Introduction to Finite Element Methods

MAK372E

Common Procedure in FEM, Discrete elements

Week 2
 Remember that:

• The finite element method (FEM) is a numerical technique for solving a wide
range of complex physical phenomena, particularly those exhibiting
geometrical and material nonlinearities (such as those that are often
encountered in the physical and engineering sciences).

• These problems can be structural in nature, thermal (or thermo-mechanical),

electrical, magnetic, acoustic etc. plus any combination of.

• It is used most frequently to tackle problems that aren’t readily amenable to

analytical treatments.

• Continuous domains (geometries) are decomposed into discrete, connected

regions (or finite elements). An assembly of element-level equations is
subsequently solved, in order to establish the response of the complete
domain to a particular set of boundary conditions.
Governing equations for various physical phenomena:
Engineering Applications of The Finite Element Method
• The finite element method was developed originally for the analysis of aircraft structures.
• However, the general nature of its theory makes it applicable to a wide variety of boundary
and initial value problems in engineering.
• A boundary value problem is one in which a solution is sought in the domain (or region) of
a body subject to the satisfaction of prescribed boundary (edge) conditions on the
dependent variables or their derivatives.
• Three major categories of boundary value problems, namely
(1) equilibrium or steady state or time-independent problems,
(2) eigenvalue problems, and
(3) propagation or transient problems.

In an equilibrium problem, we need to find the steady-state displacement or stress distribution if it is a

solid mechanics problem, temperature or heat flux distribution if it is a heat transfer problem, and
pressure or velocity distribution if it is a fluid mechanics problem.

In eigenvalue problems also, time will not appear explicitly. They may be considered as extensions of
equilibrium problems in which critical values of certain parameters are to be determined in addition to the
corresponding steady-state configurations. In these problems, we need to find the natural frequencies or
buckling loads and mode shapes if it is a solid mechanics or structures problem, stability of laminar flows if
it is a fluid mechanics problem, and resonance characteristics if it is an electrical circuit problem.

The propagation or transient problems are time-dependent problems. This type of problem arises, for
example, whenever we are interested in finding the response of a body under time-varying force in the
area of solid mechanics and under sudden heating or cooling in the field of heat transfer.
 General Description of The Finite Element Method

With reference to static structural problems, the step-by-step procedure can be stated as
follows:
Step 1: Divide the structure into discrete elements (discretization):
The first step in the finite element method is to divide the structure or solution region into subdivisions
or elements. Hence, the structure is to be modeled with suitable finite elements. The number, type,
size, and arrangement of the elements are to be decided.

Step 2: Select a proper interpolation or displacement model:

Since the displacement solution of a complex structure under any specified load conditions cannot be
predicted exactly, we assume some suitable solution within an element to approximate the unknown
solution. The assumed solution must be simple from a computational standpoint, but it should satisfy
certain convergence requirements. In general, the solution or the interpolation model is taken in the
form of a polynomial.

Step 3: Derive element stiffness matrices and load vectors:

From the assumed displacement model, the stiffness matrix [K](e) and the load vector {f} (e) of element e
are to be derived by using a suitable variational principle, a weighted residual approach (such as the
Galerkin method), or equilibrium conditions (direct method).

Step 4: Assemble element equations to obtain the overall equilibrium equations.

Since the structure is composed of several finite elements, the individual element stiffness matrices and
load vectors are to be assembled in a suitable manner and the overall equilibrium equations have to be
formulated as [K] : Global Stiffness Matrix, {d}: Global displacement vector
{F} : Global force vector
General Description of The Finite Element Method

Step 5: Solve for the unknown nodal displacements:

The overall equilibrium equations have to be modified to account for the boundary conditions of the
problem. After the incorporation of the boundary conditions, the equilibrium equations can be solved to
find nodal quantities.
For linear problems, the vector {d} can be solved very easily. However, for nonlinear problems, the
solution has to be obtained in a sequence of steps, with each step involving the modification of the
stiffness matrix [K] and/or the load vector {f}.

Step 6: Compute element strains and stresses:

From the known nodal displacements {d}; if required, the element strains and stresses can be computed
by using the necessary equations of solid or structural mechanics.

The terminology used in the previous six steps has to be modified if we want to extend the
concept to other fields. For example,
structure --> continuum or domain
field variable --> displacement
characteristic matrix --> stiffness matrix
element resultants --> element stress/strains.
Derivation of Finite Element Equations Using a Direct Approach

• The element stiffness (or characteristic) matrices and load (characteristic) vectors and the
finite element equations can be derived by using a direct approach.

