Middle East Technical University

Department of Mechanical Engineering

ME 413 Introduction to Finite Element Analysis

Chapter 5
Two-Dimensional Formulation

These notes are prepared by

Dr. Cüneyt Sert
http://www.me.metu.edu.tr/people/cuneyt
csert@metu.edu.tr

These notes are prepared with the hope to be useful to those who want to learn and teach FEM. You are free to use
them. Please send feedbacks to the above email address.
5-1
 What Are We Going to Learn?
• Compared to 1D, major differences in 2D FEM formulation are

• application of IBP.
• master elements and shape functions for triangular and quadrilateral elements.
• Jacobian transformation.
• boundary integral evaluation.

Triangular Quadrilateral
element element

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-2
 Model DE in 2D
• Poisson equation in 2D is
−𝛻 ⋅ 𝑎 𝛻𝑢 = 𝑓
where 𝑎(𝑥, 𝑦) and 𝑓(𝑥, 𝑦) are known functions and
𝑢(𝑥, 𝑦) is the unknown.
• For a problem in the 𝑥𝑦 plane of the Cartesian coordinate system, gradient operator is
𝜕 𝜕
𝛻 = 𝑖 +𝑗
𝜕𝑥 𝜕𝑦
• In the 𝑥𝑦 plane Poisson equation becomes
𝜕 𝜕 𝜕𝑢 𝜕𝑢
− 𝑖 +𝑗 ⋅ 𝑎 𝑖+𝑎 𝑗 =𝑓
𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦

𝛻 𝑎 𝛻𝑢

𝜕 𝜕𝑢 𝜕 𝜕𝑢
− 𝑎 − 𝑎 =𝑓
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-3
 Model DE in 2D
• If function 𝑎 is constant over the problem domain, Poisson eqn. becomes
𝜕2𝑢 𝜕2𝑢
−𝑎𝛻 ⋅ 𝛻𝑢 = 𝑓 → −𝛻 2 𝑢 = 𝑔 → − + =𝑔
𝜕𝑥 2 𝜕𝑥 2
• Homogeneous form of this equation is called the Laplace’s equation
𝜕2𝑢 𝜕2𝑢
−𝛻 2 𝑢 =0 → − 2
+ 2 =0
𝜕𝑥 𝜕𝑥

• Poisson equation models many physical phenomena such as

• potential flow
• heat conduction
• groundwater flow
• transverse deflection of plates
• electrostatics and magnetostatics

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-4
 Obtaining Weak Form in 2D
𝜕 𝜕𝑢 𝜕 𝜕𝑢
Model DE : − 𝑎 − 𝑎 =𝑓
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦

• Weighted residual integral statement of this DE is

𝜕 𝜕𝑢 𝜕 𝜕𝑢
𝑤 − 𝑎 − 𝑎 − 𝑓 𝑑Ω = 0
Ω 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦

• 2nd order derivatives of 𝑢 can be reduced to 1st order using the following general
equations

𝜕𝐹 𝜕𝑤
𝑤 𝑑Ω = − 𝐹 𝑑Ω + 𝑤𝐹 𝑛𝑥 𝑑Γ
Ω 𝜕𝑥 Ω 𝜕𝑥 Γ

𝜕𝐹 𝜕𝑤
𝑤 𝑑Ω = − 𝐹 𝑑Ω + 𝑤𝐹𝑛𝑦 𝑑Γ
Ω 𝜕𝑦 Ω 𝜕𝑦 Γ

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-5
 Obtaining Weak Form in 2D
𝜕 𝜕𝑢 𝜕 𝜕𝑢
−𝑤 𝑎 −𝑤 𝑎 − 𝑤𝑓 𝑑Ω = 0
Ω 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦

𝜕𝑢 𝜕𝑤 𝜕𝑢 𝜕𝑢 𝜕𝑤 𝜕𝑢
𝑎 𝑑Ω − 𝑤𝑎 𝑛𝑥 𝑑Γ 𝑎 𝑑Ω − 𝑤𝑎 𝑛𝑦 𝑑Γ
Ω 𝜕𝑥 𝜕𝑥 Γ 𝜕𝑥 Ω 𝜕𝑦 𝜕𝑦 Γ 𝜕𝑦

• Elemental weak form is

𝜕𝑢 𝜕𝑤 𝜕𝑢 𝜕𝑤 𝜕𝑢 𝜕𝑢
𝑎 + 𝑑Ω = 𝑤𝑓 𝑑Ω + 𝑤 𝑎 𝑛𝑥 + 𝑎 𝑛𝑦 𝑑Γ
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 Ω𝑒 Γ𝑒 𝜕𝑥 𝜕𝑦

𝑞𝑛 : SV of the problem

where 𝑛𝑥 and 𝑛𝑦 are the Cartesian components of the unit outward normal of Γ 𝑒 .

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-6
 2D Formulation (cont’d)
• Approximate solution over an element is
𝑁𝐸𝑁

𝑢𝑒 = 𝑢𝑗𝑒 𝑆𝑗𝑒 (𝑥, 𝑦)

𝑗=1

• For linear triangular and quadratic elements 𝑁𝐸𝑁 is 3 and 4, respectively.

3-node triangular 4-node quadrilateral

element element
𝑁𝐸𝑁 = 3 𝑁𝐸𝑁 =4

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-7
 2D Formulation (cont’d)
• To get the 𝑖 𝑡ℎ equation of element e
• substitute approximate 𝑢𝑒 into the elemental weak form and
• select 𝑤 = 𝑆𝑖𝑒

𝑁𝐸𝑁 𝑁𝐸𝑁
𝜕 𝜕𝑆𝑖𝑒 𝜕 𝜕𝑆𝑖𝑒
𝑎 𝑢𝑗𝑒 𝑆𝑗𝑒 + 𝑢𝑗𝑒 𝑆𝑗𝑒 𝑑Ω = 𝑆𝑖𝑒 𝑓 𝑑Ω + 𝑆𝑖𝑒 𝑞𝑛 𝑑Γ
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 Ω𝑒 Γ𝑒
𝑗=1 𝑗=1

