THE FINITE ELEMENT METHOD THEORY IN DEPTH

Published by FEA Academy

https://www.fea-academy.com
The Finite Element Method Theory in Depth © 2023 Dominique Madier
All Rights Reserved.

Legal Disclaimer and Exclusion of Liability

The author reserves the right to make changes in information contained in this book without prior
notice. The concepts, methods, and examples presented in this text are for illustrative and
educational purposes only and are not intended to be exhaustive or to apply to any particular
engineering problem. The author assumes no liability or responsibility to any person or company for
direct or indirect damages resulting from the use of any information contained herein.
Intellectual Property Rights
Any reproduction of any extract of this book by any means whatsoever, including photocopying,
microfilm or scanning is strictly prohibited without the written permission of the author.
Unauthorized reproduction of this publication will be considered as a copyright infringement.
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

TABLE OF CONTENTS
1. INTRODUCTION ..................................................................................................... 7

2. THE EQUILIBRIUM EQUATION ................................................................................... 9

3. THE DISPLACEMENT METHOD..................................................................................11

3.1. THE THREE CONDITIONS .................................................................................................... 11
3.2. THE STIFFNESS MATRIX ..................................................................................................... 12
3.2.1. Element Stiffness Matrix .............................................................................................................. 12
3.2.2. Global Stiffness Matrix ................................................................................................................. 12

3.3. THE LINEAR SPRING MODEL ............................................................................................... 15

3.4. APPLICATION TO THE TWO-SPRING SYSTEM............................................................................ 17
3.5. APPLICATION TO THE FOUR-SPRING SYSTEM ........................................................................... 22
3.6. APPLICATION TO A PARALLEL-SPRING SYSTEM ......................................................................... 24

4. THE PRINCIPLE OF MINIMUM POTENTIAL ENERGY ........................................................26

5. ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES ................................................28

5.1. DEGREES OF FREEDOM ...................................................................................................... 28
5.2. SHAPE FUNCTIONS ........................................................................................................... 30
5.3. 1D TRUSS ELEMENT ......................................................................................................... 31
5.3.1. Summary of the Displacement Method ....................................................................................... 31
5.3.2. Truss Element in Axial Loading ..................................................................................................... 32
Degrees of Freedom ............................................................................................................................ 32
Shape Functions ................................................................................................................................... 32
Equilibrium ........................................................................................................................................... 33
Strain-Displacement & Stress-Strain Relationships ............................................................................. 34
Element Stiffness Matrix ...................................................................................................................... 34
5.3.3. Truss Element in Torsion .............................................................................................................. 35
Degrees of Freedom ............................................................................................................................ 35
Shape Functions ................................................................................................................................... 35

3
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Strain-Angle & Stress-Strain Relationships .......................................................................................... 35

Element Stiffness Matrix ...................................................................................................................... 36
5.3.4. Complete Stiffness Matrix of the Truss Element .......................................................................... 37
5.3.5. Application to the Two-Truss Assembly ....................................................................................... 37
5.3.6. Generic Form of the Truss Element Stiffness Matrix ................................................................... 39
Internal Strain Energy WI ..................................................................................................................... 39
External Work WE................................................................................................................................. 40
Strain .................................................................................................................................................... 40
The Total Potential Energy & the Virtual Work Principle .................................................................... 41
Stiffness Matrix .................................................................................................................................... 42

5.4. 1D BEAM ELEMENT ......................................................................................................... 43

5.4.1. Euler-Bernoulli Beam Theory ....................................................................................................... 43
Degrees of Freedom ............................................................................................................................ 43
Equilibrium ........................................................................................................................................... 46
Shape Functions ................................................................................................................................... 48
Strain-Displacement & Stress-Strain Relationships ............................................................................. 50
Element Stiffness Matrix ...................................................................................................................... 51
5.4.2. Timoshenko Beam Theory ............................................................................................................ 52
Equilibrium ........................................................................................................................................... 52
Strain-Displacement & Stress-Strain Relationships ............................................................................. 53
Shape Functions ................................................................................................................................... 53
Element Stiffness Matrix ...................................................................................................................... 55
5.4.3. Bending and Torsion ..................................................................................................................... 58
Degrees of Freedom ............................................................................................................................ 58
Element Stiffness Matrix ...................................................................................................................... 59
5.4.4. Full Capabilities Beam .................................................................................................................. 60
Degrees of Freedom ............................................................................................................................ 60
Element Stiffness Matrix ...................................................................................................................... 61

5.5. 2D ELEMENTS ................................................................................................................ 63

5.5.1. Overview....................................................................................................................................... 63
5.5.2. Membrane: The Constant Strain Triangle (CST) ........................................................................... 65
The Concepts of Plane Stress and Plane Strain.................................................................................... 65
The Concept of Two-Dimensional State of Stress and Strain .............................................................. 66
Degrees of Freedom ............................................................................................................................ 69
Shape Functions ................................................................................................................................... 69
Strain-Displacement & Stress-Strain Relationships ............................................................................. 73

4
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element Stiffness Matrix ...................................................................................................................... 75

Matrix Assembly .................................................................................................................................. 79
Explicit Expression of the CST Stiffness Matrix .................................................................................... 79
5.5.3. Membrane: The Linear Strain Triangle (LST) ................................................................................ 82
Degrees of Freedom ............................................................................................................................ 82
Shape Functions ................................................................................................................................... 82
Strain-Displacement & Stress-Strain Relationships ............................................................................. 85
Element Stiffness Matrix ...................................................................................................................... 86
5.5.4. Thin-Plate ..................................................................................................................................... 87
Assumptions and Basic Equations of the Kirchhoff Theory ................................................................. 87
Degrees-of-freedom ............................................................................................................................ 91
Shape Functions ................................................................................................................................... 93
Curvature-Displacement & Moment-Curvature Relationships ........................................................... 96
Element Stiffness Matrix ...................................................................................................................... 97
5.5.5. Isoparametric Formulation ........................................................................................................... 98
The Natural Coordinate System ........................................................................................................... 98
Degrees of Freedom ............................................................................................................................ 99
Shape Functions ................................................................................................................................... 99
Strain-Displacement & Stress-Strain Relationships ........................................................................... 100
Element Stiffness Matrix .................................................................................................................... 106
5.5.6. Thick-Plate .................................................................................................................................. 107
Assumptions and Basic Equations of the Mindlin Theory ................................................................. 107
Degrees-of-freedom .......................................................................................................................... 110
Shape Functions ................................................................................................................................. 111
Curvature-Displacement, Shear-Strain & Moment-Curvature Relationships.................................... 111
Element Stiffness Matrix .................................................................................................................... 116
5.5.7. Shell ............................................................................................................................................ 119
Definition ........................................................................................................................................... 119
Degrees of Freedom .......................................................................................................................... 119
Element Stiffness Matrix .................................................................................................................... 120

5.6. 3D SOLID ELEMENT ....................................................................................................... 122

5.6.1. Linear Hexahedral Element ........................................................................................................ 122
Degrees-of-freedom .......................................................................................................................... 122
Shape Functions ................................................................................................................................. 123
Strain-Displacement & Stress-Strain Relationships ........................................................................... 124
Element Stiffness Matrix .................................................................................................................... 127
5.6.2. Quadratic Hexahedral Element .................................................................................................. 129
Degrees of freedom ........................................................................................................................... 129
5
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Shape Functions ................................................................................................................................. 129

Element Stiffness Matrix .................................................................................................................... 131

6. HOW IS THE STIFFNESS MATRIX ASSEMBLED? ........................................................... 132

6.1. MATRIX ASSEMBLY ........................................................................................................ 132
6.2. TAKING ADVANTAGE OF SPARSITY AND SYMMETRY ................................................................ 139
6.3. BANDED MATRIX .......................................................................................................... 140
6.4. SKYLINE MATRIX STORAGE .............................................................................................. 142

7. HOW ARE FEM EQUATIONS SOLVED? ..................................................................... 144

7.1. DIRECT SOLUTION .......................................................................................................... 144
7.2. ITERATIVE SOLUTION ...................................................................................................... 148

6
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

THE FINITE ELEMENT METHOD THEORY IN DEPTH

Mathematics is not about numbers,

equations, computations, or algorithms:
It is about understanding.
William Paul Thurston

1. INTRODUCTION
As a mechanical engineer, you most likely will not develop your own finite element software, you
will typically use commercial finite element software to solve complex structural problems. You may
ask yourself why you should learn the theory of Finite Element Method (FEM).
It is certainly important to develop a practical understanding of FEA. However, it is also essential to
understand exactly what the solver is doing in the background. If you have an issue with an analysis
or an unexpected problem, you may need to understand what is happening in the background. This
document therefore presents the theoretical aspect of the finite element method.
In chapter 5 of “Practical Finite Element Analysis for Mechanical Engineers”, I present the Finite
Element Method theory in fifty pages without showing all the demonstrations, just as an
introduction. Here, I want to go deeper and not only present you with the results, but I also want to
show you in detail where the things are coming from.
This document presents how to obtain the stiffness matrices of various topologies of elements. Then,
I will present the matrix assembly and decomposition processes. Finally, I discuss the methods of
solving simultaneous linear equations.
For a good understanding of stiffness matrix computation, you should know the basic equations of
the theory of elasticity. They are presented in many texts so I will not discuss these equations.
However, I recommend, especially, the book of Timoshenko and Goodier, Theory of Elasticity, 3rd
Edition, 1970.

BY DOMINIQUE MADIER – FEA ACADEMY DIRECTOR 7

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

There are three basic sets of equations that the reader has to know:
1. The differential equations of equilibrium formulated in terms of the stresses acting on a body.
2. The strain-displacement and compatibility differential equations.
3. The stress-strain or material constitutive laws.
In Appendix C of “A First Course in the Finite Element Method, 4th Edition, 2006”, Professor Daryl L.
Logan did a remarkable and complete summary of these sets of equations. I will use these equations
for the derivation of the element’s stiffness matrices.
The theoretical concepts presented here are tackled step by step, from simplicity to complexity. If
you are a beginner with the FEM theory, I strongly recommend you read this document from the
beginning, following the concepts from the simplest topologies of 1D elements to the more complex
2D and 3D elements.

BY DOMINIQUE MADIER – FEA ACADEMY DIRECTOR 8

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

2. THE EQUILIBRIUM EQUATION

With the finite element method, a continuous system described by differential equations is idealized
by a discrete system called the mesh. The discrete system is a collection of elements connected at
nodes. A structural problem with an infinite number of degrees of freedom is converted into a
problem with a finite number of degrees of freedom, making the problem solvable by a computer.
When the nodes are displaced, the elements have an elastic response within their domain, providing
a representation of the elastic properties of the complete system.

Reality: Continuous system subjected to continuous phenomena

Structural
idealization

Idealization: Discrete System

FIG 1 – FROM CONTINUOUS SYSTEM TO DISCRETE SYSTEM

THE EQUILIBRIUM EQUATION 9
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

For continuum mechanics problems, the unknown quantities of the mathematical model are the
nodal displacements. The finite element method is based on the following matrix equation, which
defines the state of equilibrium of forces acting on a structure:

[𝑴]{𝒖̈ } + [𝑩]{𝒖̇ } + [𝑲]{𝒖} = 𝑷(𝒕) Eq 1

[M] Mass Matrix {ü } Acceleration Vector P(t) Applied Load
[B] Damping Matrix {u̇ } Velocity Vector
[K] Stiffness Matrix {u} Displacement Vector

Variations of this equation of equilibrium are used to solve different types of structural problems:

GENERAL EQUATION OF EQUILIBRIUM

M ü + B u̇ + K u =P(t)

Static Analysis Buckling Analysis

K u =P K +  Kd f = P

Transient & Frequency

Vibration Analysis
Response Analysis
K − M f =0 M ü + B u̇ + K u =P(t)

 is the ith eigenvalue.

{} is the ith eigenvector.
[Kd ] is the Differential Stiffness Matrix (called also geometric stiffness).
FIG 2 - VARIATIONS OF THE GENERAL EQUATION OF EQUILIBRIUM

To solve structural problems, both the displacement method and the minimum potential energy
approach are employed. We will have a look at these methods, and we will dive in detail into the
formulation of the stiffness matrix for various topologies of elements. Finally, we will have a look at
the process of matrix assembly and decomposition as well as the methods to solve the equation of
equilibrium.
THE EQUILIBRIUM EQUATION 10
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

3. THE DISPLACEMENT METHOD

The displacement method (also called the stiffness method) assumes displacements at the nodes as
the unknowns of the problem. I use the linear spring to explain the stiffness matrix concept, because
it is a simple element that facilitates the illustration of basic concepts. I also present how to assemble
the stiffness matrix of a system comprising several springs.
First, the following introduces three conditions that must always be satisfied in any analysis.

3.1. THE THREE CONDITIONS

In a linear static analysis, the following three conditions must be satisfied:
1. The equilibrium of forces and moments
2. The compatibility of deformations
3. The Stress-Strain relations

Equilibrium
Let recall the Newton’s second law of motion:
• The resultant force acting on a body is equal to the rate of change of linear momentum (mass
d
times velocity): ∑ F = dt (mv)
• The resultant moment is equal to the rate of change of angular momentum (moment of
d
inertia times angular velocity): ∑ M = dt (Iω)

A consequence of Newton’s second law is that static equilibrium requires that the sum of forces and
moments disappears at all points of the structure. Therefore, for a static body, the equilibrium
equations are written as:

∑ 𝐹⃗ =0
Eq 2
⃗⃗⃗ = 0
∑𝑀

Compatibility of deformations
The compatibility of deformations (also called the strain-displacement relations) requires that the
displacements are continuous across the body.

Stress-Strain relations
The generalized Hooke’s law states that for a linear material:
{𝜎} = [𝐸]{𝜀} Eq 3

THE DISPLACEMENT METHOD - THE THREE CONDITIONS 11

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Where:
• {σ}={σx σy σz τxy τyz τzx } is the stress vector
• {ε}={ϵx ϵy ϵz γxy γyz γzx } is the strain vector
• [E] is the 6x6 matrix of elastic constants.

3.2. THE STIFFNESS MATRIX

The three above conditions are used to generate a system of equations in which the displacements
are unknown. In the displacement method, the key step is the formulation of the element stiffness
matrix. The stiffness matrix [K] is used to relate the forces acting on the structure and the
displacements resulting from these forces:

{𝐹} = [𝐾]{𝑢} Eq 4
Where:
• {F} are the forces acting on the structure
• [K] is the stiffness matrix [kij]
• Each kij term is the force of a constraint at coordinate i due to a unit displacement at j with
all other displacements set equal to zero
• {u} are the displacements resulting from {F}
Each element in a finite element model is represented by its own stiffness matrix [K]element
To prevent rigid body motions, boundary conditions are applied, and the system of linear equations
is solved for the unknown {u}.

3.2.1. Element Stiffness Matrix

The stiffness matrix of a structural system can be derived by various methods. For each element
composing the structural mesh, the stiffness matrix is computed in the element local axis and is
transformed to a global coordinate system common to all elements. The stiffness matrix is an
inherent property of the structure. The properties of the stiffness matrix are:
• The stiffness matrix is square
• The stiffness matrix is symmetric
• All diagonal terms are positive

3.2.2. Global Stiffness Matrix

Once the element stiffness matrices are calculated, they are assembled into a unique global stiffness
matrix depending on how the elements are connected. As we will see in paragraph 6 page 132,
stiffness at a node is obtained by summing the stiffness of all elements connected to that node.

THE DISPLACEMENT METHOD - THE STIFFNESS MATRIX 12

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Assembling the global stiffness matrix of a structure therefore consists of placing each degree of
freedom for each element in its proper position in the global stiffness matrix of the whole structure.

THE DISPLACEMENT METHOD - THE STIFFNESS MATRIX 13

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Mesh the structure with 1D, 2D and/or 3D elements

Assign geometrical properties to the elements

Assign material properties to the elements

Compute the element stiffness matrix [ke] in the local

coordinate system

Convert the element local stiffness matrix [ke] in the global

coordinate system

Place the element stiffness matrix into the global stiffness

matrix at the location corresponding to the element’s degrees
of freedom

Are all
No
elements
over?

Ye

Store the global stiffness matrix

FIG 3 – ASSEMBLY OF STIFFNESS MATRIX FROM LOCAL TO GLOBAL

THE DISPLACEMENT METHOD - THE STIFFNESS MATRIX 14

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

3.3. THE LINEAR SPRING MODEL

To understand the basic concepts on which the stiffness method is based, consider the simple,
intuitive example of the linear spring. Consider the linear spring Fig 4 having a length L, a stiffness
“k” and subjected to a nodal tensile force P along its axis.

k
G1 G2 x

u1 u2
L
P G1 G2 P

fx1 fx2

FIG 4 – THE LINEAR SPRING MODEL

The spring deformation is given as:  = u2 – u1

The Hooke’s law relates the acting force P to the displacement: P = K = k(u2 – u1)
Moreover, the equilibrium condition implies:
fx1 = -P and fx2 = P
So, the force-displacements relationships are:
fx1 = – k(u2 – u1) = k(u1 – u2)
fx2 = k(u2 – u1) = k(–u1 + u2)
The force-displacements relationships can be written in matrix form:
𝑓𝑥1 𝑘 −𝑘 𝑢1
{ }=[ ]{ } Eq 5
𝑓𝑥2 −𝑘 𝑘 𝑢2

The coefficient matrix in the above equation is called the local stiffness matrix Kelement of the spring
element:

[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 = [ 𝑘 −𝑘
] Eq 6
−𝑘 𝑘

THE DISPLACEMENT METHOD - THE LINEAR SPRING MODEL 15

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The force vector acting on the element is:

𝑓𝑥1
[𝐹]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 = { } Eq 7
𝑓𝑥2
Then, for a system discretized with N elements, the global stiffness matrix and the global force vector
can be assembled using the nodal force equilibrium, force/deformation, and compatibility equations.
The stiffness matrix of the system composed of N elements expressed in the global coordinates is
then:
𝑁

[𝐾] = ∑[𝐾]𝑒 Eq 8
𝑒=1

The force vector expressed in the global coordinates is:

𝑁

[𝐹] = ∑[𝐹]𝑒 Eq 9
𝑒=1

Let’s apply these results to a two-spring system in series.

THE DISPLACEMENT METHOD - THE LINEAR SPRING MODEL 16

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

3.4. APPLICATION TO THE TWO-SPRING SYSTEM

Consider the following two-spring system composed of two spring k1 and k2 subjected to a force P2
at one extremity and a force P3 in the middle. The extremity at grid 1 is fixed:

Element 1, k1 Element 2, k2
G3 G2 x
G1

P3 P2

FIG 5 – THE TWO-SPRING SYSTEM

So, the following parameters,

• fx1: force reacted at grid 1 of element 1
• fx2: force reacted at grid 2 of element 2
• fx3: force reacted at grid 3 common to elements 1 & 2
• u1: displacement at grid 1 along the x-axis
• u2: displacement at grid 2 along the x-axis
• u3: displacement at grid 3 along the x-axis
can be related in matrix form for each element using the Eq 5:
𝑓𝑥1 𝑘 −𝑘1 𝑢1
For element 1: { }=[ 1 ]{ } Eq 10
𝑓𝑥3 −𝑘1 𝑘1 𝑢3

𝑓𝑥 𝑘 −𝑘2 𝑢3
For element 2: { 3} = [ 2 ]{ } Eq 11
𝑓𝑥2 −𝑘2 𝑘2 𝑢2

The compatibility of displacements requires that: uelement

3
1
=uelement
3
2

The nodal equilibrium equation at each node is then:

P1 = fX1element 1
P2 = fX2element 2 Eq 12
P3 = fX3 element 1 + fX3 element 2

The three above equations can be written as force-displacement equations using Eq 10 & Eq 11:
P1 = k1u1 – k1u3
P2 = -k2u3 + k2u2 Eq 13
P3 = (-k1u1 + k1u3) + (k2u3 – k2u2)

THE DISPLACEMENT METHOD - APPLICATION TO THE TWO-SPRING SYSTEM 17

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The three above equations written in matrix form give:

𝑃1 𝑘1 0 −𝑘1 𝑢1
{𝑃2 } = [ 0 𝑘2 −𝑘2 ] {𝑢2 } Eq 14
𝑃3 −𝑘1 −𝑘2 𝑘1 + 𝑘2 𝑢3
So, the global stiffness matrix of the two-spring system is:
𝑘1 0 −𝑘1
[𝐾] = [ 0 𝑘2 −𝑘2 ] Eq 15
−𝑘1 −𝑘2 𝑘1 + 𝑘2

However, the global stiffness matrix can also be assembled by superposition.

Considering the two-spring system from Fig 5, the elemental stiffness matrix for each element can
be written as:
For element 1: For element 2:
𝑘1 −𝑘1 u1 𝑘2 −𝑘2 u3
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1 = [ ] [𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2 = [ ]
−𝑘1 𝑘1 u3 −𝑘2 𝑘2 u2
u1 u3 u3 u2

If we write the expanded stiffness matrix in global format for each element, we obtain:
𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1
𝑘1 0 −𝑘1 𝑢1 𝑓𝑥−𝑁𝑜𝑑𝑒 1
For element 1: [ 0 0 0 ] {𝑢2 } = { 0 }
−𝑘1 0 𝑘1 𝑢3 𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1
𝑓𝑥−𝑁𝑜𝑑𝑒 3

0 0 0 𝑢1 0
𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2
For element 2: [0 𝑘2 −𝑘2 ] {𝑢2 } = {𝑓𝑥−𝑁𝑜𝑑𝑒 2 }
0 −𝑘2 𝑘2 𝑢3 𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2
𝑓𝑥−𝑁𝑜𝑑𝑒 3

From the force equilibrium equations:

𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1
𝑓𝑥−𝑁𝑜𝑑𝑒 1
0 𝑃1
𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2
{ 0 } + {𝑓𝑥−𝑁𝑜𝑑𝑒 2 } = {𝑃2 } Eq 16
𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1 𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2 𝑃3
𝑓𝑥−𝑁𝑜𝑑𝑒 3 𝑓𝑥−𝑁𝑜𝑑𝑒 3

We obtain,
𝑃1 𝑘1 0 −𝑘1 𝑢1
{𝑃2 } = [ 0 𝑘2 𝑢
−𝑘2 ] { 2 } Eq 17
𝑃3 −𝑘1 −𝑘2 𝑘1 + 𝑘2 𝑢3
Which is the same result as Eq 14.

THE DISPLACEMENT METHOD - APPLICATION TO THE TWO-SPRING SYSTEM 18

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

However, we can avoid the expansion of each element stiffness matrix by using a more direct form
of the stiffness matrix. The global stiffness matrix may be constructed by directly adding terms
associated with the degrees of freedom in Kelement 1 and Kelement 2 to their corresponding degrees
of freedom:
For element 1: For element 2:
𝑘1 −𝑘1 u1 𝑘2 −𝑘2 u3
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1 = [ ] [𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2 = [ ]
−𝑘1 𝑘1 u3 −𝑘2 𝑘2 u2
u1 u3 u3 u2

u1 u2 u3
𝑘1 0 −𝑘1 u1
Eq 18
[𝐾] = [ 0 𝑘2 −𝑘2 ] u2
−𝑘1 −𝑘2 𝑘1 + 𝑘2 u3

So, in order to solve the two-spring system:

𝑃1 𝑘1 0 −𝑘1 𝑢1
{𝑃2 } = [ 0 𝑘2 −𝑘2 ] {𝑢2 } Eq 19
𝑃3 −𝑘1 −𝑘2 𝑘1 + 𝑘2 𝑢3
we have to verify if the stiffness matrix is not singular. In other words, we have to make sure that its
determinant is not null.
𝑘1 0 −𝑘1
𝑘 −𝑘2 0 −𝑘2 0 𝑘2
𝑑𝑒𝑡[𝐾] = | 0 𝑘2 −𝑘2 | = 𝑘1 | 2 | −0| | − 𝑘1 | |
−𝑘2 𝑘1 + 𝑘2 −𝑘1 𝑘1 + 𝑘2 −𝑘1 −𝑘2
−𝑘1 −𝑘2 𝑘1 + 𝑘2
Eq 20

𝑑𝑒𝑡[𝐾] = 𝑘1 [𝑘2 (𝑘1 + 𝑘2 ) − 𝑘2 𝑘2 ] − 𝑘1 (𝑘1 𝑘2 ) = 𝑘1 𝑘2 𝑘1 − 𝑘1 𝑘1 𝑘2 = 0 Eq 21

The global stiffness matrix of every system is singular because the structural system is free to move
as a rigid body. To solve the system of equations, some constraints, called boundary conditions, must
be applied. Two types can be applied: homogeneous and non-homogeneous boundary conditions.

THE DISPLACEMENT METHOD - APPLICATION TO THE TWO-SPRING SYSTEM 19

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Homogeneous boundary conditions

The boundary conditions are related to grid 1 which is clamped (homogeneous boundary condition).
So, we can consider that u1 = 0. The equations describing the elongation of the spring system become:
𝑃1 𝑘1 0 −𝑘1 0
{𝑃2 } = [ 0 𝑘2 −𝑘2 ] {𝑢2 } Eq 22
𝑃3 −𝑘1 −𝑘2 𝑘1 + 𝑘2 𝑢3
For homogeneous boundary conditions, we can delete the row and column corresponding to the
zero-displacement degree of freedom. The system becomes:
𝑃 𝑘 −𝑘2 𝑢2
{ 2} = [ 2 ] {𝑢 } Eq 23
𝑃3 −𝑘2 𝑘1 + 𝑘2 3

The determinant of the above stiffness matrix is now non-null, so this system has a solution, u2 and
u3 can be calculated. The reacted force P1 at grid 1 can also be calculated from Eq 22:

P1 = -k1u3 Eq 24
Non-Homogeneous boundary conditions
Let’s consider a non-homogeneous boundary condition at grid 1: u1 = 
The equations describing the elongation of the spring system becomes:
𝑃1 𝑘1 0 −𝑘1 𝛿
{𝑃2 } = [ 0 𝑘2 −𝑘2 ] {𝑢2 } Eq 25
𝑃3 −𝑘1 −𝑘2 𝑘1 + 𝑘2 𝑢3
Where,  is a known displacement and u2 & u3 are the unknown displacements.
For non-homogeneous boundary conditions, we cannot directly delete the row and column
corresponding to the  displacement degree of freedom, we have to transfer the terms from the
stiffness matrix to the left-side force vector before solving the unknown displacements.
P1 = k1 - k1u3
P2 = k2u2 – k2u3 Eq 26
P3 = -k1 - k2u2 +(k1+k2)u3
P1 is a reaction from the support that has moved an amount of  The known  term is moved on the
left side:
𝑃2 𝑘 −𝑘2 𝑢2
{ }=[ 2 ] {𝑢 } Eq 27
𝑃3 + 𝑘1 𝛿 −𝑘2 𝑘1 + 𝑘2 3

THE DISPLACEMENT METHOD - APPLICATION TO THE TWO-SPRING SYSTEM 20

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

U2 and U3 can be calculated. Then the reacted force P1 at grid 1 can also be calculated from:

P1 = k1 - k1u3.

THE DISPLACEMENT METHOD - APPLICATION TO THE TWO-SPRING SYSTEM 21

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

3.5. APPLICATION TO THE FOUR-SPRING SYSTEM

Consider the following four-spring system:

Element 1 Element 2 Element 3 Element 4

G1 G2 G3 G4 G5
P5
k1 k2 k3 k4


FIG 6 – THE FOUR-SPRING SYSTEM

The elemental stiffness matrices are:

u1 u2 u2 u3
𝑘 −𝑘1 𝑘 −𝑘2
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1 =[ 1 ] [𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2 =[ 2 ]
−𝑘1 𝑘1 −𝑘2 𝑘2

u3 u4 u4 u5
𝑘 −𝑘3 𝑘 −𝑘4
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 3 =[ 3 ] [𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 4 =[ 4 ]
−𝑘3 𝑘3 −𝑘4 𝑘4

Using superposition, the global stiffness matrix is:

Element 1 Element 2 Element 3 Element 4
u1 u2 u3 u4 u5
𝑘1 −𝑘1 0 0 0
−𝑘1 𝑘1 + 𝑘2 −𝑘2 0 0
[𝐾] = 0 −𝑘2 𝑘2 + 𝑘3 −𝑘3 0
0 0 −𝑘3 𝑘3 + 𝑘4 −𝑘4
[ 0 0 0 −𝑘4 𝑘4 ]

The global force-displacement equations are:

k1 -k 1 0 0 0 𝑢1 fx1
-k 1 k 1 +k 2 -k 2 0 0 𝑢2 fx2
0 -k 2 k 2 +k 3 -k 3 0 𝑢3 = fx3 Eq 28
0 0 -k 3 k 3 +k 4 -k 4 𝑢4 fx4
[0 0 0 -k 4 {
k4 ] 5𝑢 } {fx5 }

THE DISPLACEMENT METHOD - APPLICATION TO THE FOUR-SPRING SYSTEM 22

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Applying the boundary conditions (u1=0 and u5=) and the known forces (fx2=fx3=fx4=0 & fx5=P5), we
obtain:
𝑘1 −𝑘1 0 0 0 0 𝑓𝑥1
−𝑘1 𝑘1 + 𝑘2 −𝑘2 0 0 𝑢2 0
0 −𝑘2 𝑘2 + 𝑘3 −𝑘3 0 𝑢3 = 0 Eq 29
0 0 −𝑘3 𝑘3 + 𝑘4 −𝑘4 𝑢4 0
[ 0 0 0 −𝑘4 𝑘4 ] { 𝛿 } { 𝑃5 }

𝑘1 + 𝑘2 −𝑘2 0 0 𝑢2 0
−𝑘2 𝑘2 + 𝑘3 −𝑘3 0 𝑢3 0
[ ]{ } = { } Eq 30
0 −𝑘3 𝑘3 + 𝑘4 −𝑘4 𝑢4 0
0 0 −𝑘4 𝑘4 𝛿 𝑃5

For non-homogeneous boundary conditions, we cannot directly delete the row and column
corresponding to the  displacement degree of freedom. We have to transfer the terms from the
stiffness matrix to the right-side force vector before solving the unknown displacements. It gives the
following solvable system:
𝑘1 + 𝑘2 −𝑘2 0 𝑢2 0
[ −𝑘2 𝑘2 + 𝑘3 𝑢
−𝑘3 ] { 3 } = { 0 } Eq 31
0 −𝑘3 𝑘3 + 𝑘4 𝑢4 𝑘4 𝛿

We can calculate the unknown displacements u2, u3 and u4 from this system and also calculate the
reacted force at grid 1 by fx1 = -k1u2.