• In this method, direct physical reasoning relevant to the problem (such as consideration
of equilibrium of the system) is used to establish the element properties (characteristic
matrices and vectors) in terms of pertinent variables.

• The direct approach is applicable only to problems involving simple types of elements;
hence most practical (complex) problems cannot be solved using this approach.

• However, a study of direct methods enhances our understanding of the physical

interpretation of the finite element method. It is a good starting point for understanding
the finite element formulation.

• The direct approach is presented in this section by considering simple problems from the
areas of elastic systems and heat transfer.
Spring Element

• Spring element is the simplest finite element available to be used for the modeling of
axially loaded structural units such as bars/trusses.

• A linear static spring is a mechanical device capable of supporting axial loading only and
constructed such that under elastic limit, the deformation of the spring is directly
proportional to the applied load.

Idealization of an axially loaded bar with non-uniform cross section with spring elements:
Spring Element

From inspection, we can see

that there are two degrees of
freedom in this model, u1 and u2
Sign convention for nodal displacements and forces
(+) -->

• Reference points 1 and 2 are located at the ends of the element. These reference
points are called the nodes of the spring element.
• The local nodal forces are f1x and f2x for the spring element associated with the local
axis x.
• The local axis acts in the direction of the spring so that we can directly measure
displacements and forces along the spring.
• The local nodal displacements are u1 and u2 for the spring element.
• These nodal displacements are called the degrees of freedom at each node.
• Positive directions for the forces and displacements at each node are taken in the
positive x direction as shown from node 1 to node 2 in the figure.
• The symbol k is called the spring constant or stiffness of the spring.
Spring Element

Linear spring subjected to tensile forces Deformed spring

We now derive the spring element stiffness matrix. By the sign convention for nodal forces
and equilibrium

Rewriting

Local stiffness
By a single matrix
matrix for the
equation: element
Spring Element

Note
T
1. The element stiffness matrix is “symmetric”, i.e. k̂ = k̂
2. The element stiffness matrix is singular, i.e.,

det ( k̂ ) = k 2 − k 2 = 0

The consequence is that the matrix is NOT invertible. It is not possible to invert it
to obtain the displacements. Why?
The spring is not constrained in space and hence it can attain multiple positions
in space for the same nodal forces
e.g.,
f̂1x   2 - 2 1 - 2
 =   = 
f̂ 2x  - 2 2  2  2 
f̂1x   2 - 2 3 - 2
 =   = 
f̂ 2x  - 2 2  4  2 
Example of a Spring Assemblage

We will determine the total structure stiffness matrix for a spring assemblage by using the
force/displacement matrix relationships.

Two-spring assemblage

Here we fix node 1 and apply axial forces for F3x at node 3 and F2x at node 2. The stiffnesses of
spring elements 1 and 2 are k1 and k2 , respectively.

The x axis is the global axis of the assemblage. The local x axis of each element coincides with
the global axis of the assemblage.
For element 1 For element 2

The superscripts in parentheses above u refers to the element number to which they are related.
The subscript to the right identifies the node of displacement and that u3 is the node 3 displacement of the total or global
spring assemblage.
Example of a Spring Assemblage

Free-body diagrams of each element and node using the established sign conventions
for element nodal forces

Based on the free-body diagrams of each node as shown above and the fact that external forces
must equal internal forces at each node, we can write nodal equilibrium equations at nodes 3,
2, and 1 as

where F1x results from the external

applied reaction at the fixed support.

Matrix Form
Example of a Spring Assemblage

the global nodal force matrix global nodal displacement matrix

total or global or system stiffness matrix

Example of a Spring Assemblage
Assembling the Total Stiffness Matrix by Superposition (Alternative)

This method is based on proper superposition of the individual element stiffness matrices
making up a structure.

the uis written above the columns and next to the rows in the [k]s indicate the degrees of
freedom associated with each element row and column.
To superimpose the element matrices, we must expand them to the order (size) of the total
structure (spring assemblage) stiffness matrix so that each element stiffness matrix is
associated with all the degrees of freedom of the structure.
To expand each element stiffness matrix to the order of the total stiffness matrix, we simply
add rows and columns of zeros for those displacements not associated with that particular
element.
Example of a Spring Assemblage
Assembling the Total Stiffness Matrix by Superposition (Alternative)