• Arrange to get

𝑁𝐸𝑁
𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒 𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒
𝑎 + 𝑑Ω 𝑢𝑗𝑒 = 𝑆𝑖𝑒 𝑓 𝑑Ω + 𝑆𝑖𝑒 𝑞𝑛 𝑑Γ
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 Ω𝑒 Γ𝑒
𝑗=1

𝐾𝑖𝑗𝑒 𝐹𝑖𝑒 𝑄𝑖𝑒

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-8
 2D Formulation (cont’d)
• 𝑁𝐸𝑁 × 𝑁𝐸𝑁 elemental system is

𝐾 𝑒 𝑢𝑒 = 𝐹 𝑒 + {𝑄 𝑒 }

𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒 𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒

𝐾𝑖𝑗𝑒 = 𝑎 + 𝑑Ω 𝐹𝑖𝑒 = 𝑆𝑖𝑒 𝑓 𝑑Ω 𝑄𝑖𝑒 = 𝑆𝑖𝑒 𝑞𝑛 𝑑Γ
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 Ω𝑒 Γ𝑒

• To evaluate these integrals

• triangular and quadrilateral master elements will be introduced.
• shape functions will be written in master element coordinates.
• 2D Jacobian transformation will be used.
• GQ integration will be used.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-9
 2D Quadrilateral Master Element

Actual quadrilateral element Master quadrilateral element

1 𝜂
(-1,1) (1,1)
2 4 4 3
𝑦 𝜉

1 2
𝑥 3 (-1,-1) (1,-1)

• Master quadrilateral element is a square of size 2x2.

• Its nodes are always numbered in a CCW order starting with (-1,-1) corner.
• Nodes of the actual element are also numbered in CCW order. It does NOT matter
which node is selected as the first one.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-10
 Shape Functions of 2D Quadrilateral Master Element
• General form of Lagrange type 2D shape functions over a 4-node quadrilateral
element is
𝑆 = 𝐴 + 𝐵𝜉 + 𝐶𝜂 + 𝐷𝜉𝜂
• Unknown constants 𝐴, 𝐵, 𝐶 and 𝐷 can be found using the fact that shape functions
satisfy the Kronecker-Delta property
1 if 𝑖 = 𝑗
𝑆𝑗 𝜉𝑖 , 𝜂𝑖 =
0 if 𝑖 ≠ 𝑗

1 𝜂
𝑆1 = (1 − 𝜉)(1 − 𝜂)
4 4 3
1
𝑆2 = (1 + 𝜉)(1 − 𝜂) 𝜉
• Shape functions are 4
1
𝑆3 = (1 + 𝜉)(1 + 𝜂)
4
1 1 2
𝑆4 = (1 − 𝜉)(1 + 𝜂)
4

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-11
 2D Triangular Master Element

Actual triangular element Master triangular element

2 𝜂
(0,1)
1 3
𝑦

1 2 𝜉
𝑥 (0,0) (1,0)
3

• Master triangular element is a right triangle with an area of 0.5.

• Its nodes are always numbered in a CCW order, starting with (0,0) corner.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-12
 Shape Functions of 2D Triangular Master Element
• General form of Lagrange type 2D shape functions over a 3-node triangular element is
𝑆 = 𝐴 + 𝐵𝜉 + 𝐶𝜂
• Unknown constants 𝐴, 𝐵 and 𝐶 can be found using the Kronecker-Delta property of
the shape functions

𝜂
𝑆1 = 1 − 𝜉 − 𝜂 3

• Shape functions are 𝑆2 = 𝜉

𝑆3 = 𝜂 𝜉
1 2

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-13
 Jacobian Transformation in 2D
• 𝜕𝑆/𝜕𝑥 and 𝜕𝑆/𝜕𝑦 derivatives appear in the integrals of Slide 5-9.
• These derivatives need to be expressed in terms of 𝜕𝑆/𝜕𝜉 and 𝜕𝑆/𝜕𝜂 derivatives.
• This requires the transformation between (𝑥, 𝑦) and (𝜉, 𝜂) coordinates.

Actual element in (𝑥, 𝑦) Master element in (𝜉, 𝜂)

1 𝜂
𝑥 = 𝑥(𝜉, 𝜂) 4 3
2 4 𝑦 = 𝑦(𝜉, 𝜂) 𝜉
𝑦

𝑥 1 2
3

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-14
 Jacobian Transformation in 2D (cont’d)
• Remember that in 1D 𝑥(𝜉) relation was
ℎ𝑒 𝑥1𝑒 + 𝑥2𝑒
𝑥= 𝜉+
2 2
• This relation can also be expressed as
𝑁𝐸𝑁
1−𝜉 𝑒 1+𝜉 𝑒
𝑥= 𝑥𝑗𝑒 𝑆𝑗 → 𝑥= 𝑥1 + 𝑥2
2 2
𝑗=1

• This works due to the Kronecker-Delta property of the shape functions

• 𝜉 = −1 is mapped to 𝑥 = 𝑥1𝑒
• 𝜉 = 1 is mapped to 𝑥 = 𝑥2𝑒

• Same logic can also be used in 2D to get 𝑥(𝜉, 𝜂) and 𝑥(𝜉, 𝜂) relations.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-15
 Jacobian Transformation in 2D (cont’d)
𝑁𝐸𝑁 𝑁𝐸𝑁

𝑥= 𝑥𝑗𝑒 𝑆𝑗 and 𝑦= 𝑦𝑗𝑒 𝑆𝑗

𝑗=1 𝑗=1

• These can be used for both quadrilateral and triangular elements.