THE DISPLACEMENT METHOD - APPLICATION TO THE FOUR-SPRING SYSTEM 23

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

3.6. APPLICATION TO A PARALLEL-SPRING SYSTEM

Element 2, k2
G3
G2
Element 1, k1
G1 G2 P x

Element 3, k3
G2 x
G4

FIG 7 – THE PARALLEL-SPRING SYSTEM

The elemental stiffness matrices for each element are:

u1 u2
𝑘 −𝑘1 u1
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 1 =[ 1 ]
−𝑘1 𝑘1 u2

u2 u3
𝑘 −𝑘2 u2
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 2 =[ 2 ]
−𝑘2 𝑘2 u3

u2 u4
𝑘 −𝑘3 u2
[𝐾]𝑒𝑙𝑒𝑚𝑒𝑛𝑡 3 =[ 3 ]
−𝑘3 𝑘3 u4

Using the concept of superposition, the global stiffness matrix is:

u1 u2 u3 u4
𝑘1 −𝑘1 0 0 u1
−𝑘 𝑘1 + 𝑘2 + 𝑘3 −𝑘2 −𝑘3 u2
[𝐾] = [ 1 ] Eq 32
0 −𝑘2 𝑘2 0 u3
0 −𝑘3 0 𝑘3 u4

Applying the boundary conditions: u1 = u3 = u4 =0

Applying the known force fx2 = P
Applying the compatibility of displacements: u2element 1 = u2element 2 = u2element 3 = u2

THE DISPLACEMENT METHOD - APPLICATION TO A PARALLEL-SPRING SYSTEM 24

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The global force-displacement system becomes:

𝑘1 −𝑘1 0 0 0 𝑓𝑥1
−𝑘 𝑘1 + 𝑘2 + 𝑘3 −𝑘2 −𝑘3 𝑢2 𝑃
[ 1 ]{ } = { } Eq 33
0 −𝑘2 𝑘2 0 0 𝑓𝑥3
0 −𝑘3 0 𝑘3 0 𝑓𝑥4

So, solving for the unknown displacement u2:

(k1 + k2 + k3)u2 = P
𝑃
𝑢2 =
𝑘1 + 𝑘2 + 𝑘3

And, solving for the reacted forces gives:

fx1 = -k1u2 fx3 = -k2u2 fx4 = -k3u2

THE DISPLACEMENT METHOD - APPLICATION TO A PARALLEL-SPRING SYSTEM 25

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

4. THE PRINCIPLE OF MINIMUM POTENTIAL ENERGY

One of the alternative methods often used to derive the stiffness matrix for an element is based on
the principle of minimum potential energy. This method has the advantage of being more general
than the methods involving nodal and element equilibrium equations accommodated with the
stress-strain law for the element. The principle of minimum potential energy is more adaptable for
the determination of element equations for complicated elements with large numbers of degrees of
freedom such as the plane stress-strain element, plate bending element, and solid element. Let’s see
how this principle works for the simple linear spring model.

The total potential energy p of a system is the sum of the internal strain energy U and the potential
energy of the external forces We:
p = U + W e Eq 34

• The Internal strain energy U represents the capacity of the internal forces to do work through
deformation of the structure.
• The potential energy of the external forces We represents the capacity of the external forces
to do work through deformation of the structure.
Reminder: the force-displacement relationship for a linear spring is F = kx
The differential internal work (or strain energy) dU in the spring is the internal force multiplied by
the change in displacement:
dU = Fdx = (kx)dx Eq 35
The total strain energy is:
𝑋 𝑋
1
𝑈 = ∫ 𝑑𝑈 = ∫ (𝑘𝑥)𝑑𝑥 = 𝑘𝑥 2 Eq 36
0 0 2

The potential energy of the external forces is the work done by the external forces, so: W = -Fx
Therefore, the total potential energy is:
1
𝜋𝑝 = 𝑘𝑥 2 − 𝐹𝑥 Eq 37
2

Statically, to find the equilibrium of the spring corresponds to find a value of x yielding a stationary
value of the potential energy p. This condition is obtained for:
𝜕𝜋𝑝
=0 Eq 38
𝜕𝑥

THE PRINCIPLE OF MINIMUM POTENTIAL ENERGY 26

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The variation of p can be written as:

𝜕𝜋𝑝 𝜕𝜋𝑝 𝜕𝜋𝑝
𝛿𝜋𝑝 = 𝛿𝑢1 + 𝛿𝑢2 + ⋯ + 𝛿𝑢 Eq 39
𝜕𝑢1 𝜕𝑢2 𝜕𝑢𝑛 𝑛

To satisfy p/x =0, all coefficients associated with ui must be null independently:
𝜕𝜋𝑝
=0 𝑓𝑜𝑟 𝑖 = 1 𝑡𝑜 𝑛 Eq 40
𝜕𝑢𝑖
Consider the linear spring subjected to nodal forces from Fig 8 and let’s derive the spring stiffness
matrix using the principle of minimum potential energy:

k
fx1 G1 G2 fx2

x
L

FIG 8 – THE LINEAR SPRING SYSTEM

The total potential energy p of the above spring is then:

1
𝜋𝑝 = 𝑘(𝑢2 − 𝑢1 )2 − 𝑓𝑥1 𝑢1 − 𝑓𝑥2 𝑢2
2
1 Eq 41
= 𝑘(𝑢22 − 2𝑢1 𝑢2 + 𝑢12 ) − 𝑓𝑥1 𝑢1 − 𝑓𝑥2 𝑢2
2
So,
𝜕𝜋𝑝 𝑘
= (−2𝑢2 + 2𝑢1 ) − 𝑓𝑥1 = 0
𝜕𝑢1 2
Eq 42
𝜕𝜋𝑝 𝑘
= (2𝑢2 − 2𝑢1 ) − 𝑓𝑥2 = 0
𝜕𝑢2 2
Therefore,
fx1 = k(u1 – u2)
Eq 43
fx2 = k(–u1 + u2)
In matrix form, we obtain:
𝑓𝑥1 𝑘 −𝑘 𝑢1
{ }=[ ]{ } Eq 44
𝑓𝑥2 −𝑘 𝑘 𝑢2

Which is the same result as the one obtained with the displacement method (Eq 5).
THE PRINCIPLE OF MINIMUM POTENTIAL ENERGY 27
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5. ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES

The previously presented principle of minimum potential energy (see paragraph 4, page 26) is used
by the FEA software to assemble the stiffness matrix of a structure. In a real problem, it becomes
rapidly complex, since the structures are modeled using different topologies of elements (1D, 2D,
and 3D). Before tackling the topic of stiffness matrices, let’s look at the degrees of freedom and shape
function concepts. It is assumed that you are already familiar with matrix algebra.

5.1. DEGREES OF FREEDOM

In FEA, the degrees of freedom (DoFs) are the fundamental variables. In mechanical structural
analysis, they are usually translations and rotations at the nodes. It is important to keep in mind that
not all types of degrees of freedom are necessarily transmitted by all types of elements. Degrees of
freedom are fundamental to the understanding of how loads can be applied, how boundary
conditions restrain the model, and how the different types of elements can be connected. A
translational degree of freedom stipulates that a force can be transmitted through the nodes, while
a rotational degree of freedom stipulates that a moment can be transmitted though the nodes. Table
1 below summarizes the common structural types of elements, along with their active degrees of
freedom:
TABLE 1 - DOF’S FOR COMMON STRUCTURAL ELEMENTS
ELEMENT DEGREES OF FREEDOM
1 Translation: Tx
Truss u x
1 Rotation: Rx
y
3 Translations: Tx Ty Tz v
Beam u x
3 Rotations: Rx Ry Rz
w
z

y
3 Translations: Tx Ty Tz v
Shell u x
2 In-plane rotations: Rx Ry
w

v
Solid 3 Translations: Tx Ty Tz u
w

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - DEGREES OF FREEDOM 28

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Based on Table 1, you can deduce which elements you can connect or not, at the risk of creating
mechanisms: two connected structural beams, for example, will behave like a fully welded
connection, since the beam elements transmit three translations (one axial and two shear planes)
and three rotations (one torsional moment and two bending moments). However, a beam element
connected to a truss element may behave like a pinned joint because the truss element cannot
transmit all rotations.
An FEA analyst must understand the concept of degrees of freedom very well to select the proper
elements among all the elements in the FEA software library.
The size of the stiffness matrix of each element depends on the number of degrees of freedom
associated with each node of the element. Indeed, for an element composed of n nodes having N
degrees of freedom per node, the elemental stiffness matrix order is n x N. This concept of degrees
of freedom number per element type is fundamental, given that it drives the size of the problem and
then the computing time. The more degrees of freedom in the problem, the greater the number of
equations to be solved and the more computation time it takes to find the solution.
TABLE 2 – SIZE OF THE ELEMENTAL STIFFNESS MATRICES
NUMBER OF
NUMBER OF
DEGREES OF
NODES PER MATRIX
SIZE ELEMENT TYPE ORDER FREEDOM PER
ELEMENT SIZE
NODE
n
N
1D Truss Line Linear 2 2 4x4
1D Beam Line Linear 2 6 12 x 12
2D Membrane Triangular Linear 3 2 6x6
2D Membrane Triangular Quadratic 6 2 12 x 12
2D Membrane Quadrilateral Linear 4 2 8x8
2D Membrane Quadrilateral Quadratic 8 2 16 x 16
2D Shell Triangular Linear 3 5 15 x 15
2D Shell Triangular Quadratic 6 5 30 x 30
2D Shell Quadrilateral Linear 4 5 20 x 20
2D Shell Quadrilateral Quadratic 8 5 40 x 40
3D Solid Tetrahedron Linear 4 3 12 x 12
3D Solid Tetrahedron Quadratic 10 3 30 x 30
3D Solid Hexahedron Linear 8 3 24 x 24
3D Solid Hexahedron Quadratic 20 3 60 x 60

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - DEGREES OF FREEDOM 29

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

When you know the number of nodes that compose your model and the topology of the elements,
you can predict the total number of degrees of freedom for your problem. Given today’s computing
resources, a model with less than 500,000 degrees of freedom is considered a small problem, and a
medium problem is one with between 500,000 and three million degrees of freedom. Large models
comprise more than three million degrees of freedom.

5.2. SHAPE FUNCTIONS

The discretization of the problem into several nodes defined by the FEA analyst means that the
computer will solve the problem at these particular nodes. The nodal results must also be defined in
the elements which connect the nodes. For this purpose, a mathematical object called a shape
function is used to represent assumed behavior for a given element. How well each assumed shape
function matches the true behavior directly affects the solution’s accuracy.
To represent the deformation of elements under loading, we must therefore choose a mathematical
function for each element. For each element, an appropriate shape function is defined to represent
the distribution of displacement within the elements, using an appropriate mathematical function.
The most commonly used functions are polynomials.
Depending on the number of degrees of freedom of the elements, we derive different shape
functions to express the shape of the assumed displacement fields over the domain of the elements.
It is important to choose functions that guarantee the continuity of displacements for all degrees of
freedom at each node of the elements. In other words, the shape functions must prevent openings,
overlaps, and jumps.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - SHAPE FUNCTIONS 30

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.3. 1D TRUSS ELEMENT

5.3.1. Summary of the Displacement Method
The displacement method is used in this paragraph to derive the 1D element’s stiffness matrices. Fig
9 summarizes the steps of the method:

Identify the degrees of freedom of the element

Define the shape functions to describe the

displacement fields through the element

Write the equilibrium equations

Write the strain-displacement & stress-strain

relationships

Derive the Element Stiffness Matrix

FIG 9 – DISPLACEMENT METHOD TO DERIVE THE STIFFNESS MATRICES OF ONE-DIMENSIONAL ELEMENTS

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 31

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.3.2. Truss Element in Axial Loading

Consider a simple truss element of uniform cross section A, length L and made of a homogeneous
material with a Young Modulus E. The truss is subjected to an axial load.
Degrees of Freedom
The axial translations u1 and u2 are the only displacements at grids 1 and 2. So, this element has two
degrees of freedom.

L
P1 G1 G2 P2 X

u1 u2
A

FIG 10 - TRUSS ELEMENT IN AXIAL LOADING

Shape Functions
The truss element resists axial loading only along the local x direction. A linear displacement variation
along the x axis of the truss is assumed because a linear function with specified endpoints has a
unique path. So, the displacement function u to represent the axial displacement throughout the
element is written:

u(x) = a1 + a2x Eq 45
The total number of coefficients ai is equal to the total number of degrees of freedom associated
with the element (here, there are two degrees of freedom, in total: one axial displacement at each
of the two nodes of the element).
Let’s express u as a function of the nodal displacements u1 and u2. This will permit the application of
boundary conditions to nodal displacements and then relate the nodal displacements to the nodal
forces. To do so, we evaluate u at each node and solve for a1 and a2.
u(0) = u1 = a1
Eq 46
u(L)=u2 = u1 + a2L
This system gives:
𝑢2 − 𝑢1
𝑢= 𝑥 + 𝑢1 Eq 47
𝐿
In matrix form:
𝑥 𝑥 𝑢1
𝑢 = [1 − , ]{ } Eq 48
𝐿 𝐿 𝑢2
ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 32
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝑢1
𝑢 = [𝑁1 , 𝑁2 ] {𝑢 } Eq 49
2

Where,
𝑥
𝑁1 = 1 −
𝐿
𝑥 Eq 50
𝑁2 =
𝐿
are the shape functions. These shape functions Ni’s express the shape of the assumed displacement
over the domain of the element (x coordinate) when the i th element degree of freedom has unit
value, and all other degrees of freedom are zero.

u = a 1 + a 2x

x
U1 U2

FIG 11 - DISPLACEMENT FUNCTION FOR TRUSS ELEMENT

x
N1 =1-
1 L
0
L

FIG 12 - SHAPE FUNCTION N1 OVER DOMAIN OF TRUSS ELEMENT

x
N2 =
L 1
0
L

FIG 13 - SHAPE FUNCTION N2 OVER DOMAIN OF TRUSS ELEMENT

In this case, N1 and N2 are linear functions that have the following behavior: N1=1 at node 1 & N1=0
at node 2 whereas N2=0 at node 1 & N2 =1 at node 2. So, for any axial coordinate along the truss
element, N1 + N2 = 1.
Equilibrium
To satisfy the static equilibrium:
P1 + P2 = 0  P2 = -P1

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 33

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Strain-Displacement & Stress-Strain Relationships

The Strain/Displacement relation can be written as:
𝑑𝑢 𝑢2 − 𝑢1
𝜀𝑋 = = Eq 51
𝑑𝑥 𝐿
The Stress/Strain relation gives:

x = Ex Eq 52

Element Stiffness Matrix

The force to stress relation is:
−𝑃1
𝜎𝑋1 =
𝐴
𝑃2 Eq 53
𝜎𝑋2 =
𝐴
So,
𝐸𝐴 𝐸𝐴
−𝑃1 = 𝑢2 − 𝑢 Eq 54
𝐿 𝐿 1
And
𝐸𝐴 𝐸𝐴
𝑃2 = 𝑢2 − 𝑢 Eq 55
𝐿 𝐿 1
Rewrite the two equations above in matrix form:
𝑃 𝐸𝐴/𝐿 −𝐸𝐴/𝐿 𝑢1
{ 1} = [ ]{ } Eq 56
𝑃2 −𝐸𝐴/𝐿 𝐸𝐴/𝐿 𝑢2

So, the 2 x 2 truss element axial stiffness matrix is:

𝐸𝐴/𝐿 −𝐸𝐴/𝐿 Eq 57
[𝐾]𝑇𝑟𝑢𝑠𝑠 𝐴𝑥𝑖𝑎𝑙 = [ ]
−𝐸𝐴/𝐿 𝐸𝐴/𝐿

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 34

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.3.3. Truss Element in Torsion

Degrees of Freedom
Consider a simple truss element of uniform cross section A, length L and made of a homogeneous
material with a Shear Modulus G. The truss is subjected to a torsional load T. The rotations 1 and 2
are the only displacements at grids 1 and 2. So, this element has two degrees of freedom.

L
T1 T2 x
G1 G2
x1 x2
J
FIG 14 - TRUSS ELEMENT IN TORSION
Shape Functions
A linear angle-of-twist variation along the x axis of the truss is assumed:

 = a 1 + a 2x Eq 58
So, expressing a1 and a2 in terms of unknown nodal angles of twist, we obtain:
𝜃2 − 𝜃1
𝜃= 𝑥 + 𝜃1 Eq 59
𝐿
In matrix form:
𝑥 𝑥 𝜃1
𝜃 = [1 − ]{ }
𝐿 𝐿 𝜃2
𝑢 Eq 60
𝑢 = [𝑁1 𝑁2 ] {𝑢1 }
2

Where the shape functions are:

𝑥
𝑁1 = 1 −
𝐿
𝑥 Eq 61
𝑁2 =
𝐿

Strain-Angle & Stress-Strain Relationships

The relationship between the shear strain  and the angle of twist  is obtained from the deformation
of the bar as shown in Fig 15:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 35

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

y
dx

R r
x
O
A
max
B
z
d

FIG 15 – SHEAR STRAIN AND ANGLE OF TWIST FOR THE TRUSS ELEMENT

Based on the assumption that all radial lines remain straight during torsional deformation, the arc
̂ is given by:
length AB
̂ = 𝛾𝑚𝑎𝑥 𝑑𝑥 = 𝑅𝑑𝜃
𝐴𝐵
𝑑𝜃 Eq 62
𝛾𝑚𝑎𝑥 = 𝑅
𝑑𝑥
So, at any radial position r, we have:
𝑑𝜃
𝛾=𝑟 Eq 63
𝑑𝑥
Using the equation expressing a1 and a2 in terms of unknown nodal angles of twist, we have:
𝑟
𝛾= (𝜃 − 𝜃1 ) Eq 64
𝐿 2
The shear stress  is related to the shear strain  using the shear modulus G of the material: =G
Element Stiffness Matrix
We know also that the shear stress is related to the applied torque by:
𝜏𝐽
𝑇= Eq 65
𝑅
Where J is the torsional constant. Then, with the nodal torque sign convention of Fig 14 we have:
𝜏𝐽 𝐺𝛾𝐽 𝐺𝑅𝐽 𝐺𝐽
𝑇2 = = = (𝜃2 − 𝜃1 )  𝑇2 = (𝜃2 − 𝜃1 ) Eq 66
𝑅 𝑅 𝑅𝐿 𝐿
And,
𝐺𝐽
𝑇1 = −𝑇2 = (𝜃 − 𝜃2 ) Eq 67
𝐿 1

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 36

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

In matrix format:
𝑇 𝐺𝐽/𝐿 −𝐺𝐽/𝐿 𝜃1
{ 1} = [ ]{ } Eq 68
𝑇2 −𝐺𝐽/𝐿 𝐺𝐽/𝐿 𝜃2

So, the 2 x 2 truss element torsional stiffness matrix is:

𝐺𝐽/𝐿 −𝐺𝐽/𝐿
[𝐾] 𝑇𝑟𝑢𝑠𝑠 𝑇𝑜𝑟𝑠𝑖𝑜𝑛 = [ ] Eq 69
−𝐺𝐽/𝐿 𝐺𝐽/𝐿

5.3.4. Complete Stiffness Matrix of the Truss Element

Globally, the truss element has two degrees of freedom at each node: the axial translation uX and
the rotation x. Since the element has two nodes, the truss element has four degrees of freedom.
So, we need a 4 x 4 matrix to describe the stiffness of the truss element. The complete stiffness
matrix of the truss element is:
𝑃1 𝐸𝐴/𝐿 0 −𝐸𝐴/𝐿 0 𝑢1
𝑇 0 𝐺𝐽/𝐿 0 −𝐺𝐽/𝐿 𝜃1
{ 1} = [ ] {𝑢 } Eq 70
𝑃2 −𝐸𝐴/𝐿 0 𝐸𝐴/𝐿 0 2
𝑇2 0 −𝐺𝐽/𝐿 0 𝐺𝐽/𝐿 𝜃2

𝐸𝐴/𝐿 0 −𝐸𝐴/𝐿 0
0 𝐺𝐽/𝐿 0 −𝐺𝐽/𝐿 Eq 71
[𝐾] 𝑇𝑟𝑢𝑠𝑠 =[ ]
−𝐸𝐴/𝐿 0 𝐸𝐴/𝐿 0
0 −𝐺𝐽/𝐿 0 𝐺𝐽/𝐿

5.3.5. Application to the Two-Truss Assembly

Now, let’s consider the two-truss assembly clamped at one side and subjected to an axial load P on
the other side.

G1 G2
G3 P
(E1 ; A1) (E2 ; A2) X

L1 L2

u 1, F 1 u 2, F 2 u 3, F 3

FIG 16 – THE TWO-TRUSS ASSEMBLY

Based on the previous derivation of stiffness matrix for a truss element:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 37

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝐸1 𝐴1 𝐸1 𝐴1
−
𝐹 𝑢1 𝐿1 𝐿1 𝑢1
{ 1 } = [𝐾1 ] {𝑢 } = { }
𝐹2 2 𝐸 1 𝐴1 𝐸1 𝐴1 𝑢2
−
[ 𝐿1 𝐿1 ]
𝐸2 𝐴2 𝐸2 𝐴2 Eq 72
−
𝐹 𝑢2 𝐿2 𝐿2 𝑢2
{ 2 } = [𝐾2 ] {𝑢 } = {𝑢 }
𝐹3 3 𝐸2 𝐴2 𝐸2 𝐴2 3
−
[ 𝐿2 𝐿2 ]
By assembling the two stiffness matrices by superposition, the resulting global stiffness matrix is:
𝐸1 𝐴1 𝐸1 𝐴1
− 0
𝐿1 𝐿1
𝐹1 𝐸1 𝐴1 𝐸1 𝐴1 𝐸2 𝐴2 𝐸2 𝐴2 𝑢1
{𝐹2 } = − + − {𝑢2 } Eq 73
𝐹 𝐿1 𝐿1 𝐿2 𝐿2 𝑢3
3
𝐸2 𝐴2 𝐸2 𝐴2
0 −
[ 𝐿2 𝐿2 ]

By applying the external loads to the system:

𝐸1 𝐴1 𝐸1 𝐴1
− 0
𝐿1 𝐿1
0 𝐸1 𝐴1 𝐸1 𝐴1 𝐸2 𝐴2 𝐸2 𝐴2 𝑢1
{0} = − + − {𝑢2 } Eq 74
𝑃 𝐿1 𝐿1 𝐿2 𝐿2 𝑢3
𝐸2 𝐴2 𝐸2 𝐴2
0 −
[ 𝐿2 𝐿2 ]

By imposing the boundary conditions, since the node 1 is fixed, u1 = 0. We can remove row 1 and
column 1 from the global stiffness matrix.
E 1 A1 E 2 A2 E2 A 2
+ -
0 L1 L2 L2 𝑢2
{ }= { } Eq 75
P E 2 A2 E2 A2 𝑢3
-
[ L2 L2 ]
The above system is composed of two equations with two unknowns (u1 and u2) and the matrix is
not singular (det[K]≠0). So, the system has a solution. One way to solve the system would be to invert
[K]. In practice, inverting the stiffness matrix is not an efficient method. Usually, the finite element
solvers use a matrix decomposition procedure to solve the system (see paragraph 7 page 144).

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 38

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.3.6. Generic Form of the Truss Element Stiffness Matrix

To conclude this section on the truss element stiffness matrix, the generic form of the stiffness matrix
using the virtual work principle is presented. The virtual work principle is an alternative to the
displacement method for the derivation of an element stiffness matrix. This method is used for more
complex elements such as 2D and 3D elements. However, employing the method with a simple 1D
element is a good way to familiarize yourself with it.
Let’s consider a generic 1D truss element subjected to an external distributed load q(x) (axial load
per unit length). The element has a constant section A and length L and is made of a homogeneous
material having a Young modulus E.

q(x)
x
G1 G2

U1 U2
A

L
FIG 17 - 1D TRUSS ELEMENT UNDER EXTERNAL DISTRIBUTED LOAD

First, we know that the external work WE done by the forces q(x) is stored in the truss element as
internal strain energy WI. So, let’s define WE and WI.
Internal Strain Energy WI
We know that the strain energy density at each location of the linear-elastic truss element subjected
to a one-dimensional state of stress  and strain  is:
1
𝑈= 𝜎𝜀 Eq 76
2
Then, the total strain energy on the total volume V of the truss element is:
1
𝑊𝐼 = ∫ 𝜎𝜀𝑑𝑉 Eq 77
𝑉2

Since we have =E=constant over the section A of the truss element and since dV=A.dx, the above
equation can be written as:
1
𝑊𝐼 = ∫ . 𝜀. 𝐸. 𝜀. 𝐴. 𝑑𝑥 Eq 78
𝐿2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 39

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

External Work WE
The external work of q(x) forces applied on an infinitesimal length dx at the distance x is: q(x).dx.u(x).
So, the total external work on the length L of the truss element is:

𝑊𝐸 = ∫𝑞(𝑥). 𝑢(𝑥). 𝑑𝑥 Eq 79
𝐿

Strain
We also know that the strain and displacements are related by the equation:
𝑑
𝜀(𝑥) = 𝑢(𝑥) Eq 80
𝑑𝑥
As shown above, the variation of u(x) along the truss element is defined by its shape functions. We
demonstrated that these shape functions for the truss elements are such that the variation of
displacement u(x) in the truss element is linear and can be written as:

u(x) = N1(x).u1 + N2(x).u2 Eq 81

where the shape functions N1 and N2 are defined as:
𝑥
𝑁1 = 1 −
𝐿
𝑥 Eq 82
𝑁2 =
𝐿
In matrix form:
𝑥 𝑥 𝑢1
𝑢(𝑥) = [1 − ]{ } Eq 83
𝐿 𝐿 𝑢2
𝑢
u(x)=[N1 N2 ] { 1 } Eq 84
𝑢2

𝑢(𝑥) = [𝑁]{𝑢} Eq 85
where
• [N] is the shape function vector
• {u} is the nodal displacement vector
So, if we come back to the strain-displacement equation:

𝑑 𝑑
𝜀(𝑥) = 𝑢(𝑥) = [ [𝑁]] {𝑢} Eq 86
𝑑𝑥 𝑑𝑥

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 40

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

1 1
𝜀(𝑥) = [− , ] {𝑢} Eq 87
𝐿 𝐿
where [B] is called the Strain-Displacement Matrix:
1 1
[𝐵] = [− , ] Eq 88
𝐿 𝐿
And
1
−
[𝐵]𝑇 = { 𝐿} Eq 89
1
𝐿
So,

𝜀(𝑥) = [𝐵]{𝑢} = [𝑢][𝐵]𝑇 Eq 90

The Total Potential Energy & the Virtual Work Principle

The total potential energy of the truss element is given by:

p = WI – WE Eq 91
The virtual work principle states that the total work done by all forces acting on the truss element in
static equilibrium is null for any infinitesimal virtual displacement u:

p = WI – WE = 0 Eq 92

From Eq 78 we have,
1
𝛿𝑊𝐼 = ∫(𝜀. 𝐸. 𝛿𝜀 + 𝛿𝜀. 𝐸. 𝜀). 𝐴. 𝑑𝑥 = ∫𝜀. 𝐸. 𝛿𝜀. 𝐴. 𝑑𝑥 Eq 93
2 𝐿 𝐿

And from Eq 79,

𝛿𝑊𝐸 = ∫𝑞(𝑥). 𝛿𝑢(𝑥). 𝑑𝑥 Eq 94

𝐿

Then, as per the equations of WE and WI derived above, we have:

∫𝜀. 𝐸. 𝛿𝜀. 𝐴. 𝑑𝑥 = ∫𝑞(𝑥). 𝛿𝑢(𝑥). 𝑑𝑥 Eq 95

𝐿 𝐿

So, using Eq 85 & Eq 90 in Eq 95, we obtain:

[𝛿𝑢] ∫𝐴. 𝐸. [𝐵]𝑇 . [𝐵]. {𝑢} . 𝑑𝑥 = [𝛿𝑢] ∫𝑞𝑥 {𝑁}. 𝑑𝑥 Eq 96

𝐿 𝐿

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 41

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Where [u] is the vector of nodal virtual displacements.

Then,

[∫𝐴. 𝐸. [𝐵]𝑇 . [𝐵]. 𝑑𝑥] . {𝑢} = ∫𝑞𝑥 . {𝑁}. 𝑑𝑥 Eq 97

𝐿 𝐿

This can be written as:

[K]{u} = {P} Eq 98
Where {u} is the displacement vector and the truss element stiffness matrix [K] in generic format is:

[𝐾] = ∫𝐴. 𝐸. [𝐵]𝑇 . [𝐵]. 𝑑𝑥 Eq 99

𝐿

Since the nodal forces {F} produce the same external work as the distributed load q(x), the work-
equivalent nodal force vector is:

{𝐹} = ∫𝑞𝑥 . {𝑁}. 𝑑𝑥 Eq 100

𝐿

Stiffness Matrix
Using the generic form of the truss element stiffness Eq 99, calculate the truss element stiffness
matrix derived previously:

[𝐾] = ∫𝐴. 𝐸. [𝐵]𝑇 . [𝐵]. 𝑑𝑥 Eq 101

𝐿

1
− 1 1
[𝐾] = ∫ 𝐴. 𝐸. { 𝐿} . [− ] . 𝑑𝑥 Eq 102
𝐿
1 𝐿 𝐿
𝐿
1 1
2
− 2
[𝐾] = ∫ 𝐴. 𝐸. [ 𝐿 𝐿 ] . 𝑑𝑥
Eq 103
𝐿
1 1
− 2
𝐿 𝐿2
𝐸𝐴
[𝐾] = 2 [ 1 −1] ∫𝑑𝑥 Eq 104
𝐿 −1 1 𝐿
𝐸𝐴 1 −1
[𝐾] = [ ] Eq 105
𝐿 −1 1
This is the same result as Eq 57.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D TRUSS ELEMENT 42

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.4. 1D BEAM ELEMENT

The beam element is by far the most complex element in finite element analysis because it can
transfer all types of loading: axial load, transverse shear, bending moments and torsion. The beam
element has therefore all six degrees of freedom: three translations and three rotations.
So, I will present the development of the beam stiffness matrix step by step. First, I will present the
beam stiffness matrix based on Euler-Bernoulli beam theory (bending deformation only) and then I
will present the beam stiffness matrix based on Timoshenko beam theory (including transverse shear
deformation). Finally, I will introduce the torsional stiffness as well as the axial stiffness to complete
the stiffness matrix development.