For element 1, we rewrite the

stiffness matrix in expanded form

Similarly, for element 2

Considering force
equilibrium at each
node results in
Example of a Spring Assemblage
Assembling the Total Stiffness Matrix by Superposition (Alternative)

Simplifying
Example of a Spring Assemblage
Assembling the Total Stiffness Matrix by Superposition (Alternative)

For this simple example, it is easy to expand the element stiffness matrices and then
superimpose them to arrive at the total stiffness matrix. However, for problems involving a
large number of degrees of freedom, it will become tedious to expand each element stiffness
matrix to the order of the total stiffness matrix. To avoid this expansion of each element
stiffness matrix, we suggest a direct, or shortcut, form of the direct stiffness method to obtain
the total stiffness matrix.
For the spring assemblage example, the rows and columns of each element stiffness matrix
are labeled according to the degrees of freedom associated with them as follows:

[K ] is then constructed simply by directly adding terms associated with degrees of freedom in
[k(1)] and [k(2)] into their corresponding identical degree-of-freedom locations in [K ] as follows:
Example of a Spring Assemblage
Boundary Conditions

• Assemblage or [K ] will be singular; that is, the determinant of [K ] will be zero, and its
inverse will not exist.
• This means the structural system is unstable. Without our specifying adequate kinematic
constraints or support conditions, the structure will be free to move as a rigid body and
not resist any applied loads.
• Thus, we must specify boundary (or support) conditions.
• In general, the number of boundary conditions necessary to make [K ] nonsingular is equal
to the number of possible rigid body modes.

Here we have u1 = 0 because node 1 is fixed.

where F1x is the unknown reaction and

F2x and F3x are known applied loads.
Example of a Spring Assemblage
Boundary Conditions

Writing the second and third of these equations in matrix form, we have

(*)

For homogeneous boundary conditions, the equations (*) could have been obtained directly
by deleting the row and column of original matrix equation corresponding to the zero-
displacement degrees of freedom. Here row 1 and column 1 are deleted because one is really
multiplying column 1 of [K] by u1= 0 . However, F1x is not necessarily zero and can be
determined once u2 and u3 are solved for.
After solving (*) for u2 and u3 , we have
Example of a Spring Assemblage
Boundary Conditions

Now that u2 and u3 are known, we can obtain the reaction F1x as

For all homogeneous boundary conditions, we can delete the rows and columns
corresponding to the zero-displacement degrees of freedom from the original set of
equations and then solve for the unknown displacements. This procedure is useful for
hand calculations.
 Example of a Spring Assemblage
Non-Homogenous Boundary Conditions

let u1 = δ, where δ is a known displacement

(**)
written in
expanded form

Considering the second and third

equations, because they have known
right-side nodal forces:
Transforming the known δ
term to the right side
Example of a Spring Assemblage
Non-Homogenous Boundary Conditions

Matrix Form

When dealing with nonhomogeneous boundary conditions, we cannot initially delete row
1 and column 1 of global equation (**), corresponding to the nonhomogeneous boundary
condition, because we are multiplying each element by a nonzero number..

Notice that if the displacement is known at a node (say u1 = δ ), then the force F1x at the
node in the same direction as the displacement is not initially known and is determined
using the global equation after solving for the unknown nodal displacements.
Properties of the Global Stiffness Matrix

We summarize some properties of the global stiffness matrix that are also
applicable to the generalization of the finite element method:

1. [K] is square, as it relates the same number of forces and displacements.

2. [K] is symmetric, as is each of the element stiffness matrices. If you are familiar with
structural mechanics, you will not find this symmetry property surprising.
3. [K] is singular (its determinant is equal to zero), and thus, no inverse exists until sufficient
boundary conditions are imposed to remove the singularity and prevent rigid body
motion.
4. The main diagonal terms of [K] are always positive. Otherwise, a positive nodal force Fi
could produce a negative displacement di—a behavior contrary to the physical behavior
of any actual structure.
5. [K] is positive semidefinite (that is {x}T[K]{x} > 0 for all non-zero vector {x} with real
numbers).
Note on Boundary Conditions:

In general, specified support conditions are treated mathematically by partitioning the

global equilibrium equations as follows:

where we let { d1} be the unconstrained or free displacements and {d2} be the specified
displacements.

• { F1} are the known nodal forces and {F2} are the unknown nodal forces at the specified
displacement nodes.
• {F2} is found after {d1} is determined.
• We assume that [K11] is no longer singular, thus allowing for the determination of {d1}.
Example 1:
Singular!
Example 2: