Dynamics of Structures

A Two Day Course at

Syncrude Canada Ltd.

A E Elwi, PhD, PEng

Professor of Civil Engineering
University of Alberta
April 2004
 Outline

• Basic dynamic theory for single degree of

freedom systems
• Formulation, concepts of mass, stiffness and damping
• Free Vibrations
• Dynamic loads
– Harmonic loads
– General dynamic loads such as blast and impact loads
• Forced responses of single degree of freedom systems to
harmonic loads
• Resonance

3/1/2018 2
 Outline (continued)
• Basic theory for multiple degree of freedom systems.
• Stiffness and mass matrices
• Modes of vibration
• Damping in the context of multiple degree of freedom systems.
• Solution techniques for multiple degree of freedom systems
– Modal superposition
– Time step (explicit and implicit techniques)
• Dynamic measurements and processing
• Filtering techniques.
• Examples and a mini workshop

3/1/2018 3
 Dynamic forces
Most structures and many engineering systems are subject
to some form or another of a dynamic force; a force that
varies with time. If the dynamic force is substantial it can
cause the system to respond in a dynamic manner; the
system moves with time. In the context of an operation
such as Syncrude, dynamic forces come from a variety of
sources:
– Machine imposed forces:
• Harmonic forces
• Periodic non-harmonic forces
– Operation forces such as impact of falling ore
– Possible blast loads
– Wind loads

3/1/2018 4
 Motion of the System
• Naturally the system responds to dynamic forces by moving in time. Thus
the motion of the system is described in terms its:
– Displacement. Alternately we call it the deformed shape or position.
We will refer to it with the generic name x

d x
– Velocity: v = x = m/sec
d t
 2
d x
– Acceleration: a= x = m/sec/sec
2
d t

• All three are functions of time t, x = x(t), v = v(t), a = a(t)

• The displacement governs the strains and hence the stresses and the
finally the structural integrity of the system.
• The velocity and acceleration govern the threshold of perception of
motion and thus the safety and level of comfort of the operators for
example.

3/1/2018 5
 Resistance to motion
• Resistance to motion comes from three sources:
– Energy storage in the form of elastic strain energy, which is
a function of geometry and material properties of the system
as well as the deformation
– Energy dissipation or the so called damping, which comes
from several sources such as friction, plastic deformation,
drag, cracking, etc; none of which is mathematically friendly.
Therefore, we usually represent dissipation with a viscous
damping force that is a function of the system velocity.
– Inertia of the system, which is s function of its mass and
mass distribution as well as the acceleration of the system

3/1/2018 6
 Single Degree of Freedom Systems
(SDOF)
• The simplest dynamic system is a single
degree of freedom system consisting of a
mass, m, a spring with an elastic stiffness k,
and a dashpot viscous damper with a damping
coefficient c.
c

k
m

3/1/2018 7
 Forces on a SDOF System
• There are four forces applied to any single degree of freedom
system:
• Elastic force Fe, which is a function of the elastic properties of
the system represented with a spring of stiffness k and the
displacement of the system x:
– Fe = kx
• Damping force, Fd, which is assumed to arise from a viscous
damper with a damping coefficient c and the system velocity v:
– Fd = cv
• Inertial force, Fi, which is a function of the mass of the system m
and the acceleration a:
– Fi = ma
• Forcing function (THE DYNAMIC LOAD), which we will call F

3/1/2018 8
 Equation of Motion of a SDOF System
• All four forces are functions of time and together they are
assumed to obey Newton’s second law of motion. So that:
– Fi + Fd + Fe = F or
– ma + cv + kx = F or
.. .
– m x + c x + kx = F .

• This is the differential equation of motion of a SDOF system. it is

non-homogeneous initial-value second-order linear ordinary
differential equation with constant coefficients.
• In order to solve it (obtain the motion) you need to know the two
initial conditions, which are the initial displacement, xo, and
velocity, vo at time t = 0.

3/1/2018 9
 Free motion of a SDOF System
• If one sets the right hand side of the equation
to zero (no dynamic load) and then gives the
system an initial displacement and velocity, xo
and vo, The system obviously has an initially
stored strain energy and initial damping force.
It may even have an initial inertial force.
• The system will then move to keep all three
forces in equilibrium.
• This motion may or may not be a vibration
motion depending on the properties of the
system.

3/1/2018 10
 • The equation of motion is thus written as: .. .
m x + c x + kx = 0.
..
• the characteristic equation is written as: m r 2 + cr + k = 0.