• 𝑥𝑗𝑒 and 𝑦𝑗𝑒 are the coordinates of the corner points of the elements.

e.g. Example 5.1: Obtain 𝑥(𝜉, 𝜂) and 𝑦(𝜉, 𝜂) relations for the following element.
𝑦
2 Corner coordinates are
1 Corner 1 : (5, 6)
Corner 2 : (0, 7)
Corner 3 : (2, 0)
𝑥
3

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-16
 Example 5.1 (cont’d)
3

𝑥= 𝑥𝑗𝑒 𝑆𝑗 = 5 1 − 𝜉 − 𝜂 + 0 𝜉 + 2 𝜂 = 5 − 5𝜉 − 3𝜂
𝑗=1
3

𝑦= 𝑦𝑗𝑒 𝑆𝑗 = 6 1 − 𝜉 − 𝜂 + 7 𝜉 + 0 𝜂 = 6 + 𝜉 − 6𝜂
𝑗=1
𝜂
3
• Every point on the master element can be mapped to
a point on the actual element using these relations. P(0.5,0.5)

• For example point P with 𝜉, 𝜂 = (0.5, 0.5) maps to 𝜉

1 2
𝑥 = 5 − 5 0.5 − 3 0.5 =1 𝑦
𝑦 = 6 + 0.5 − 6 0.5 = 3.5 2
1
• Both points P and Q are the mid-points of the Q(1,3.5)
faces opposite to node 1. 𝑥
3
METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-17
 Jacobian Transformation in 2D (cont’d)
• With the link between (𝑥, 𝑦) and (𝜉, 𝜂) coordinates, 𝜕𝑆/𝜕𝑥 and 𝜕𝑆/𝜕𝑦 derivatives can
be linked to 𝜕𝑆/𝜕𝜉 and 𝜕𝑆/𝜕𝜂.

𝜕𝑆 𝜕𝑆 𝜕𝑥 𝜕𝑆 𝜕𝑦 𝜕𝑆 𝜕𝑥 𝜕𝑦
= + 𝜕𝑆
𝜕𝜉 𝜕𝑥 𝜕𝜉 𝜕𝑦 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝑥
→ = 𝜕𝑆
𝜕𝑆 𝜕𝑆 𝜕𝑥 𝜕𝑆 𝜕𝑦 𝜕𝑆 𝜕𝑥 𝜕𝑦
= + 𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝑦
𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝜕𝜂
Jacobian matrix
[𝐽𝑒 ]

𝑑𝑥
• In 1D Jacobian was 𝐽𝑒 = .
𝑑𝜉

• In 2D Jacobian is a matrix.
• In general [𝐽𝑒 ] is different for each element of the FE mesh.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-18
 Jacobian Transformation in 2D (cont’d)
• For the integrals of slide 5-9 what we actually need is
𝜕𝑆 𝜕𝜉 𝜕𝜂 𝜕𝑆
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝜉
𝜕𝑆 = 𝜕𝜉 𝜕𝜂 𝜕𝑆
𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝜂
Inverse of the
Jacobian matrix
𝐽𝑒 −1

• Since we know 𝑥 and 𝑦 as a function of 𝜉 and 𝜂, but not the other way, it is NOT
practical to calculate 𝐽𝑒 −1 directly.
• Instead we first calculate 𝐽𝑒 and then take its inverse.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-19
 Example 5.2
e.g. Example 5.2: Obtain [𝐽𝑒 ] and 𝐽𝑒 −1 of the element that we studied in exercise 5.1.

𝑦
2 Corner coordinates are
1
Corner 1 : (5, 6)
Corner 2 : (0, 7)
Corner 3 : (2, 0)
𝑥
3

𝜕𝑥 𝜕𝑦 𝜕𝑆𝑗 𝜕𝑆𝑗𝜕𝑆1 𝜕𝑆2 𝜕𝑆3

𝑥𝑗𝑒 𝑦𝑗𝑒 𝑥1𝑒 𝑦1𝑒
𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉
𝐽𝑒 = = = 𝑥2𝑒 𝑦2𝑒
𝜕𝑥 𝜕𝑦 𝜕𝑆𝑗 𝜕𝑆𝑗 𝜕𝑆1 𝜕𝑆2 𝜕𝑆3
𝑥𝑗𝑒 𝑒
𝑦𝑗 𝑥3𝑒 𝑦3𝑒
𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝜂

• This general 𝐽𝑒 calculation formula applies to both triangular and quadrilateral

elements.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-20
 Example 5.2 (cont’d)
• For a triangular element shape functions are
𝑆1 = 1 − 𝜉 − 𝜂 , 𝑆2 = 𝜉 , 𝑆3 = 𝜂
• For a triangular element derivatives of the shape functions are
𝜕𝑆1 𝜕𝑆2 𝜕𝑆3
𝜕𝜉 𝜕𝜉 𝜕𝜉 −1 1 0
=
𝜕𝑆1 𝜕𝑆2 𝜕𝑆3 −1 0 1
𝜕𝜂 𝜕𝜂 𝜕𝜂
• Jacobian of the element is
5 6
−1 1 0 −5 1
𝐽𝑒 = 0 7 =
−1 0 1 −3 −6
2 0

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-21
 Example 5.2 (cont’d)
• Inverse of the Jacobian matrix is
𝑒
1 𝐽22 𝑒
−𝐽12
𝐽𝑒 −1 = 𝑒 𝑒 𝑒
𝐽 −𝐽21 𝐽11

𝑒 𝑒 𝑒 𝑒
𝐽𝑒 = 𝐽11 𝐽22 − 𝐽12 𝐽21
= −5 −6 − 1 −3 = 33

1 −6 −1 −0.182 0.030
𝐽𝑒 −1 = =
33 3 −5 0.091 −0.152

• Note that in 1D 𝐽𝑒 was equal to the ratio of actual element’s length to master
element’s length.
• Similarly for a 3-node triangular element 𝐽𝑒 is equal to the ratio of actual element’s
area to master element’s area. In this exercise the area ratio is 16.5/(0.5) = 33 .

