Introduction to Finite Element Method

Dr. Suhas P. Deshmukh

Professor,
Sinhgad Academy of Engineering,
Kondhawa (Bk) Pune

Todays Outline
Free Vibration Problems

Formulation of Eigen Value Problem

FEM Formulation

Time-dependent Problems
Combination of Galerkin FEM and FDM
Convergence and Stability of FD Scheme
Problems with Material Non-linearity
Direct Solution Technique

Dynamic Analysis
Dynamics is a special branch of mechanics where inertia of
accelerating masses must be considered in the force-deflection
relationships.
In order to describe motion of the mass system, a component
with distributed mass is approximated by a finite number of
mass points.
Every structure is associated with certain frequencies and mode
shapes of free vibration (without continuous application of load),
based on the distribution of mass and stiffness in the structure.
Any time-dependent external load acting on the structure,
whose frequency matches with the natural frequencies of the
structure, causes resonance and produces large displacements
leading to failure of the structure.

Free Vibration

For a simple spring of stiffness 'k' and a lumped mass 'm' under steady
state undamped condition of oscillation without external force, the
force equilibrium condition of the system is given by
+ = 0

Displacement in vibration is a simple harmonic motion and can be

represented by a sinusoidal function of time as
= 0 sin

Substituting in differential equation

+ = 2 + 0 sin = 0

In general, for a system with on' degrees of freedom, stiffness 'k' and
mass 'm' are represented by stiffness matrix [K] and mass matrix [M]
respectively.

Free Vibration
Here, [M] is the mass matrix of the entire structure and is of
the same order, say n x n, as the stiffness matrix [K]. This is
also obtained by assembling element mass matrices in a manner
exactly identical to assembling element stiffness matrices.
This is a typical eigenvalue problem, with 2 as eigenvalues and
{u} as eigenvectors. A structure with 'n' DOF will therefore
have 'n' eigenvalues and 'n' eigenvectors. Some eigenvalues may
be repeated and some eigenvalues may be complex, in pairs.

The equation can be represented in the standard form,

=
In dynamic analysis,, indicates ith natural frequency and
indicates ith natural mode of vibration. A natural mode is a
qualitative plot of nodal displacements.

Free Vibration Problems

In every natural mode of vibration, all the points on the
component will reach their maximum values at the same
time and will pass through zero displacements at the same
time.
Thus, in a particular mode, all the points of a component
will vibrate with the same frequency and their relative
displacements are indicated by the components of the
corresponding eigenvector.

Free Vibration

Free Vibration Problems

The equation of motion of free vibrations 2 = {0} is a

system of homogeneous equations (right side vector zero) and hence
does not give unique numerical solution.

Mode shape is a set of relative displacements in various degrees of

Methods of Normalization

freedom, while the structure is vibrating in a particular frequency and is

usually expressed in normalised form
The maximum value of anyone component of the eigenvector is equated to 1' and,
so, all other components will have a value less than or equal to '1'.

The length of the vector is equated to 1' and values of all components are divided
by the length of this vector so that each component will have a value less than or
equal to 1'.
The eigenvectors are usually normalised so that

= 1

 For a positive definite symmetric stiffness matrix of

size n x n, the eigenvalues are all real and
eigenvectors are orthogonal

Solution for any dynamic analysis is an iterative

process and, hence, is time consuming. Geometric
model of the structure for dynamic analysis can be
significantly simplified, giving higher priority for
proper representation of distributed mass.

Mass Matrix
Mass matrix [M] differs from the stiffness matrix
in many ways:
The mass of each element is equally distributed at all
the nodes of that element

Mass, being a scalar quantity, has same effect along

the three translational degrees of freedom (u, v and w)
and is not shared
Mass, being a scalar quantity, is not influenced by the
local or global coordinate system. Hence, no
transformation matrix is used for converting mass
matrix from element (or local) coordinate system to
structural (or global) coordinate system.

Mass Matrix
Two different approaches of evaluating mass matrix [M] are
commonly considered.
Lumped mass matrix

Total mass of the element is assumed equally distributed at all

the nodes of the element in each of the translational degrees of
freedom. Lumped mass is not used for rotational degrees of
freedom. Off-diagonal elements of this matrix are all zero. This
assumption excludes dynamic coupling that exists between
different nodal displacements. Lumped mass matrices [M] of
some elements are given here.

Mass Matrix

Please note that the same lumped mass is considered in each translational
degree of freedom (without proportional sharing of mass between them) at
each node.

Mass Matrix

Mass Matrix

Mass Matrix

Consistent mass matrix of a Truss element along its axis (in local
coordinate system)

Mass Matrix

Mass Matrix

Mass Matrix

Mass Matrix

Mass Matrix

Numerical

The first eigenvector implies rigid body motion of the shaft. One component
(u1 in this example) is equated to '1' and other displacement components (U2
and U3 in this example) are obtained as ratios w.r.t. that component,
following one method of normalisation. Alternatively, they may also be
expressed in other normalised forms.

Numerical 2

Numerical 3

Numerical 4

Summary

Summary

Non-linear Analysis
The finite element formulation discussed so far was based on
the following assumptions:
Stress-strain relation is linear and
Strain displacement relation is also linear

It resulted into stiffness equation k = F in which [k] and {F}

were independent of displacement . Hence after forming the
stiffness equation we could get required displacements by
solving the set of linear equations only once.
In engineering we come across many problems in which stress
strain and strain displacement relations are not linear.
As the computer facility is increasing, the researchers are
taking up rigorous analysis of structures incorporating actual
stress strains curves and changes in geometry due to loading.

Non-linear Problems
Various non-linear problems in finite element analysis
may be grouped into the following three categories,
the basis being the sources of non-linearities:
Material Non-Linearity Problems
Geometric Non-Linearity Problems and
Both material and Geometric Non-linearity Problems.

Material Nonlinearity
The stress-strain relation for the material i.e. the
constitutive law may not be linear and may be some times
time-dependent too.
Apart from these basic nonlinear relations, there are time
dependent complex constitutive relations like plasticity,
creep which make the problem non-linear.

Geometric Non-Linearity Problems

In many problems strains displacement
relations are not linear. They need
consideration of actual strain
displacement relations as shown

Large deflection problems like the analysis of tension structures and post
buckling studies of beams, plates and shells also fall under this category.

Analysis of Nonlinear Material Problems

For many materials non-linear stress
strain curves have been obtained
experimentally by conducting uniaxial
tests. These results are enough to
carry out non linear analysis of
homogeneous materials.
When a structure is loaded stresses
are different at different points.
Hence Youngs Modulus is different
from point to point. For finite element
analysis a single value is assumed for
an element and element to element
the value may be different

Analysis of Nonlinear Material Problems

The following three methods are available in literature
for handling material non-linear problems:
1. Incremental procedure
2. Iterative procedure and
3. Mixed procedure

Analysis of Nonlinear Material Problems

For many materials non-linear stress strain curves
have been obtained experimentally by conducting
uniaxial tests. These results are enough to carry out
non linear analysis of homogeneous materials.
When a structure is loaded stresses are different
at different points. Hence Youngs Modulus is
different from point to point. For finite element
analysis a single value is assumed for an element and
element to element the value may be different

Direct Solution Technique

Thank You