5.4.1. Euler-Bernoulli Beam Theory

Degrees of Freedom
The basic definition of a beam is: a long structural member subjected to a transverse loading
producing a significant bending effect. The bending effect is characterized by two degrees of
freedom: a transverse displacement uiy and a rotation iz at each end of the beam. So, let’s consider
a simple beam element of:
• A uniform cross-section A
• A principal moment of inertia Iz
• A length L
• Made of a homogeneous material: Young modulus E and shear modulus G

y Homogeneous material [E ; G]

1 2
z x
G1 G2

Cross-section [A ; Iz]
u1y u2y

FIG 18 – BEAM ELEMENT DEGREES OF FREEDOM

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 43

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

M1 M2
z x
G1 G2

f1y f2y

FIG 19 – BEAM ELEMENT POSITIVE FORCES AND MOMENTS

The following figure shows the sign convention used in simple beam theory for positive shear forces
V and bending moments M:

M M

V V
L

FIG 20 – BEAM THEORY SIGN CONVENTION FOR SHEAR FORCES AND BENDING MOMENTS

In the Euler-Bernoulli beam theory, the differential equation governing the elementary linear-elastic
beam behavior is based on the following assumption:

The planar cross-sections perpendicular to the longitudinal axis of the beam before
deformation remain plane and perpendicular after deformation.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 44

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

y
Undeformed beam Longitudinal axis
w(x)
–
A C –
x
G1 – G2
B D

Longitudinal axis
G1 Deformed beam due to w(x) –
G2
–
A’ C’
–
B’ D’

FIG 21 – BEAM DEFORMATION ACCORDING TO THE EULER-BERNOULLI THEORY

The Euler-Bernoulli assumption states that both planes A-B and A’-B’ are perpendicular to the
longitudinal axis. This assumption is valid only if a pure couple or a constant moment occurs in the
beam.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 45

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Equilibrium

w(x)
M V M+dM

1 2
dx
V+dV

FIG 22 – EQUILIBRIUM OF THE DIFFERENTIAL BEAM ELEMENT

The equilibrium of a differential element of the beam is:

• Force equilibrium: Fy = 0

V – (V+dv) – w(x)dx = 0

-w(x)dx – dV = 0 Eq 106
dV
W(X) = -
dX

• Moment equilibrium: Mz = 0

𝑑𝑥
−𝑀 − 𝑉𝑑𝑥 + 𝑤(𝑥)𝑑𝑥 ( ) + 𝑀 + 𝑑𝑀 = 0
2

𝑑𝑥 Eq 107
−𝑉𝑑𝑥 + 𝑤(𝑥)𝑑𝑥 ( ) + 𝑑𝑀 = 0
2

𝑑𝑥 𝑑𝑀
−𝑉 + 𝑤(𝑥) ( ) + = 0
2 𝑑𝑥

The limit of the above equation as dx tends to 0 gives:

𝑑𝑀 Eq 108
𝑉 =
𝑑𝑥

From Eq 106 & Eq 108:

𝑑2𝑀 Eq 109
𝑤(𝑥) = −
𝑑𝑥 2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 46

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

y

v(x)
x
x

FIG 23 – BEAM CURVATURE

The curvature of the beam is related to the moment by:

1 𝑀
= Eq 110
𝜌 𝐸𝐼
For small slopes, the curvature is given by the second derivative of the transverse displacement v(x):

1 𝑑2𝑣
= Eq 111
𝜌 𝑑𝑥 2
So,

𝑑2𝑣 𝑀 Eq 112
=
𝑑𝑥 2 𝐸𝐼
Then

𝑑2𝑣 Eq 113
𝑀 = 𝐸𝐼
𝑑𝑥 2
Substituting Eq 113 into Eq 109:
𝑑2 𝑑2𝑣
𝑤(𝑥) = − (𝐸𝐼 ) Eq 114
𝑑𝑥 2 𝑑𝑥 2

Only for nodal forces and moments, the above equation becomes (EI=constant):
𝑑4𝑣
𝐸𝐼 4 = 0 Eq 115
𝑑𝑥

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 47

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Shape Functions

The beam element under bending deformation has four degrees of freedom: a transverse
displacement and a rotation at each node. So, the transverse displacement variation through the
element length can be described by a cubic function:
v(x) = a1x3 + a2x2 + a3x + a4 Eq 116

This function can be expressed as a function of the nodal degrees of freedom u1y, u2y, 1 and 2:

v(0) = u1y = a4 v(L) = u2y = a1L3 + a2L2 + a3L + a4

Eq 117
𝑑𝑣(0) 𝑑𝑣(𝐿)
= 𝜃1 = 𝑎3 = 𝜃2 = 3𝑎1 𝐿2 + 2𝑎2 𝐿 + 𝑎3
𝑑𝑥 𝑑𝑥

So, let’s resolve the above system:

a4 = u1y

a3 = 1
Eq 118
u2y = a1L3 + a2L2 + a3L + u1y

2 = 3a1L2 + 2a2L + a3 = 3a1L2 + 2a2L + 1

So,

3a1L2 = 2 - 1 - 2a2L
Eq 119
𝜃2 − 𝜃1 − 2𝑎2 𝐿
𝑎1 =
3𝐿2

Substituting the above expression of a1 in the equation of u2y, we obtain:

3 1 Eq 120
𝑎2 = 2
(𝑢2𝑦 − 𝑢1𝑦 ) − (2𝜃1 + 𝜃2 )
𝐿 𝐿

Substituting the above expression of a2 in the equation of a1, we obtain:

2 1 Eq 121
𝑎1 = − (𝑢2𝑦 − 𝑢1𝑦 ) + (𝜃 + 𝜃2 )
𝐿3 𝐿2 1

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 48

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Then substituting the expressions of a1, a2. a3 and a4 in Eq 116, we obtain:

2 1
𝑣(𝑥) = [ 3
(𝑢1𝑦 − 𝑢2𝑦 ) + 2
(𝜃1 + 𝜃2 )] 𝑥 3
𝐿 𝐿 Eq 122
3 1
+ [− 2 (𝑢1𝑦 − 𝑢2𝑦 ) − (2𝜃1 + 𝜃2 )] 𝑥 2 + 𝜃1 𝑥 + 𝑢1𝑦
𝐿 𝐿

The above equation in matrix form can be written:

v = [N]{u} Eq 123
where,
𝑢1𝑦
𝜃
{𝑢} = {𝑢 1 } Eq 124
2𝑦
𝜃2

And,
[N] = [N1 N2 N3 N4]
With the shape functions Ni for a beam element:
1
𝑁1 = (2𝑥 3 − 3𝑥 2 𝐿 + 𝐿3 )
𝐿3
1
𝑁2 = 3 (𝑥 3 𝐿 − 2𝑥 2 𝐿2 + 𝑥𝐿3 )
𝐿
1 Eq 125
𝑁3 = 3 (−2𝑥 3 + 3𝑥 2 𝐿)
𝐿
1
𝑁4 = 3 (𝑥 3 𝐿 − 𝑥 2 𝐿2 )
𝐿

When evaluated at node 1, N1 = 1 (with x=0) and when evaluated at node 2, N1 = 0 (with x=L). On the
other hand, since N2 is associated to 1 in the equation v=[N]{U}, we have:
𝑑𝑁2 1
= 3 (3𝑥𝐿 − 4𝑥𝐿2 + 𝐿3 ) Eq 126
𝑑𝑥 𝐿

When evaluated at node 1, dN2/dx = 1 (with x=0). The shape functions N3 and N4 have the same
behavior.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 49

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Strain-Displacement & Stress-Strain Relationships

A’ A

After deformation Before deformation

-y dv
θ=
dx

B u B’

FIG 24 – ANGLE OF ROTATION OF SECTION A-B

The axial strain/displacement relationship is:

𝑑𝑢 Eq 127
𝜀(𝑥, 𝑦) =
𝑑𝑥
Where u is an axial displacement function. On the other hand, the transverse displacement relates
to the axial displacement by:

𝑑𝑣 Eq 128
𝑢 = −𝑦
𝑑𝑥
So, combining Eq 127 & Eq 128, we obtain:

𝑑2𝑣 Eq 129
𝜀(𝑥, 𝑦) = −𝑦 2
𝑑𝑥

Finally, the elementary beam theory teaches us that the bending moment and the shear force are
related to the transverse function by the equations:

𝑑2𝑣 Eq 130
𝑀(𝑥) = 𝐸𝐼
𝑑𝑥 2

𝑑3𝑣 Eq 131
𝑉(𝑥) = 𝐸𝐼 3
𝑑𝑥

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 50

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element Stiffness Matrix

Since the Euler-Bernoulli beam element exhibits two degrees of freedom at each node, its stiffness
is described with a 4x4 matrix.

Combining Eq 122, Eq 130 & Eq 131 and using the sign conventions from Fig 19 & Fig 20, we have:
𝑑3 𝑣(0) 𝐸𝐼
𝑓1𝑦 = 𝑉(0) = 𝐸𝐼 = 3 (12𝑢1𝑦 + 6𝐿𝜃1 − 12𝑢2𝑦 + 6𝐿𝜃2 )
𝑑𝑥 3 𝐿
𝑑 2 𝑣(0) 𝐸𝐼
𝑀1 = −𝑀(0) = −𝐸𝐼 = 3 (6𝐿𝑢1𝑦 + 4𝐿2 𝜃1 − 6𝐿𝑢2𝑦 + 2𝐿2 𝜃2 )
𝑑𝑥 2 𝐿
Eq 132
𝑑 3 𝑣(𝐿) 𝐸𝐼
𝑓2𝑦 = −𝑉(𝐿) = −𝐸𝐼 = 3 (−12𝑢1𝑦 − 6𝐿𝜃1 + 12𝑢2𝑦 − 6𝐿𝜃2 )
𝑑𝑥 3 𝐿
𝑑 2 𝑣(𝐿) 𝐸𝐼
𝑀2 = 𝑀(𝐿) = 𝐸𝐼 = 3 (6𝐿𝑢1𝑦 + 2𝐿2 𝜃1 − 6𝐿𝑢2𝑦 + 4𝐿2 𝜃2 )
𝑑𝑥 2 𝐿

So, in matrix form, the above system becomes:

𝑓1𝑦 12 6𝐿 −12 6𝐿 𝑢1𝑦

𝑀1 𝐸𝐼 6𝐿 4𝐿2 −6𝐿 2𝐿2 𝜃1
= 3[ ] {𝑢 } Eq 133
𝑓2𝑦 𝐿 −12 −6𝐿 12 −6𝐿 2𝑦
{ 𝑀2 } 6𝐿 2𝐿2 −6𝐿 4𝐿2 𝜃2

Where the stiffness matrix is:

12 6𝐿 −12 6𝐿
𝐸𝐼 6𝐿 4𝐿2 −6𝐿 2𝐿2
[𝐾]𝐵𝑒𝑎𝑚 = 3[ ] Eq 134
𝐿 −12 −6𝐿 12 −6𝐿
6𝐿 2𝐿2 −6𝐿 4𝐿2

The stiffness matrix [K]Beam relates the transverse forces and the bending moments to transverse
displacements and rotations, neglecting the axial effects. This result has been obtained assuming
that the beam is long and slender which means that the dimension ratio L/h (h is the depth of the
beam) is large. Using the Eq 133 to predict the displacements is adequate only for these assumptions.
For a short deep beam, the transverse shear deformation is usually more significant. In this case, the
Timoshenko beam theory has to be considered.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 51

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.4.2. Timoshenko Beam Theory

Equilibrium
The Timoshenko beam theory states that the plane sections do not remain perpendicular to the
neutral axis of the beam after bending. Fig 25 shows a section of the differential beam dx after
deformation with the cross-section no longer perpendicular to the neutral axis due to the shear force
inducing an additional rotation term .

(x) M + M/x (x)

Neutral axis of the beam
after deformation
V+V/x

Timoshenko Beam dx

V
M
v(x)

Euler-Bernoulli Beam

FIG 25 – TIMOSHENKO BEAM DEFORMATION

So, the total deflection of the beam at point x consists of two parts: one due to bending and one due
to shear force. The slope of the deflected beam at point x is:

𝑑𝑣 Eq 135
= 𝜃(𝑥) + 𝛼(𝑥)
𝑑𝑥

Where (x) is the rotation due to bending moment and (x) is the rotation due to the transverse
shear. Note that all the deflections are assumed to be small.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 52

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Strain-Displacement & Stress-Strain Relationships

The relation between the bending moment and the bending curvature is:

𝑑𝜃(𝑥) Eq 136
𝑀(𝑥) = 𝐸𝐼
𝑑𝑥
The relation between the shear force and the rotation due to shear is:

𝑉(𝑥) = 𝑘𝑠 𝐴𝐺𝛼(𝑥) Eq 137

Where ksA is the shear area. The shear area varies with the cross-section shapes. Refer to a
mechanical handbook for the various values of ks. Note that most of the FEA pre-processors are able
to compute ks based on the geometrical properties of the modeled cross-sections.
The shear strain yz of the beam is represented by the difference between dv/dx and :

𝑑𝑣 Eq 138
𝛾𝑦𝑧 = –𝜃 = 𝛼
𝑑𝑥

Shape Functions
The transverse shear displacement of the beam is still described by the cubic shape function Eq 116
previously defined for the Euler-Bernoulli beam:

v(x) = a1x3 + a2x2 + a3x + a4 Eq 139

The transverse shear strain  is chosen consistent with the cubic polynomial for v(x), such that  is a
constant: =c. From Eq 106, Eq 136, Eq 137 & Eq 138:
𝑑𝑀 𝑑 𝑑𝜃(𝑥)
𝑉 = = (𝐸𝐼 ) = 𝑘𝑠 𝐴𝐺𝛼(𝑥) = 𝑘𝑠 𝐴𝐺𝛾𝑦𝑧 = 𝑘𝑠 𝐴𝐺𝑐 Eq 140
𝑑𝑥 𝑑𝑥 𝑑𝑥
So,

𝐸𝐼 𝑑 2 𝜃(𝑥)
𝑐 = Eq 141
𝑘𝑠 𝐴𝐺 𝑑𝑥 2
From Eq 135 & Eq 116:
𝑑𝑣 𝑑𝑣
𝜃(𝑥) = − 𝛾 = –𝑐
𝑑𝑥 𝑑𝑥
(x) = 3a1x2 + 2a2x + a3 – c
𝑑𝜃(𝑥)
= 6𝑎1 𝑥 + 2𝑎2
𝑑𝑥

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 53

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝑑 2 𝜃(𝑥) Eq 142
= 6𝑎1
𝑑𝑥 2
By combining Eq 141 & Eq 142:

6𝐸𝐼𝑎1
𝑐 = Eq 143
𝑘𝑠 𝐴𝐺
Then, (x) is expressed as a polynomial in x as follow:
(x) = 3a1x2 + 2a2x + a3 – c
6𝐸𝐼 Eq 144
𝜃(𝑥) = 3𝑎1 𝑥 2 + 2𝑎2 𝑥 + 𝑎3 − 𝑎
𝑘𝑠 𝐴𝐺 1
With the transverse shear term  defined as:
𝐸𝐼
𝛽= Eq 145
𝑘𝑠 𝐴𝐺
The (x) equation is written:

(x) = (3x2 - 6) a1 + 2a2x + a3 Eq 146

From Eq 146 and the cubic shape function Eq 116, it is possible to compute the coefficients a1 to a4
in terms of nodal displacements u1y, u2y and rotations 1,  2 using the same approach as the one
used for the Euler-Bernoulli assumption. The equations of the ai coefficients are:
2(𝑢1𝑦 − 𝑢2𝑦 ) + 𝐿(𝜃1 + 𝜃2 )
𝑎1 =
𝐿3 + 12𝛽𝐿
3𝐿(𝑢2𝑦 − 𝑢1𝑦 ) − (2𝐿2 + 6𝛽)𝜃1 + ( 6𝛽 − 𝐿2 )𝜃2
𝑎2 =
𝐿3 + 12𝛽𝐿 Eq 147
12𝛽(𝑢2𝑦 − 𝑢1𝑦 ) + (𝐿3
+ 6𝐿𝛽)𝜃1 − 6𝛽𝜃2
𝑎3 =
𝐿3 + 12𝛽𝐿
a4 = u1y

Substituting the ai coefficients into the beam shape function Eq 116:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 54

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

2(𝑢1𝑦 − 𝑢2𝑦 ) + 𝐿(𝜃1 + 𝜃2 ) 3

𝑣(𝑥) = 𝑥
𝐿3 + 12𝛽𝐿
3𝐿(𝑢2𝑦 − 𝑢1𝑦 ) − (2𝐿2 + 6𝛽)𝜃1 + ( 6𝛽 − 𝐿2 )𝜃2 2
+ 𝑥 Eq 148
𝐿3 + 12𝛽𝐿
12𝛽(𝑢2𝑦 − 𝑢1𝑦 ) + (𝐿3 + 6𝐿𝛽)𝜃1 − 6𝛽𝜃2
+ 𝑥 + 𝑢1𝑦
𝐿3 + 12𝛽𝐿

Element Stiffness Matrix

Similarly to the Euler-Bernoulli beam stiffness matrix derivation, we have:
𝑓1𝑦 = 𝑉(0)
𝑀1 = −𝑀(0)
Eq 149
𝑓2𝑦 = −𝑉(𝐿)
𝑀2 = 𝑀(𝐿)
From Eq 131:
𝑑3𝑣
𝑉(𝑥) = 𝐸𝐼 3 Eq 150
𝑑𝑥
From Eq 116:

v(x) = a1x3 + a2x2 + a3x + a4 Eq 151

So,
𝑑3𝑣
= 6𝑎1 Eq 152
𝑑𝑥 3
Then,

V = 6EIa1 = V(0) = V(L) Eq 153

From Eq 130:
𝑑2𝑣
𝑀(𝑥) = 𝐸𝐼 2 Eq 154
𝑑𝑥
The second derivative of the shape function Eq 116 is:
𝑑2𝑣
= 6𝑎1 𝑥 + 2𝑎2 Eq 155
𝑑𝑥 2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 55

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So,
M(0) = 2a2EI
Eq 156
M(L) = EI(6a1L+2a2)

By substituting Eq 153 & Eq 156 into Eq 149:

𝑓1𝑦 = 𝑉(0) = 6EIa1
𝑓2𝑦 = −𝑉(𝐿) = -6EIa1
Eq 157
𝑀1 = −𝑀(0) = -2a2EI
𝑀2 = 𝑀(𝐿) = EI(6a1L+2a2)

Which gives:
12(𝑢1𝑦 − 𝑢2𝑦 ) + 6𝐿(𝜃1 + 𝜃2 )
𝑓1𝑦 = 𝐸𝐼
𝐿3 + 12𝛽𝐿
12(𝑢2𝑦 − 𝑢1𝑦 ) − 6𝐿(𝜃1 + 𝜃2 )
𝑓2𝑦 = 𝐸𝐼
𝐿3 + 12𝛽𝐿
Eq 158
6𝐿(𝑢1𝑦 − 𝑢2𝑦 ) + (4𝐿2 + 12𝛽)𝜃1 + (2𝐿2 − 12𝛽)𝜃2
𝑀1 = 𝐸𝐼
𝐿3 + 12𝛽𝐿
6𝐿(𝑢1𝑦 − 𝑢2𝑦 ) + (2𝐿2 − 12𝛽)𝜃1 + (4𝐿2 + 12𝛽)𝜃2
𝑀2 = 𝐸𝐼
𝐿3 + 12𝛽𝐿

In matrix form, the Eq 158 is written:

𝑓1𝑦
𝑀1
𝑓2𝑦
{𝑀2 }
12 6𝐿 −12 6𝐿 𝑢1𝑦 Eq 159
𝐸𝐼 6𝐿 4𝐿2 + 12𝛽 −6𝐿 2
2𝐿 − 12𝛽 𝜃1
= 3 [ ] {𝑢 }
𝐿 + 12𝛽𝐿 −12 −6𝐿 12 −6𝐿 2𝑦
6𝐿 2𝐿2 − 12𝛽 −6𝐿 4𝐿2 + 12𝛽 𝜃2

Since  represents the transverse shear term, we can easily see that if we neglect the transverse
shear deformation by setting =0, we obtain the matrix of the Euler-Bernoulli beam from Eq 134.

It is usual to define a non-dimensional shear correction term  to see the shear effect more easily:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 56

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

12𝐸𝐼 12𝛽
𝜑= = 2 Eq 160
𝑘𝑠 𝐴𝐺𝐿2 𝐿

So, the beam stiffness matrix for the Timoshenko beam is written as:
12 6𝐿 −12 6𝐿
𝐸𝐼 6𝐿 (4 + 𝜑)𝐿2 −6𝐿 (2 − 𝜑)𝐿2
[𝐾]𝐵𝑒𝑎𝑚 = 3 [ ] Eq 161
𝐿 (1 − 𝜑) −12 −6𝐿 12 −6𝐿
6𝐿 (2 − 𝜑)𝐿2 −6𝐿 (4 + 𝜑)𝐿2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 57

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.4.3. Bending and Torsion

Degrees of Freedom
We will now consider the development of the beam element stiffness matrix under bending and
torsion.
The degrees of freedom at each node of the beam are:
• Vertical deflection uiy
• Bending rotations iz about the z-axis
• Torsional rotations ix about the x-axis
So, let’s consider a simple beam element of:
• A uniform cross-section A
• A principal moment of inertia Iz
• A torsional constant J
• A length L
• Made of a homogeneous material: Young modulus E and shear modulus G

y Homogeneous material [E ; G]

1z 2x 2z

G1 x
G2
1x
z Cross-section [A ; Iz ; J]
u1y u2y

FIG 26 – BEAM ELEMENT DEGREES OF FREEDOM UNDER BENDING AND TORSION

M1z M2x M2z

G1 x
G2
M1x
z
f1y f2y

FIG 27 – BEAM ELEMENT NODAL FORCES AND MOMENTS UNDER BENDING AND TORSION

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 58

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element Stiffness Matrix

The torsional stiffness matrix of the beam element is the same as the truss element in torsion derived
previously (see paragraph Truss Element in Torsion page 35):
𝑀1𝑥 𝐺𝐽 1 −1 𝜃1𝑥
{ }= [ ]{ } Eq 162
𝑀2𝑥 𝐿 −1 1 𝜃2𝑥

𝐺𝐽/𝐿 −𝐺𝐽/𝐿 Eq 163

[𝐾]𝐵𝑒𝑎𝑚 𝑇𝑜𝑟𝑠𝑖𝑜𝑛 = [ ]
−𝐺𝐽/𝐿 𝐺𝐽/𝐿

The torsional constant J depends on the cross-section shape. Refer to a mechanical book for the
computation of J. Again, the modern FEA pre-processors are able to calculate J for every type of
cross-section shape.
By combining the torsional effect of Eq 163 with the bending effect from the Euler Bernoulli beam in
bending described by Eq 134, we obtain the stiffness matrix of the beam element in bending and
torsion. With three degrees of freedom per node, the stiffness matrix is a 6x6:
𝑓1𝑦
𝑀1𝑥
𝑀1𝑧
=
𝑓2𝑦
𝑀2𝑥
{𝑀2𝑧 }
12𝐸𝐼 6𝐸𝐼 12𝐸𝐼 6𝐸𝐼
0 − 0
𝐿3 𝐿2 𝐿3 𝐿2
𝐺𝐽 𝐺𝐽
0 0 0 − 0 𝑢1𝑦 Eq 164
𝐿 𝐿
6𝐸𝐼 4𝐸𝐼 6𝐸𝐼 2𝐸𝐼 𝜃1𝑥
0 − 0 𝜃1𝑧
= 𝐿2 𝐿 𝐿2 𝐿
12𝐸𝐼 6𝐸𝐼 12𝐸𝐼 6𝐸𝐼 𝑢2𝑦
− 3 0 − 2 0 − 2 𝜃2𝑥
𝐿 𝐿 𝐿3 𝐿
𝐺𝐽 𝐺𝐽 { 𝜃2𝑧 }
0 − 0 0 0
𝐿 𝐿
6𝐸𝐼 2𝐸𝐼 6𝐸𝐼 4𝐸𝐼
[ 0 − 0
𝐿2 𝐿 𝐿2 𝐿 ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 59

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.4.4. Full Capabilities Beam

Degrees of Freedom
Now we can consider the full capabilities beam element stiffness matrix with six degrees of freedom:
• Axial displacements: uix
• Transverse displacements: uiy & uiz
• Bending rotations: iy & iz
• Torsional rotations: ix

y
Homogeneous material [E ; G]
1y 2y
u1z u2z
1z 2x
u1x x
u2x
1x 2z
z Cross-section [A ; Iz ; J]
u1y u2y

FIG 28 – FULL CAPABILITIES BEAM ELEMENT DEGREES OF FREEDOM

M1y M2y
f1z f2z
M1z M2x
f1x x
f2x
M1x M2z
z
f1y f2y

FIG 29 – FULL CAPABILITIES BEAM ELEMENT NODAL FORCES AND MOMENTS

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 60

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element Stiffness Matrix

The axial stiffness is given by Eq 57:
𝑓1𝑥 𝐸𝐴/𝐿 −𝐸𝐴/𝐿 𝑢1𝑥
{ }=[ ]{ } Eq 165
𝑓2𝑥 −𝐸𝐴/𝐿 𝐸𝐴/𝐿 𝑢2𝑥

The stiffness in bending and shear in plane <x,y> is given by Eq 133:

𝑓1𝑦 12 6𝐿 −12 6𝐿 𝑢1𝑦
𝑀1𝑧 𝐸𝐼𝑧 6𝐿 4𝐿2 −6𝐿 2𝐿2 𝜃1𝑧
= 3 [ ] {𝑢 } Eq 166
𝑓2𝑦 𝐿 −12 −6𝐿 12 −6𝐿 2𝑦
{𝑀2𝑧 } 6𝐿 2𝐿2 −6𝐿 4𝐿2 𝜃2𝑧

The stiffness in bending and shear in plane <x,z> is given by Eq 133:

𝑓1𝑧 12 6𝐿 −12 6𝐿 𝑢1𝑧
𝑀1𝑦 𝐸𝐼𝑦 6𝐿 4𝐿2 −6𝐿 2𝐿2 𝜃1𝑦
= 3 [ ] {𝑢 } Eq 167
𝑓2𝑧 𝐿 −12 −6𝐿 12 −6𝐿 2𝑧
𝑀
{ 2𝑦 } 6𝐿 2𝐿2 −6𝐿 4𝐿2 𝜃2𝑦

The torsional stiffness is given by Eq 162:

𝐺𝐽 𝐺𝐽
𝑀 −
{ 1𝑥 } = [ 𝐿 𝐿 ] {𝜃1𝑥 }
𝑀2𝑥 𝐺𝐽 𝐺𝐽 𝜃2𝑥 Eq 168
−
𝐿 𝐿

By direct superposition of Eq 165, Eq 166, Eq 167 & Eq 168, the 12 x 12 beam element stiffness matrix
with full capabilities is:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 61

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

u1x u1y u1z 1x 1y 1z u2x u2y u2z 2x 2y 2z

EA EA
0 0 0 0 0 - 0 0 0 0 0
L L
12EIz 6EIz 12EIz 6EIz
0 0 0 0 0 - 0 0 0
L3 L2 L3 L2
12EIy 6EIy 12EIy 6EIy
0 0 3 0 - 2 0 0 0 - 3 0 - 0
L L L L2
GJ GJ
0 0 0 0 0 0 0 0 0 0
L L
6EIy 4EIy 6EIy 2EIy
0 0 - 0 0 0 0 0 0
L2 L L2 L
6EIz 4EIz 6EIz 2EIz
0 0 0 0 0 - 0 0 0
[K]Beam = L2 L L2 L
EA EA
- 0 0 0 0 0 0 0 0 0 0
L L
12EIz 6EIz 12EIz 6EIz
0 - 3 0 0 0 - 2 0 3 0 0 0 -
L L L L2
12EIy 6EIy 12EIy 6EIy
0 0 - 3 0 2 0 0 0 3 0 0
L L L L2
GJ GJ
0 0 0 - 0 0 0 0 0 0 0
L L
6EIy 2EIy 6EIy 4EIy
0 0 - 2 0 0 0 0 2 0 0
L L L L
6EIz 2EIz 6EIz 4EIz
0 0 0 0 0 - 0 0 0
[ L2 L L2 L ]

Eq 169

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 1D BEAM ELEMENT 62

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5. 2D ELEMENTS
5.5.1. Overview
The 2D elements are defined by at least three nodes up to height nodes. The elements are connected
at common nodes and along common edges to form continuous structures such as those shown in
Fig 30.

FIG 30 – 2D CONTINUOUS STRUCTURES

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 63

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The derivation of the two-dimensional element stiffness matrix will be presented across the
following steps:
1. Membrane: Constant Strain Triangle (CST) element. This 2D element is the simplest among
the available 2D elements. It is called CST because it has a constant strain throughout it. The
stiffness matrix will be derived using the minimum potential energy.
2. Membrane: Linear Strain Triangle (LST) element. This second step will illustrate the
advantages of high-order elements.
3. Plate bending element. This third step will present the derivation of stiffness matrix for the
common plate bending finite element. Two types of plates will be considered: the thin-plates
described by the Kirchhoff theory and the thick-plates described by the Mindlin theory.
4. Shell element. Finally, the stiffness matrix of the shell element will be derived. It is a
rectangular element used for dealing with the combination of membrane and bending
effects.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 64

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5.2. Membrane: The Constant Strain Triangle (CST)

First, let’s refresh your memory of the following two basic concepts:
• The concepts of plane stress and plane strain
• The concept of a two-dimensional state of stress and strain

The understanding of these concepts is very important for the development of the stiffness matrix
of the 2D elements. Note that only the equations are recalled for these two concepts. For the details
about their derivations, refer to a book about the elasticity theory. For the derivation of the 2D
element stiffness matrix, the same pattern as the one used for the 1D elements will be followed.
The Concepts of Plane Stress and Plane Strain
Plane stress: it is a state of stress in which the normal stress and the shear stress perpendicular to
the plane are null. In the Fig 31, the plate in the plane <x,y> subjected to the tension load P at the
edge of the surface and in the plane is under a state of plane stress. The normal stress Z and the
shear stress XZ and YZ are null. A thin member, with a small z dimension compared to the in-plane
x and y dimensions, and a member on which loads act only in the <x,y> plane can be considered to
be under plane stress.
y

t
x

FIG 31 – PLANE STRESS

Plane strain: it is a state of strains in which the strain Z normal to the <x,y> plane and the shear
strains XZ and YZ are null. The plane strain assumption is used for long bodies, in the z direction, with
constant cross-section, subjected to loads acting only in the plane <x,y>. The dam is a good example
of plane strain state.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 65

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

z
x
FIG 32 – PLANE STRAIN

The Concept of Two-Dimensional State of Stress and Strain

Let’s consider the two-dimensional state of stress illustrated in Fig 33 on an infinitesimal element
<dx,dy>.