• which has roots of the form: -c

( ) - mk
2
c
r= 
2m 2m

• Since the three quantities m, c and k. are positive

independent physical quantities, the form of the solution will
obviously depend on whether the quantity under the square
root is negative or not.
• If not, the roots are real and the solution decays
exponentially and there will be no vibration. We call that
system over-damped.
• If it is negative the roots are imaginary and the solution is
periodic (harmonic), which we call a vibration. We call that
system lightly damped.

3/1/2018 11
 Free Vibration with Zero Damping
• This is the most interesting
case.
• If one sets the damping
coefficient c=0, the solution
will take the form
• In which w is called the x = A sin(wt ) + B cos(wt )
system angular velocity
• and A and B are constants k
w=
that depend on the initial m
conditions.
• We call this response a
simple harmonic motion

3/1/2018 12
 Simple Harmonic Motion
x = r cos(wt - q )
• After some manipulation
 
2
the solution takes the v
r = x + o

2
form:
w 
o

 v 
q = tan  
-1 o
• Clearly this motion is a w x 

o
periodic motion
(vibration) with 1.5

– an Amplitude of r
1
r
0.5

– a Natural Period T = 2p/w 0

x
0 5 10 15 20 25
-0.5
– A Frequency f = 1/T or w/2p -1 T

– and can be plotted against -1.5

time (sec)
time as:

3/1/2018 13
 Simple Harmonic Motion
• Clearly now we have certain insight into the system:
• If you increase the mass, m, the angular velocity w
becomes smaller, the natural period, T, becomes larger
and frequency, f, becomes smaller, i.e. the system slows
down.
• If you increase the stiffness, k, of the system, the angular
velocity becomes larger, the natural period becomes
smaller and the frequency becomes larger, i.e. the
system speeds up.
• The amplitude of the motion depends largely on the
initial conditions.

3/1/2018 14
 Lightly Damped Vibration
• Most structural and mechanical systems c
x = e 2 m t [ A sin(w D t ) + B cos(w D t )]
-
are lightly damped: c2 < 4km

• In this case the roots of the characteristic

()
2
equation are complex and the motion k c
takes the form: wD = -
m 2m

• Here we have a combination of a simple

1.2
harmonic motion multiplied by a decay 1
function and can be plotted against time 0.8
as: 0.6
0.4
0.2

x
0

Here wD is called the damped angular

-0.2 0
• -0.4
50 100 150

velocity. The system still has a constant -0.6

-0.8
natural period TD, which is longer than T -1
and a corresponding frequency fD which Time (sec)
is smaller than f.

3/1/2018 15
 Lightly Damped Systems
• As the damping coefficient c increases wD slows down and when it is equal
to zero it is said that we have critical damping.
• The critical damping coefficient cc is written as:
cc2 = 4km or cc = 2mw
• This is important because we usually describe the damping of a system as
a fraction of the critical damping coefficient and we call it the damping ratio:

z=c/cc

• Examples:
– Lightly cracked concrete systems: z= 2-3%
– Heavily cracked concrete systems z= 5-10%
– Undamaged steel structures z= 1-2%
– When z >1.0 there is no vibration at all

3/1/2018 16
 Units
• Try to stay consistent:
– Masses:
• linear motion: kg Angular motion: kgm2
– Stiffness:
• linear stiffness: N/m Radial (angular) stiffness Nm/radian
– Damping Coefficients
• N/(m/sec) (Force per unit velocity) Nm/(radian/sec)
– Motion:
• Displacement: m radians
• Velocity: m/sec radians/sec
• Acceleration: m/sec2 radians/sec2
– Frequency Hz (Hertz) or cycles per second which is the same.
Sometimes it is given in RPM (revolutions per minute)

3/1/2018 17
 Forced Motion of a SDOF System
• If we rewrite the full equation of motion in terms of the damping
ration it takes the standard form:

.. .
x + 2zw x+w =F
2
/m

• The right hand side could take a number of forms

– Harmonic forces such as arise mostly from reciprocating machines and
rotary machines
– Periodic but non-harmonic signals such as arise from presses and
machines that involve regular impact forces.
– General dynamic forces, which could be short term such as blast loads,
impact of falling objects and seismic forces, or long term such as wind
loads or loads arising from gravity flow of fluids or granular material.