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-22
 Example 5.3
e.g. Example 5.3: Obtain [𝐽𝑒 ] and 𝐽𝑒 of the element shown below.
𝑦
2 Corner coordinates are
1 Corner 1 : (5, 6)
Corner 2 : (0, 7)
Corner 3 : (0, 0)
𝑥 Corner 4 : (2, 0)
3 4

𝜕𝑆1 𝜕𝑆2 𝜕𝑆3 𝜕𝑆4 𝑥1𝑒 𝑦1𝑒 𝜂−1 1−𝜂 𝜂+1 −1 − 𝜂 5 6

𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝑥2𝑒 𝑦2𝑒 4 4 4 4 0 7
𝐽𝑒 = =
𝜕𝑆1 𝜕𝑆2 𝜕𝑆3 𝜕𝑆4 𝑥3𝑒 𝑦3𝑒 𝜉−1 −𝜉 − 1 𝜉+1 1−𝜉 0 0
𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝜂 𝑥4𝑒 𝑦4𝑒 4 4 4 4 2 0

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-23
 Example 5.3 (cont’d)
1 3𝜂 − 7 −𝜂 + 1
𝐽𝑒 =
4 3𝜉 − 3 −𝜉 − 13
• Determinant of 𝐽𝑒 is
𝑒 𝑒 𝑒 𝑒
1
𝐽𝑒 = 𝐽11 𝐽22 − 𝐽12 𝐽21 = 2𝜉 − 21𝜂 + 47
8
• Note that this time both the Jacobian matrix and its determinant are functions of 𝜉
and 𝜂.
• Integral of 𝐽𝑒 over the master element will give the area of the actual element.
1 1
1
2𝜉 − 21𝜂 + 47 𝑑𝜉 𝑑𝜂 = 23.5
𝜂=−1 𝜉=−1 8
𝐽𝑒
Actual element’s
area

• This can be generalized as follows which will be used in GQ integration

𝑑𝑥𝑑𝑦 = 𝐽𝑒 𝑑𝜉𝑑𝜂 (True for both triangular and

Ω Ω𝑒 quadrilateral elements)
METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-24
 Calculation of Integrals Over a Master Element
• Now the integrals of Slide 5-9 can be evaluated on a master element using GQ.
𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒 𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒
𝐾𝑖𝑗𝑒 = 𝑎 + 𝑑Ω , 𝐹𝑖𝑒 = 𝑆𝑖𝑒 𝑓 𝑑Ω
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 Ω𝑒

Use 𝑥 = 𝑥𝑗𝑒 𝑆𝑗 and

𝑦 = 𝑦𝑗𝑒 𝑆𝑗 to convert
𝑥 and 𝑦 of function 𝑎
to 𝜉 and 𝜂. 𝑑𝑥𝑑𝑦 = 𝐽𝑒 𝑑𝜉𝑑𝜂
Ω𝑒 Ω𝑒𝑚𝑎𝑠𝑡𝑒𝑟
Same for 𝑓 of 𝐹𝑖𝑒 .

𝜕𝑆 𝑒 𝜕𝑆 𝜕𝜉 𝜕𝑆 𝜕𝜂 𝜕𝑆 𝑒 𝜕𝑆 𝜕𝜉 𝜕𝑆 𝜕𝜂
= + = +
𝜕𝑥 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝑦 𝜕𝜉 𝜕𝑦 𝜕𝜂 𝜕𝑦

𝜕𝑆 𝑒 −1
𝜕𝑆 𝑒 −1
𝜕𝑆 𝑒 −1
𝜕𝑆 𝑒 −1
= 𝐽 11 + 𝐽 12 = 𝐽 21 + 𝐽 22
𝜕𝜉 𝜕𝜂 𝜕𝜉 𝜕𝜂

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-25
 Gauss Quadrature Over Quadrilateral Elements
• For a quadrilateral master element both 𝜉 and 𝜂 change between -1 and 1.
• [-1, 1] are the limits used in 1D GQ integration.
• Therefore for 2D quadrilateral elements 1D GQ tables can be used.

• Consider the evaluation of the following integral using NGP points in both 𝜉 and 𝜂
directions.
1 1
𝐼= 𝑔 𝑑𝜉 𝑑𝜂
𝜂=−1 𝜉=−1
In 2D there are 𝑁𝐺𝑃2
GQ points
𝑁𝐺𝑃 𝑁𝐺𝑃 𝑁𝐺𝑃2

𝐼= 𝑔 𝜉𝑚 , 𝜂𝑛 𝑊𝑚 𝑊𝑛 or 𝐼= 𝑔 𝜉𝑘 , 𝜂𝑘 𝑊𝑘
𝑛=1 𝑚=1 𝑘=1

Sum for 𝜂
Combined sum 𝑊𝑘 = 𝑊𝑚 𝑊𝑛
Sum for 𝜉
for 𝜉 and 𝜂

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-26
 Gauss Quadrature Over Quadrilateral Elements (cont’d)

1D GQ Integration 2D GQ Integration Over Quads

NGP 𝝃𝒌 𝑾𝒌 NGP 𝝃𝒌 𝜼𝒌 𝑾𝒌
1 0.0 2 1 0.0 0.0 4
−1/ 3 1 −1/ 3 −1/ 3 1
2
1/ 3 1 1/ 3 −1/ 3 1
4
−1/ 3 1/ 3 1
− 0.6 5/9
3 0.0 8/9 1/ 3 1/ 3 1
0.6 5/9 − 0.6 − 0.6 25/81
0.0 − 0.6 40/81
𝜂 0.6 − 0.6 25/81
− 0.6 0.0 40/81
7 8 9 9 0.0 0.0 64/81
9 point GQ 0.6 0.0 40/81
over a quad. 4 5 6 𝜉 − 0.6 0.6 25/81
element 0.0 0.6 40/81
1 2 3
0.6 0.6 25/81
METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-27
 Gauss Quadrature Over Triangular Elements
• For a triangular master element limits for 𝜉 and 𝜂 are [0, 1] and [0, 1-𝜉], respectively.
1 1−𝜉
𝑓 𝑑𝜉𝑑𝜂 = 𝑓 𝑑𝜂𝑑𝜉
Ω𝑒 𝜉=0 𝜂=0

• Therefore a new GQ table is necessary.

2D GQ Integration Over Triangles

NGP 𝝃𝒌 𝜼𝒌 𝑾𝒌 𝜂
1 1/3 1/3 0.5
0.5 0.0 1/6
3 0.0 0.5 1/6 4 point GQ over
3
0.5 0.5 1/6 a triangular
element 1
1/3 1/3 - 27/96
0.6 0.2 25/96 4 2 𝜉
4
0.2 0.6 25/96
0.2 0.2 25/96

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-28
 Example 5.4
e.g. Example 5.4: Calculate the first entry of the following elemental force vector
𝐹𝑖𝑒 = 𝑥 2 𝑆𝑖 𝑑𝑥𝑑𝑦
Ω𝑒

over the following element using 4 point GQ integration.

