 − 2 − 1) · 5 · |𝐽| − 1 · ⎢ 2 ⎥ · (1 − 1 − 2) · 5 · |𝐽|⎟
2 ⎜ 3 ⎢6⎥ 6 6 3 ⎢6⎥ 3 6 3 ⎢3⎥ 6 3 ⎟
⎜ ⎢2⎥ ⎢1⎥ ⎢1⎥ ⎟
⎜⎝ ⎢⎣ 3 ⎥⎦ ⎢⎣ 6 ⎥⎦ ⎢⎣ 6 ⎥⎦ ⎟⎠

⎡ −0.8136 ⎤
⎢ ⎥
𝑟 = ⎢ − 0.8156 ⎥
⎢⎣ −1.6271 ⎥⎦

Step 6: Transformation

(1) We can use outer-normal to determine directional cosine. But, we have to

normalize it first

⎡ ⎤ ⎡ cos𝛾 ⎤
⎢ 0.7682 ⎥ ⎢ 1 ⎥
∥𝑛∥ ⎢⎢ 0.5122 ⎥⎥ ⎢⎢ ⎥
∧ 𝑛
𝑛 = = = cos𝛾2
⎥
⎣ 0.3841 ⎦ ⎣ cos𝛾3 ⎦

8
 We also need two tangents

⎡ −0.5547 ⎤ ⎡ cos𝛼1 ⎤
∧ 𝑡1 ⎢ ⎥ ⎢ ⎥
∥𝑡1 ∥ ⎢ 0.8320 ⎥ ⎢ ⎥
𝑡1 = = = cos𝛼2
⎢⎣ 0 ⎥
⎦ ⎢⎣ cos𝛼3 ⎥⎦

| | ⎡ ⎤ ⎡ ⎤
| 𝑒1 𝑒2 𝑒3 | ⎢ −0.3196 ⎥ ⎢ cos𝛽1 ⎥
| |
= || 0.7682 0.5122 0.3841 || = ⎢ − 0.2131 ⎥ = ⎢ cos𝛽2 ⎥
∧ ∧
𝑡2 ' = 𝑛 × 𝑡 1
| −0.5547 | ⎢ ⎥ ⎢ ⎥
| 0.8320 0 | ⎢⎣ 0.9233 ⎥⎦ ⎢⎣ cos𝛽 ⎥⎦
| | 3

Define transformation matrix

⎡ cos𝛼1 cos𝛼2 cos𝛼3 ⎤

⎢ ⎥
𝑇 = ⎢ cos𝛽1 cos𝛽2 cos𝛽3 ⎥
⎢ cos𝛾 cos𝛾 cos𝛾 ⎥
⎣ 1 2 3 ⎦

Transformation for node ③

⎡ ⎤ ⎡ −1.25 ⎤
𝑇⎢ ⎥ ⎢ ⎥
0
𝑟𝑔𝑙𝑜𝑏𝑎𝑙3 𝑇 𝑙𝑜𝑐𝑎𝑙3
=𝑇 𝑟 =𝑇 ⎢ 0 ⎥ = ⎢ − 0.8333 ⎥
⎢⎣ −1.6271 ⎥⎦ ⎢⎣ −0.625 ⎥⎦