3/1/2018 18
 Harmonic Forcing Functions
• Such a force is described often as a sine function of time as:
F(t)=Fosin(wt)
• It has a magnitude Fo and an angular velocity w in radians /sec.
Often, however, one describes the machine velocity in RPM so that

w = 2p*RPM/60
• The ratio of the applied angular velocity w to the natural angular
velocity of the system, w is most important. We will call that ratio

b = w/w

A fast forcing function would have b greater than 1.0, while a slow
forcing function would have b less than 1.0

3/1/2018 19
 • The motion has two parts
– A Transient part, which fades if there is damping present in the system.
– A Steady State part, which may be affected by damping but will always be there
as long as the forcing function is active.
• The steady state part which is the interesting one under light damping takes
the form:

F 1  
x = o
sin  w t - q 

(1 - b )
 - 
SS s

(2zb
K
)
2
2
2 +

• The first term is the static deformation.

• The second term is a dynamic magnification factor
• The third term is a time function, which is obviously linked to the forcing
function angular velocity

3/1/2018 20
 Dynamic Magnification Factor
• If one plots the dynamic
magnification factor
DMF against b you
something like that:
• Notice that for slow
systems the DMF is
always greater than 1.0,
while for fast systems,
say b>1.42 the DMF
becomes small very
quickly.
• At b=1.0 we get what we
call a resonance
condition, where DMF
becomes very large
especially for very lightly
damped systems

3/1/2018 21
 Beating
• As b approaches 1.0 an
interesting phenomenon
takes place where the
20.0000
vibration magnitude
builds up and dies down 15.0000

only to repeat the pattern

displacement(mm)
10.0000
regularly. The motion is 5.0000
called a beating motion 0.0000
and is a sure indicator -5.00000.0000 0.2000 0.4000 0.6000 0.8000 1.0000 1.2000
that somewhere in your
-10.0000
building or your machine
-15.0000
a resonance condition is
taking place and could -20.0000

lead to a localized failure time(sec)

at one point or another.

3/1/2018 22
 Simple General Dynamic Forces
• Simple general dynamic forces are those that can be easily described
mathematically such as forces that rise on a straight line or single triangular
signals, or single sudden rectangular pulses.
• In these cases the equation of motion can be integrated directly and many
books such as Biggs or the Handbook of Vibration include charts that give
you the dynamic load factor as a function of the duration of the load for
example relative to the system natural period of vibration.

• At the end of such single pulses, the motion is a free vibration motion with
the initial conditions being the conditions at the end of the pulse.
• These cases are important to some of you because they approximate the
effect of a falling object such a rock on a vibrating screen.

3/1/2018 23
 Impact Forces
• Impact forces are rapidly rising and falling pulses of a very short duration
relative to the natural period of the system. These are best described in terms
of the so called Duhamel Integral.
• The concept here is that a pulse introduces a change in momentum.

td
I (The Impulse) = m D v = 0 F(t )dt

• The integral sign merely indicates the area under the load time diagram, which
we call the impulse.
• The term Dx is a change in velocity. If the motion is traced beyond the impulse
duration in the form of a free vibration motion it would take the form:

( ( ))
x(t ) = 1 I sin w t - t d
mw

3/1/2018 24
 Continuous Systems
• Simple beams, cantilevers and columns and even beam columns
with simple boundary conditions are examples of continuous
systems. The motion of such systems is governed by a fourth order
partial differential equation of the form:

EI Y + m Y = F (x, t )
iv

• Here Y is the motion and is a function of time as well as position

along the beam or column.
• Solutions to equations such as these are plentiful in old text books
for simple loads and simple boundary conditions. The more
interesting solutions, however, deal specifically with free vibrations of
such systems.

3/1/2018 25
Free Vibration of Simple Continuous Systems

• Some of the more important properties of the solution

are as follows:
1. Each solutions is composed of a mode of vibration, which is a
continuous deflected shape that satisfy the boundary
conditions. We call each of these a mode of vibration.
2. There is an infinite number of modes of vibration for any single
continuous structure.
3. Each mode has a specific angular velocity and a
corresponding natural period.
4. The slowest mode with the largest natural period is called the
fundamental mode and it is the one that usually dominates the
response of the system under generalized excitations.

3/1/2018 26
 A simple beam with a distributed mass
• The solution for a simple beam with length L
and mass m per unit length, moment of inertia I
and made from a material with E Young’s
modulus and with zero initial velocity has the
form:
 npx 
( )
Y ( x, t ) = Y cos w t sin  
n n
 L 
• The term Yn is called the modal amplitude for
mode number n and can be obtained from the
initial conditions.
• The term wn is the angular velocity for mode n
and has the form below. Note that as the mode
becomes more complex it becomes also faster.

 np 
2

wn = 
EI
 l  m

3/1/2018 27
 Multiple Degree of Freedom Systems (MDOF)
• In real life most structural systems are continuous
systems. However, modern structural analysis theories
such as Direct Stiffness and Finite Element approaches
allow us to describe many of our systems in the form of
discrete degrees of freedom, joined by continuous
elements.
• The fundamental properties of the structural response do
not change, however, they are only approximated.
• In the SDOF system we had one degree of freedom, x.
In a MDOF system you may have as many as you need
to approximate the system, say N.

3/1/2018 28
 Stiffness Matrix
• In a SDOF system the relation between the force and the DOF is
described in terms of the spring stiffness ,k. In a MDOF system, the
spring is replaced with a matrix of coefficients which we call the
structure stiffness matrix and we will call [K]. It is a square matrix
with a size of NxN.
• The stiffness matrix is a mathematical model holding information on
the geometry of the structure, the sizes of the element, cross section
information, material properties and boundary conditions.
• If we allow large deformations it also holds information on the
current state of stress in the structure.
• It tells us how the structure deforms when forces are applied to it.
• Finally it is normally positive definite meaning that all the elements
on the main diagonal are positive and large relative to off diagonal
elements

3/1/2018 29
 The Mass Matrix
• In a SDOF system the mass is a single number. In a MDOF system,
each degree of freedom could have a mass. We group all the
masses in a matrix, which we call the mass matrix [M].
• The mass matrix is also a square matrix of a size NxN.
• The most general finite element description allows six degrees of
freedom to each node; three translational and three rotational.
• There are two types of mass matrices; Lumped and Consistent
• The lumped mass matrix is a diagonal matrix that associates all
body masses to all translational degrees of freedom. Only large
rotary masses are associated with rotational degrees of freedom as
you see fit.
• The consistent mass matrix is more complex and allows the
association of masses with rotary degrees of freedom even under
small deformations. Unless you really want it will complicate your
life.

3/1/2018 30
 Damping and the Forcing Functions

• Likewise damping is described in terms of a

damping matrix [C]. We will talk about it in detail
later. For now assume a square matrix with a
size of NxN

• You may apply any number of forcing functions

to the DOFs. All these functions are grouped
together in a vector we will call the force vector
{F}, which has a size of Nx1

3/1/2018 31
 The Equation of Motion of a MDOF System

• The equation of motion of MDOF systems takes a form very similar

to the one for SDOF systems:

       
[M ] D + [C ] D + [K ] D =  F 
       

• Here, there is an associated vector of accelerations, velocities and

displacements (which we are going to call D instead of x) with each
DOF.
• The equation describes a mathematical model of the response of
the structure. In a full course we would call it a coupled set of linear
ordinary second order differential equations. The size of each matrix
NxN is and the size of each vector is Nx1, where N is the number of
degrees of freedom.
• Clearly solving such a set of equations is not a simple exercise.

3/1/2018 32
 Free Vibration of a MDOF System with zero Damping
• If one sets the forcing functions on
      
the right hand side and the
damping matrix to zero, the [M ] D + [K ] D =  0 
equation of motion takes the form:      
• In order to obtain a solution we
must make an assumption. Here
we assume that the solution will {D}= cos(wt ){f }
take a form similar to that of a
SDOF system so that:
• Here {f} is a vector that holds the
   = - w 2 cos(wt )f 
•
magnitudes of each DOF.
If we differential twice we get an
D 
expression for the accelerations
as:
• If we substitute and simplify we
get the equation of motion in the
form: [[ K ] - ω 2
[ M ] ]{ f}cos ( ωt )= 0

3/1/2018 33
• The cosine term cannot be equal to zero. The rest of the equation is
a homogenous set of linear algebraic equations, which can only
have a nonzero solution if and only if the determinant of the matrix of
coefficients is equal to zero.

• Setting the determinant equal to zero results in an N dimensional

polynomial in w2. This gives us a set of N positive and real solutions
for w2. With each w there is an associated vector of displacement
shapes {f} which we are going to call a mode. Each mode has a
distinct shape but no known magnitude.

• Thus, we have N modes of vibration, each has a natural period and

a frequency. We call each pair, {f}i and wi, a eigen pair.

• The mode with the slowest w is called the fundamental mode. And
under most generalized loads, it dominates the behaviour.

3/1/2018 34
Example
m=600 units
K=600 units
• The most famous m=600 units
K=1200 units
example used in almost m=600 units
K=1800 units
every text book is the
three story frame:   3000 - 1200 0  600 0 0   d 1  0 
  2    
 - 1200 1800 - 600  - w  0 600 0   d 2  = 0 
  0 - 600 600   0 0 600   d 3  0 
• The equation of motion
takes the form: w 1 = 0.645 rad/sec w 2 = 1.514 rad/sec w 3 = 2.508 rad/sec
d 1 0.215 d 1  0.505  d 1  0.836 
• It turns out to have three            
d 2 = 0.493 d 2 =  0.683  d 2 = - 0.539
solutions: d  0.843 d  - 0.527  d   0.102 
 31  32  33
• If we draw the modes
they would take the form:

3/1/2018 35
 Properties of the modes
• The eigen pairs have many properties. The most important of
which is that the modes are distinct and orthogonal with respect to
both the mass and stiffness matrices.
• Mathematically we say that for two different modes i and j:
– <f>i[M]{f}j=0 and <f>i[k]{f}j=0
• This allows us to normalize the modes to make some sense out of
a bunch of numbers. This is done as follows:

{f }= {f } i
f i [M ]{f }i = 1.0 Please note that the
f [M ]{f }
so that
i
i
units of these modes
i
are not in

f i [K ]{f }i = w i2
displacements but in
It follows that kg-1/2
for linear masses

3/1/2018 36
 Actual Practice
• Theoretically, any motion of a MDOF system can be composed from the
summation of the modes each multiplied by a modal magnitude and a time
function for example:

• {D} = S{Ai {f}i cos(wit-qi)}

• The magnitudes associated with slow modes are generally high, and
become very small rapidly for faster modes.
• The reason is that as the mode becomes more complex it demands more
energy to excite. The magnitude, therefore, cannot be too large unless it
resonates with a forcing function. In this case the effect becomes more
acoustic than structural.
• An acoustic effect is not to be discounted, because it is iwithin the threshold
perception and can be quite disquieting.

3/1/2018 37
 Actual Practice
• The implications are that when you attack a MDOF system using a
computer code, you need only obtain a few modes. How many to obtain
depends on your experience with this type of a structure:
– For responses of systems to seismic and wind loads, only the first two or three
modes are important.

– For responses to harmonic and periodic loads you need the modes with a
frequency around that of the forcing function say from b=0.5 to b=1.5. To zero in
on those, some programs will ask you about the shift. The shift is intended to
position you close to the operating function frequency. Experiment with it. If you
are not sure, it is always safe to evaluate the first few modes and find out where
you are.

– Other programs will ask you about the cutoff frequency which is self explanatory.

– Modes mean nothing in so far as the initial response to blast loads is concerned.
You will however, need to evaluate more modes than usual; say the first twenty
or so for subsequent free vibration motion after the end of the blast.

3/1/2018 38
The UE1 Centrifuge Building

• A full model of the

centrifuge building
looks like this:

Mass % Participation Factor

Frequency
Mode % error x y z x y z
(Hz)

1 0.94 0.0001 1.3829 78.1979 0.0001 -226.955 1,703.78 -1.7758

2 1.277 0.0017 86.0335 3.2368 0 1,790.12 346.6372 0.4678

3 1.4682 0.0037 8.1156 0.8929 0.0001 -549.805 182.0595 -1.9487

4 1.7602 0.022 0.646 10.8707 0.0002 -155.117 635.2502 -2.6162

5 1.8545 0.0162 0.0006 0.0632 0 4.6617 -48.4364 0.0974

6 1.8574 0.0085 0.0002 0.0175 0 2.7186 -25.5177 0.0573

3/1/2018 39
3/1/2018 40
 Damping
• Having obtained the modes you need, you may wish to take a stab at evaluating
damping. There is no sure or scientific way to evaluate damping for a structure that
you have not built yet!!.
• However, one can estimate the damping associated with the modes as a starting
point. In this case evaluate the damping ratios associated with the first two modes.
The rule of thumb is that as the mode becomes faster damping increases. So if you
assume that the first mode has 1% damping the second could be 3%, maybe or
whatever.
• Past this point most practitioners invoke Rayleigh type damping, in which the
damping matrix is a function of the mass and stiffness matrices:

[C] = a1[M] + a2[K]