𝑦 Corner coordinates are

Corner 1 : (0, 0)
4 3 Corner 2 : (5, 0)
𝑥 Corner 3 : (3, 2)
1 2
Corner 4 : (0, 2)

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-29
 Example 5.4 (cont’d)
• We need to calculate
𝐹1𝑒 = 𝑥 2 𝑆1 𝑑𝑥𝑑𝑦
Ω𝑒
• Switching to master element coordinates the integral becomes
1 1
𝑒 2
1
𝐹1 = 𝑥 𝜉, 𝜂 (1 − 𝜉)(1 − 𝜂) 𝐽𝑒 𝑑𝜉𝑑𝜂
−1 −1 4
𝑑𝑥𝑑𝑦
𝑆1

• We first need 𝑥 as a function of 𝜉 and 𝜂.

4

𝑥= 𝑥𝑗𝑒 𝑆𝑗 = 0 𝑆1 + 5 𝑆2 + 3 𝑆3 + 0 𝑆4
𝑗=1
1 1
𝑥= 5 1+𝜉 1−𝜂 + 3 1+𝜉 1+𝜂
4 4
1
𝑥 = (4 + 4𝜉 − 𝜂 − 𝜉𝜂)
2

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-30
 Example 5.4 (cont’d)
• Next we need to calculate the Jacobian and its determinant (similar to Slide 5-23)

𝜂−1 1−𝜂 𝜂+1 −1 − 𝜂 0 0 4−𝜂

0
4 4 4 4 5 0 2
𝐽𝑒 = =
𝜉−1 −𝜉 − 1 𝜉+1 1−𝜉 3 2 −1 − 𝜉
0 2 1
4 4 4 4 2

4−𝜂 −1 − 𝜉 4−𝜂
𝐽𝑒 = 1 − 0 =
2 2 2
• The integral becomes
1 1 2
1 1 4−𝜂
𝐹1𝑒 = 4 + 4𝜉 − 𝜂 − 𝜉𝜂 1−𝜉 1−𝜂 𝑑𝜉𝑑𝜂
−1 −1 2 4 2
𝑔(𝜉,𝜂)

1 1
𝐹1𝑒 = 𝑔 𝑑𝜉𝑑𝜂
−1 −1

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-31
 Example 5.4 (cont’d)
• 4 point GQ over the quadrilateral master element will be
𝐹1𝑒 = 𝑔 𝜉1 , 𝜂1 𝑊1 + 𝑔 𝜉2 , 𝜂2 𝑊2 + 𝑔 𝜉3 , 𝜂3 𝑊3 + 𝑔 𝜉4 , 𝜂4 𝑊4

where points and weights are provided in Slide 5-27

• The result will be

𝐹1𝑒 = 1.3320 + 4.971 + 0.1492 + 0.5569 = 6.9993

Notes:
• In general 𝐽𝑒 is a function of 𝜉 and 𝜂 and it needs to be evaluated at GQ points.
• In this example we did not need 𝑦(𝜉, 𝜂) bacause 𝑓 was not a function of 𝑦.
• In this example we did not calculate the inverse of 𝐽𝑒 because force vector does not
contain shape function derivatives. Stiffness matrix calculation will need it.

• Is the above result exact? What is the exact value? What will 1 point and 9 point
integrations give?

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-32
 Calculation of 𝑄
• {𝑄} integrals need to be evaluated only for the real boundary faces where NBC or MBC
is specified.
• Consider the following problem with a mesh of 3 elements and 6 nodes.

e=3 EBC

NBC 4 5

e=1
e=2
1 2 3

NBC

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-33
 Calculation of 𝑄 (cont’d)
6

• There are 4 elements faces at NBC or MBC boundaries. 3 EBC

1 5
• The assembled global {𝑄} will be (first local node NBC 4
of each element is shown, the others are located 1 2
in a CCW order) 1 1
1 2 3

𝑄1 𝑄11 NBC

𝑄2 𝑄21 + 𝑄12 • 𝑄3 , 𝑄5 and 𝑄6 are not necessary

𝑄3 𝑄22 because PVs are known at these
𝑄= =
𝑄4 𝑄41 + 𝑄13 nodes.
𝑄5 𝑄31 + 𝑄32 + 𝑄23 • Only the circled ones are necessary.
𝑄6
𝑄23

• Note : In this example we do not have an internal node (a node that is not located at a
boundary), but for those nodes sum of 𝑄𝑖𝑒 ’s will be zero.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-34
 Calculation of 𝑄 (cont’d)
• For the Poisson equation {𝑄 𝑒 } is calculated as (Slide 5-9)

𝑄𝑖𝑒 = 𝑆𝑖 𝑞𝑛 𝑑𝑠
Γ𝑒
𝜕𝑢 𝜕𝑢
𝑞𝑛 = 𝑎 𝑛 +𝑎 𝑛
𝜕𝑥 𝑥 𝜕𝑦 𝑦

• Γ 𝑒 is the boundary of the element and it is composed of 𝑁𝐸𝑁 straight lines.

• For a triangular element the integral can be decomposed into 3 parts.

𝑄𝑖𝑒 = 𝑆𝑖 𝑞𝑛 𝑑𝑠 + 𝑆𝑖 𝑞𝑛 𝑑𝑠 + 𝑆𝑖 𝑞𝑛 𝑑𝑠
𝑓1 𝑓2 𝑓3
𝑓3 : Face 3

1 1
1 2 𝑠
2 e = 2 𝑠 + +
𝑠
3
3 3
METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-35
 Calculation of 𝑄 (cont’d) 6

• Consider the 3 element problem of Slide 5-34. 3 EBC

1 5
• We need 𝑄11 , 𝑄21 , 𝑄41 , 𝑄12 , 𝑄13 NBC 4
1 2
1 1
1 2 3
• e=1 :
NBC
𝑄11 = 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠
𝑓1 𝑓2 𝑓3 𝑓4
𝑆1 is zero on faces 2 and 3