Y

YX

XY
y

X X
dy
x

XY dx

YX

Y

FIG 33 – TWO-DIMENSIONAL STATE OF STRESS

• dx has normal stress y acting in the x-direction
• dy has normal stress x acting in the y-direction
• The shear stress XY acts on the x edge in the y-direction
• The shear stress YX acts on the y edge in the x-direction
• The equilibrium of the element results in XY= YX

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 66

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

This state of stress is represented by the column matrix:

𝜎𝑋
{𝜎} = { 𝜎𝑌 } Eq 170
𝜏𝑋𝑌
The displacements and rotations of the infinitesimal element <dx,dy> in the <x,y> plane is shown in
Fig 34. The properties of the general two-dimensional state of strain at some point of a structure are:
• At point G, the element is displaced by an amount u in the x direction and by an amount v in
the y direction.
• Along the line GH, the element extends an additional amount (u/x)dx in the x direction
• Along the line GJ, the element extends an additional amount (v/y)dy in the y direction
• Point H moves upward by (v/x)dx with respect to G
• Point J moves to the right by (u/y)dy with respect to G
So, the normal and shear strain-displacement relationships are defined as follow.
X: Normal strain along the x-axis. It is the change in length per unit length of material fibers originally
parallel to the x-axis when the element undergoes deformation:
𝜕𝑢
𝜀𝑥 = Eq 171
𝜕𝑥

Y: Normal strain along the y-axis. It is the change in length per unit length of material fibers originally
parallel to the y-axis when the element undergoes deformation:
𝜕𝑣
𝜀𝑦 = Eq 172
𝜕𝑦

XY: Shear strain. It is the change of the original right angle made between dx and dy when the
element undergoes deformation:
𝜕𝑢 𝜕𝑣
𝛾𝑥𝑦 = + Eq 173
𝜕𝑦 𝜕𝑥

The strains given by the three above equations are represented by the vector column matrix:
𝜀𝑋
{𝜖} = { 𝜀𝑌 } Eq 174
𝛾𝑋𝑌

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 67

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

∂u
dy
∂y
I
∂v
v+ dy J
∂y
∂u
y,v
∂y

x,u H ∂v
dy dx
∂x
G
v
∂v
∂x
u ∂u
u+ dx
dx ∂x

FIG 34 – DISPLACEMENTS AND ROTATIONS OF AN INFINITESIMAL ELEMENT IN THE <X,Y> PLANE

We know from the theory of elasticity that the stress-strain relationship for an isotropic body is:

{} = [D]{} Eq 175

Where:
• {} and {} are given by Eq 170 and Eq 174 respectively.
• [D] is the constitutive matrix.
The constitutive matrix [D] for plane stress and plane strain conditions is given by (E is the Young
modulus and  is the Poisson’s ratio):
• Plane Stress: Z = XZ = YZ = 0
1  0
𝐸  1 0
[𝐷] = [ 1 − ] Eq 176
1−2
0 0
2

• Plane Strain: Z = XZ = YZ = 0

1−  0
𝐸  1− 0
[𝐷] = [ 1 − 2] Eq 177
(1 + )(1 − 2)
0 0
2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 68

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Degrees of Freedom
The basic definition of the membrane triangular element is a planar element having three nodes.
Each node has two degrees of freedom representing the x and y translations of the node. Therefore,
the triangular element has 3 nodes times 2 translations per node equal 6 degrees of freedom. The
nodal displacements will be represented by u in x-direction and v in the y-direction.
• (xp ; yp) are the known nodal coordinates of node p in the plane <x,y>
• (xq ; yq) are the known nodal coordinates of node q in the plane <x,y>
• (xr ; yr) are the known nodal coordinates of node r in the plane <x,y>
• (up ; vp) are the unknown degrees of freedom of node p in the plane <x,y>
• (uq ; vq) are the unknown degrees of freedom of node q in the plane <x,y>
• (ur ; vr) are the unknown degrees of freedom of node r in the plane <x,y>
vr

y,v
Node r (xr ; yr) ur
vp

vq
up
Node p (xp ; yp)
uq
Node q (xq ; yq)
x,
u
FIG 35 – NODES AND DEGREES OF FREEDOM OF THE CONSTANT STRAIN TRIANGULAR ELEMENT

The nodal displacement vector is written:

up
vp
δp
u
{δ}= {δq } = v q Eq 178
q
δr ur
{ vr }

Shape Functions
Since the CST element has six nodal unknowns, six constants ai are needed to describe the complete
displacement field. So, the linear displacement functions describing the displacements at any interior
point (xi ; yi) of the element are:
u(x,y) = a1 + a2x + a3y
Eq 179
v(x,y) = a4 + a5x + a6y

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 69

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

These linear functions guarantee the compatibility of displacements along the edges of the elements
and at the nodes shared by adjacent elements.
A general displacement function  is then defined from Eq 179:
a1
a2
u(x,y) 1 x y 0 0 0 a3
{Γ }= { }=[ ] Eq 180
v(x,y) 0 0 0 1 x y a4
a5
{a6 }
To solve for the ai’s, the known coordinates of the nodal points are substituted into Eq 179:
up = u(xp ; yp) = a1 + a2xp + a3yp
uq = u(xq ; yq) = a1 + a2xq + a3yq
ur = u(xr ; yr) = a1 + a2xr + a3yr
Eq 181
vp = v(xp ; yp) = a4 + a5xp + a6yp
vq = v(xq ; yq) = a4 + a5xq + a6yq
vr = v(xr ; yr) = a4 + a5xr + a6yr

Let’s solve for the ai’s beginning with the three first equations of Eq 181 in matrix form:
𝑢𝑝 1 𝑥𝑝 𝑦𝑝 𝑎1
{𝑢𝑞 } = [1 𝑥𝑞 𝑦𝑞 ] {𝑎2 } Eq 182
𝑢𝑟 1 𝑥𝑟 𝑦𝑟 𝑎3

{u} = [X]{a} Eq 183

So, we can solve for ai’s by:
{a} = [X]-1{u} Eq 184
The inverse of [X] is calculated by the method of cofactors:
𝛼𝑝 𝛼𝑞 𝛼𝑟
−1
1
[𝑋] = [ 𝛽𝑝 𝛽𝑞 𝛽𝑟 ] Eq 185
|𝑋| 𝛾 𝛾𝑞 𝛾𝑟
𝑝

Where X is the determinant of [X]:

X = xp(yq – yr) + xq(yr – yp) + xr(yp – yq) Eq 186
and

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 70

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

p = xqyr - yqxr q = ypxr - xpyr r = xpyq - ypxq

p = yq - yr q = yr - yp r = yp - yq Eq 187
p = xr - xq q = xp - xr r = xq - xp

So, Eq 184 is written:

𝑎1 𝛼𝑝 𝛼𝑞 𝛼𝑟 𝑢𝑝
1
{𝑎2 } = [ 𝛽𝑝 𝛽𝑞 𝛽𝑟 ] {𝑢𝑞 } Eq 188
𝑎3 |𝑋| 𝛾𝑝 𝛾𝑞 𝛾𝑟 𝑢𝑟

Similarly, the last three ai’s of Eq 181 are solved:

𝑎4 𝛼𝑝 𝛼𝑞 𝛼𝑟 𝑣𝑝
1
{ 𝑎5 } = [ 𝛽𝑝 𝛽𝑞 𝛽𝑟 ] {𝑣𝑞 } Eq 189
𝑎6 |𝑋| 𝛾 𝛾𝑞 𝛾𝑟 𝑣𝑟
𝑝

Expressing Eq 179 in matrix form:

𝑎1
{𝑢} = [1 𝑥 𝑎
𝑦] { 2 } Eq 190
𝑎3

And substituting Eq 188 into Eq 190 and expanding:

𝛼𝑝 𝛼𝑞 𝛼𝑟 𝑢𝑝
1
{𝑢} = [1 𝑥 𝑦] [ 𝛽𝑝 𝛽𝑞 𝛽𝑟 ] {𝑢𝑞 } Eq 191
|𝑋| 𝛾 𝛾𝑞 𝛾𝑟 𝑢𝑟
𝑝

𝛼𝑝 𝑢𝑝 + 𝛼𝑞 𝑢𝑞 + 𝛼𝑟 𝑢𝑟
1
{𝑢} = [1 𝑥 𝑦] { 𝛽𝑝 𝑢𝑝 + 𝛽𝑞 𝑢𝑞 + 𝛽𝑟 𝑢𝑟 } Eq 192
|𝑋|
𝛾𝑝 𝑢𝑝 + 𝛾𝑞 𝑢𝑞 + 𝛾𝑟 𝑢𝑟

Eq 192 gives the u(x,y) displacement function:

1
𝑢(𝑥, 𝑦) = {(𝛼𝑝 + 𝛽𝑝 𝑥 + 𝛾𝑝 𝑦)𝑢𝑝 + (𝛼𝑞 + 𝛽𝑞 𝑥 + 𝛾𝑞 𝑦)𝑢𝑞
|𝑋| Eq 193
+ (𝛼𝑟 + 𝛽𝑟 𝑥 + 𝛾𝑟 𝑦)𝑢𝑟 }

Similarly, replacing the u degree of freedom by the v degree of freedom in Eq 193, we obtain the
v(x,y) displacement function:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 71

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

1
𝑣(𝑥, 𝑦) = {(𝛼𝑝 + 𝛽𝑝 𝑥 + 𝛾𝑝 𝑦)𝑣𝑝 + (𝛼𝑞 + 𝛽𝑞 𝑥 + 𝛾𝑞 𝑦)𝑣𝑞
|𝑋| Eq 194
+ (𝛼𝑟 + 𝛽𝑟 𝑥 + 𝛾𝑟 𝑦)𝑣𝑟 }

From Eq 193 & Eq 194, the following shape functions are considered:
1
𝑁𝑝 = (𝛼 + 𝛽𝑝 𝑥 + 𝛾𝑝 𝑦)
|𝑋| 𝑝
1
𝑁𝑞 = (𝛼 + 𝛽𝑞 𝑥 + 𝛾𝑞 𝑦) Eq 195
|𝑋| 𝑞
1
𝑁𝑟 = (𝛼 + 𝛽𝑟 𝑥 + 𝛾𝑟 𝑦)
|𝑋| 𝑟

Using the shapes functions Eq 195 into Eq 193 & Eq 194, the following equations are obtained:
u(x,y) = Npup + Nquq + Nrur
Eq 196
v(x,y) = Npvp + Nqvq + Nrvr

The general displacement function  is then written as:

𝑢(𝑥, 𝑦) 𝑁𝑝 𝑢𝑝 + 𝑁𝑞 𝑢𝑞 + 𝑁𝑟 𝑢𝑟
{𝛤} = { } = { } Eq 197
𝑣(𝑥, 𝑦) 𝑁𝑝 𝑣𝑝 + 𝑁𝑞 𝑣𝑞 + 𝑁𝑟 𝑣𝑟
𝑢𝑝
𝑣𝑝
𝑁𝑝 0 𝑁𝑞 0 𝑁𝑟 0 𝑢𝑞
{𝛤} = [ ] 𝑣 Eq 198
0 𝑁𝑝 0 𝑁𝑞 0 𝑁𝑟 𝑞
𝑢𝑟
{ 𝑣𝑟 }
So,
{} = [N]{} Eq 199

Where,
𝑁𝑝 0 𝑁𝑞 0 𝑁𝑟 0
{𝑁} = [ ] Eq 200
0 𝑁𝑝 0 𝑁𝑞 0 𝑁𝑟

The shape functions Np, Nq & Nr represent the shape of the general displacement function  over the
surface of the element.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 72

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Strain-Displacement & Stress-Strain Relationships

Let’s express the element strains and stresses in terms of the unknown nodal displacements. It is
known that the strains of the two-dimensional element are:
𝜕𝑢
𝜕𝑥
𝜀𝑋 𝜕𝑣
{𝜖} = { 𝜀𝑌 } = Eq 201
𝛾𝑋𝑌 𝜕𝑦
𝜕𝑢 𝜕𝑣
+
{𝜕𝑦 𝜕𝑥}
From Eq 196:
𝜕𝑢 𝜕
= (𝑁 𝑢 + 𝑁𝑞 𝑢𝑞 + 𝑁𝑟 𝑢𝑟 ) Eq 202
𝜕𝑥 𝜕𝑥 𝑝 𝑝

However, in the CST element it is assumed that:

• up = u(xp,yp) = constant
• uq = u(xq,yq) = constant
• ur = u(xr,yr) = constant
The derivatives of up, uq and ur with respect of the variable x are null, so:
𝜕𝑢 𝜕𝑁𝑝 𝜕𝑁𝑞 𝜕𝑁𝑟
= 𝑢𝑝 + 𝑢𝑞 + 𝑢 Eq 203
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝑟

From Eq 195, the derivatives of the shape functions are:

𝜕𝑁𝑝 1 𝜕 𝛽𝑝
= (𝛼𝑝 + 𝛽𝑝 𝑥 + 𝛾𝑝 𝑦) = Eq 204
𝜕𝑥 |𝑋| 𝜕𝑥 |𝑋|
𝜕𝑁𝑞 𝛽𝑞
= Eq 205
𝜕𝑥 |𝑋|
𝜕𝑁𝑟 𝛽𝑟
= Eq 206
𝜕𝑥 |𝑋|
Substituting Eq 204, Eq 205 & Eq 206 into Eq 203:
𝜕𝑢 1
= (𝛽 𝑢 + 𝛽𝑞 𝑢𝑞 + 𝛽𝑟 𝑢𝑟 ) Eq 207
𝜕𝑥 |𝑋| 𝑝 𝑝
Similarly:
𝜕𝑣 1
= (𝛾 𝑣 + 𝛾𝑞 𝑣𝑞 + 𝛾𝑟 𝑣𝑟 ) Eq 208
𝜕𝑦 |𝑋| 𝑝 𝑝

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 73

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝜕𝑢 𝜕𝑣 1
+ = (𝛾 𝑢 + 𝛽𝑝 𝑣𝑝 + 𝛾𝑞 𝑢𝑞 + 𝛽𝑞 𝑣𝑞 + 𝛾𝑟 𝑢𝑟 + 𝛽𝑟 𝑣𝑟 ) Eq 209
𝜕𝑦 𝜕𝑥 |𝑋| 𝑝 𝑝

Using Eq 178, Eq 207, Eq 208 & Eq 209 into Eq 201:

𝑢𝑝
𝛽𝑝 0 𝛽𝑞 0 𝛽𝑟 0 𝑣𝑝
1 𝑢𝑞
{𝜖} = [0 𝛾𝑝 0 𝛾𝑞 0 𝛾𝑟 ]
𝑣𝑞 Eq 210
|𝑋|
𝛾𝑝 𝛽𝑝 𝛾𝑞 𝛽𝑞 𝛾𝑟 𝛽𝑟 𝑢
𝑟
{ 𝑣𝑟 }
𝛿𝑝
{𝜖} = [[𝐵𝑝 ] [𝐵𝑞 ] [𝐵𝑟 ]] {𝛿𝑞 } Eq 211
𝛿𝑟

{𝜖} = [𝐵]{𝛿} Eq 212

Where [B] is:

[𝐵] = [[𝐵𝑝 ] [𝐵𝑞 ] [𝐵𝑟 ]] Eq 213

[B] is a matrix independent of the x and y coordinates. It depends only on the element nodal
coordinates.
𝛽 0
1 𝑝
[𝐵𝑝 ] = [0 𝛾𝑝 ]
|𝑋|
𝛾𝑝 𝛽𝑝
𝛽𝑞 0
1
[𝐵𝑞 ] = [0 𝛾𝑞 ] Eq 214
|𝑋|
𝛾𝑞 𝛽𝑞

1 𝛽𝑟 0
[𝐵𝑟 ] = [0 𝛾𝑟 ]
|𝑋|
𝛾𝑟 𝛽𝑟

From Eq 175:

{} = [D]{}

Where [D] is given by Eq 176 for a plane stress problem and by Eq 177 for a plane strain problem.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 74

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Using Eq 175 and Eq 212, the stress in the element is:

{} = [D][B]{} Eq 215

Note that {} is constant in the CST element.

Element Stiffness Matrix
The principle of minimum potential energy is used to derive the stiffness matrix of the CST element.
The total potential energy of the element is given by:

𝜋𝑝 = 𝑈 + 𝛺𝑏 + 𝛺𝑝 + 𝛺𝑠 Eq 216
Where,
• U is the strain energy
• b is the potential energy of the body forces
• p is the potential energy of the concentrated forces
• s is the potential energy of distributed loads

The strain energy U is given by:

1
𝑈 = ∭{𝜀}𝑇 {𝜎} 𝑑𝑉
2 Eq 217
𝑉

Using Eq 175:
1
𝑈 = ∭{𝜀}𝑇 [𝐷]{𝜀} 𝑑𝑉
2 Eq 218
𝑉

Substituting Eq 212 into Eq 218:

1
𝑈 = ∭[𝐵]𝑇 {𝛿}𝑇 [𝐷][𝐵]{𝛿} 𝑑𝑉
2 Eq 219
𝑉

The potential energy b of the body forces is given by:

𝛺𝑏 = − ∭{𝛤}𝑇 {𝛷}𝑑𝑉
Eq 220
𝑉

Where,
• {} is the general displacement function defined by Eq 199.
• {} is the body weight density matrix in units of [F]/[L3]
ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 75
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Substituting Eq 199 into Eq 220:

𝛺𝑏 = − ∭[𝑁]𝑇 {𝛿}𝑇 {𝛷}𝑑𝑉

Eq 221
𝑉

The potential energy p of the concentrated forces is given by:

p = -{}T{P} Eq 222
Where,
• {} is the nodal displacement vector
• {P} is the concentrated external loads vector

The potential energy s of the distributed loads is given by:

𝛺𝑠 = − ∬{𝛤𝑠 }𝑇 {𝑄𝑠 }𝑑𝑆

Eq 223
𝑆

Where,
• {s} is the field of surface displacements through which the distributed loads act
• {Qs} is the distributed loads in units od [F]/[L2]
• S is the surface over which the distributed loads {Qs} act
Similar to Eq 199, we can express {s} in relation with the shape function matrix [Ns] evaluated along
the surface where the distributed loads act:

{s} = [Ns]{} Eq 224

So, Eq 223 becomes:

𝛺𝑠 = − ∬[𝑁𝑠 ]𝑇 {}𝑇 {𝑄𝑠 }𝑑𝑆

Eq 225
𝑆

Then, substituting Eq 219, Eq 221, Eq 222 & Eq 225 into Eq 216:

1
𝜋𝑝 = ∭[𝐵]𝑇 {𝛿}𝑇 [𝐷][𝐵]{𝛿} 𝑑𝑉 − ∭[𝑁]𝑇 {𝛿}𝑇 {𝛷}𝑑𝑉 − {}𝑇{𝑃} − ∬[𝑁𝑠 ]𝑇 {}𝑇 {𝑄𝑠 }𝑑𝑆
2 Eq 226
𝑉 𝑉 𝑆

Since the nodal displacement vector {} is not dependent of the x and y coordinates, {}T can be taken
out of the integrals:
1
𝜋𝑝 = {𝛿}𝑇 ∭[𝐵]𝑇 [𝐷][𝐵]{𝛿} 𝑑𝑉 − {𝛿}𝑇 ∭[𝑁]𝑇 {𝛷}𝑑𝑉 − {}𝑇{𝑃} − {}𝑇 ∬[𝑁𝑠 ]𝑇 {𝑄𝑠 }𝑑𝑆
2 Eq 227
𝑉 𝑉 𝑆

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 76

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The total load {f} applied on the element is the summation of the body force loads, the concentrated
forces and the distributed forces used to compute the potential energies from Eq 221, Eq 222 & Eq
225.
So, the total load vector applied on the element is:
{𝑓} = ∭[𝑁]𝑇 {𝛷}𝑑𝑉 + {𝑃} + ∬[𝑁𝑠 ]𝑇 {𝑄𝑠 }𝑑𝑆
Eq 228
𝑉 𝑆

Substituting Eq 228 into Eq 227:

1
𝜋𝑝 = {𝛿}𝑇 ∭[𝐵]𝑇 [𝐷][𝐵]{𝛿} 𝑑𝑉 − {𝛿}𝑇 {𝑓}
2 Eq 229
𝑉

Statically, finding the equilibrium of the element corresponds to find a value of the nodal
displacement vector {} yielding a stationary value of the potential energy p. This condition is
obtained for:
𝜕𝜋𝑝
=0 Eq 230
𝜕(𝛿)

Computing the derivative of Eq 229 with respect to the nodal displacement vector {} and
substituting into Eq 230:

𝜕𝜋𝑝
= [∭[𝐵]𝑇 [𝐷][𝐵] 𝑑𝑉] {𝛿} − {𝑓} = 0 Eq 231
𝜕(𝛿)
𝑉

[∭[𝐵]𝑇 [𝐷][𝐵] 𝑑𝑉] {𝛿} = {𝑓} Eq 232

𝑉

From Eq 232, the stiffness matrix appears:

[𝐾] = ∭[𝐵]𝑇 [𝐷][𝐵] 𝑑𝑉

Eq 233
𝑉

For an element with constant thickness t, Eq 233 becomes:

[𝐾] = 𝑡 ∬[𝐵]𝑇 [𝐷][𝐵]𝑑𝑥𝑑𝑦

Eq 234
𝑆

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 77

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

However, for the constant strain triangular element, all the terms of the integral are not a function
of x or y. Then:

[K] = tA[B]T[D][B] Eq 235

Where,
• A is the area of the triangle and is given by: A = X/2
• [B] is given by Eq 213
• [D] is given by Eq 176 for a plane stress problem and by Eq 177 for a plane strain problem.

The stiffness matrix [K] is a function of:

• The nodal coordinates via [B] and A
• The mechanical properties E and  via [D]

Finally, the stiffness matrix of the constant strain triangular element is a 6 x 6 matrix written as:
[𝑘𝑝𝑝 ] [𝑘𝑝𝑞 ] [𝑘𝑝𝑟 ]
[𝐾] = [[𝑘𝑞𝑝 ] [𝑘𝑞𝑞 ] [𝑘𝑞𝑟 ]] Eq 236
[𝑘𝑟𝑝 ] [𝑘𝑟𝑞 ] [𝑘𝑟𝑟 ]

Where each submatrice is given by:

For i = p, q or r 𝑡|𝑋|
[𝑘𝑖𝑗 ] = [𝐵𝑖 ]𝑇 [𝐷][𝐵𝑗 ] Eq 237
and j = p, q or r 2

Where [Bi] and [Bj] are defined by Eq 214. Note that each submatrice [kij] is a 2 x 2 matrix.
So, the 6 x 6 matrix [k] of the CST element is:

vr
y,v
Node r ur
vp

vq
up
Node p
uq
Node q
x,
u
FIG 36 – THE CST ELEMENT’S COORDINATE SYSTEM AND DOF`S

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 78

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Node p Node q Node r

up vp uq vq ur vr
𝑘𝑝𝑝𝑥 𝑘𝑝𝑝𝑥𝑦 𝑘𝑝𝑞𝑥 𝑘𝑝𝑞𝑥𝑦 𝑘𝑝𝑟𝑥 𝑘𝑝𝑟𝑥𝑦
[ ] [ ] [ ] up
𝑘𝑝𝑝𝑦𝑥 𝑘𝑝𝑝𝑦 𝑘𝑝𝑞𝑦𝑥 𝑘𝑝𝑞𝑦 𝑘𝑝𝑟𝑦𝑥 𝑘𝑝𝑟𝑦
vp
𝑘𝑞𝑝𝑥 𝑘𝑞𝑝𝑥𝑦 𝑘𝑞𝑞𝑥 𝑘𝑞𝑞𝑥𝑦 𝑘𝑞𝑟𝑥 𝑘𝑞𝑟𝑥𝑦 uq
[𝑘] = [ ] [ ] [ ] Eq 238
𝑘𝑞𝑝𝑦𝑥 𝑘𝑞𝑝𝑦 𝑘𝑞𝑞𝑦𝑥 𝑘𝑞𝑞𝑦 𝑘𝑞𝑟𝑦𝑥 𝑘𝑞𝑟𝑦 vq
𝑘𝑟𝑝𝑥 𝑘𝑟𝑝𝑥𝑦 𝑘𝑟𝑞𝑥 𝑘𝑟𝑞𝑥𝑦 𝑘𝑟𝑟𝑥 𝑘𝑟𝑟𝑥𝑦 ur
[ ] [ ] [ ] vr
[ 𝑘𝑟𝑝𝑦𝑥 𝑘𝑟𝑝𝑦 𝑘𝑟𝑞𝑦𝑥 𝑘𝑟𝑞𝑦 𝑘𝑟𝑟𝑦𝑥 𝑘𝑟𝑟𝑦 ]

Matrix Assembly
The direct stiffness method can be used to obtain the global structure stiffness matrix [K] for a
structure meshed with N CST elements:
𝑁

[𝐾] = ∑[𝑘]𝑒 Eq 239

𝑒=1

Where [k]e is the stiffness matrix of each element of the discretization. Each [k] e is computed with
Eq 236.
Explicit Expression of the CST Stiffness Matrix
In this last paragraph about the CST element the 6 x 6 stiffness matrix is explicitly written using the
Eq 235.

[𝐾] = 𝑡𝐴[𝐵]𝑇 [𝐷][𝐵] Eq 240

Where,
• A is the area of the triangle element
• t is the constant thickness of the element
• [D] is the constitutive matrix given by Eq 177
• [B] is given by Eq 213
We will write the CST element stiffness matrix considering the plane strain assumption. By
substituting the matrices [D] and [B] into Eq 240, we have:
βp 0 γp
0 γp βp 1-  0 βp 0 βq 0 βr 0
tE βq 0 γq  1- 0
[𝐾]= [ 0 γp 0 γq 0 γr ]
4A(1+ν)(1-2ν) 0 γq βq 0 1-2] [ Eq 241
0 γp βp γq βq γr βr
βr 0 γr 2
[0 γr βr ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 79

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The explicit constant strain triangle element stiffness matrix for the plane strain assumption is
obtained by multiplying the matrices in Eq 241.
The stiffness matrix [K] is a function of:
• the x and y nodal coordinates thanks to the ’s and ’s
• The material properties E and 
• The thickness t of the element
• The area A of the element
Note that the stiffness matrix is transformed for the plane stress assumption by replacing the
following terms in Eq 241:
• Replace (1 - ) by 1
• Replace (1 - 2)/2 by ( 1- )/2
• Replace (1 + )(1 - 2) by (1 - 2)

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 80

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝑡𝐸
[𝐾] =
4𝐴(1 + 𝜈)(1 − 2𝜈)
1 − 2𝜈 1 − 2𝜈 1 − 2𝜈
𝛽𝑝2 (1 − 𝜈) + 𝛾𝑝2 ( ) 𝛽𝑝 𝛾𝑝 𝜈 + 𝛽𝑝 𝛾𝑝 ( ) 𝛽𝑝 𝛽𝑞 (1 − 𝜈) + 𝛾𝑝 𝛾𝑞 ( )
2 2 2
1 − 2𝜈 1 − 2𝜈
𝛾𝑝2 (1 − 𝜈) + 𝛽𝑝2 ( ) 𝛽𝑞 𝛾𝑝 𝜈 + 𝛽𝑝 𝛾𝑞 ( )
2 2
𝑋 1 − 2𝜈
𝛽𝑞2 (1 − 𝜈) + 𝛾𝑞2 ( )
2

[ 𝑆 𝑌 𝑀
1 − 2𝜈 1 − 2𝜈 1 − 2𝜈
𝛽𝑝 𝛾𝑞 𝜈 + 𝛽𝑞 𝛾𝑝 ( ) 𝛽𝑝 𝛽𝑟 (1 − 𝜈) + 𝛾𝑝 𝛾𝑟 ( ) 𝛽𝑝 𝛾𝑟 𝜈 + 𝛽𝑟 𝛾𝑝 ( )
2 2 2
1 − 2𝜈 1 − 2𝜈 1 − 2𝜈
𝛾𝑝 𝛾𝑞 (1 − 𝜈) + 𝛽𝑝 𝛽𝑞 ( ) 𝛽𝑟 𝛾𝑝 𝜈 + 𝛽𝑝 𝛾𝑟 ( ) 𝛾𝑝 𝛾𝑟 (1 − 𝜈) + 𝛽𝑝 𝛽𝑟 ( )
2 2 2
1 − 2𝜈 1 − 2𝜈 1 − 2𝜈
𝛽𝑞 𝛾𝑞 𝜈 + 𝛽𝑞 𝛾𝑞 ( ) 𝛽𝑞 𝛽𝑟 (1 − 𝜈) + 𝛾𝑞 𝛾𝑟 ( ) 𝛽𝑞 𝛾𝑟 𝜈 + 𝛽𝑟 𝛾𝑞 ( )
2 2 2
1 − 2𝜈 1 − 2𝜈 1 − 2𝜈
𝛾𝑞2 (1 − 𝜈) + 𝛽𝑞2 ( ) 𝛽𝑟 𝛾𝑞 𝜈 + 𝛽𝑞 𝛾𝑟 ( ) 𝛾𝑞 𝛾𝑟 (1 − 𝜈) + 𝛽𝑞 𝛽𝑟 ( )
2 2 2
1 − 2𝜈 1 − 2𝜈
𝛽𝑟2 (1 − 𝜈) + 𝛾𝑟2 ( ) 𝛽𝑟 𝛾𝑟 𝜈 + 𝛽𝑟 𝛾𝑟 ( )
2 2
1 − 2𝜈
𝛾𝑟2 (1 − 𝜈) + 𝛽𝑟2 ( ) ]
2

Eq 242

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 81

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5.3. Membrane: The Linear Strain Triangle (LST)

Let’s now consider the stiffness matrix computation for the high-order triangular element called the
Linear Strain Triangle (LST). The same approach as the CST element is used for the development of
the stiffness matrix. However, since the number of equations is double, the hand calculations are
heavy. So, we will not solve the equations for the LST element, we will just expose the equations that
will be solved numerically by a computer.
Degrees of Freedom
The LST element has six nodes with two degrees of freedom per node, giving a total of twelve degrees
of freedom for the element. In addition to the three corner nodes, three mid-nodes are added on
each side of the triangular element.

v3

y,v
v4 3 u3
v1
4 u4 v6
u6 v2
u1 v5 6
1
u5
5 u2
2
x,
u
FIG 37 – NODES AND DEGREES OF FREEDOM OF THE LINEAR STRAIN TRIANGULAR ELEMENT

The nodal displacement vector is now written as:

𝑢1
𝑣1
𝑢2
𝛿1 𝑣2
𝛿2 𝑢3
𝛿 𝑣
{𝛿} = 3 = 𝑢3 Eq 243
𝛿4 4
𝛿5 𝑣4
{𝛿6 } 𝑢5
𝑣5
𝑢6
{𝑣6 }

Shape Functions

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 82

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Since the LST element has twelve nodal unknowns, we need twelve constants a i to describe its
complete displacement field. Usually, for the triangular elements, a complete polynomial is used in
cartesian coordinates to describe the displacement field within the element. Pascal’s triangle is used
to assign properly the x-y coordinates of the polynomial to the ai’s.