• In which a1 and a2 are constants evaluated using the orthogonality properties of these
first two modes and two equations in two unknowns:
a1 + a2w12=2w1z1 and a1 + a2w22=2w2z2
• The damping ratios for other modes are evaluated using the now known a1 and a2 as:
zi =(a1 + a2wi2)/2wi

3/1/2018 41
 Example
• A system has the following properties:
 w1 = 2.0 rad/sec, w2 = 3.0 rad/sec
• The damping ratios for these two modes are:
 z1= 0.02 and z2 = 0.10
• It follows that
a1 + 4 a2 = 0.08
a1 + 9 a2 = 0.60
• When we solve for a1 and a2 we get:
 a1 = -0.336 and a2 = 0.104
• If w3 was equal to 5 rads per second, then z3 would be equal ro:
 z3=(-0.336 + 0.104*52)/(2*5) = 0.226 or 22.6%

3/1/2018 42
 Forced Motion of MDOF Systems

• There are two primary methods to tracing the

motion of MDOF systems:
– Modal superposition
• This method is particularly suited to linear elastic systems
subjected to harmonic type forcing functions.
– Time step integration
• This method is much more general in the sense that it can
handle any type of load and any system formulation from
linear-elastic to nonlinear-elastic to nonlinear-in-elastic but at
a cost.

3/1/2018 43
 Modal Superposition
• The primary assumption is that any motion of a MDOF system can
be written in terms of a summation of the modes each multiplied by
a magnitude and a time function:
{D} = [F]{Y}
• Here Y is a set of time functions as of yet unknown and [F] is a
matrix holding the modes that you have decided to use.
• If you differentiate once, you get the velocities and if you
differentiate twice you get the accelerations:
       
  = [F ]Y  and   = [F ]Y 
D D
       
• Theoretically, one should use all the modes. But since we have
agreed that only a few modes are important, you may only wish to
use those modes, say 3 or 4 or 20, which is better than 20,000.

3/1/2018 44
• Invoking the orthogonality properties of the modes we know that:
– [F]T[M][F]=[I]
– [F]T[K][F]=[w2]
• The matrix [I] is a diagonal identity matrix, while the matrix [w2] is a
diagonal matrix holding the squares of the angular velocities of the
modes you have decided to use.
• Now recall the equation of motion: [M]  + [C ]  + [K ]  =  
    
 D  D  D F 
• Substitute for the displacements, accelerations and velocities:
     
[M ][F ]Y  + [C ][F ]Y  + [k ][F ]Y  = {F }
     
• Now pre-multiply by [F]T, and invoke the orthogonality properties
and Rayleigh’s damping above. You will find a very nice
construction:

   
Y  + 2zw
  

[ ]
   + 2   = {}
 Y  w Y 
   

3/1/2018 45
• This is a set of decoupled second order linear ordinary differential
equations with constant coefficients just like the ones we used in
the SDOF system part, and for which solutions exist in every text
book.
• The right hand side represents a vector of modal forces.
• Each equation represents a modal equation of motion, which can
be written as:
 + 2
Y i z i wY
i
 + 2i
i w Y i
= i

• Once you have obtained the solutions for the functions Y, you can
now go back to the very first assumption and obtain the complete
motion as:
{D} = [F]{Y}
• Better yet obtain the modal motion for each mode as:
{D}i = {F}iYi

3/1/2018 46
 Superposing the solutions
• You may add t he solutions directly to obtain a full history of motion.
But this is of little value for the practitioner. A practitioner is
interested in maximum responses.
• The maximum modal motion for any mode is pegged at the
maximum value of the function Y, which allows you to obtain the
maximum modal displacement, bending moments, shear forces and
axial forces everywhere for every mode separately.
• For any one element in your structure you can mix the maximums
using any one of a number of methods such as summation of
squares.
• For example if at a particular element the maximum modal bending
moments from three modes are 370 kNm, 200 kNm and 15 kNm.
The design moment becomes:
•
370 2 + 200 2 + 152 = 421 kNm

3/1/2018 47
 Ritz Vectors
• The method of modal superposition is a subset of a mathematical
group of solution strategies called the Reduction methods.
• In these methods it is not necessary that one uses the modes. Any
set of vectors that are orthogonal to the mass and stiffness matrices
can do the job.
• Wilson has proposed the so called Ritz vectors, which is a set of
approximate vectors closely related to the loading functions be used
in place of the modes of vibration.
• For one thing they are easy to obtain and for another they are close
to the type and nature of applied loads.
• They must not, however, be mistaken for modes of vibration. They
are not intended for harmonic loads, but rather for seismic analysis.
• Lastly, not every program has them. Sframe does not, but Staad 3D
does and naturally Sap2000. Higher order codes such as Abaqus
would not stoop to such approximations, but Ansys may.