𝑄21 = 𝑆2 𝑞𝑛 𝑑𝑠 + 𝑆2 𝑞𝑛 𝑑𝑠 + 𝑆2 𝑞𝑛 𝑑𝑠 + 𝑆2 𝑞𝑛 𝑑𝑠
𝑓1 𝑓2 𝑓3 𝑓4

No need (f2 is internal) 𝑆2 is zero on faces 3 and 4

𝑄41 = 𝑆4 𝑞𝑛 𝑑𝑠 + 𝑆4 𝑞𝑛 𝑑𝑠 + 𝑆4 𝑞𝑛 𝑑𝑠 + 𝑆4 𝑞𝑛 𝑑𝑠
𝑓1 𝑓2 𝑓3 𝑓4

𝑆4 is zero on faces 1 and 2 No need (f3 is internal)

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-36
 Calculation of 𝑄 (cont’d)
6
• e=2 :
EBC
𝑄12 = 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 1
3
5
𝑓1 𝑓2 𝑓3 NBC 4
𝑆1 is zero on face 2 No need (f3 is 1 2
internal) 1 1
1 2 3
• e=3 :
NBC
𝑄13 = 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠
𝑓1 𝑓2 𝑓3

No need (f1 is 𝑆1 is zero on face 2

internal)

• Conclusion : Boundary integrals need to be calculated only for real boundary faces
where NBC or MBC is provided.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-37
 Calculation of 𝑄 (cont’d)
6
• Consider the common and simple case of 𝑞𝑛 = constant.
• Let’s study the calculation of 3 EBC
NBC 1 5
4
𝑞𝑛 = 𝑞𝐿
𝑄11 = 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 1
𝑓1 𝑓4 2
1 1
• 𝑆1 is a 2D shape function but it reduces to a first order 1 2 3
function over faces 1 and 4 of element 1. NBC, 𝑞𝑛 = 𝑞𝐵

4 3
f4 1

𝑠 1 2 𝑠
𝑆1 = 1 f1 𝑆1 = 1 − 1
𝐿𝑓4 𝐿𝑓1
𝑠
1 f1 2
4 𝑠 f4 1

𝐿1𝑓1
𝐿1𝑓4

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-38
 Calculation of 𝑄 (cont’d)

𝑄11 = 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠
𝑓1 𝑓4
𝐿1𝑓1 𝐿1𝑓4
𝑠 𝑠
𝑄11 = 1− 1 𝑞𝐵 𝑑𝑠 + 1 𝑞𝐿 𝑑𝑠
𝑠=0 𝐿𝑓1 𝑠=0 𝐿𝑓4

𝑞𝐵 1 𝑞𝐿
𝑄11 = 𝐿𝑓1 + 𝐿1𝑓4
2 2

• Same procedure can be followed to calculate 𝑄21 .

𝑄21 = 𝑆2 𝑞𝑛 𝑑𝑠
𝑓1 𝑠 4 3
𝑆1 = 1
𝑠 𝐿𝑓1 f4 1
𝑄21 = 1 𝑞𝐵 𝑑𝑠
𝑓1 𝐿𝑓1 1 2
1 𝑠 f1 2 f1
𝑞𝐵 1
𝑄21 = 𝐿
2 𝑓1 𝐿1𝑓1

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-39
 Calculation of 𝑄 (cont’d)
• Calculation of 𝑄41 , 𝑄12 and 𝑄13 follow the same procedure.

𝑞𝐿 1 𝑞𝐵 2 𝑞𝐿 3
𝑄41 = 𝐿𝑓4 , 𝑄12 = 𝐿 , 𝑄13 = 𝐿𝑓3
2 2 𝑓1 2

• Summary :
• From the bottom face of e=1, amount of provided SV
is 𝑞𝐵 𝐿1𝑓1 and it is divided equally to 𝑄11 and 𝑄21 . 6
𝑞𝐿 𝐿3𝑓3
• From the bottom face of e=2, amount of provided SV 3 EBC
2
is 𝑞𝐵 𝐿2𝑓1 and it is divided equally to 𝑄12 and 𝑄22 . 4
1 5
𝑞𝐿 𝐿1𝑓4 1
• From the left face of e=1, amount of provided SV 2
2 1 1
is 𝑞𝐿 𝐿1𝑓4 and it is divided equally to 𝑄11 and 𝑄41 . 1 2 3
• From the left face of e=3, amount of provided SV
𝑞𝐵 𝐿1𝑓1 𝑞𝐵 𝐿2𝑓1
is 𝑞𝐿 𝐿3𝑓3 and it is divided equally to 𝑄13 and 𝑄33 .
2 2

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-40
 Calculation of 𝑄 (cont’d)
• Assembled {𝑄} vector is

𝑞𝐵 1 𝑞𝐿 1 6
𝐿 + 𝐿
𝑄1 2 𝑓1 2 𝑓4
𝑞𝐵 1 𝑞𝐵 2 3 EBC
𝑄2 𝐿 + 𝐿 NBC
2 𝑓1 2 𝑓1
1 5
4
𝑄3 𝑞𝑛 = 𝑞𝐿
𝑄= = 𝑄3 1
𝑄4 𝑞𝐿 1 𝑞𝐿 3 2
𝑄5 𝐿𝑓4 + 𝐿𝑓3 1 1
2 2 1 2 3
𝑄6 𝑄5
NBC, 𝑞𝑛 = 𝑞𝐵
𝑄6

• Note that it is not possible to evaluate 𝑄3 , 𝑄5 and 𝑄6 exactly, and these are not
necessary due to given EBC for 𝑢3 , 𝑢5 and 𝑢6 .

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-41
 Calculation of 𝑄 (cont’d)
• Question : What if 𝑞𝑛 is not constant at an NBC boundary?
• Answer : Just evaluate the line integrals with the given variable 𝑞𝑛 .

• Question : What if the BC is not NBC but MBC? Consider the following case where
bottom BC is MBC with constant 𝛼 and 𝛽.