1
xy
x2 xy y2
x3 x2y xy2 y3
x4 x3y x2y2 xy3 y4
x5 x4y x3y2 x2y3 xy4 y5
FIG 38 – PASCAL’S TRIANGLE FOR LST ELEMENT

So, for the LST, a quadratic displacement function is selected:

u(x,y) = a1 + a2x + a3y + a4x2 + a5xy + a6y2
Eq 244
v(x,y) = a7 + a8x + a9y + a10x2 + a11xy + a12y2

The adjacent elements exhibit a displacement compatibility guarantee by a parabola defined by

three points along each side.
So, the general displacement function is written:
𝑢(𝑥, 𝑦)
{𝛤 } = { }
𝑣(𝑥, 𝑦)
𝑎1
𝑎2
1 𝑥 𝑦 𝑥2 𝑥𝑦 𝑦2 0 0 0 0 0 0 𝑎3 Eq 245
= [ 2] .
0 0 0 0 0 0 1 𝑥 𝑦 𝑥2 𝑥𝑦 𝑦
.
{𝑎12 }
In matrix form:
{} = []{a} Eq 246
With
1 𝑥 𝑦 𝑥2 𝑥𝑦 𝑦2 0 0 0 0 0 0
[] = [ ] Eq 247
0 0 0 0 0 0 1 𝑥 𝑦 𝑥2 𝑥𝑦 𝑦2

First, let’s substitute the coordinates into u and v using the Eq 245:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 83

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝑢1 1 𝑥1 𝑦1 𝑥12 𝑥1 𝑦1 𝑦12 0 0 0 0 0 0 𝑎1
𝑢2 1 𝑥2 𝑦2 𝑥22 𝑥2 𝑦2 𝑦22 0 0 0 0 0 0 𝑎2
𝑢3 . . . . . . . . . . . . 𝑎3
𝑢4 . . . . . . . . . . . . 𝑎4
𝑢5 . . . . . . . . . . . . 𝑎5
𝑢6 1 𝑥6 𝑦6 𝑥62 𝑥6 𝑦6 𝑦62 0 0 0 0 0 0 𝑎6
𝑣1 = 0 0 0 0 0 0 1 𝑥1 𝑦1 𝑥12 𝑥1 𝑦1 𝑦12 𝑎7 Eq 248
𝑣2 . . . . . . . . . . . . 𝑎8
𝑣3 . . . . . . . . . . . . 𝑎9
𝑣4 . . . . . . . . . . . . 𝑎10
𝑣5 0 0 0 0 0 0 1 𝑥5 𝑦5 𝑥52 𝑥5 𝑦5 𝑦52 𝑎11
{𝑣6 } [0 0 0 0 0 0 1 𝑥6 𝑦6 𝑥62 𝑥6 𝑦6 𝑦62 ] {𝑎12 }

Or,
{} = {X}{a} Eq 249

The ai’s are solved by:

{a} = {X}-1{} Eq 250

Then, the obtained ai’s are substituted into Eq 246 to obtain the general displacement equations in
terms of the shape functions and the nodal degrees of freedom:

{} = []{X}-1{} Eq 251

Then [N] is defined as:

[N] = []{X}-1 Eq 252

The general displacement function becomes:
{} = [N]{} Eq 253

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 84

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Strain-Displacement & Stress-Strain Relationships

Knowing that the strains of the two-dimensional element are:
𝜕𝑢
𝜕𝑥
𝜀𝑋 𝜕𝑣
{𝜖} = { 𝜀𝑌 } = Eq 254
𝛾𝑋𝑌 𝜕𝑦
𝜕𝑢 𝜕𝑣
+
{𝜕𝑦 𝜕𝑥}
So, using Eq 245 of u and v into Eq 254:
𝑎1
𝑎2
0 1 0 2𝑥 𝑦 0 0 0 0 0 0 0
𝑎
{𝜀} = [0 0 0 0 0 0 0 0 1 0 𝑥 2𝑦] .3 Eq 255
0 0 1 0 𝑥 2𝑦 0 1 0 2𝑥 𝑦 0 .
{𝑎12 }
The equation above shows a linear strain variation in the element. In matrix form:

{ = [’]{a} Eq 256

Where
0 1 0 2𝑥 𝑦 0 0 0 0 0 0 0
{Λ′ } = [0 0 0 0 0 0 0 0 1 0 𝑥 2𝑦] Eq 257
0 0 1 0 𝑥 2𝑦 0 1 0 2𝑥 𝑦 0

So, substituting Eq 250 into Eq 256, we obtain {} in terms of the nodal displacements:

{ = [’] {X}-1{} Eq 258

Considering the matrix [B] as a function of the variables x and y and the coordinates (x1,y1) to (x6,y6):

{B = [’] {X}-1 Eq 259

Note that [B] is a 3 x 12 matrix.
From Eq 259 into Eq 258 we can write:

{ = [B] {} Eq 260

Like it has been stated for the CST element:

{} = [D][B]{} Eq 261

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 85

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Where [D] is given by Eq 176 for a plane stress problem and by Eq 177 for a plane strain problem.
Element Stiffness Matrix
The stiffness matrix for the LST element can be calculated using the same equation as the one
demonstrated for the CST element:

[𝑘] = ∭[𝐵]𝑇 [𝐷][𝐵] 𝑑𝑉

Eq 262
𝑉

However, unlike the CST element, the [B] matrix of the LST element is a function of x and y but also
of the nodal coordinates. The [B] matrix is defined by:
𝛽1 0 𝛽2 0 𝛽3 0 𝛽4 0 𝛽5 0 𝛽6 0
[𝐵 ] = [ 0 𝛾1 0 𝛾2 0 𝛾3 0 𝛾4 0 𝛾5 0 𝛾6 ] Eq 263
𝛾1 𝛽1 𝛾2 𝛽2 𝛾3 𝛽3 𝛾4 𝛽4 𝛾5 𝛽5 𝛾6 𝛽6

The ’s and ’s are obtained by the inversion of the matrix [X].

So, to compute the stiffness matrix using the Eq 262, since [B] is a function of x and y, the integration
needs to be done before to multiply the matrices. The resulting stiffness matrix of the LST element
is therefore a 12 x 12 matrix.
It is very tedious to compute this matrix explicitly. This integration is however very easy for a
computer to carry out numerically.
So, the LST stiffness matrix is of the form:
𝑘11 𝑘12 𝑘13 . . . 𝑘1,12
𝑘21 𝑘22 𝑘23 . . . 𝑘2,12
𝑘31 𝑘32 𝑘33 . . . 𝑘3,12
[𝑘] = . . . . . . . Eq 264
. . . . . . .
. . . . . . .
[𝑘12,1 𝑘12,2 𝑘12,3 . . . 𝑘12,12 ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 86

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5.4. Thin-Plate
Assumptions and Basic Equations of the Kirchhoff Theory
The behavior of a plate is similar to that of a beam, they both carry transverse loads by bending
action. To start with the plates, we will focus first on the thin plates for which the behavior is
described by the Kirchhoff theory. This theory is named for the German physicist Gustav Kirchhoff.
So, considering a thin-plate in the xy plane and of thickness t measured in the z-direction submitted
to a transverse loading q.

q y
b

FIG 39 – THIN-PLATE GEOMETRY AND LOADING

The thin plate satisfies the following assumptions:

• The plate thickness is much smaller than the two dimensions a and b: t << a and b). For a thin
plate the basic assumption is that t is less than 1/10 the span of the plate. If t is more than
1/10 the span of the plate, the transverse shear has to be accounted for and the Kirchhoff
theory is no longer applicable, we are in the domain of the thick-plate theory named the
Mindlin theory (see next paragraph).
• The deflection in the transverse direction, denoted w, is much less than the thickness t
(w/t<<1).
The transverse loading q applied on the thin-plate produces a lateral deformation w in the z-direction
which is a function of x and y: it is denoted w(x,y).
The Kirchhoff assumptions are:
• A line D-D’ drawn perpendicular to the mid-surface of the plate before deformation remains
perpendicular after deformation (see Fig 40)
• The normal remain normal. That means the transverse shear strains XZ and YZ are null.
ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 87
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

• The in-plane shear strain XY is not null. In the plane <x,y>, the right angles do not remain right
angles
• The variation of thickness is null: Z = 0
• Z = 0. So, the normal stress has no influence on the in-plane stress-strain equations.
• The in-plane deformations at the mid-surface are null: u(x,y,0) = v(x,y,0) = 0
Let’s draw the behavior of a differential slice of plate before and after deformation following the
Kirchhoff assumptions.

z,w

dx
D
M
t/2 z
x,u
O
t/2

D’

z,w u = -z

Mid-
D
surface
M z
𝜕𝑤
𝛼=
𝜕𝑥
w O
x,u
D’

FIG 40 – BEHAVIOR OF THE DIFFERENTIAL SLICE OF PLATE ACCORDING TO THE KIRCHHOFF THEORY. [TOP] BEFORE
DEFORMATION [BOTTOM] AFTER DEFORMATION

As shown in Fig 40, any point M of the plate cross-section has a u displacement in the x-direction due
to the small rotation :
𝜕𝑤
𝑢 = −𝑧𝛼 = −𝑧 Eq 265
𝜕𝑥

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 88

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The behavior shown in the <x,z> plane on the Fig 40 is replicated in the <y,z> plane:
𝜕𝑤
𝑣 = −𝑧𝛼 = −𝑧 Eq 266
𝜕𝑦

So, writing the equations of the rate of change of the angular displacements of the normal known as
the curvatures of the plate:

𝜕 2𝑤 𝜕 2𝑤 𝜕 2𝑤
𝜒𝑥 = − 2 𝜒𝑦 = − 𝜒𝑥𝑦 = −2 Eq 267
𝜕𝑥 𝜕𝑦 2 𝜕𝑥𝜕𝑦

Since the in-plane strain-displacement equations are written:

𝜕 2𝑤 𝜕 2𝑤 𝜕 2𝑤
𝜀𝑥 = −𝑧 𝜀𝑦 = −𝑧 𝛾𝑥𝑦 = −2𝑧 Eq 268
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝜕𝑦

By using Eq 267 into Eq 268Eq 270:

𝜀𝑥 = −𝑧𝜒𝑥 𝜀𝑦 = −𝑧𝜒𝑦 𝛾𝑥𝑦 = −𝑧𝜒𝑥𝑦 Eq 269

Based on the assumptions of the Kirchhoff theory, the in-plane stress-strain relationships for an
isotropic material are:
𝐸 𝐸
𝜎𝑥 = (𝜀 + 𝜈𝜀𝑦 ) 𝜎𝑦 = (𝜀 + 𝜈𝜀𝑥 ) 𝜏𝑥𝑦 = 𝐺𝛾𝑥𝑦 Eq 270
1 − 𝜈2 𝑥 1 − 𝜈2 𝑦

The normal stresses  and shear stresses  are plotted in Fig 41 acting on the edges of the plate:
• The stresses X, Y, XY and YX follow a linear variation in the z-direction from the mid-surface.
• The transverse shear stresses XZ and YZ are null as per the Kirchhoff assumptions.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 89

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

q y
dy

t
YZ dx

Y
XZ
YX XY X
x

FIG 41 – STRESSES ON THE EDGES OF THE DIFFERENTIAL ELEMENT OF THE PLATE

Moreover, the bending moments MX and MY as well as the twisting moment MXY acting along the
edges of the plate can be related to the stresses of equations Eq 270 as shown in Fig 42:

z
Qx
Qy y
Mxy
My
Mx Myx

dx
dy
Myx Mx
My
Mxy
Qy
x Qx

FIG 42 – FORCES AND MOMENTS ON THE EDGES OF THE DIFFERENTIAL ELEMENT OF THE PLATE

The moments per unit length functions of x and y are:

𝑡/2 𝑡
2 𝑡/2
𝑀𝑥 = ∫ 𝑧𝜎𝑥 𝑑𝑧 𝑀𝑦 = ∫ 𝑧𝜎𝑦 𝑑𝑧 𝑀𝑥𝑦 = ∫ 𝑧𝜏𝑥𝑦 𝑑𝑧 Eq 271
−𝑡/2 𝑡
− −𝑡/2
2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 90

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So, substituting Eq 269 & Eq 270 into Eq 271, the moments are function of the curvatures by:

𝑀𝑥 = 𝐷(𝜒𝑥 + 𝜈𝜒𝑦 ) 𝑀𝑦 = 𝐷(𝜒𝑦 + 𝜈𝜒𝑥 ) 𝐷(1 − 𝜈)

𝑀𝑥𝑦 = 𝜒𝑥𝑦 Eq 272
2

Where D is the bending rigidity of the plate:

𝐸𝑡 3
𝐷= Eq 273
12(1 − 𝜈 2 )

The equilibrium of forces with respect to the z-axis as well as the equilibrium of moments around
the x and y axes permit to write the equilibrium equations in the form of the following differential
equations:
𝜕𝑄𝑥 𝜕𝑄𝑦 𝜕𝑀𝑥 𝜕𝑀𝑥𝑦 𝜕𝑀𝑦 𝜕𝑀𝑥𝑦
+ +𝑞 =0 + − 𝑄𝑥 = 0 + − 𝑄𝑦 = 0 Eq 274
𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥

So, by substituting the moment-curvature relationships Eq 272 into Eq 274 and by solving for Qx and
Qy, the governing partial differential equation for an isotropic thin-plate in bending is:
𝜕 4𝑤 𝜕 4𝑤 𝜕 4𝑤
𝐷( 4 + + 2 2 2) = 𝑞 Eq 275
𝜕𝑥 𝜕𝑦 4 𝜕𝑥 𝜕𝑦

This governing partial differential equation shows that the thin plate bending behavior depends only
on the transverse displacement w.

Degrees-of-freedom
To derive the stiffness matrix of a thin plate we will consider the four nodes element called the
quadrilateral element. Each node [j,k,m,n] of the quadrilateral element has three degrees of
freedom:
• A transverse displacement w the z-direction
• A rotation x about the x-axis
• A rotation y about the y-axis

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 91

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

z
n
w
 y
j y
 x
x m

FIG 43 – THIN-PLATE DEGREES OF FREEDOM

The nodal displacement vector at node i (for i = j, k, m or n) is:

𝑤𝑖
{𝛿𝑖 } = {𝜃𝑥𝑖 } Eq 276
𝜃𝑦𝑖

Note that the rotations  are related to the transverse displacement by:
𝜕𝑤
𝜃𝑥 =
𝜕𝑦
Eq 277
𝜕𝑤
𝜃𝑦 = −
𝜕𝑥
The negative sign for y is needed because to produce a positive rotation about the y-axis, a negative
displacement w is required.
The total quadrilateral element displacement vector is composed with 4 x 3 = 12 degrees of freedom:
𝑤𝑗
𝜃𝑥𝑗
𝜃𝑦𝑗
𝑤𝑘
𝜃𝑥𝑘 𝛿𝑗
𝜃𝑦𝑘 𝛿𝑘
{𝛿} =
𝑤𝑚 = 𝛿𝑚 Eq 278
𝜃𝑥𝑚 { 𝛿𝑛 }
𝜃𝑦𝑚
𝑤𝑛
𝜃𝑥𝑛
{ 𝜃𝑦𝑛 }

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 92

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Shape Functions
Since the thin-plate element has twelve degrees of freedom, we need twelve constants ai to describe
its complete displacement field with a polynomial function. Like for the LST elements, a complete
polynomial in cartesian coordinates is used to describe the displacement field within the element.
Pascal’s triangle is used to assign properly the x-y coordinates of the polynomial to the ai’s.

1
xy
x2 xy y2
x3 x2y xy2 y3
x4 x3y x2y2 xy3 y4
x5 x4y x3y2 x2y3 xy4 y5
FIG 44 – PASCAL’S TRIANGLE FOR THE THIN-PLATE ELEMENT

The last two terms x3y and xy3 are chosen to ensure the continuity in displacement at the boundaries
of element. The x2y2 is rejected because it cannot be paired with any other terms.
𝑤 = 𝑎1 + 𝑎2 𝑥 + 𝑎3 𝑦 + 𝑎4 𝑥 2 + 𝑎5 𝑥𝑦 + 𝑎6 𝑦 2 + 𝑎7 𝑥 3 + 𝑎8 𝑥 2 𝑦 + 𝑎9 𝑥𝑦 2 + 𝑎10 𝑦 3
+ 𝑎11 𝑥 3 𝑦 + 𝑎12 𝑥𝑦 3 Eq 279

Similarly to the 1D beam element, the constants ai’s can be computed by expressing the 12
simultaneous shape functions linking the values of w and its slopes at the four nodes of the
quadrilateral element using their appropriate values.
Starting with the evaluation of the slopes:
𝜕𝑤
= 𝑎3 + 𝑎5 𝑥 + 2𝑎6 𝑦 + 𝑎8 𝑥 2 + 2𝑎9 𝑥𝑦 + 3𝑎10 𝑦 2 + 𝑎11 𝑥 3
𝜕𝑦
+ 3𝑎12 𝑥𝑦 2
Eq 280
𝜕𝑤
= 𝑎2 + 2𝑎4 𝑥 + 𝑎5 𝑦 + 3𝑎7 𝑥 2 + 2𝑎8 𝑥𝑦 + 𝑎9 𝑦 2 + 3𝑎11 𝑥 2 𝑦
𝜕𝑥
+ 𝑎12 𝑦 3

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 93

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So, Eq 279 & Eq 280 in matrix form:

𝑤
𝜕𝑤 𝑤
𝜕𝑦 = {𝜃𝑥 } =
𝜕𝑤 𝜃𝑦
{ 𝜕𝑥 }
𝑎1 Eq
𝑎2 281
1 𝑥 𝑦 𝑥2 𝑥𝑦 𝑦2 𝑥3 𝑥2𝑦 𝑥𝑦 2 𝑦3 𝑥3𝑦 𝑥𝑦 3
𝑎3
= [0 0 1 0 𝑥 2𝑦 0 𝑥2 2𝑥𝑦 3𝑦 2 𝑥3 3𝑥𝑦 2 ] .
0 −1 0 −2𝑥 −𝑦 0 −3𝑥 2 −2𝑥𝑦 −𝑦 2 0 −3𝑥 2 𝑦 −𝑦 3 .
{𝑎12 }
Again, the general displacement function  is expressed as:
{} = []{a} Eq 282
With
[] =
1 𝑥 𝑦 𝑥2 𝑥𝑦 𝑦2 𝑥3 𝑥2𝑦 𝑥𝑦 2 𝑦3 𝑥3𝑦 𝑥𝑦 3
= [0 0 1 0 𝑥 2𝑦 0 𝑥2 2𝑥𝑦 3𝑦 2 𝑥3 3𝑥𝑦 2 ] Eq 283
0 −1 0 −2𝑥 −𝑦 0 −3𝑥 2 −2𝑥𝑦 −𝑦 2 0 −3𝑥 2 𝑦 −𝑦 3

Eq 281 is now expanded by using the appropriate values at each point of the quadrilateral element:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 94

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝑤𝑗
𝜃𝑥𝑗
𝜃𝑦𝑗
𝑤𝑘
𝜃𝑥𝑘
𝜃𝑦𝑘
{𝛿} =
𝑤𝑚 =
𝜃𝑥𝑚
𝜃𝑦𝑚
𝑤𝑛
𝜃𝑥𝑛
{ 𝜃𝑦𝑛 }
1 𝑥𝑗 𝑦𝑗 𝑥𝑗2 𝑥𝑗 𝑦𝑗 𝑦𝑗 2 𝑥𝑗 3 𝑥𝑗 2 𝑦𝑗 𝑥𝑗 𝑦𝑗 2 𝑦𝑗 3 𝑥𝑗 3 𝑦𝑗 𝑥𝑗 𝑦𝑗 3 𝑎1
0 0 1 0 𝑥𝑗 2𝑦𝑗 0 𝑥𝑗 2 2𝑥𝑗 𝑦𝑗 3𝑦𝑗 2 𝑥𝑗 3 3𝑥𝑗 𝑦𝑗 2 𝑎2
0 −1 0 −2𝑥𝑗 −𝑦𝑗 0 −3𝑥𝑗 2 −2𝑥𝑗 𝑦𝑗 −𝑦𝑗 2 0 −3𝑥𝑗 2 𝑦𝑗 −𝑦𝑗 3 𝑎3
. . 𝑎4
. . 𝑎5
𝑎6
= . .
𝑎7
. .
. . 𝑎8
. . 𝑎9
1 𝑥𝑛 𝑦𝑛 𝑥𝑛2 𝑥𝑛 𝑦𝑛 𝑦𝑛 2 𝑥𝑛 3 𝑥𝑛 2 𝑦𝑛 𝑥𝑛 𝑦𝑛 2 𝑦𝑛 3 𝑥𝑛 3 𝑦𝑛 𝑥𝑛 𝑦𝑛 3 𝑎10
𝑎11
0 0 1 0 𝑥𝑛 2𝑦𝑛 0 𝑥𝑛 2 2𝑥𝑛 𝑦𝑛 3𝑦𝑛 2 𝑥𝑛 3 3𝑥𝑛 𝑦𝑛 2
{𝑎12 }
[0 −1 0 −2𝑥𝑛 −𝑦𝑛 0 −3𝑥𝑛 2 −2𝑥𝑛 𝑦𝑛 −𝑦𝑛 2 0 −3𝑥𝑛 2 𝑦𝑛 −𝑦𝑛 3 ]

Eq 284

{𝛿} = [X]{𝑎} Eq 285

The twelve constants ai are solved by:

{𝑎} = [𝑋]−1 {𝛿} Eq 286

Substituting Eq 286 into Eq 282:

{} = [][𝑋]−1 {𝛿} Eq 287
With the shape function matrix given by:
{N} = [][𝑋]−1 Eq 288
the general displacement function  is written in function of the shape function matrix [N] and the
nodal displacement vector {}:
{} = [N]{𝛿} Eq 289

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 95

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Curvature-Displacement & Moment-Curvature Relationships

As per the Kirchhoff assumptions we have:
𝜕 2𝑤 𝜕 2𝑤 𝜕 2𝑤
𝜒𝑥 = − 2 𝜒𝑦 = − 2 𝜒𝑥𝑦 = −2 Eq 290
𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦

So, by evaluating the above second derivatives of w from Eq 279, the curvature matrix is:
𝜒𝑥
𝜒
{ 𝑦}=
𝜒𝑥𝑦
𝑎1
𝑎2
𝑎3
𝑎4
𝑎5 Eq 291
0 0 0 −2 0 0 −6𝑥 −2𝑦 0 0 −6𝑥𝑦 0
𝑎6
= [0 0 0 0 0 −2 0 0 −2𝑥 −6𝑦 0 −6𝑥𝑦]
𝑎7
2
0 0 0 0 −2 0 0 −4𝑥 −4𝑦 0 −6𝑥 −6𝑦 2 𝑎8
𝑎9
𝑎10
𝑎11
{𝑎12 }

{𝜒} = [Λ′ ]{𝑎} Eq 292

Substituting Eq 286 into Eq 292:

{𝜒} = [Λ′ ][𝑋]−1 {𝛿} Eq 293
So, by considering the gradient matrix [B]:
[𝐵] = [Λ′ ][𝑋]−1 Eq 294
The curvature matrix is given by:

{𝜒} = [𝐵]{𝛿} Eq 295

By writing Eq 272 & Eq 273 in matrix form:
1 𝜈 0 𝜒𝑥
𝑀𝑥 𝐸𝑡 3 𝜈 1 0 𝜒
{𝑀} = { 𝑀𝑦 } = 2)
[ 1 − 𝜈] { 𝑦 } Eq 296
𝑀𝑥𝑦 12(1 − 𝜈 𝜒𝑥𝑦
0 0
2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 96

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Or,

{𝑀} = [𝐷]{𝜒} Eq 297

Where [D] is the constitutive matrix for an isotropic material:
1 𝜈 0
𝐸𝑡 3 𝜈 1 0
[𝐷] = 2
[ 1 − 𝜈] Eq 298
12(1 − 𝜈 )
0 0
2

Finally, using Eq 295 into Eq 297:

{𝑀} = [𝐷][𝐵]{𝛿} Eq 299

Element Stiffness Matrix

The stiffness matrix of the four nodes thin-plate element is calculated using the usual form:

[𝑘] = ∬[𝐵]𝑇 [𝐷][𝐵]𝑑𝑥𝑑𝑦 Eq 300

Where:
• [B] is computed by Eq 294
• [D] is computed by Eq 298
The element stiffness matrix [k] is a 12 x 12 matrix.
𝑘11 𝑘12 𝑘13 . . . 𝑘1,12
𝑘21 𝑘22 𝑘23 . . . 𝑘2,12
𝑘31 𝑘32 𝑘33 . . . 𝑘3,12
[𝑘] = . . . . . . . Eq 301
. . . . . . .
. . . . . . .
[𝑘12,1 𝑘12,2 𝑘12,3 . . . 𝑘12,12 ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 97

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5.5. Isoparametric Formulation

The development of stiffness matrices for the membranes CST, LST as well as the four nodes plate
elements presented in the previous paragraphs have been done using a global coordinate system.
We have seen that the developments of the equations in terms of global coordinates is tedious.
However, an isoparametric formulation is available which leads to a simpler formulation more
adequate for programming and computing. This formulation is generally used for 2D and 3D stress
analysis. Moreover, the isoparametric formulation offers the possibility to have non-right-angles
elements or elements with curved sides. This formulation has been implemented in all the modern
FEA packages. This formulation is used for the derivation of the thick-plate element stiffness matrix
using the Mindlin assumptions. But before, the development of the stiffness matrix for the
membrane four nodes quadrilateral element using the isoparametric formulation is required.
The Natural Coordinate System
Since the quadrilateral element may be non-rectangular, an alternative coordinate system called the
natural coordinate system <, > is introduced:
• Its origin in the global coordinates <x,y> is at the average of the corner coordinates
• Regardless of the shape of the element, the physical size of the element or its orientation in
the global coordinate system <x,y>, the edges and the corners are at  = ±1 and  = ±1
•  and  are not necessarily orthogonal and have no particular orientation with respect to axes
x and y

 =1
=½
=-½ =
=-1 3 1

4 =½

y,v 

1 =-
½

2 =-1
x,u

FIG 45 – FOUR NODES QUADRILATERAL ELEMENT IN THE <, > NATURAL COORDINATE SYSTEM

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 98

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Degrees of Freedom

Let’s consider the membrane four nodes quadrilateral element having two degrees of freedom (ui,
vi) at each node in the global coordinate system <x,y>. So, the membrane quadrilateral element has
eight degrees of freedom.

v3

u3
v4 3
(x3,y3)

4 u4
(x4,y4)

y,v 
v1

u1 v2
1
(x1,y1)
u2
2
x,u (x2,y2)

FIG 46 – DEGREES OF FREEDOM OF THE MEMBRANE FOUR NODES QUADRILATERAL ELEMENT

Shape Functions
The shape functions will be defined to map the quadrilateral element of Fig 45 in the natural
coordinates <, >. Since the membrane quadrilateral element has eight degrees of freedom, we
need eight constants ai to describe its complete displacement field.
First, the element shape described in the x and y coordinates can be related to the isoparametric
coordinates  and  by:
𝑥 = 𝑎1 + 𝑎2 𝜉 + 𝑎3 𝜂 + 𝑎4 𝜉𝜂
Eq 302
𝑦 = 𝑎5 + 𝑎6 𝜉 + 𝑎7 𝜂 + 𝑎8 𝜉𝜂

As done previously, the ai’s coefficients in terms of the eight known nodal coordinates (x1,y1), (x2,y2),
(x3,y3) & (x4,y4) can be solved:
1
𝑥 = [(1 − 𝜉)(1 − 𝜂)𝑥1 + (1 + 𝜉)(1 − 𝜂)𝑥2 + (1 + 𝜉)(1 + 𝜂)𝑥3
4 Eq 303
+ (1 − 𝜉)(1 + 𝜂)𝑥4 ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 99

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

1
𝑦 = [(1 − 𝜉)(1 − 𝜂)𝑦1 + (1 + 𝜉)(1 − 𝜂)𝑦2 + (1 + 𝜉)(1 + 𝜂)𝑦3
4
+ (1 − 𝜉)(1 + 𝜂)𝑦4 ]

In matrix form:
𝑥1
𝑦1
𝑥2
𝑥 𝑁 0 𝑁2 0 𝑁3 0 𝑁4 0 𝑦2
{𝑦} = [ 1 ] Eq 304
0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 𝑥3
𝑦3
𝑥4
{𝑦4 }
With the shape functions of the above equation:
(1 − 𝜉)(1 − 𝜂)
𝑁1 =
4
(1 + 𝜉)(1 − 𝜂)
𝑁2 =
4
Eq 305
(1 + 𝜉)(1 + 𝜂)
𝑁3 =
4
(1 − 𝜉)(1 + 𝜂)
𝑁4 =
4

So, the displacement function  can now be defined by the same shape functions used to define the
element shape:

{} = [N]{} Eq 306

𝑢1
𝑣1
𝑢2
𝑢 𝑁 0 𝑁2 0 𝑁3 0 𝑁4 0 𝑣2
{ }=[ 1 ] 𝑢 Eq 307
𝑣 0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 3
𝑣3
𝑢4
{ 𝑣4 }

Strain-Displacement & Stress-Strain Relationships

To calculate the stiffness matrix [K], the strain-displacement matrix [B] has to be computed. It is
easier to formulate [B] in terms of natural coordinates  and  rather than in the cartesian

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 100

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

coordinates. Moreover, this isoparametric approach is easier to program and it significantly improves
computing time.
The strains are defined in terms of the derivatives of the displacements with respect to x and y
coordinates (for example x=u/x). However, the displacements u and v are now functions of  and
 coordinates as given by Eq 307 with the shape functions given by Eq 305.
So, to generalize the problem, let introduce a function j (could be u or v) expressed in terms of  and
 for which the derivatives are computed with respect to x and y. So, the chain rule of differentiation
is applied since it is not possible to express  and  as functions of x and y directly. For j as a function
of x and y, the chain rule gives:
𝜕𝑗 𝜕𝑗 𝜕𝑥 𝜕𝑗 𝜕𝑦
= +
𝜕𝜉 𝜕𝑥 𝜕𝜉 𝜕𝑦 𝜕𝜉
Eq 308
𝜕𝑗 𝜕𝑗 𝜕𝑥 𝜕𝑗 𝜕𝑦
= +
𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝜕𝜂

Since the function j can be the displacement functions u or v, (j/) and (j/) are computed by Eq
307 while (x/), (x/), (y/) and (y/) are computed using Eq 304. The two-remaining
unknown in the Eq 308 are (j/x) and (j/x) needed to compute the strains. To solve the Eq 308 for
(j/x) and (j/x) the Cramer’s rule is proposed.

Reminder on the Cramer’s rule:

To solve the following system for {X}:
𝐾11 𝐾12 𝐾13 𝑥1 𝑓1
[𝐾21 𝐾22 𝐾23 ] {𝑥2 } = {𝑓2 } Eq 309
𝐾31 𝐾32 𝐾33 𝑥3 𝑓3
Where:
• [K] is a 3 x 3 known matrix.
• {f} is a known vector.
• {X} is the unknown vector.