3/1/2018 48
 Direct Time Step Integration Methods
• All time step integration schemes proceed from a known point in
time and assume that the previous history of motion is known
including displacements, velocities and accelerations. All the
methods then proceed step by step to predict the motion in
successive time steps.
• Thus by definition all these methods are approximate. In other words
from a purist point of view there will be errors.
• This should not scare anybody, the question is how big is the error
and whether the error grows with time.
• The first question is easily answered accuracy must depend on the
size of the time step. The smaller the step, the more accuracy you
obtain.
• The second question is more complex. Whether the error grows
depends on the stability of the method. The available methods could
be either conditionally or unconditionally stable.

3/1/2018 49
• These are very versatile methods and they come in a variety of
flavours. The scientific flavour is based on whether the method is
implicit or explicit.
• Explicit methods write the equation of motion at the current time and
try to project approximately what the solution may be after a short
time step later and then proceed from step to step. They are easy
use, computationally very efficient and fast, but they are only
conditionally stable. This means that the time step size must be less
than a given critical size for the error to be bounded.
• Implicit methods write the equation of motion at the step you want to
move to and are therefore more accurate, but they are
computationally more demanding.
• The most commonly used explicit method is the method of Central
Differences
• The two most commonly used implicit techniques are the Newmark
Beta and Wilson Theta. Both methods are present in Sap2000 and
Staad 3D, but only the Newmark Beta is present in Sframe.

3/1/2018 50
 Central Difference Scheme
• This is the method of choice in nonlinear codes such as ABAQUS, ANSYS and
DYNA3D.
• It is based on the assumption that he acceleration is constant in a given time step.
• It requires a special starting procedure, which is transparent to the user, but,
nonetheless, is an added approximation.
• Because it is explicit it is conditionally stable. It is necessary, therefore to keep the
time step smaller than a critical value:
Dt < TN/p
• Here TN is the fastest natural period of the structure, which in any significant
structure is very hard to evaluate.
• Luckily we know that the natural period of any single element must be shorter than
the that of the structure. Therefore, you may wish to evaluate the smallest possible
natural period for the smallest element in your structure. This is easier but may
also be prohibitive.
• The accepted method is that the time step must be smaller than the time it takes
sound to travel across the smallest element.
• The speed of sound in your structure is easy to evaluate. E
=
• E is Young’s Modulus and r is the density of the material. r

3/1/2018 51
 Newmark Beta and Wilson Theta
• Newmark Beta:
– Assumes that the acceleration varies linearly in any time step.
– It is an implicit scheme, taking more computational effort but is more
accurate than the central difference scheme.
– This is the method of choice in Sframe.
– I would not want to change the default values for Newmark constants
unless you really know what you are doing.
• Wilson Theta:
– Is of the same variety, but is unconditionally stable. Thus you may use
much larger time steps than you would otherwise in the other two
methods.
– If you choose to use large time steps it only means that the error will not
grow, but there will still be an error.
– This is the method of choice in SAP2000.

3/1/2018 52
 Dynamics of Machine Foundation
• Foundations subject to significant dynamic forces can lead to some
spectacular failures.
• Large reciprocating compressors typically large size equipment supported
well above ground surface to allow clearance for piping etc. While the
dynamic forces by definition are large and ongoing , the motion tolerances
are small.
• By contrast large turbines pumps and centrifuges are typically rotary
equipment that have very little dynamic effects when they are balanced.
However, they could generate fairly large forces if they are unbalanced.
• Normally such large equipment are supported on large rigid foundations
either directly on the ground or on piles below the foundation block.
• On the other hand when the equipment is supported on flexible structural
floors, say a steel platform controlling the vibrations transmitted to the floor
becomes the objective of the practitioner.
• In the following we will discuss foundation blocks on the ground and on piles
and we willallude quickly to vibration control problems.

3/1/2018 53
 Foundation Blocks on the Ground
• Large foundation blocks for large
reciprocating compressors are typical of
such situations. The machine and the
block form a rigid massive system which
can be simulated as a single mass with
a shape and a centre of gravity.
• This mass has six degrees of freedom;
three translational and three rotational.
The three rotational degrees of freedom
are two rocking motions about horizontal
axes, while the last is a twisting motion
about the vertical axis.
• Typically we assume the motions of all
six degrees of freedom to be completely
decoupled, so that each can be a SDOF
system having a mass, a spring stiffness
and associated damping. The spring
stiffness and damping come from the
ground under the machine block.