3 EBC
NBC 1 5
4
𝑞𝑛 = 𝑞𝐿
1 2
1 1
1 2 3
MBC, 𝑞𝑛 = 𝛼𝑢 + 𝛽

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-42
 Calculation of 𝑄 (cont’d)
𝑄11 = 𝑆1 𝑞𝑛 𝑑𝑠 + 𝑆1 𝑞𝑛 𝑑𝑠 6
𝑓1 𝑓4
3 EBC
NBC 1
Same as 𝑞𝐿 1 4 5
= 𝐿𝑓4 𝑞𝑛 = 𝑞𝐿
before 2 1 2
1 1
1 2 3
𝑠 1 MBC, 𝑞𝑛 = 𝛼𝑢 + 𝛽
1− 1 𝛼𝑢𝑓1 + 𝛽 𝑑𝑠
𝑓1 𝐿𝑓1

1
𝑢𝑓1 1 𝑠 𝑠 1
𝑢21 𝑢𝑓1 1
= 1 − 1 𝑢1 + 1 𝑢2
𝑢11 𝐿𝑓1 𝐿𝑓1
𝑆1 1 1
1 𝑢 2 − 𝑢1
𝑠 𝑢𝑓1 = 𝑢11 + 𝑠
1 f1 2 𝐿1𝑓1

𝐿1𝑓1

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-43
 Calculation of 𝑄 (cont’d)
𝐿1𝑓1 1 1
𝑠 𝑢 2 − 𝑢1 𝑞𝐿 1
𝑄11 = 1− 1 𝛼 𝑢11 + 1 𝑠 + 𝛽 𝑑𝑠 + 𝐿𝑓4
𝑠=0 𝐿𝑓1 𝐿𝑓1 2

6
𝛽 1 𝛼𝐿1𝑓1 𝛼𝐿1𝑓1
𝑞𝐿 1
𝑄11 = 𝐿𝑓1 + 𝑢11 + 𝑢12 + 𝐿𝑓4 EBC
2 3 6 2 NBC
3
1 5
4
𝑞𝑛 = 𝑞𝐿
Contribution of the MBC Contribution of the 1 2
of the bottom face NBC of the left face 1 1
1 2 3
MBC, 𝑞𝑛 = 𝛼𝑢 + 𝛽
• To calculate 𝑄21 a similar integral is evaluated but this time with 𝑆2 .
𝐿1𝑓1
1
𝑠 1
𝑢12 − 𝑢11
𝑄2 = 1 𝛼 𝑢1 + 1 𝑠 + 𝛽 𝑑𝑠
𝑠=0 𝐿𝑓1 𝐿𝑓1
1 1
𝛽 𝛼𝐿𝑓1 1 𝛼𝐿𝑓1 1
𝑄21 = 𝐿1𝑓1 + 𝑢1 + 𝑢
2 6 3 2

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-44
 Calculation of 𝑄 (cont’d)
6
• Calculation of 𝑄12 is just the same as 𝑄11 .
2 2 EBC
𝛽 𝛼𝐿𝑓1 𝛼𝐿𝑓1 2 3
𝑄12 = 𝐿2𝑓1 + 𝑢12 + 𝑢 NBC 1 5
2 3 6 2 𝑞𝑛 = 𝑞𝐿
4
1 2
1 1
1 2 3
• Assembled {𝑄} is MBC, 𝑞𝑛 = 𝛼𝑢 + 𝛽

𝛽 1 𝛼𝐿1𝑓1 1 𝛼𝐿1𝑓1 1 𝑞𝐿 1
𝐿 + 𝑢 + 𝑢 + 𝐿 Circled terms
𝑄1 2 𝑓1 3 1 6 2 2 𝑓4 need to be
𝑄2 𝛽 1 𝛼𝐿1𝑓1 1 𝛼𝐿1𝑓1 1 𝛽 2 𝛼𝐿2𝑓1 2 𝛼𝐿2𝑓1 2 transferred to
𝐿 + 𝑢 + 𝑢 + 𝐿 + 𝑢 + 𝑢 the [𝐾]
𝑄3 2 𝑓1 6 1 3 2 2 𝑓1 3 1 6 2
𝑄= = 𝑄3 matrix. So it is
𝑄4
𝑄5 𝑞𝐿 1 𝑞𝐿 3 better to
𝐿𝑓4 + 𝐿𝑓3 write them
𝑄6 2 2 using global
𝑄5
indices.
𝑄6

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-45
 Example 5.5
e.g. Example 5.5: Determine the temperture distribution over the following 2D
geometry. Obtain unknown nodal temperatures. Thermal conductivity of the
medium is 1.3 𝑊 (𝑚𝐾) .
First local corners of the elements are shown with ‘‘1’’s inside the elements.

6
Node coordinates [m]
𝑇 = 100 ℃
4 Node 1 : (0, 0)
1 5
Insulated 4 1 Node 2 : (0.5, 0)
2
3
Node 3 : (1, 0)
1
1 1
𝑥 Node 4 : (0, 0.5)
1 2 3 Node 5 : (0.5, 0.5)
𝑑𝑇 Node 6 : (0, 1)
−𝑘 = −ℎ(𝑇 − 𝑇∞ )
𝑑𝑦
𝑊
ℎ = 5 𝑚2𝐾 , 𝑇∞ = 20 ℃

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-46
 Example 5.5 (cont’d)
• Governing DE is
−𝛻 ⋅ 𝑘𝛻𝑇 = 0
• Elemental weak form (Slide 5-6) is

𝜕𝑇 𝜕𝑤 𝜕𝑇 𝜕𝑤
𝑘 + 𝑑Ω = 𝑤𝑓 𝑑Ω + 𝑤𝑞𝑛 𝑑Γ
Ω 𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 Ω𝑒 Γ𝑒

𝜕𝑇 𝜕𝑇
𝜕𝑆𝑖𝑒 𝜕𝑆𝑗𝑒 𝜕𝑆𝑖𝑒 𝜕𝑆𝑗𝑒 𝑞𝑛 = 𝑘
𝜕𝑥
𝑛𝑥 + 𝑘 𝑛
𝜕𝑦 𝑦
𝐾𝑖𝑗𝑒 = 𝑘 + 𝑑Ω
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦

𝐹𝑖𝑒 = 𝑓𝑆𝑖 𝑑Ω
Ω𝑒

• We can start by calculating the Jacobian matrix of each element.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-47
 Example 5.5 (cont’d)
0 0 𝑦
−1 1 0 0.5 0
• e=1 : 𝐽1 = 0.5 0 =
−1 0 1 0 0.5
0 0.5
𝐽1 = 0.25 1
4
2 0 2 1
𝐽1 −1 =
0 2 1 3
0.5 0.5 1 1 𝑥
−1 1 0 −0.5 0
• e=2 : 𝐽2 = 0 0 =
−1 0 1 0 −0.5
0.5 0
𝐽2 = 0.25
−2 0
𝐽2 −1 =
0 −2

• Elements 3 and 4 have the same shape and size as element 1 and their first local node
is at the right angle corner. Therefore their Jacobian matrices are the same.