To solve the system Eq 309 for {X}, the following Cramer’s rule is applied in two steps:
• Step 1: Create the matrix [Li] as the matrix [K] with column i replaced by the vector {f}
• Step 2: Compute:
|𝐿𝑖 |
𝑥𝑖 = Eq 310
|𝐾|

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 101

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Where,
• Li is the determinant of matrix [Li]
• K is the determinant of matrix [K]

So, by applying the Cramer’s rule, the system Eq 309 for {X} is solved:
𝑓1 𝐾12 𝐾13 𝐾11 𝑓1 𝐾13
|𝑓2 𝐾22 𝐾23 | |𝐾21 𝑓2 𝐾23 |
|𝐿1 | 𝑓3 𝐾32 𝐾33 |𝐿2 | 𝐾31 𝑓3 𝐾33
𝑥1 = = 𝑥2 = =
|𝐾| 𝐾11 𝐾12 𝐾13 |𝐾| 𝐾11 𝐾12 𝐾13
| 21 𝐾22 𝐾23 |
𝐾 | 21 𝐾22 𝐾23 |
𝐾
𝐾31 𝐾32 𝐾33 𝐾31 𝐾32 𝐾33
Eq 311
𝐾11 𝐾12 𝑓1
|𝐾21 𝐾22 𝑓2 |
|𝐿3 | 𝐾31 𝐾32 𝑓3
𝑥3 = =
|𝐾| 𝐾11 𝐾12 𝐾13
|𝐾21 𝐾22 𝐾23 |
𝐾31 𝐾32 𝐾33

Using the Cramer’s rule, Eq 308 is solved for (j/x) and (j/x):
𝜕𝑗 𝜕𝑦 𝜕𝑥 𝜕𝑗
𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉
| | | |
𝜕𝑗 𝜕𝑦 𝜕𝑥 𝜕𝑗
𝜕𝑗 𝜕𝜂 𝜕𝜂 𝜕𝑗 𝜕𝜂 𝜕𝜂
= = Eq 312
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦
𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉
| | | |
𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦
𝜕𝜂 𝜕𝜂 𝜕𝜂 𝜕𝜂

Where the denominator matrix is called the Jacobian matrix [J]. So, the Jacobian matrix is given by:
𝜕𝑥 𝜕𝑦
𝜕𝜉 𝜕𝜉
[𝐽] = Eq 313
𝜕𝑥 𝜕𝑦
[𝜕𝜂 𝜕𝜂]

Let’s continue with the expression of the element strains already used in previous paragraphs:

𝜖 = [𝐵]{𝛿} Eq 314

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 102

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Here, [B] is expressed as a function of  and . The usual strain-displacement relationships are:
𝜕𝑢
𝜕𝑥
𝜀𝑋 𝜕𝑣
{𝜖} = { 𝜀𝑌 } = Eq 315
𝛾𝑋𝑌 𝜕𝑦
𝜕𝑢 𝜕𝑣
+
{𝜕𝑦 𝜕𝑥}

Let’s introduce an operator matrix to re-write the Eq 315:

𝜕( )
0
𝜕𝑥
𝜀𝑋 𝜕( ) 𝑢
{ 𝜀𝑌 } = 0 { } Eq 316
𝛾𝑋𝑌 𝜕𝑦 𝑣
𝜕( ) 𝜕( )
[ 𝜕𝑦 𝜕𝑥 ]

Where the operator matrix is

𝜕( )
0
𝜕𝑥
𝜕( )
[Ε] = 0 Eq 317
𝜕𝑦
𝜕( ) 𝜕( )
[ 𝜕𝑦 𝜕𝑥 ]

In the operator matrix epsilon from Eq 317, ()/x and ()/y are the partial derivatives of any
variable we included in the parentheses.
Coming back to Eq 312 the determinants in the numerators are evaluated:
𝜕( ) 1 𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
= [ − ]
𝜕𝑥 |𝐽| 𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂
Eq 318
𝜕( ) 1 𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( )
= [ − ]
𝜕𝑦 |𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉

In the Eq 318 above, |J| is the determinant of the Jacobian matrix given by Eq 313.

So, by substituting Eq 318 into Eq 316, the strains are expressed in terms of the natural coordinates
 and :

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 103

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
[ − ] 0
𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂
𝜀𝑋
1 𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( ) 𝑢
{ 𝜀𝑌 } = 0 [ − ] { } Eq 319
𝛾𝑋𝑌 |𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉 𝑣
𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( ) 𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
[ − ] [ − ]
[ 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂 ]

By using Eq 307 the strains of Eq 319 are expressed in terms of the shape functions:
𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
[ − ] 0
𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂
𝜀𝑋
1 𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( )
{ 𝜀𝑌 } = 0 [ − ]
𝛾𝑋𝑌 |𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉
𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( ) 𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
[ − ] [ − ]
[ 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂 ]
𝑢1 Eq 320
𝑣1
𝑢2
𝑁 0 𝑁2 0 𝑁3 0 𝑁4 0 𝑣2
∗[ 1 ] 𝑢
0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 3
𝑣3
𝑢4
{ 𝑣4 }
Or in compact matrix form:
{𝜀} = [Ψ][𝑁]{𝛿} Eq 321

Where [] is another operator matrix given by:

𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
[ − ] 0
𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂
1 𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( )
[Ψ] = 0 [ − ] Eq 322
|𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉
𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( ) 𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( )
[ − ] [ − ]
[ 𝜕𝜉 𝜕𝜂 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂 ]

[N] is the 2 x 8 shape function matrix:

𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 0
[𝑁] = [ ] Eq 323
0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 104

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

{} is the degrees of freedom vector:

𝑢1
𝑣1
𝑢2
𝑣
{𝛿} = 𝑢2 Eq 324
3
𝑣3
𝑢4
{ 𝑣4 }

So, as per Eq 314, [B] is the 3 x 8 matrix defined as:

[𝐵] = [Ψ][𝑁] Eq 325

By substituting Eq 322 for [] into Eq 325 and Eq 305 & Eq 323 for the shape functions into Eq 325,
[B] is:
1
[𝐵(𝜉, 𝜂)] = [𝐵 𝐵2 𝐵3 𝐵4 ] Eq 326
|𝐽| 1
Where the submatrices [Bi] are:
𝜕𝑁𝑖 𝜕𝑁𝑖
𝛼 −𝛽 0
𝜕𝜉 𝜕𝜂
𝜕𝑁𝑖 𝜕𝑁𝑖
For i = 1, 2, 3, 4 [𝐵𝑖 ] = 0 𝛾 −𝜆 Eq 327
𝜕𝜂 𝜕𝜉
𝜕𝑁𝑖 𝜕𝑁𝑖 𝜕𝑁𝑖 𝜕𝑁𝑖
𝛾 −𝜆 𝛼 −𝛽
[ 𝜕𝜂 𝜕𝜉 𝜕𝜉 𝜕𝜂 ]
1
𝛼 = [(𝜉 − 1)𝑦1 − (𝜉 + 1)𝑦2 + (𝜉 + 1)𝑦3 + (1 − 𝜉)𝑦4 ]
4
1
𝛽 = [(𝜂 − 1)𝑦1 + (1 − 𝜂)𝑦2 + (1 + 𝜂)𝑦3 − (1 + 𝜂)𝑦4 ]
4
1 Eq 328
𝛾 = [(𝜂 − 1)𝑥1 + (1 − 𝜂)𝑥2 + (1 + 𝜂)𝑥3 − (1 + 𝜂)𝑥4 ]
4
1
𝜆 = [(𝜉 − 1)𝑥1 − (𝜉 + 1)𝑥2 + (𝜉 + 1)𝑥3 + (1 − 𝜉)𝑥4 ]
4
So, to evaluate [B], the determinant of the Jacobian matrix which is a polynomial of  and  is
evaluated. Then, from Eq 313 for [J] and Eq 303 for x and y, the determinant of the Jacobian matrix
is calculated:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 105

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

0 1−𝜂 𝜂−𝜉 𝜉−1 𝑥1

1 𝜂−1 0 𝜉 + 1 −𝜉 − 𝜂 𝑥2
|𝐽| = [𝑥1 𝑥2 𝑥3 𝑥4 ] [ ] {𝑥 } Eq 329
8 𝜉 − 𝜂 −𝜉 − 1 0 𝜂+1 3
1−𝜉 𝜉+𝜂 −𝜂 − 1 0 𝑥4

So, |J| and [B] are function of ξ and η and of the known global coordinates x1, x2, x3, x4, y1, y2, y3,
y4. Due to the stress-strain relationship, the stress matrix σ is also a function of ξ and η:

𝜎 = [𝐷][𝐵]{𝛿} Eq 330

Element Stiffness Matrix

The stiffness matrix [k] is evaluated using the well-known equation:

[𝑘] = ∭[𝐵]𝑇 [𝐷][𝐵] 𝑑𝑉

Eq 331
𝑉

So, for a planar element having a constant thickness t, Eq 331 becomes:

[𝑘] = ∬[𝐵]𝑇 [𝐷][𝐵]. 𝑡. 𝑑𝑥𝑑𝑦

Eq 332
𝐴

To express the above equation in terms of <,> coordinates, since [B] is a function of  and , the
integration is done with respect to  and . So, using the determinant of the Jacobian matrix, the
transformation of x and y variables into the <,> coordinates is performed. From a theorem of
integral calculus, the generic form for a function j is written as (see a mathematics book for the
complete proof):

∬ 𝑗(𝑥, 𝑦)𝑑𝑥𝑑𝑦 = ∬ 𝑗(𝜉, 𝜂)|𝐽|𝑑𝜉𝑑𝜂

Eq 333
𝐴 𝐴

So, using Eq 333 into Eq 347, the iso-parametric coordinates is:

+1 +1
[𝑘] = ∫ ∫ [𝐵]𝑇 [𝐷][𝐵]𝑡|𝐽|𝑑𝜉𝑑𝜂 Eq 334
−1 −1

The above integration to determine the stiffness matrix is done numerically using the Gaussian
quadrature method. Refer to a mathematics book for details about this method. Note that the
stiffness matrix computed with Eq 334 is an 8 x 8 matrix.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 106

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5.6. Thick-Plate
Assumptions and Basic Equations of the Mindlin Theory

This theory is named for the American mechanical engineer Raymond David Mindlin. The Mindlin
theory of plates is an extension of the Kirchhoff plate theory that considers shear deformations
through-the thickness of a plate. The Mindlin theory is to the plates what the theory of Timoshenko
is to the beams.
1D Beam 2D Plate Cross-Sections
Shear deformation EULER- Right angles
KIRCHHOFF
neglected BERNOULLI preserved
Shear deformation Right angles
TIMOSHENKO MINDLIN
taken into account NOT preserved

Similarly to the Timoshenko theory for the beams, the transverse shear deformation effects are
included in the Mindlin plate theory. It means that the plane sections do not remain perpendicular
to the mid-surface after deformation. In other words, the planes initially normal to the mid-surface
may experience different rotations than the mid-surface itself. The main consequence is that the
transverse strains xz and yz are no more null and the rotations of the cross-sections are treated as
independent variables.

Before

Mid-

After deformation

Mid-

Kirchhoff Mindlin

FIG 47 – SHEAR DEFORMATION IN A MINDLIN PLATE

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 107

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Let’s draw the behavior of a differential slice of plate after deformation following the Mindlin
assumptions. Fig 48 shows the displacements of a differential slice of plate in the <x,z> and <y,z>
planes. x and y are the rotations of lines initially perpendicular to the middle surface. The top view
shows the definition of positive rotations viewed normal to <x,y> plane. The middle view shows the
displacements in <x,z> plane. The bottom view shows the displacements in <y,z> plane.
The Mindlin assumptions are:
• The plate thickness is approximately one tenth of the planar dimensions
• The normal to the mid-surface remains straight but not necessarily perpendicular to the mid-
surface
• The displacement field across the plate thickness is linear
• The plate thickness does not change during deformation
• The normal stress Z through the thickness is ignored (plane stress condition)

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 108

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

y

y,v

x,u x
z,w

z,w y
u = zy

+O
w
x,u

𝜕𝑤
𝜕𝑥

z,w v = -zx

x
M

+O
w
y,v

𝜕𝑤
𝜕𝑦

FIG 48 – BEHAVIOR OF THE DIFFERENTIAL SLICE OF PLATE ACCORDING TO THE MINDLIN THEORY
ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 109
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Degrees-of-freedom
To derive the stiffness matrix of the thick-plate, the same four nodes element as in thin-plate is
considered. Each node [j,k,m,n] of the quadrilateral element has three degrees of freedom:
• A transverse displacement w in the z-direction
• A rotation θx about the x-axis
• A rotation θy about the y-axis

z
n
w
 y
j y
 x
x m

FIG 49 – THICK-PLATE DEGREES OF FREEDOM

The nodal displacement vector at node i (for i = j, k, m or n) is:

𝑤𝑖
𝜃
{𝛿𝑖 } = { 𝑥𝑖 } Eq 335
𝜃𝑦𝑖
The total quadrilateral element displacement vector is composed with 4 x 3 = 12 degrees of freedom:

Eq 336

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 110

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Shape Functions
Using the isoparametric formulation of the shape functions defined in the previous paragraph:
(1 − 𝜉)(1 − 𝜂)
𝑁1 =
4
(1 + 𝜉)(1 − 𝜂)
𝑁2 =
4
Eq 337
(1 + 𝜉)(1 + 𝜂)
𝑁3 =
4
(1 − 𝜉)(1 + 𝜂)
𝑁4 =
4
And, considering the four nodes isoparametric element, the variations of displacement w and
rotations x and y within the thick-plate element are expressed in relation of nodal values as:
𝑤1
𝜃𝑥1
𝜃𝑦1
𝑤2
𝜃𝑥2
𝑤 𝑁1 0 0 𝑁2 0 0 𝑁3 0 0 𝑁4 0 0 𝜃𝑦2
{𝜃𝑥 } = [ 0 𝑁1 0 0 𝑁2 0 0 𝑁3 0 0 𝑁4 0] 𝑤 Eq 338
3
𝜃𝑦 0 0 𝑁1 0 0 𝑁2 0 0 𝑁3 0 0 𝑁4 𝜃
𝑥3
𝜃𝑦3
𝑤4
𝜃𝑥4
{𝜃𝑦4 }
where,
4 4 4

𝑤 = ∑ 𝑁𝑖 𝑤𝑖 𝜃𝑥 = ∑ 𝑁𝑖 𝜃𝑥𝑖 𝜃𝑦 = ∑ 𝑁𝑖 𝜃𝑦𝑖 Eq 339

𝑖=1 𝑖=1 𝑖=1

Curvature-Displacement, Shear-Strain & Moment-Curvature Relationships

Thus, according to the Mindlin plate theory, the deformations u and v parallel to the undeformed
mid-surface, at a distance z from the centroidal axis are given by:

𝑢 = 𝑧𝜃𝑦 𝑣 = −𝑧𝜃𝑥 Eq 340

Where θx and θy are the rotations of the normal line initially perpendicular to the mid-surface of the
plate with respect to the x and y axes.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 111

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The curvatures are given by:

𝜕𝜃𝑦 𝜕𝜃𝑥 𝜕𝜃𝑦 𝜕𝜃𝑥
𝜒𝑥 = 𝜒𝑦 = − 𝜒𝑥𝑦 = − Eq 341
𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥

Similarly to the thin-plate, the moments and the curvatures are related by Eq 272 & Eq 273. So, the
bending effect in the thick-plate of thickness t measured in the z-direction is written:
1 𝜈 0 𝜒𝑥
𝑀𝑥 𝐸𝑡 3 𝜈 1 0 𝜒
{ 𝑀𝑦 } = 2
[ 1 − 𝜈] { 𝑦 } Eq 342
𝑀𝑥𝑦 12(1 − 𝜈 ) 𝜒𝑥𝑦
0 0
2
Or,
{𝑀} = [𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 {𝜒} Eq 343

So, since the transverse shear effect is considered in the Mindlin assumption, the shear strains are
expressed by:

𝜕𝑤 𝜕𝑤
𝛾𝑥𝑧 = 𝜃𝑦 + 𝛾𝑦𝑧 = −𝜃𝑥 + Eq 344
𝜕𝑥 𝜕𝑦

As per the Theory of Elasticity, the shear effect in the thick plate is given by:
𝑄𝑥 𝐺𝑡𝜅 0 𝛾𝑥𝑧
{𝑄 } = [ ]{ } Eq 345
𝑦 0 𝐺𝑡𝜅 𝛾𝑦𝑧

Where G is the shear modulus given by:

𝐸
𝐺= Eq 346
2(1 + 𝜈)

In the Mindlin theory, the shear strain is assumed constant across the thickness of the plate.
However, it is known to be parabolic across the thickness of the plate. To account for the inaccuracy
in the shear strain, a shear correction factor  is applied so that the correct amount of internal energy
is predicted by the theory. The value of  is usually taken to be π2/12 or 5/6 (refer to a mechanic
book). Eq 345 is also written:

{𝑄} = [𝐷]𝑆ℎ𝑒𝑎𝑟 {𝛾} Eq 347

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 112

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So, the complete moment-shear-strain relationship for the thick-plate is:

𝑀𝑥
𝑀𝑦
𝑀𝑥𝑦
𝑄𝑥
{ 𝑄𝑦 }
𝜕𝜃𝑦
𝜕𝑥
1 𝜈 0 𝜕𝜃𝑥
0 0 − Eq 348
𝐸𝑡 3 𝜈 1 0 𝜕𝑦
[ 1 − 𝜈] 0 0
2
12(1 − 𝜈 ) 𝜕𝜃𝑦 𝜕𝜃𝑥
= 0 0 0 0 −
2 𝜕𝑦 𝜕𝑥
0 0 0 𝐸𝑡 𝜅 0 𝜕𝑤
[ ]
[ 0 0 0 2(1 + 𝜈) 0 𝜅 ] 𝜃𝑦 + 𝜕𝑥
𝜕𝑤
−𝜃𝑥 +
{ 𝜕𝑦 }
So,
1 𝜈 0 𝜒𝑥
𝑀𝑥 𝐸𝑡 3 0 0
𝜈 1 0 𝜒𝑦
𝑀𝑦 2
[ 1 − 𝜈] 0 0
12(1 − 𝜈 ) 𝜒𝑥𝑦
𝑀𝑥𝑦 = 0 0 0 0 Eq 349
2 𝛾𝑥𝑧
𝑄𝑥 𝐸𝑡
0 0 0 𝜅 0 𝛾
{ 𝑄𝑦 } [ ] { 𝑦𝑧 }
[ 0 0 0 2(1 + 𝜈) 0 𝜅]

Or,
{𝑀} [𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0] {𝜒}
{ }=[ ]{ } Eq 350
{𝑄} [0] [𝐷]𝑆ℎ𝑒𝑎𝑟 {𝛾}

{𝑀} [𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0]

{ }=[ ] {𝜀} Eq 351
{𝑄} [0] [𝐷]𝑆ℎ𝑒𝑎𝑟

Where,
𝜒𝑥
𝜒𝑦 𝑤𝑖
{𝜀} = 𝜒𝑥𝑦 = [𝐵]{𝛿𝑖 } = [𝐵] {𝜃𝑥𝑖 } Eq 352
𝛾𝑥𝑧 𝜃𝑦𝑖
{ 𝛾𝑦𝑧 }

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 113

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The constitutive matrix is then:

1 𝜈 0
𝐸𝑡 3 0 0
𝜈 1 0
[ 1−𝜈 ] 0 0
[𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0] 2
12(1 − 𝜈 )
[ ]= 0 0 0 0 Eq 353
[0] [𝐷]𝑆ℎ𝑒𝑎𝑟 2
0 0 0 𝐸𝑡 𝜅 0
[ ]
[ 0 0 0 2(1 + 𝜈) 0 𝜅]

With:
• [B] the strain displacement matrix
• {} the nodal displacement vector
By substituting Eq 352 into Eq 351:
{𝑀} [𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0]
{ }=[ ] [𝐵]{𝛿𝑖 } Eq 354
{𝑄} [0] [𝐷]𝑆ℎ𝑒𝑎𝑟

To evaluate [B], using Eq 339, Eq 341 and Eq 344. the curvatures and the shear strains are:
4
𝜕𝑁𝑖
4 4 𝜒𝑥𝑦 = ∑ 𝜃𝑦𝑖
𝜕𝑁𝑖 𝜕𝑁𝑖 𝜕𝑦
𝑖=1
𝜒𝑥 = ∑ 𝜃𝑦𝑖 𝜒𝑦 = ∑ −𝜃𝑥𝑖 4
𝜕𝑥 𝜕𝑦 𝜕𝑁𝑖
𝑖=1 𝑖=1
− ∑ 𝜃𝑥𝑖
𝜕𝑥 Eq 355
𝑖=1
4 4 4 4
𝜕𝑁𝑖 𝜕𝑁𝑖
𝛾𝑥𝑧 = ∑ 𝑤𝑖 + ∑ 𝜃𝑦𝑖 𝑁𝑖 𝛾𝑦𝑧 = ∑ 𝑤𝑖 − ∑ 𝜃𝑥𝑖 𝑁𝑖
𝜕𝑥 𝜕𝑦
𝑖=1 𝑖=1 𝑖=1 𝑖=1

The Eq 355 written in matrix form is:

𝜕𝑁𝑖
0 0
𝜕𝑥
𝜕𝑁𝑖
𝜒𝑥 0 − 0
𝜕𝑦
𝜒𝑦 𝑤𝑖
𝜕𝑁𝑖 𝜕𝑁𝑖
𝜒𝑥𝑦 = 0 − 𝜃
{ 𝑥𝑖 }
𝜕𝑥 𝜕𝑦 𝜃 Eq 356
𝛾𝑥𝑧 𝑦𝑖
𝛾
{ 𝑦𝑧 } 𝜕𝑁𝑖
0 𝑁𝑖
𝜕𝑥
𝜕𝑁𝑖
−𝑁𝑖 0
[ 𝜕𝑦 ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 114

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝜒𝑥
𝜒𝑦 𝑤𝑖
{𝜀} = 𝜒𝑥𝑦 = [𝐵] {𝜃𝑥𝑖 }
𝛾𝑥𝑧 𝜃𝑦𝑖
{ 𝛾𝑦𝑧 }

From Eq 356, we obtain the 5 x 3 strain displacement matrix [B]:

𝜕𝑁𝑖
0 0
𝜕𝑥
𝜕𝑁𝑖
0 − 0
𝜕𝑦
𝜕𝑁𝑖 𝜕𝑁𝑖
[𝐵] = 0 − Eq 357
𝜕𝑥 𝜕𝑦
𝜕𝑁𝑖
0 𝑁𝑖
𝜕𝑥
𝜕𝑁𝑖
−𝑁𝑖 0
[ 𝜕𝑦 ]

Then, for the four nodes quadrilateral element, using the nodal displacement vector from Eq 336,
the strain displacement matrix becomes the following 5 x 12 matrix:

[𝐵] = [[𝐵𝑗 ] [𝐵𝑘 ] [𝐵𝑚 ] [𝐵𝑛 ]] Eq 358

Where each [Bi] for i = j, k, m, n is the 5 x 3 matrix given by Eq 357. [B] is then written as:
[𝐵] =
𝜕𝑁𝑗 𝜕𝑁𝑘 𝜕𝑁𝑚 𝜕𝑁𝑛
0 0 0 0 0 0 0 0
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝑁𝑗 𝜕𝑁𝑘 𝜕𝑁𝑚 𝜕𝑁𝑛
0 − 0 0 − 0 0 − 0 0 − 0
𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑦
𝜕𝑁𝑗 𝜕𝑁𝑗 𝜕𝑁𝑘 𝜕𝑁𝑘 𝜕𝑁𝑚 𝜕𝑁𝑚 𝜕𝑁𝑛 𝜕𝑁𝑛 Eq
0 − 0 − 0 − 0 − 359
𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦
𝜕𝑁𝑗 𝜕𝑁𝑘 𝜕𝑁𝑚 𝜕𝑁𝑛
0 𝑁𝑗 0 𝑁𝑘 0 𝑁𝑚 0 𝑁𝑛
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝑁𝑗 𝜕𝑁𝑘 𝜕𝑁𝑚 𝜕𝑁𝑛
−𝑁𝑗 0 −𝑁𝑘 0 −𝑁𝑚 0 −𝑁𝑛 0
[ 𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑦 ]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 115

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So, using Eq 354, the bending and shear terms are:

[𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0] [𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0]

[ ] [𝐵] = [ ] [[𝐵𝑗 ] [𝐵𝑘 ] [𝐵𝑚 ] [𝐵𝑛 ]] Eq 360
[0] [𝐷]𝑆ℎ𝑒𝑎𝑟 [0] [𝐷]𝑆ℎ𝑒𝑎𝑟

For i = j, k, m, n, the ith submatrice in the Eq 360 is:

−𝜈𝑡 2 𝜕𝑁𝑖 𝑡 2 𝜕𝑁𝑖
0
1 − 𝜈 𝜕𝑦 1 − 𝜈 𝜕𝑥
−𝑡 2 𝜕𝑁𝑖 𝜈𝑡 2 𝜕𝑁𝑖
0
1 − 𝜈 𝜕𝑥 1 − 𝜈 𝜕𝑦
[𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0] 𝐸𝑡 −𝑡 2 𝜕𝑁𝑖 𝑡 2 𝜕𝑁𝑖
[ ] [𝐵𝑖 ] = 0 Eq 361
[0] [𝐷]𝑆ℎ𝑒𝑎𝑟 12(1 + 𝜈) 2 𝜕𝑥 2 𝜕𝑦
𝜕𝑁𝑖
6𝜅 0 6𝜅𝑁𝑖
𝜕𝑥
𝜕𝑁𝑖
6𝜅 −6𝜅𝑁𝑖 0
[ 𝜕𝑦 ]

So, the bending and shear terms in Eq 361 can be separated and written as:
[𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0]
[ ] [𝐵𝑖 ] =
[0] [𝐷]𝑆ℎ𝑒𝑎𝑟
−𝜈𝑡 2 𝜕𝑁𝑖 𝑡 2 𝜕𝑁𝑖 0 0 0
0 0 0 0
1 − 𝜈 𝜕𝑦 1 − 𝜈 𝜕𝑥
0 0 0
𝐸𝑡 −𝑡 2 𝜕𝑁𝑖 𝜈𝑡 2 𝜕𝑁𝑖 𝜕𝑁
= 0 + 6𝜅 𝑖
0 𝑁𝑖
12(1 + 𝜈) 1 − 𝜈 𝜕𝑥 1 − 𝜈 𝜕𝑦 𝜕𝑥
−𝑡 2 𝜕𝑁𝑖 𝑡 2 𝜕𝑁𝑖 𝜕𝑁𝑖
0 −𝑁𝑖 0
[[ 2 𝜕𝑥 2 𝜕𝑦 ] [ 𝜕𝑦 ]]

Eq 362
Element Stiffness Matrix
The stiffness matrix of the four nodes thick-plate element is calculated using the usual form:
[𝐷] [0]
[𝐾] = ∬[𝐵]𝑇 [ 𝐵𝑒𝑛𝑑𝑖𝑛𝑔 ] [𝐵]𝑑𝑥𝑑𝑦 Eq 363
[0] [𝐷]𝑆ℎ𝑒𝑎𝑟

Where [B]T[D]{B} is computed as:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 116

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝜕𝑁𝑖 𝜕𝑁𝑖
0 0 0
𝜕𝑥 𝜕𝑦
𝜕𝑁𝑖 𝜕𝑁𝑖
[𝑘̅] = 0 − − 0 −𝑁𝑖
𝜕𝑦 𝜕𝑥
𝜕𝑁𝑖 𝜕𝑁𝑖
0 𝑁𝑖 0
[ 𝜕𝑥 𝜕𝑦 ]𝑖=𝑗,𝑘,𝑚,𝑛

𝜕𝑁𝑖
0 0
𝜕𝑥
1 𝜈 0 𝜕𝑁𝑖
3 0 0 0 − 0
𝐸𝑡 𝜈 1 0 𝜕𝑦
[ 1 − 𝜈] 0 0
2
12(1 − 𝜈 ) 𝜕𝑁𝑖 𝜕𝑁𝑖
x 0 0 0 0 x 0 −
2 𝜕𝑥 𝜕𝑦
0 0 0 𝐸𝑡 𝜅 0 𝜕𝑁𝑖
[ ] 0 𝑁𝑖
[ 0 0 0 2(1 + 𝜈) 0 𝜅] 𝜕𝑥
𝜕𝑁𝑖
−𝑁𝑖 0
[ 𝜕𝑦 ]𝑖=𝑗,𝑘,𝑚,𝑛

Eq 364
So, [𝑘̅] is a 12 x 12 matrix:
[𝑘̅11 ] [𝑘̅11 ] [𝑘̅13 ] [𝑘̅14 ]
[𝐷]𝐵𝑒𝑛𝑑𝑖𝑛𝑔 [0] [𝑘̅21 ] [𝑘̅22 ] [𝑘̅23 ] [𝑘̅24 ]
[𝑘̅] = [𝐵]𝑇 [ ] [𝐵] = Eq 365
[0] [𝐷]𝑆ℎ𝑒𝑎𝑟 [𝑘̅31 ] [𝑘̅32 ] [𝑘̅33 ] [𝑘̅34 ]
[[𝑘̅41 ] [𝑘̅42 ] [𝑘̅43 ] [𝑘̅44 ]]

Where each [𝑘̅𝛼𝛽 ] is a 3 x 3 matrix defined as:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 117

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝐸𝑡
[𝑘̅𝛼𝛽 ] =
12(1 + 𝜈)
𝜕𝑁𝛼 𝜕𝑁𝛽 𝜕𝑁𝛼 𝜕𝑁𝛽 𝜕𝑁𝛼 𝜕𝑁𝛼
6𝜅 ( + ) −6𝜅 𝑁 6𝜅 𝑁
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑦 𝛽 𝜕𝑥 𝛽
𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽 𝜈𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽
( ) − ( )
𝜕𝑁𝛼 1 − 𝜈 𝜕𝑦 𝜕𝑦 1 − 𝜈 𝜕𝑦 𝜕𝑥
−6𝜅 𝑁
𝜕𝑦 𝛽 𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽 𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽
𝑋 + ( ) + 6𝜅𝑁𝛼 𝑁𝛽 − ( )
( 2 𝜕𝑥 𝜕𝑥 ) ( 2 𝜕𝑥 𝜕𝑦 )
𝜈𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽 𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽
− ( ) ( )
𝜕𝑁𝛼 1 − 𝜈 𝜕𝑥 𝜕𝑦 1 − 𝜈 𝜕𝑥 𝜕𝑥
6𝜅 𝑁
𝜕𝑥 𝛽 𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽 𝑡 2 𝜕𝑁𝛼 𝜕𝑁𝛽
− ( ) + ( ) + 6𝜅𝑁𝛼 𝑁𝛽
[ ( 2 𝜕𝑦 𝜕𝑥 ) ( 2 𝜕𝑦 𝜕𝑦 )]
Eq 366

So, the stiffness matrix of the four nodes thick-plate element is calculated as:

[𝐾] = ∬[𝑘̅]𝑑𝑥𝑑𝑦 Eq 367

To express the above equation in terms of the natural coordinates <,> in the isoparametric
formulation, since [B] is a function of  and , the integration with respect to  and  is possible using
the determinant of the Jacobian matrix to transform the x and y variables into the <,> coordinates.
+1 +1
[𝐾] = ∫ ∫ [𝑘̅]|𝐽|𝑑𝜉𝑑𝜂 Eq 368
−1 −1

The integral can be easily computed numerically by the FE solver.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 118

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.5.7. Shell
Definition
The shell element can carry loads in almost all directions: it combines the membrane and plate
elements capabilities. The shell element undergoes in-plane, bending and twisting deformations. It
can be used to model large structures like the roof of a building, cylindrical and hemispherical tanks,
fuselage of aircraft…
In order to derive the stiffness matrix of the shell element we can use the usual method presented
for membranes and plates in previous paragraphs which consist in defining the shape functions, the
strain-displacement and stress-strain relationships and then obtaining the element matrix. However,
since the shell element combines the properties of the membrane and the plate elements previously
discussed in detail, we will use a short cut method to derive the shell element stiffness matrix.
Degrees of Freedom
Let’s consider the degrees of freedom of the membrane and the plate elements in a local element
coordinate system:

Shell Element = Membrane Element + Plate Element

5 DOF’s per node 2 DOF’s per node 3 DOF’s per node

z 3

y
2
x
4 w w
= +
v v
u 1 u 
  y
 y x
x
FIG 50 – DEGREES OF FREEDOM OF SHELL, MEMBRANE AND PLATE ELEMENT

The nodal displacement vector at node i (for i = 1, 2, 3, 4) is:

𝑢𝑖
𝑣𝑖
{𝛿𝑖 } = 𝑤𝑖 Eq 369
𝜃𝑥𝑖
{𝜃𝑦𝑖 }

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 119

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The total quadrilateral shell element displacement vector is composed with 5 x 4 = 20 degrees of
freedom:
𝛿1
𝛿
{𝛿} = { 2 } Eq 370
𝛿3
𝛿4

Element Stiffness Matrix

The stiffness matrix for the four nodes membrane element corresponding to the u and v DOF’s is a 8
x 8 matrix:
Node 1 Node 2 Node 3 Node 4
𝑚 𝑚 𝑚 𝑚
𝑘11 𝑘12 𝑘13 𝑘14

𝑚 𝑚 𝑚 𝑚
𝑘21 𝑘22 𝑘23 𝑘24
[𝑘𝑒 ]𝑀𝑒𝑚𝑏𝑟𝑎𝑛𝑒 = Eq 371
𝑚 𝑚 𝑚 𝑚
𝑘31 𝑘32 𝑘33 𝑘34

𝑚 𝑚 𝑚 𝑚
[𝑘41 𝑘42 𝑘43 𝑘44 ]

𝑚
Where each submatrice [𝑘𝑖𝑗 ] is a 2 x 2 membrane element matrix.