3/1/2018 54
 Machine block and ground properties
• The solutions worked out for the properties of the ground are based on differential equation
solutions to the governing partial differential equations of a an infinite hemispherical half space.
So the first thing you want to do is transform the foot print of the machine to an equivalent circle
of a radius r:
r = A/ p for vertical and horizontal motions
r = 2IP/p for Twisting about the vertical axis
r = 4 4I/p for rocking about the horizontal axes
Centre of mass of
block and machine
Height above ground: h
Equivalent radius: r

• This allows you to determine the mass effects. You will need:
– Translational masses which are the total mass of machine and block, m
– Torsional mass which is the torsional moment of Inertia Jf=mr2/2
– Rocking mass, which is the rocking moment of inertia Jq=m(3r2 + 4h2)/12

3/1/2018 55
 Ground and mass properties
• Then you want to estimate the ground modulus of rigidity, Poisson’s
ratio and material density. You should ask the geotechnical
consultant.
• The following table gives mass an stiffness values based on the
work of Ritchart, Hall and Woods:

Motion Mass m Spring stiffness k Damping c

Vertical m 4Gr/(1 - n ) 3.4r 2( r G) 1/2/(1 - n )

Horizontal m 32 Gr(1- n )/(7 -8 n ) 18.4(1 -n )r2( r G) 1/2/(7-8 n )

Rocking Jq = m(3r 2+4h 2)/12 8Gr3/3(1 - N ) 0.8r 4( r G) 1/2/((1 - n )(1+3(1 -n )Jq/r r5))

Torsional Jf = mr 2/2 16 Gr3/3 4(JfGr3/3) 1/2/(1+2J f/ r r5)

3/1/2018 56
 Example
• Two cylinder horizontally opposed compressor. This means that only twisting and rocking about
one axis are present
• G = 69 MPa, n=0.42, r=2160 kg/m3
• Footprint of foundation 3.66 m x 10.7 m
– r = 4.04 m for torsion
– r = 4.67 m for rocking
• Torsional Inertia: Jf = 2.34x106 kgm2 Torsional spring: kf= 2.42x1010 Nm/radian

1 kf
• Torsional frequency: f f = 2p = 16.2 Hz
Jf
• Rocking Inertia: Jq = 2.78x106 kgm2 Rocking spring: Kq=3.23x1010 Nm/radian

f q = 2p k q
1
• Rocking frequency: = 17.2 Hz
Jq
• Now if the Machine is operating at 630 RPM, (10.5 Hz), the corresponding dynamic
magnification factors would be:
bf = 1/(1-(10.5/16.2)2) = 1.72
bq = 1/(1-(10.5/17.2)2) = 1.59

3/1/2018 57
 Pile Foundations
• The foundation block over piles is normally as large as that on ground
if not even larger. Therefore, what we have a rigid mass setting on
piles, which can be treated as a set of springs.
• If one now uses sets of inclined piles to resist horizontal loads, the
problem reduces to evaluating the axial stiffness of the corresponding
groups of piles.
• Things are not always as simple as that, because of the interaction
between piles in a group. Particularly, when you are dealing with
friction piles. Much research is going on now on that topic. A list of
references has been distributed.
• There are a number of approaches:
– The simplest is the “equivalent cantilever method”, which assigns the
pile group stiffness and damping properties. The system is thus reduced
to a series of SDOF systems much similar to foundation blocks on ground.
– More accurately, you may wish to construct a finite element model
including springs that model the relation between piles and the ground as
well as in-between piles relations.

3/1/2018 58
 Vibration Control

• We are not going to talk about that in a great

detail. However the field is broken up as follows:

Motion of Motion of support

mass on a under mass, Such as
support in seismic design

Active control Passive control

Including a
feedback loop,
which adjusts Base Isolators Tuned Mass Dampers
the control
parameters

3/1/2018 59
 Base isolators under rotating masses
• Rotary machines such as pumps and centrifuges
normally impart very little dynamic force to supporting
systems. But when unbalanced, the effects can be
devastating. Normally yu need a base isolator. An
unbalanced rotating mass can be modeled as follows:
•
•
•
•
•
•
3/1/2018 60