𝐽3 = 𝐽4 = 𝐽1

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-48
 Example 5.5 (cont’d)
• Elemental systems can now be calculated.
𝑒 𝜕𝑆 𝑒 𝑒 𝜕𝑆 𝑒
𝜕𝑆 𝑖 𝑗 𝜕𝑆𝑖 𝑗
𝐾𝑖𝑗𝑒 = 𝑘 + 𝑑Ω
Ω𝑒 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦

𝜕𝑆𝑖 𝑒 −1 𝜕𝑆𝑖 𝑒 −1 𝜕𝑆𝑗 𝑒 −1 𝜕𝑆𝑗 𝑒 −1

𝐾𝑖𝑗𝑒 = 𝑘 𝐽 11 + 𝐽 12 𝐽 11 + 𝐽 12
Ω𝑒 𝜕𝜉 𝜕𝜂 𝜕𝜉 𝜕𝜂
𝜕𝑆𝑖 𝑒 −1 𝜕𝑆𝑖 𝑒 −1 𝜕𝑆𝑗 𝑒 −1 𝜕𝑗 𝑒 −1
+ 𝐽 21 + 𝐽 22 𝐽 21 + 𝐽 22 𝐽𝑒 𝑑Ω
𝜕𝜉 𝜕𝜂 𝜕𝜉 𝜕𝜂

𝐹𝑖𝑒 = 0

e=1 :
1.3 −0.65 −0.65
𝐾 1 = −0.65 0.65 0
−0.65 0 0.65

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-49
 Example 5.5 (cont’d)
e=2 :
1.3 −0.65 −0.65
𝐾 2 = −0.65 0.65 0
−0.65 0 0.65

e=3 and 4 : 𝐾 3 = 𝐾 4 = [𝐾 1 ]
𝑦

6
𝑇 = 100 ℃
• Now the {𝑄} vector should be calculated. 4
1 5
NBC 4 1
• Only contribution will come from the two 𝑞𝑛 = 0 2
MBC faces at the bottom. 1 3
1 1
𝑥
1 2 3
𝑑𝑇
𝑞𝑛 = −𝑘 = −ℎ(𝑇 − 𝑇∞ )
𝑑𝑦
𝛼 = −5, 𝛽 = 100
METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-50
 Example 5.5 (cont’d)

Contributions of the MBC at 3

1
face 1 of e=1 to 𝑄1 and 𝑄2 . 1 1
1 2 3
𝛽𝐿1𝑓1 𝛼𝐿1𝑓1 𝛼𝐿1𝑓1 𝛽𝐿1𝑓1 𝛼𝐿1𝑓1 𝛼𝐿1𝑓1
+ 𝑇 + 𝑇 + 𝑇 + 𝑇
2 3 1 6 2 2 6 1 3 2

Contribution of the MBC at

face 1 of e=3 to 𝑄2 and 𝑄3 .
The contribution to 𝑄3 is not 3
1
required because node 3 is an 1 1
1 2 3
EBC node.
𝛽𝐿3𝑓1 𝛼𝐿3𝑓1 𝛼𝐿3𝑓1
+ 𝑇 + 𝑇 No need
2 3 2 6 3

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-51
 Example 5.5 (cont’d)
100 −5(0.5) −5(0.5)
0.5 + 𝑇1 + 𝑇2
2 3 6
100 −5(0.5) −5(0.5) 100 −5(0.5) −5(0.5)
0.5 + 𝑇1 + 𝑇2 + 0.5 + 𝑇2 + 𝑇3
𝑄= 2 6 3 2 3 6
𝑄3
0
𝑄5 Pay attention
𝑄6

25 − 0.8333𝑇1 − 0.4167𝑇2
50 − 0.4167𝑇1 − 1.6667𝑇2 − 0.4167𝑇3
𝑄3
𝑄=
0
𝑄5
𝑄6

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-52
 Example 5.5 (cont’d)
• Global system is
1.3 −0.65 0 −0.65 0 0 𝑇1 25 − 0.8333𝑇1 − 0.4167𝑇2
2.6 −0.65 0 −1.3 0 𝑇2 50 − 0.4167𝑇1 − 1.6667𝑇2 − 0.4167𝑇3
0.65 0 0 0 𝑇3 𝑄3
=
2.6 −1.3 −0.65 𝑇4 0
sym. 2.6 0 𝑇5 𝑄5
0.65 𝑇6 𝑄6

• Take the unknonws due to MBC from the {𝑄} vector into the [𝐾] matrix.
2.1333 −0.2333 0 −0.65 0 0 𝑇1 25
−0.2333 4.2667 −0.2333 0 −1.3 0 𝑇2 50
0 −0.65 0.65 0 0 0 𝑇3 𝑄3
=
−0.65 0 0 2.6 −1.3 −0.65 𝑇4 0
0 −1.3 0 −1.3 2.6 0 𝑇5 𝑄5
0 0 0 −0.65 0 0.65 𝑇6 𝑄6

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-53
 Example 5.5 (cont’d)
• Apply reduction for the known 𝑇3 , 𝑇5 and 𝑇6 .

2.1333 −0.2333 −0.65 𝑇1 25

−0.2333 4.2667 0 𝑇2 = 50 + 0.2333 100 + 1.3(100)
−0.65 0 2.6 𝑇4 0 + 1.3 100 + +0.65(100)

• As seen MBC’s do not destroy the symmetry of the reduced system.

• Solve for the unknown primary variables

𝑇1 43.3
𝑇2 = 50.0 ℃
𝑇4 85.8

• Constant 𝑇 lines should be parallel to the EBC

boundary and they should be perpendicular
to the insulated boundary.

METU – Dept. of Mechanical Engineering – ME 413 Int. to Finite Element Analysis – Lecture Notes of Dr. Sert 5-54