The stiffness matrix for the four nodes plate element corresponding to the w, x and y DOF’s is a
12 x 12 matrix:
Node 1 Node 2 Node 3 Node 4
𝑝 𝑝 𝑝 𝑝
𝑘11 𝑘12 𝑘13 𝑘14

𝑝 𝑝 𝑝 𝑝
𝑘21 𝑘22 𝑘23 𝑘24
[𝑘𝑒 ]𝑃𝑙𝑎𝑡𝑒 = Eq 372
𝑝 𝑝 𝑝 𝑝
𝑘31 𝑘32 𝑘33 𝑘34

𝑝 𝑝 𝑝 𝑝
[𝑘41 𝑘42 𝑘43 𝑘44 ]

𝑝
Where each submatrice [𝑘𝑖𝑗 ] is a 3 x 3 plate element matrix. So, the stiffness matrix of the shell
element is written as a 20 x 20 matrix:
[𝑘𝑒 ]𝑀𝑒𝑚𝑏𝑟𝑎𝑛𝑒 0
[𝑘𝑒 ]𝑆ℎ𝑒𝑙𝑙 = [ ] Eq 373
0 [𝑘𝑒 ]𝑃𝑙𝑎𝑡𝑒

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 120

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

However, this 20 x 20 matrix is not convenient for the matrix transformation process which consists
in transforming the matrix from the local coordinate system to the global coordinate system. So, an
extended 24 x 24 matrix including the z degree of freedom is used as shown below. Note that the
stiffness corresponding to z are zero. From a practical point of view, the stiffness corresponding to
z is set to a small value using a specific parameter for the shell elements, not a null value, to avoid
singularities and obtain a solution.
Node 1 Node 2 Node 3 Node 4

𝑚 𝑚 𝑚 𝑚
𝑘11 0 0 𝑘12 0 0 𝑘13 0 0 𝑘14 0 0
𝑝 𝑝 𝑝 𝑝
0 𝑘11 0 0 𝑘12 0 0 𝑘13 0 0 𝑘14 0
0 0 0 0 0 0 0 0 0 0 0 0
𝑚 𝑚 𝑚 𝑚
𝑘21 0 0 𝑘22 0 0 𝑘23 0 0 𝑘24 0 0
𝑝 𝑝 𝑝 𝑝
0 𝑘21 0 0 𝑘22 0 0 𝑘23 0 0 𝑘24 0
[𝑘𝑒 ]𝑆ℎ𝑒𝑙𝑙 0 0 0 0 0 0 0 0 0 0 0 0
= 𝑚 𝑚 𝑚 𝑚 Eq 374
𝑘31 0 0 𝑘32 0 0 𝑘33 0 0 𝑘34 0 0
𝑝 𝑝 𝑝 𝑝
0 𝑘31 0 0 𝑘32 0 0 𝑘33 0 0 𝑘34 0
0 0 0 0 0 0 0 0 0 0 0 0
𝑚 𝑚 𝑚 𝑚
𝑘41 0 0 𝑘42 0 0 𝑘43 0 0 𝑘44 0 0
𝑝 𝑝 𝑝 𝑝
0 𝑘41 0 0 𝑘42 0 0 𝑘43 0 0 𝑘43 0
[ 0 0 0 0 0 0 0 0 0 0 0 0]

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 2D ELEMENTS 121

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.6. 3D SOLID ELEMENT

Let’s consider the three-dimensional element, also called the solid element. This element is used to
analyze general 3D bodies like dams, thick wall structures, solid forging parts like fittings or engine
parts. As mentioned in chapter about “The Library of Elements” in “Practical Finite Element Analysis
for Mechanical Engineers”, the basic solid elements are the tetrahedron with 4 or 10 nodes. A more
advanced solid element is the hexahedron, also called the brick element. In this paragraph, the
stiffness matrix is derived for the 8 nodes and the 20 nodes hexahedron elements. The isoparametric
formulation presented for a 2D element page 98 is used.

5.6.1. Linear Hexahedral Element

Degrees-of-freedom
The linear hexahedral element has eight corner nodes. Each node has three translational degrees of
freedom u, v, w. So, the linear hexahedral element has 8 x 3 = 24 degrees of freedom.

(x8,y8,z8)
8

(x4,y4,z4)
4 7
z (x7,y7,z7)
(x3,y3,z3)
w 3
5
y (x5,y5,z5)
v
1
(x1,y1,z1) 6
x u (x6,y6,z6)

2
(x2,y2,z2)

FIG 51 – DEGREES OF FREEDOM OF THE LINEAR HEXAHEDRAL ELEMENT

The nodal displacement vector at node i (for i = 1 to 8) is:

𝑢𝑖
{𝛿𝑖 } = { 𝑣𝑖 } Eq 375
𝑤𝑖
The total linear hexahedral element displacement vector is composed with 8 x 3 = 24 degrees of
freedom:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 122

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝛿1
𝛿2
𝛿3
𝛿
{𝛿} = 4 Eq 376
𝛿5
𝛿6
𝛿7
{𝛿8 }

Shape Functions
The isoparametric natural coordinates <,,> is used to define the shape functions and to facilitate
the stiffness matrix derivation. The element faces are defined by =±1, =±1, =±1 as shown in Fig
52.

8
4
z 7
3

y

1 5
x
6
2 

FIG 52 –LINEAR HEXAHEDRAL ELEMENT IN THE <, ,> NATURAL COORDINATE SYSTEM

In the natural coordinate system <,,>, the corner nodes have the following coordinates:
TABLE 3 – NODES NATURAL COORDINATES
Node 0 0 0
1 -1 -1 -1
2 1 -1 -1
3 1 -1 1
4 -1 -1 1
5 -1 1 -1
6 1 1 -1
7 1 1 1
8 -1 1 1

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 123

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So, we need 24 coefficients to define the functions used to describe the element geometry for x, y
and z in terms of the generalized degrees of freedom:
𝑥 = 𝑎1 + 𝑎2 𝜉 + 𝑎3 𝜂 + 𝑎4 𝜓 + 𝑎5 𝜉𝜂 + 𝑎6 𝜂𝜓 + 𝑎7 𝜓𝜉 + 𝑎8 𝜉𝜂𝜓
𝑦 = 𝑎9 + 𝑎10 𝜉 + 𝑎11 𝜂 + 𝑎12 𝜓 + 𝑎13 𝜉𝜂 + 𝑎14 𝜂𝜓 + 𝑎15 𝜓𝜉 + 𝑎16 𝜉𝜂𝜓
Eq 377
𝑧 = 𝑎17 + 𝑎18 𝜉 + 𝑎19 𝜂 + 𝑎20 𝜓 + 𝑎21 𝜉𝜂 + 𝑎22 𝜂𝜓 + 𝑎23 𝜓𝜉
+ 𝑎24 𝜉𝜂𝜓

So, same as the Eq 304 extended to include the z coordinate:

𝑥 8 𝑁𝑖 0 0 𝑥𝑖
{𝑦} = ∑ ([ 0 𝑁𝑖 0 ] {𝑦𝑖 }) Eq 378
𝑧 𝑖=1 0 0 𝑁𝑖 𝑧𝑖

Where Ni are the shapes functions for i=1 to 8, with 0 = ±1, 0 = ±1, 0 = ±1 as per Table 3:
1
𝑁𝑖 = (1 + 𝜉𝜉0 )(1 + 𝜂𝜂0 )(1 + 𝜓𝜓0 ) Eq 379
8

The displacement functions in terms of the generalized degrees of freedom are of the same form as
the functions used to describe the element geometry by Eq 377. So, the displacement functions
including u, v, w is written:
𝑢 8 𝑁𝑖 0 0 𝑢𝑖
{ 𝑣 } = ∑ ([ 0 𝑁𝑖 0 ] { 𝑣𝑖 }) Eq 380
𝑤 𝑖=1 0 0 𝑁𝑖 𝑤𝑖

The shape function matrix is a 3 x 24 matrix.

Strain-Displacement & Stress-Strain Relationships
Let’s consider the three-dimensional infinitesimal element (dx,dy,dz) in a cartesian coordinate
system as shown in Fig 53. It represents the state of stress on three perpendicular faces of a body.
• The normal stresses X, Y and Z are perpendicular to the faces.
• The shear stresses act in the plane of each face: XY = YX, YZ = ZY, ZX = XZ

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 124

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

z,w

Z

ZY
ZX
YZ
XZ
Y YX XY
y,v dz
X
dx
dy x,u

FIG 53 –STATE OF STRESS ON A SOLID ELEMENT

The element strain-displacement relationships are (u, v, w are the displacements associated with the
x, y, z directions):
𝜕𝑢 𝜕𝑣 𝜕𝑤
𝜀𝑥 = 𝜀𝑦 = 𝜀𝑧 = Eq 381
𝜕𝑥 𝜕𝑦 𝜕𝑧

The shear strains are:

𝜕𝑢 𝜕𝑣 𝜕𝑣 𝜕𝑤 𝜕𝑤 𝜕𝑢
𝛾𝑥𝑦 = 𝛾𝑦𝑥 = + 𝛾𝑦𝑧 = 𝛾𝑧𝑦 = + 𝛾𝑧𝑥 = 𝛾𝑥𝑧 = + Eq 382
𝜕𝑦 𝜕𝑥 𝜕𝑧 𝜕𝑦 𝜕𝑥 𝜕𝑧

The stress-strain relationships for an isotropic material are:

{𝜎} = [𝐷]{𝜀} Eq 383

Where {} is the stress vector, {} is the strain vector and [D] is the constitutive matrix:

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 125

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝜎𝑥
𝜎𝑦
𝜎𝑧
𝜏𝑥𝑦
𝜏𝑦𝑧
{ 𝜏𝑧𝑥 }
1−𝜈 𝜈 𝜈 0 0 0
1−𝜈 𝜈 0 0 0 Eq 384
𝜀𝑥
1−𝜈 0 0 0 𝜀𝑦
1 − 2𝜈
𝐸 0 0 𝜀𝑧
= 2 𝛾𝑥𝑦
(1 + 𝜈)(1 − 2𝜈) 1 − 2𝜈
0 𝛾𝑦𝑧
2 { 𝛾𝑧𝑥 }
1 − 2𝜈
[ 𝑆 𝑌 𝑀
2 ]

To define the strain-displacement matrix [B], Eq 381 & Eq 382 are used. These relations are expressed
in terms of global coordinates, including a differentiation with respect to z. So, the Jacobian matrix
presented for the isoparametric formulation of the four nodes quadrilateral element (see Eq 313)
needs to be expanded:
𝜕𝑥 𝜕𝑦 𝜕𝑧
𝜕𝜉 𝜕𝜉 𝜕𝜉
𝜕𝑥 𝜕𝑦 𝜕𝑧
[𝐽] = Eq 385
𝜕𝜂 𝜕𝜂 𝜕𝜂
𝜕𝑥 𝜕𝑦 𝜕𝑧
[𝜕𝜓 𝜕𝜓 𝜕𝜓]

So, Eq 312 is also expanded:

𝜕𝑗 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑗 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑗
𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉 𝜕𝜉
| 𝜕𝑗 𝜕𝑦 𝜕𝑧 | | 𝜕𝑥 𝜕𝑗 𝜕𝑧 | | 𝜕𝑥 𝜕𝑦 𝜕𝑗 |
| 𝜕𝜂 𝜕𝜂 𝜕𝜂 | | 𝜕𝜂 𝜕𝜂 𝜕𝜂 | | 𝜕𝜂 𝜕𝜂 𝜕𝜂 | Eq 386
𝜕𝑗 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑗 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑗
𝜕𝑗 𝜕𝜓 𝜕𝜓 𝜕𝜓 𝜕𝑗 𝜕𝜓 𝜕𝜓 𝜕𝜓 𝜕𝑗 𝜕𝜓 𝜕𝜓 𝜕𝜓
= = =
𝜕𝑥 |𝐽| 𝜕𝑦 |𝐽| 𝜕𝑧 |𝐽|

So, from Eq 386:

𝜕( ) 1 𝜕( ) 𝜕𝑦 𝜕𝑧 𝜕𝑦 𝜕𝑧 𝜕𝑦 𝜕( ) 𝜕𝑧 𝜕( ) 𝜕𝑧 𝜕𝑧 𝜕( ) 𝜕𝑦 𝜕( ) 𝜕𝑦
= [ ( − )− ( − )+ ( − )] Eq 387
𝜕𝑥 |𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 126

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝜕( ) 1 𝜕𝑥 𝜕( ) 𝜕𝑧 𝜕( ) 𝜕𝑧 𝜕( ) 𝜕𝑥 𝜕𝑧 𝜕𝑥 𝜕𝑧 𝜕𝑧 𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( )
= [ ( − )− ( − )+ ( − )]
𝜕𝑦 |𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂

𝜕( ) 1 𝜕𝑥 𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕( ) 𝜕𝑦 𝜕𝑥 𝜕( ) 𝜕𝑥 𝜕( ) 𝜕( ) 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦
= [ ( − )− ( − )+ ( − )]
𝜕𝑧 |𝐽| 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂 𝜕𝜉 𝜕𝜂 𝜕𝜓 𝜕𝜓 𝜕𝜂

So, writing the strain-displacement equations Eq 381 & Eq 382 in matrix form:
𝜕( )
0 0
𝜕𝑥
𝜕( )
0 0
𝜀𝑥 𝜕𝑦
𝜀𝑦 𝜕( )
𝜀𝑧 0 0 𝑢
𝜕𝑧
𝛾𝑥𝑦 = 𝜕( ) 𝜕( ) {𝑣 } Eq 388
𝛾𝑦𝑧 0 𝑤
𝜕𝑦 𝜕𝑥
{ 𝛾𝑧𝑥 } 𝜕( ) 𝜕( )
0
𝜕𝑧 𝜕𝑦
𝜕( ) 𝜕( )
[ 𝜕𝑧 0
𝜕𝑥 ]

So, the operator matrix [] is defined by substituting Eq 387 into Eq 388 and the strains expressed
in term of the natural coordinates <,,> are obtained. Then, Eq 388 is expressed in terms of the
shape functions in compact matrix form:

{𝜀} = [Ψ][𝑁]{𝛿} Eq 389

So, the strain-displacement matrix is given by:

[𝐵] = [Ψ][𝑁] Eq 390

Where,
• [] is the 6 x 3 operator matrix from Eq 388
• [N] is the 3 x 24 shape function matrix from Eq 379 & Eq 380

[B] is now a function of , and . Note that [B] is of order 6 x 24.
Element Stiffness Matrix
The stiffness matrix of the linear hexahedral element is calculated using the usual equation:
+1 +1 +1
[𝑘] = ∫ ∫ ∫ [𝐵]𝑇 [𝐷][𝐵]|𝐽|𝑑𝜉𝑑𝜂𝑑𝜓 Eq 391
−1 −1 −1

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 127

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Where,
• [B] is computed by Eq 390
• [D] is computed by Eq 384
• [k] is a 24 x 24 matrix
Again, it is best to evaluate [k] by numerical integration using a FE solver.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 128

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

5.6.2. Quadratic Hexahedral Element

Degrees of freedom
The quadratic hexahedral element has eight corner nodes and twelve mid-side nodes for a total of
20 nodes (see Fig 54). Like for the linear hexahedral element, each node has three translational
degrees of freedom u, v, w. So, the linear hexahedral element has 20 x 3 = 60 degrees of freedom.

7
19
15
3
14 11
8
20
4 10
16 6
13 18 2 
12
5 9
17
1


FIG 54 – DEGREES OF FREEDOM OF THE QUADRATIC HEXAHEDRAL ELEMENT

Shape Functions

So, a total of 60 coefficients are needed to define the functions used to describe the element
geometry for x, y and z in terms of the generalized degrees of freedom:
𝑥 = 𝑎1 + 𝑎2 𝜉 + 𝑎3 𝜂 + 𝑎4 𝜓 + 𝑎5 𝜉𝜂 + 𝑎6 𝜂𝜓 + 𝑎7 𝜓𝜉 + 𝑎8 𝜉 2 + 𝑎9 𝜂2 + 𝑎10 𝜓2
+ 𝑎11 𝜉 2 𝜂 + 𝑎12 𝜉𝜂2 + 𝑎13 𝜂2 𝜓 + 𝑎14 𝜂𝜓 2 + 𝑎15 𝜓2 𝜉 + 𝑎16 𝜓𝜉 2
+ 𝑎17 𝜉𝜂𝜓 + 𝑎18 𝜉 2 𝜂𝜓 + 𝑎19 𝜉𝜂2 𝜓 + 𝑎20 𝜉𝜂𝜓2
𝑦 = 𝑎21 + 𝑎22 𝜉 + 𝑎23 𝜂 + 𝑎24 𝜓 + 𝑎25 𝜉𝜂 + 𝑎26 𝜂𝜓 + 𝑎27 𝜓𝜉 + 𝑎28 𝜉 2 + 𝑎29 𝜂2 Eq 392
+ 𝑎30 𝜓2
+ 𝑎31 𝜉 2 𝜂 + 𝑎32 𝜉𝜂2 + 𝑎33 𝜂2 𝜓 + 𝑎34 𝜂𝜓2 + 𝑎35 𝜓2 𝜉 + 𝑎36 𝜓𝜉 2
+ 𝑎37 𝜉𝜂𝜓 + 𝑎38 𝜉 2 𝜂𝜓 + 𝑎39 𝜉𝜂2 𝜓 + 𝑎40 𝜉𝜂𝜓2

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 129

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

𝑧 = 𝑎41 + 𝑎42 𝜉 + 𝑎43 𝜂 + 𝑎44 𝜓 + 𝑎45 𝜉𝜂 + 𝑎46 𝜂𝜓 + 𝑎47 𝜓𝜉 + 𝑎48 𝜉 2 + 𝑎49 𝜂2

+ 𝑎50 𝜓2
+ 𝑎51 𝜉 2 𝜂 + 𝑎52 𝜉𝜂2 + 𝑎53 𝜂2 𝜓 + 𝑎54 𝜂𝜓2 + 𝑎55 𝜓 2 𝜉 + 𝑎56 𝜓𝜉 2
+ 𝑎57 𝜉𝜂𝜓 + 𝑎58 𝜉 2 𝜂𝜓 + 𝑎59 𝜉𝜂2 𝜓 + 𝑎60 𝜉𝜂𝜓2

In the natural coordinate system <,,>, the corner and mid-side nodes have the following
coordinates:
TABLE 4 – NODES NATURAL COORDINATES
Node 0 0 0
1 +1 -1 -1
2 +1 -1 -1
3 +1 +1 -1
4 +1 +1 +1
5 -1 -1 +1
6 -1 -1 -1
7 -1 +1 -1
8 -1 +1 +1
9 +1 -1 0
10 +1 0 -1
11 +1 +1 0
12 +1 0 +1
13 -1 -1 0
14 -1 0 -1
15 -1 +1 0
16 -1 0 +1
17 0 -1 +1
18 0 -1 -1
19 0 +1 -1
20 0 +1 +1

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 130

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

So, the shape functions are given by:

For the corner 1

𝑁𝑖 = [(1 + 𝜉𝜉0 )(1 + 𝜂𝜂0 )(1 + 𝜓𝜓0 )][𝜉𝜉0 + 𝜂𝜂0 + 𝜓𝜓0
nodes 8 Eq 393
i = 1 to 8 − 2]

For the mid-side 1

nodes 𝑁𝑖 = [(1 + 𝜉𝜉0 )(1 + 𝜂𝜂0 )(1 + 𝜓2 )] Eq 394
4
i = 9, 11, 13, 15

For the mid-side 1

nodes 𝑁𝑖 = [(1 + 𝜉𝜉0 )(1 + 𝜂2 )(1 + 𝜓𝜓0 )] Eq 395
4
i = 10, 12, 14, 16

For the mid-side 1

nodes 𝑁𝑖 = [(1 + 𝜉 2 )(1 + 𝜂𝜂0 )(1 + 𝜓𝜓0 )] Eq 396
4
i = 17 to 20

Element Stiffness Matrix

The derivation of the stiffness matrix for the quadratic hexahedral element follows the same steps
presented for the linear hexahedral element except that the shape functions from Eq 393 to Eq 396
are used.
The stiffness matrix is computed numerically using Eq 391 and is a 60 x 60 matrix.

ELEMENT STIFFNESS MATRIX FOR VARIOUS TOPOLOGIES - 3D SOLID ELEMENT 131

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

6. HOW IS THE STIFFNESS MATRIX ASSEMBLED?

When computing an FEA with software, the actual assembly of the stiffness matrix is invisible to the
FEA analyst, but it is useful to know how this occurs.
The finite element model is a system composed of several elements connected at nodes. The global
stiffness matrix of the model is built by assembling each individual element’s stiffness matrix.
At the beginning of this chapter, the way the stiffness matrices of individual elements are computed,
based on element properties (degrees of freedom, geometry, and materials), was presented. This
section will explain how these elemental matrices are assembled into the global stiffness matrix for
the entire model.

6.1. MATRIX ASSEMBLY

The assembly of elements consists of placing each elemental matrix in an appropriate position in the
global matrix, based on degrees of freedom. The procedure is best understood through a simple
example.
Let’s consider a simple rectangle plate discretized using four constant strain triangular (CST)
elements as shown in Fig 55. Six nodes are involved in this discretization. Each CST element has two
degrees of freedom at each node, so the total number of degrees of freedom for the plate is 6 x 2 =
12. Then, the stiffness matrix for the plate meshed with four CST elements will have a size of 12 x 12.
The elements and the nodes are numbered as shown in Fig 55.

v1 v2 v3
Node Node Node
#1 u1 #2 u2 #3 u3
Element Element
v4 #2 v #35 v6
Element Element
#1 #4
u4 u5 u6
Node Node Node
#4 #5 #6

FIG 55 – SIMPLE RECTANGLE PLATE DISCRETIZED WITH FOUR CST ELEMENTS

The form of the stiffness matrix corresponding to the CST elements meshing is shown in Fig 56: two
columns and two lines per element for a total of twelve degrees of freedom for the rectangular plate.

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 132

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Node 1 Node 2 Node 3 Node 4 Node 5 Node 6

u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6
Node u1
1 v1
Node u2
2 v2
Node u3
3 v3
Node u4
4 v4
Node u5
5 v5
Node u6
6 v6
FIG 56 – STIFFNESS MATRIX FORM OF THE RECTANGLE PLATE MESHED WITH FOUR CST ELEMENTS

Let’s closely examine how this global stiffness matrix is populated, which consists of passing the elements of
the mesh one by one and identifying the degrees of freedom involved in the construction of the elemental
stiffness matrix.

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 133

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element 1: The element #1 is connected at the nodes #1, #4 & #5. So, the degrees of freedom
involved in the construction of the element #1 stiffness matrix are: (u1,v1), (u4,v4) & (u5,v5). Using the
6 x 6 stiffness matrix of the CST element derived in a previous paragraph (see Eq 238), we have the
following matrix for the element #1 (note that the nodes are considered anti-clock wise; this
approach will be used for all elements):

Node #1 Node #4 Node #5

u1 v1 u4 v4 u5 v5
𝑘11𝑥 𝑘11𝑥𝑦 𝑘14𝑥 𝑘14𝑥𝑦 𝑘15𝑥 𝑘15𝑥𝑦
[ ] [ ] [ ] u1
𝑘11𝑦𝑥 𝑘11𝑦 𝑘14𝑦𝑥 𝑘14𝑦 𝑘15𝑦𝑥 𝑘15𝑦
v1
𝑘41𝑥 𝑘41𝑥𝑦 𝑘44𝑥 𝑘44𝑥𝑦 𝑘45𝑥 𝑘45𝑥𝑦 u4
[𝑘]𝐸𝑙𝑒𝑚𝑒𝑛𝑡 #1 = [ ] [ ] [ ] Eq 397
𝑘41𝑦𝑥 𝑘41𝑦 𝑘44𝑦𝑥 𝑘44𝑦 𝑘45𝑦𝑥 𝑘45𝑦 v4
𝑘51𝑥 𝑘51𝑥𝑦 𝑘54𝑥 𝑘54𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦 u5
[ ] [ ] [ ] v5
[ 𝑘51𝑦𝑥 𝑘51𝑦 𝑘54𝑦𝑥 𝑘54𝑦 𝑘55𝑦𝑥 𝑘55𝑦 ]

Now, the key question is, at which location of the global stiffness matrix the element #1 matrix will
be located? The answer lies in the list of degrees of freedom involved with the element #1 matrix
formulation: (u1,v1), (u4,v4) & (u5,v5). The Fig 57 shows the elemental matrix Eq 397 written at the
appropriate locations of the global stiffness matrix based on the degrees of freedom of the element
#1.
Node 1 Node 2 Node 3 Node 4 Node 5 Node 6
u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6
𝑒1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1
Node u1 𝑘11𝑥 𝑘11𝑥𝑦 𝑘14𝑥 𝑘14𝑥𝑦 𝑘15𝑥 𝑘15𝑥𝑦
1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1
v1 𝑘11𝑦𝑥 𝑘11𝑦 𝑘14𝑦𝑥 𝑘14𝑦 𝑘15𝑦𝑥 𝑘15𝑦
Node u2
2
v2
Node u3
3
v3
𝑒1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1
Node u4 𝑘41𝑥 𝑘41𝑥𝑦 𝑘44𝑥 𝑘44𝑥𝑦 𝑘45𝑥 𝑘45𝑥𝑦
4 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1
v4 𝑘41𝑦𝑥 𝑘41𝑦 𝑘44𝑦𝑥 𝑘44𝑦 𝑘45𝑦𝑥 𝑘45𝑦
𝑒1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1
Node u5 𝑘51𝑥 𝑘51𝑥𝑦 𝑘54𝑥 𝑘54𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦
5 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1 𝑒1
v5 𝑘51𝑦𝑥 𝑘51𝑦 𝑘54𝑦𝑥 𝑘54𝑦 𝑘55𝑦𝑥 𝑘55𝑦
Node u6
6
v6
FIG 57 – GLOBAL STIFFNESS MATRIX POPULATED WITH ELEMENT #1 MATRIX

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 134

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element 2: The element #2 is connected at the nodes #1, #5 & #2 (again, note that the nodes are
considered anti-clock wise). So, the degrees of freedom involved in the construction of the element
#2 stiffness matrix are: (u1,v1), (u5,v5) & (u2,v2). Using the 6 x 6 stiffness matrix of the CST element
derived in a previous paragraph (see Eq 238), we have the following matrix for the element #2:

Node #1 Node #5 Node #2

u1 v1 u5 v5 u2 v2
𝑘11𝑥 𝑘11𝑥𝑦 𝑘15𝑥 𝑘15𝑥𝑦 𝑘12𝑥 𝑘12𝑥𝑦
[ ] [ ] [ ] u1
𝑘11𝑦𝑥 𝑘11𝑦 𝑘15𝑦𝑥 𝑘15𝑦 𝑘12𝑦𝑥 𝑘12𝑦
v1
𝑘51𝑥 𝑘51𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦 𝑘52𝑥 𝑘52𝑥𝑦 u5
[𝑘]𝐸𝑙𝑒𝑚𝑒𝑛𝑡 #2 = [ ] [ ] [ ] Eq 398
𝑘51𝑦𝑥 𝑘51𝑦 𝑘55𝑦𝑥 𝑘55𝑦 𝑘52𝑦𝑥 𝑘52𝑦 v5
𝑘21𝑥 𝑘21𝑥𝑦 𝑘25𝑥 𝑘25𝑥𝑦 𝑘22𝑥 𝑘22𝑥𝑦 u2
[ ] [ ] [ ] v2
[ 𝑘21𝑦𝑥 𝑘21𝑦 𝑘25𝑦𝑥 𝑘25𝑦 𝑘22𝑦𝑥 𝑘22𝑦 ]

The Fig 58 shows the elemental matrix Eq 398 written at the appropriate locations of the global
stiffness matrix based on the degrees of freedom of the element #2.
Node 1 Node 2 Node 3 Node 4 Node 5 Node 6
u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6
𝑒2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2
Node u1 𝑘11𝑥 𝑘11𝑥𝑦 𝑘12𝑥 𝑘12𝑥𝑦 𝑘15𝑥 𝑘15𝑥𝑦
1 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2
v1 𝑘11𝑦𝑥 𝑘11𝑦 𝑘12𝑦𝑥 𝑘12𝑦 𝑘15𝑦𝑥 𝑘15𝑦
𝑒2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2
Node u2 𝑘21𝑥 𝑘21𝑥𝑦 𝑘22𝑥 𝑘22𝑥𝑦 𝑘25𝑥 𝑘25𝑥𝑦
2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2
v2 𝑘21𝑦𝑥 𝑘21𝑦 𝑘22𝑦𝑥 𝑘22𝑦 𝑘25𝑦𝑥 𝑘25𝑦
Node u3
3
v3
Node u4
4
v4
𝑒2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2
Node u5 𝑘51𝑥 𝑘51𝑥𝑦 𝑘52𝑥 𝑘52𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦
5 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2 𝑒2
v5 𝑘51𝑦𝑥 𝑘51𝑦 𝑘52𝑦𝑥 𝑘52𝑦 𝑘55𝑦𝑥 𝑘55𝑦
Node u6
6
v6
FIG 58 – GLOBAL STIFFNESS MATRIX POPULATED WITH ELEMENT #2 MATRIX

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 135

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element 3: The element #3 is connected at the nodes #2, #5 & #3. So, the degrees of freedom
involved in the construction of the element #3 stiffness matrix are: (u2,v2), (u5,v5) & (u3,v3). Using the
6 x 6 stiffness matrix of the CST element derived in a previous paragraph (see Eq 238), we have the
following matrix for the element #3:

Node #2 Node #5 Node #3

u2 v2 u5 v5 u3 v3
𝑘22𝑥 𝑘22𝑥𝑦 𝑘25𝑥 𝑘25𝑥𝑦 𝑘23𝑥 𝑘23𝑥𝑦
[ ] [ ] [ ] u2
𝑘22𝑦𝑥 𝑘22𝑦 𝑘25𝑦𝑥 𝑘25𝑦 𝑘23𝑦𝑥 𝑘23𝑦
v2
𝑘52𝑥 𝑘52𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦 𝑘53𝑥 𝑘53𝑥𝑦 u5
[𝑘]𝐸𝑙𝑒𝑚𝑒𝑛𝑡 #3 = [ ] [ ] [ ] Eq 399
𝑘52𝑦𝑥 𝑘52𝑦 𝑘55𝑦𝑥 𝑘55𝑦 𝑘53𝑦𝑥 𝑘53𝑦 v5
𝑘32𝑥 𝑘32𝑥𝑦 𝑘35𝑥 𝑘35𝑥𝑦 𝑘33𝑥 𝑘33𝑥𝑦 u3
[ ] [ ] [ ] v3
[ 𝑘32𝑦𝑥 𝑘32𝑦 𝑘35𝑦𝑥 𝑘35𝑦 𝑘33𝑦𝑥 𝑘33𝑦 ]

The Fig 58 shows the elemental matrix Eq 399 written at the appropriate locations of the global
stiffness matrix based on the degrees of freedom of the element #3.
Node 1 Node 2 Node 3 Node 4 Node 5 Node 6
u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6
Node u1
1
v1
𝑒3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3
Node u2 𝑘22𝑥 𝑘22𝑥𝑦 𝑘23𝑥 𝑘23𝑥𝑦 𝑘25𝑥 𝑘25𝑥𝑦
2 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3
v2 𝑘22𝑦𝑥 𝑘22𝑦 𝑘23𝑦𝑥 𝑘23𝑦 𝑘25𝑦𝑥 𝑘25𝑦
𝑒3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3
Node u3 𝑘32𝑥 𝑘32𝑥𝑦 𝑘33𝑥 𝑘33𝑥𝑦 𝑘35𝑥 𝑘35𝑥𝑦
3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3
v3 𝑘32𝑦𝑥 𝑘32𝑦 𝑘33𝑦𝑥 𝑘33𝑦 𝑘35𝑦𝑥 𝑘35𝑦
Node u4
4
v4
𝑒3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3
Node u5 𝑘52𝑥 𝑘52𝑥𝑦 𝑘53𝑥 𝑘53𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦
5 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3 𝑒3
v5 𝑘52𝑦𝑥 𝑘52𝑦 𝑘53𝑦𝑥 𝑘53𝑦 𝑘55𝑦𝑥 𝑘55𝑦
Node u6
6
v6
FIG 59 – GLOBAL STIFFNESS MATRIX POPULATED WITH ELEMENT #3 MATRIX

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 136

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Element 4: The element #4 is connected at the nodes #3, #5 & #6. So, the degrees of freedom
involved in the construction of the element #4 stiffness matrix are: (u3,v3), (u5,v5) & (u6,v6). Using the
6 x 6 stiffness matrix of the CST element derived in a previous paragraph (see Eq 238), we have the
following matrix for the element #4:

Node #3 Node #5 Node #6

u3 v3 u5 v5 u6 v6
𝑘33𝑥 𝑘33𝑥𝑦 𝑘35𝑥 𝑘35𝑥𝑦 𝑘36𝑥 𝑘36𝑥𝑦
[ ] [ ] [ ] u3
𝑘33𝑦𝑥 𝑘33𝑦 𝑘35𝑦𝑥 𝑘35𝑦 𝑘36𝑦𝑥 𝑘36𝑦
v3
𝑘53𝑥 𝑘53𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦 𝑘56𝑥 𝑘56𝑥𝑦 u5
[𝑘]𝐸𝑙𝑒𝑚𝑒𝑛𝑡 #4 = [ ] [ ] [ ] Eq 400
𝑘53𝑦𝑥 𝑘53𝑦 𝑘55𝑦𝑥 𝑘55𝑦 𝑘56𝑦𝑥 𝑘56𝑦 v5
𝑘63𝑥 𝑘63𝑥𝑦 𝑘65𝑥 𝑘65𝑥𝑦 𝑘66𝑥 𝑘66𝑥𝑦 u6
[ ] [ ] [ ] v6
[ 𝑘63𝑦𝑥 𝑘63𝑦 𝑘65𝑦𝑥 𝑘65𝑦 𝑘66𝑦𝑥 𝑘66𝑦 ]

The Fig 60 shows the elemental matrix Eq 400 written at the appropriate locations of the global
stiffness matrix based on the degrees of freedom of the element #4.
Node 1 Node 2 Node 3 Node 4 Node 5 Node 6
u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6
Node u1
1
v1
Node u2
2
v2
𝑒4 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4
Node u3 𝑘33𝑥 𝑘33𝑥𝑦 𝑘35𝑥 𝑘35𝑥𝑦 𝑘36𝑥 𝑘36𝑥𝑦
3 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4
v3 𝑘33𝑦𝑥 𝑘33𝑦 𝑘35𝑦𝑥 𝑘35𝑦 𝑘36𝑦𝑥 𝑘36𝑦
Node u4
4
v4
𝑒4 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4
Node u5 𝑘53𝑥 𝑘53𝑥𝑦 𝑘55𝑥 𝑘55𝑥𝑦 𝑘56𝑥 𝑘56𝑥𝑦
5 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4
v5 𝑘53𝑦𝑥 𝑘53𝑦 𝑘55𝑦𝑥 𝑘55𝑦 𝑘56𝑦𝑥 𝑘56𝑦
𝑒4 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4
Node u6 𝑘63𝑥 𝑘63𝑥𝑦 𝑘65𝑥 𝑘65𝑥𝑦 𝑘66𝑥 𝑘66𝑥𝑦
6 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4 𝑒4
v6 𝑘63𝑦𝑥 𝑘63𝑦 𝑘65𝑦𝑥 𝑘65𝑦 𝑘66𝑦𝑥 𝑘66𝑦
FIG 60 – GLOBAL STIFFNESS MATRIX POPULATED WITH ELEMENT #4 MATRIX

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 137

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Fig 61 shows a summary of the elements with their associated degrees of freedom (ui,vi) for i = 1 to
6:
Node 1 Node 2 Node 3 Node 4 Node 5 Node 6
u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6
Node u1 0 0 0 0
E1,E2 E2 E1 E1,E2
1 0 0 0 0
v1
Node u2 0 0 0 0
E2 E2,E3 E3 E2,E3
2 0 0 0 0
v2
Node u3 0 0 0 0
E3 E3,E4 E3,E4 E4
3 0 0 0 0
v3
Node u4 0 0 0 0 0 0
E1 E1 E1
4 v4 0 0 0 0 0 0

Node u5
E1,E2 E2,E3 E3,E4 E1 E1,E2,E3,E4 E4
5
v5
Node u6 0 0 0 0 0 0
E4 E4 E4
6 0 0 0 0 0 0
v6
FIG 61 – GLOBAL STIFFNESS MATRIX POPULATED WITH FOUR ELEMENTS OF THE DISCRETIZATION

The global structure stiffness matrix [K] for the rectangle plate meshed with four constant strain
triangular elements is obtained using the direct stiffness method:
4

[𝐾 ] = ∑[𝑘]𝑒 Eq 401
𝑒=1

Where [k]e is the stiffness matrix of each element of the discretization. For example, for the position
framed in bold in Fig 61, the involved degrees of freedom are (u3,v3) and (u5,v5) for the elements #3
and #4. So the coefficients of the matrix at this position are:

𝑘35𝑥 𝑘35𝑥𝑦 𝐸𝑙𝑒𝑚𝑒𝑛𝑡 #3 𝑘35𝑥 𝑘35𝑥𝑦 𝐸𝑙𝑒𝑚𝑒𝑛𝑡 #4

[ ] +[ ] Eq 402
𝑘35𝑦𝑥 𝑘35𝑦 𝑘35𝑦𝑥 𝑘35𝑦

𝑒3 𝑒4 𝑒3 𝑒4
𝑘35𝑥 + 𝑘35𝑥 𝑘35𝑥𝑦 + 𝑘35𝑥𝑦
[ 𝑒3 𝑒4 𝑒3 𝑒4 ] Eq 403
𝑘35𝑦𝑥 + 𝑘35𝑦𝑥 𝑘35𝑦 + 𝑘35𝑦

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - MATRIX ASSEMBLY 138

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Fig 61 shows that half of the positions in the matrix have null values. Moreover Fig 61 shows the
symmetry of the stiffness matrix. So, only the upper portion of the stiffness matrix needs to be
computed and stored to perform the structural analysis.

6.2. TAKING ADVANTAGE OF SPARSITY AND SYMMETRY

The global stiffness matrix for a structural FEM is a square matrix having the size (Total Number of
degrees of freedom) x (Total Number of degrees of freedom). In a complex medium- or large-sized
structural analysis, it is common to have a few million degrees of freedom, in total.
Let’s consider a small model with 100,000 degrees of freedom. The number of entries in the stiffness
matrix will be 100,000 x 100,000 = 1010. Therefore, assuming that each entry is saved in a double
precision variable of eight bytes, the memory required to store this matrix is 80 gigabytes.
Manipulating a global stiffness matrix of this size is not possible.
However, there are two properties of the global stiffness matrix that can be exploited to reduce its
size in memory:
• Matrix sparsity
• Matrix symmetry
In structural analysis, due to the method used to connect elements, many stiffness matrix [K] entries
are null. We call such a stiffness matrix sparse. Since a zero value does not contribute to the solution,
resolution algorithms ignore the null terms, and they do not need to be stored. Another interesting
property of the stiffness matrix is its symmetry: the upper terms (above the diagonal) are identical
to the lower terms (below the diagonal). Therefore, only half of the stiffness matrix must be stored—
usually, the diagonal and upper terms. These two fundamental properties of the global stiffness
matrix reduce the required memory and allow the FEA software to use smart schemes for storing the
stiffness matrix, generally in one of the following two formats:
• Banded Matrix
• Skyline Matrix Storage

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - TAKING ADVANTAGE OF SPARSITY AND SYMMETRY 139
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

6.3. BANDED MATRIX

In matrix theory, a banded matrix is a sparse matrix whose non-zero entries are confined to a
diagonal band, comprising the main diagonal and zero or more diagonals on either side. Fig 5 41
illustrates the banded matrix concept. Each black dot represents a term of the stiffness matrix. The
left view shows the matrix before bandwidth optimization. The right view shows the same matrix
with an optimized bandwidth.

FIG 62 – (LEFT) ORIGINAL SPARE MATRIX (RIGHT) MATRIX WITH OPTIMIZED BANDWIDTH

Banded matrices are common in many fields of scientific computing that require the solving of large
systems. A banded matrix has non-zero values only above and below the diagonal. From a
computational perspective, banded matrices are always preferred to similarly dimensioned square
matrices, because the work involved in performing operations such as multiplication decreases
significantly, which greatly reduces computation time. One of the most efficient algorithms for
minimizing bandwidth is the Cuthill–McKee algorithm, named after Elizabeth Cuthill and James
McKee. This algorithm uses a method based on renumbering the nodes that constitute the mesh.

From a practical point of view, the maximum bandwidth for a given matrix can be found by
considering each element in a mesh and finding the element with the largest difference in node ID.
To understand the concept, let’s consider the following simple example of a truss structure
composed of six elements and eight nodes, with node numbering as shown in Fig 63.
1 2 3 4

5 6 7 8

FIG 63 – ORIGINAL TRUSS STRUCTURE BEFORE OPTIMIZATION

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - BANDED MATRIX 140

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The stiffness matrix for this node arrangement is (the non-zero terms are denoted by a X and the
blank terms are null):
𝑋 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋 𝑋
[𝐾𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 ] = 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋 Eq 404
𝑋 𝑋 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋 𝑋
[ 𝑋 𝑋 𝑋]

Looking at the matrix above, it is noted that the non-zero terms extend five columns beyond the
main diagonal. It is possible to reduce the bandwidth by numbering the nodes differently. So, let’s
renumber the nodes as follow:
1 3 5 7

2 4 6 8

FIG 64 – NODES NUMBERING FOR BANDWIDTH OPTIMIZATION

The new stiffness matrix becomes:

𝑋 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋
𝑋 𝑋 𝑋 𝑋 𝑋
[𝐾𝑂𝑝𝑡𝑖𝑚𝑖𝑧𝑒𝑑 ] = 𝑋 𝑋 𝑋 𝑋 𝑋
Eq 405
𝑋 𝑋 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋 𝑋 𝑋
𝑋 𝑋 𝑋
[ 𝑋 𝑋 𝑋 𝑋]

The new node numbering gives a reduced bandwidth with three columns beyond the main diagonal.
In the commercial FEA codes, this operation is automatically performed without any action from the
user being required.

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - BANDED MATRIX 141

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

6.4. SKYLINE MATRIX STORAGE

A symmetric sparse matrix can be stored using a matrix format called skyline matrix storage. This
format significantly reduces storage space and requires less computation.
To understand the skyline storage format, let’s look at another simple example. Consider the
following N x N = 6 x 6 sparse stiffness matrix:

𝐾11 0 𝐾13 0 0 𝐾16

𝐾22 0 𝐾24 0 0
𝐾33 𝐾34 0 0
[𝐾] = Eq 406
𝐾44 0 𝐾46
𝐾55 𝐾56
[ 𝑠 𝑦 𝑚 𝐾66 ]

Since the stiffness matrix is symmetric, only the upper portion is considered. The envelope of [K] is
defined as follows: from each diagonal entry move up the corresponding column until the last non-
zero entry is found. The remaining zero entries are removed. This envelope, called the skyline
template, defines the skyline profile of the matrix. A sparse matrix that can be profitably stored in
this form is called a skymatrix. Note that the skyline template may include zero entries.

𝐾11 0 𝐾13 0 0 𝐾16 Skyline template

𝐾22 0 𝐾24 0 0
𝐾33 𝐾34 0 0
[𝐾] =
𝐾44 0 𝐾46
𝐾55 𝐾56
[ 𝑠 𝑦 𝑚 𝐾66 ]

Only entries in the skyline template must be stored, since null terms that are outside it will not
influence the results. A one-dimensional array is used to store the terms of the skyline template:

s : [K11, K22, K13, 0, K33, K24, K34, K44, K55, K16, 0, 0, K46, K56, K66] Eq 407

The skyline array Eq 407 is complemented with an integer array p containing the addresses of the
diagonal terms in the skyline array s. The array p has (N+1) entries. The (i+1)th entry of p contains the
location of ith diagonal entry of [K] in s. So, for the example matrix:

p : [0, 1, 2, 5, 8, 9, 15] Eq 408

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - SKYLINE MATRIX STORAGE 142

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The terms of p for which the displacement component is prescribed are identified by a negative
diagonal location value in the p array. For example, if u4 and u6 are the prescribed displacement
components in the example, then:

p : [0, 1, 2, 5, -8, 9, -15] Eq 409

So, finally, the symmetric squared stiffness matrix Eq 406 originally containing 36 terms is stored in
the following reduced format:

S = {p , s}
Eq 410
S = {[0, 1, 2, 5, -8, 9, -15] , [K11, K22, K13, 0, K33, K24, K34, K44, K55, K16, 0, 0, K46,
K56, K66]}

HOW IS THE STIFFNESS MATRIX ASSEMBLED? - SKYLINE MATRIX STORAGE 143

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

7. HOW ARE FEM EQUATIONS SOLVED?

A finite element analysis software is a set of algorithms, whose main function is to solve equilibrium
equations written in matrix form. Computational techniques and hardware evolve so quickly that
FEA software are frequently updated. The user can choose from a large library of resolution methods
to solve a problem. Among them, two main techniques stand out:
• The direct method, based on LDLT decomposition
• The iterative method, based on substitution of an assumed solution into a system of
equations to obtain a better estimate of an unknown

7.1. DIRECT SOLUTION

With the direct method, the equilibrium equation {P} = [K]{u} is solved for unknown displacements
{u}, without inverting [K]. LDLT decomposition, the matrix form of Gaussian elimination, is an efficient
and accurate method, comprised of three steps:
• Step 1: Factorization: [K] = [L].[D].[L]T
• Step 2: Forward Substitution: {P} = [L].{y}
• Step 3: Backward Substitution: {y} = [D].[L]T.{u}

HOW ARE FEM EQUATIONS SOLVED? - DIRECT SOLUTION 144

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Step 1: Factorization
The stiffness matrix [K] is replaced by the product of three matrices [L], [D] and [L]T:

[K] = [L].[D].[L]T Eq 411

[L] is the unit lower triangular matrix
[D] is the diagonal matrix
[L] is the unit upper triangular matrix
The factorization concept is illustrated for a n x n matrix:

𝑲𝟏𝟏 𝐾12 𝐾13 𝐾14 . . 𝐾1𝑛

𝐾12 𝑲𝟐𝟐 𝐾23 𝐾24 . . 𝐾2𝑛
𝐾13 𝐾23 𝑲𝟑𝟑 𝐾34 . . 𝐾3𝑛
𝐾14 𝐾24 𝐾34 𝑲𝟒𝟒 . . 𝐾4𝑛 =
. . . . . . .
. . . . . . .
[ 𝐾1𝑛 𝐾2𝑛 𝐾3𝑛 𝐾4𝑛 . . 𝑲𝒏𝒏 ]

𝟏 0 0 0 . . 0 𝐷11 0 0 0 . . 0 𝟏 𝐿21 𝐿31 𝐿41 . . 𝐿𝑛1

𝐿21 𝟏 0 0 . . 0 0 𝐷22 0 0 . . 0 0 𝟏 𝐿32 𝐿42 . . 𝐿𝑛2
𝐿31 𝐿32 𝟏 0 . . 0 0 0 𝐷33 0 . . 0 0 0 𝟏 𝐿43 . . 𝐿𝑛3
𝐿41 𝐿42 𝐿43 𝟏 . . 0 0 0 0 𝐷44 . . 0 0 0 0 𝟏 . . 𝐿𝑛4
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
[𝐿𝑛1 𝐿𝑛2 𝐿𝑛3 𝐿𝑛4 . . 𝟏] [ 0 0 0 0 . . 𝐷𝑛𝑛 ] [0 0 0 0 . . 𝟏 ]
Eq 412

The decomposition means that the stiffness matrix [K] is separated into the product of three simple
matrices. The decomposition process consists in finding the factors for [D] and [L].
To compute the factors of the decomposition, the Cholesky decomposition is used. This method is
very efficient in terms of memory storage capacity, computational cost and speed. The factors of the
Cholesky decomposition are computed as:
𝑗−1
2
𝐷𝑗𝑗 = 𝐾𝑗𝑗 − ∑ 𝐿𝑗𝑘 𝐷𝑘𝑘 Eq 413
𝑘=1
𝑗−1
1
For i > j 𝐿𝑖𝑗 = (𝐾𝑖𝑗 − ∑ 𝐿𝑖𝑘 𝐿𝑗𝑘 𝐷𝑘𝑘 ) Eq 414
𝐷𝑗𝑗
𝑘=1

HOW ARE FEM EQUATIONS SOLVED? - DIRECT SOLUTION 145

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

The implementation of the Cholesky decomposition is:

• The subscript i is the row index and the subscript j is the column index
• The decomposition is computed column by column
• The decomposition starts by the evaluation of the diagonal term of the column using Eq 413
• The other elements are evaluated row by row using Eq 414

Step 2: Forward Substitution

From the factorization performed at step 1, the equilibrium equation can be written:

[P] = [K].{u} = [L].[D].[L]T.{u} = [L].{y} Eq 415

Where,
{y} = [D].[L]T.{u} Eq 416
So {y} is computed by starting with the first equation in the system:
[P] = [L].{y}
𝑓1 𝟏 0 0 0 . . 0 𝑦1
𝑓2 𝐿21 𝟏 0 0 . . 0 𝑦2
𝑓3 𝐿31 𝐿32 𝟏 0 . . 0 𝑦3
Eq 417
𝑓4 = 𝐿41 𝐿42 𝐿43 𝟏 . . 0 𝑦4
. . . . . . . . .
. . . . . . . . .
{𝑓𝑛 } [𝐿𝑛1 𝐿𝑛2 𝐿𝑛3 𝐿𝑛4 . . 𝟏] 𝑛 }
{ 𝑦

So, starting from the first equation:

f1 = y1
And continuing with the next lines using the results obtained from the previous lines:
f2 = L21.y1 + y2  y2 = f2 - L21.y1
f3 = L31.y1 + L32.y2 + y3  y3 = f3 - L31.y1 - L32.y2
…and so on.
So, the solution for {y} is given by:
𝑖−1

𝑦𝑖 = 𝑓𝑖 − ∑ 𝐿𝑖𝑗 . 𝑦𝑗 For i = 2 to n Eq 418

𝑗=1

HOW ARE FEM EQUATIONS SOLVED? - DIRECT SOLUTION 146

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Step 3: Backward Substitution

Having computed {y} at step 2, it is easy to compute the unknown displacements {u} from {y} =
[D].[L]T.{u} by starting with the last equation:
{y} = [D].[L]T.{u}

𝑦1
𝑦2
𝑦3
𝑦4
.
.
{𝑦𝑛 }
𝐷11 0 0 0 . . 0 𝟏 𝐿21 𝐿31 𝐿41 . . 𝐿𝑛1 𝑢1 Eq 419
0 𝐷22 0 0 . . 0 0 𝟏 𝐿32 𝐿42 . . 𝐿𝑛2 𝑢2
0 0 𝐷33 0 . . 0 0 0 𝟏 𝐿43 . . 𝐿𝑛3 𝑢3
= 0 0 0 𝐷44 . . 0 0 0 0 𝟏 . . 𝐿𝑛4 𝑢4
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
[ 0 0 0 0 . . 𝐷𝑛𝑛 ] [0 0 0 0 . . 𝟏 ] {𝑢 𝑛}

Starting from the last equation, i=n:

yn = Dnn.un Eq 420

un is easily found,

𝑦𝑛
𝑢𝑛 = Eq 421
𝐷𝑛𝑛

Then, moving backward in the Eq 419, {u} is solved by:

𝑢𝑖
𝑁
1 For i = (n-1) to 1 Eq 422
= (𝑦 − ∑ 𝐿𝑗𝑖 𝑢𝑗 )
𝐷𝑖𝑖 𝑖
𝑗=𝑖+1

HOW ARE FEM EQUATIONS SOLVED? - DIRECT SOLUTION 147

 THE FINITE ELEMENT METHOD THEORY IN DEPTH

7.2. ITERATIVE SOLUTION

An alternative to the direct method is an iterative method used to solve a linear system of equations,
similar to the Jacobi method. Called the Gauss–Seidel method for the German mathematicians Carl
Friedrich Gauss and Philipp Ludwig von Seidel, it is also known as the Liebmann method, or the
method of successive displacement. Though it can be applied to any matrix with non-zero elements
on the diagonals, convergence is only guaranteed if the matrix is either diagonally dominant, or
symmetric and positive definite. The Gauss-Seidel method can be used to solve a square system of n
equations with unknown displacements {u}:
{P} = [K]{u}
𝑓1 𝑲𝟏𝟏 𝐾12 𝐾13 𝐾14 . . 𝐾1𝑛 𝑢1
𝑓2 𝐾21 𝑲𝟐𝟐 𝐾23 𝐾24 . . 𝐾2𝑛 𝑢2
𝑓3 𝐾31 𝐾32 𝑲𝟑𝟑 𝐾34 . . 𝐾3𝑛 𝑢3
Eq 423
𝑓4 = 𝐾41 𝐾42 𝐾43 𝑲𝟒𝟒 . . 𝐾4𝑛 𝑢4
. . . . . . . . .
. . . . . . . . .
{𝑓𝑛 } [ 𝐾𝑛1 𝐾𝑛2 𝐾𝑛3 𝐾𝑛4 . . 𝑲𝒏𝒏 ] {𝑢𝑛 }

First, the stiffness matrix [K] is decomposed into a lower triangular component [L] Lower and an upper
triangular component [L]Upper:

[K] = [L]Lower + [L]Upper

𝑲𝟏𝟏 𝐾12 𝐾13 𝐾14 . . 𝐾1𝑛
𝐾21 𝑲𝟐𝟐 𝐾23 𝐾24 . . 𝐾2𝑛
𝐾31 𝐾32 𝑲𝟑𝟑 𝐾34 . . 𝐾3𝑛
𝐾41 𝐾42 𝐾43 𝑲𝟒𝟒 . . 𝐾4𝑛
. . . . . . .
. . . . . . .
[ 𝐾𝑛1 𝐾𝑛2 𝐾𝑛3 𝐾𝑛4 . . 𝑲𝒏𝒏 ]
𝑲𝟏𝟏 0 0 0 . . 0 𝟎 𝐾12 𝐾13 𝐾14 . . 𝐾1𝑛
𝐾21 𝑲𝟐𝟐 0 0 . . 0 0 𝟎 𝐾23 𝐾24 . . 𝐾2𝑛
𝐾31 𝐾32 𝑲𝟑𝟑 0 . . 0 0 0 𝟎 𝐾34 . . 𝐾3𝑛
= 𝐾41 𝐾42 𝐾43 𝑲𝟒𝟒 . . 0 + 0 0 0 𝟎 . . 𝐾4𝑛
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
[ 𝐾𝑛1 𝐾𝑛2 𝐾𝑛3 𝐾𝑛4 . . 𝑲𝒏𝒏 ] [0 0 0 0 . . 𝟎 ]
Eq 424

So, the equilibrium equation is written:

{P} = {u} [L]Lower + {u} [L]Upper Eq 425

HOW ARE FEM EQUATIONS SOLVED? - ITERATIVE SOLUTION 148
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

{u} [L]Lower = {P} - {u} [L]Upper

Then, the iterative process is defined as:

{u}(k+1) [L]Lower = {P} - {u}(k) [L]Upper Eq 426

Where:
• {u}(k) is the kth iteration of {u}
• {u}(k+1) is the (k+1)th iteration of {u}
• The iteration process starts with an estimate.
The Gauss-Seidel method solves the left-hand side of Eq 426 for {u} using the previous value for {u}
on the right-hand side. This is written as:

{𝑢}(𝑘+1) = [𝐿]−1
𝐿𝑜𝑤𝑒𝑟 ({𝑃} − [𝐿]
𝑈𝑝𝑝𝑒𝑟
{𝑢}(𝑘) ) Eq 427
Taking advantage of the triangular form of [L], the elements of {u}(k+1) are computed sequentially
using forward substitution:
𝑛 𝑛
For i = 1, 2, …, (𝑘+1) 1 (𝑘+1) (𝑘)
𝑢𝑖 = [𝑓𝑖 − ∑ 𝐾𝑖𝑗 𝑢𝑗 − ∑ 𝐾𝑖𝑗 𝑢𝑗 ] Eq 428
n 𝐾𝑖𝑖
𝑗=1 𝑗=𝑖+1

The iteration process continues until the solution converges:

‖𝑢(𝑘+1) − 𝑢(𝑘) ‖ Eq 429

≤ 𝜀
‖𝑢(𝑘) ‖

where  is the tolerance for convergence control. The analyst must specify the maximum number of
iterations to be performed, and a convergence criterion must be defined to determine when the
solution converges. Since the iterative method must begin with an initial value, solvers typically use
a preconditioning algorithm. However, there is no guarantee that a solution will result within the
number of iterations specified by the user. For a large problem with a high number of degrees of
freedom and, therefore, a large number of equations, the iterative method requires less memory
than the direct method. Ultimately, the solution’s accuracy depends on the convergence tolerance:
a smaller tolerance will result in a more accurate solution but may take more iterations and,
consequently, more computing time.

HOW ARE FEM EQUATIONS SOLVED? - ITERATIVE SOLUTION 149

 GET THE BOOK

650+ PAGES
FEA BEST PRACTICAL METHODS & GUIDELINES
100+ EXAMPLES
Available in Hardcover and eBook
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

INDEX

B L
Banded Matrix................................................... 140
Linear Spring ....................................................... 15
Beam
Bending ........................................................... 58
Full capabilities ............................................... 60 M
Torsion ............................................................ 58
Matrix Assembly ............................................... 132
Beam Theory
Matrix Sparsity ................................................. 139
Euler-Bernoulli ................................................ 43
Membrane CST ................................................... 65
Timoshenko .................................................... 52
Membrane LST ................................................... 82

C
P
Compatibility of Deformation ............................. 11
Principle of Potential Energy .............................. 26

D
S
Degrees of Freedom............................................ 28
Shape Functions ................................................. 30
Displacement Method .................................. 11, 31
Truss Element............................................32, 35
Shell .................................................................. 119
E Skyline Storage ................................................. 142
Solid Element.................................................... 122
Equilibrium .......................................................... 11
Solving the FEM Equations ............................... 144
Equilibrium equation ............................................ 9
Direct Solution ............................................. 144
Iterative Solution.......................................... 148
H Sparsity ............................................................. 139
Stiffness Matrix..............................................12, 28
Hexahedral Element 2D Element ..................................................... 63
Linear ............................................................ 122 Beam Element ................................................ 43
Quadratic ...................................................... 129 Membrane CST ............................................... 65
Membrane LST ............................................... 82
I Shell .............................................................. 119
Solid Element ............................................... 122
Isoparametric Formulation ................................. 98 Thick Plate .................................................... 107

INDEX 152
 THE FINITE ELEMENT METHOD THEORY IN DEPTH

Thin Plate ........................................................ 87 Thin Plate ............................................................ 87

Truss Element ................................................. 31 Truss Element
Stress-Strain Relations ........................................ 11 Axial Loading .................................................. 32
Complete Stiffness ......................................... 37
Generic Form.................................................. 39
T
Torsion ........................................................... 35
Thick Plate ......................................................... 107

INDEX 153