STRUCTURAL

DYNAMICS
T K Datta
 Foreword

 Slides are prepared with a lot of text material to help young teachers to teach the
course for the first time.
 Slides are comprehensive and almost self-contained.
 Slides contain solved problems.
 Slides can be used to teach a first course on structural dynamics and earthquake
engineering.
 The lecture notes based on which slides are prepared are available in SCRIBD.
Lecture 1
 Oscillation (Vibration) Lecture 1-1

to and fro
motion
W
u t

 Inherent damping in the system makes free oscillation diminish with time
 Oscillation makes displacement time dependent i.e., u(t); hence, u (t ) and u(t )
 u(t ) gives rise to inertia force mu(t ) according to Newton’s second law of motion
 u(t ) acts opposite to the direction of u(t)
 All the three quantities are important in oscillatory body signifying the dynamics of
the system
 Oscillation has a static equilibrium position; it has also a stable position
 Under certain conditions, oscillation can be unbounded, bounded and steady
 Oscillation of elastic bodies creates, to and fro movement of the molecular particles
 Lecture 1-2
Simplest Oscillation (SHM)
X

A Amplitude, A
 x
Time

Time period, T NOTE

2
  t 
T
Displacement
x  A sin t
Velocity dx
x   A cos t
dt
Acceleration
d2 x
x  2   A 2 sint   2 x

dt
 Lecture 1-3
Classifying different dynamic loads

Nature Sources of Dynamic Loads Long Duration

(T > 5 min - 10 min)
o Earthquakes
o Wind o Oceanic Waves
o Wave o Cyclones
o Land Slides/ Avalanches o Machine Vibrations
o Caving in of Earth o Vehicular Movements
o Volcanic Eruptions Impact type o Volcanic Eruptions
(T<1 sec)
o Missile Impact
o Falling mass/ Collision
o Sudden blast of short duration

Man made Sources Short Duration

(T< 1 min)
o Blast/ Terrorist attack
o Collisions o Earthquakes
o Vehicular Movement/ track Train o Land Slides/ Avalanches
Dynamics o Caving in of Earth
o Machines/ Impact/ Accidents o Blast of longer durations
 Types of Oscillation Lecture 1-4

Harmonic

Amplitude

u(t) = Asin (t

Phase
+)
T

Periodic
 Lecture 1-5

Random
Irregular u

u
t

Impact
u
t
u
T T T T

t
 Lecture 1-6
Mathematical Model of SDOF

m
u k1
24 EI
K
l3 l EI m1

c1
 Lecture 1-7
Inertia Force


Mu t What is inertia?
It is the resistance of an object
Inertial Force
to change its state of motion
(magnitude and direction)
Understanding Mass in a better light…
REST IS A STATE OF MOTION WITH ZERO
VELOCITY
Newton’s First Law of
An object at rest stays at rest and an object in motion stays in motion with the same
Motion
speed and in the same direction unless acted upon by an unbalanced force.
All objects have the tendency to resist changes in their state of motion
Mass moment of
Mass as a measure of
amount of inertia
F  ma I
Inertia
TI  J


Direction is opposite to that of

motion
 Lecture 1-8
Damping Force
NOTE
Cu Dissipative
It is hard to quantify explicitly
different factors for energy loss.
Thus an approximate model
Force maybe chosen…

Viscous
Damping FD  u n Exponentially

Displacement
decaying (for viscous
Dissipative Force

For practical purposes, in the damping)

analysis of buildings, a linear
relationship maybe assumed, thus Tim
e
FD  cu
This constant of
proportionality, C
is called the
Velocity damping constant.

In reality, the dissipative force is a frequency-

dependent quantity.
 Lecture 1-9
Elastic/ Restoring Force

Ku
Elastic Force

This maybe familiar to you from the static

Elastic Force
analysis…
Nevertheless, this is also an integral part
of the dynamic force equilibrium

Displacement

For a conventional building we

NOTE will assume it to have a linear
relationship.
 Lecture 1-10
Examples of Structural Dynamics
 Buildings and other land based tall structures vibrating under earthquake
 Some types of structures vibrating in cyclonic wind
 Aero elastic instability of flexible structure such as galloping of conductor cables, flutter of
suspension bridges, torsional vibration of wings of aeroplane; vortex induced oscillation of tall
chimneys etc.
 Vibration of oceanic structures (oil platforms) due to wave induced forces
 Turbo generator structures (foundations)
 Crusher frame foundation
 Vibration of industrial structures for machine vibrations
 Vibration of bridges due to vehicular movement/ track-train dynamics
 Vibration of underground structures like, liquid storage tanks, pipelines, tunnels due to propagating
earthquake waves
 Sloshing in liquid tanks due to earthquake
 Lecture 1-11
Implication In The Structural Design

 Earthquake resistant design of structures

 Development of Impact factor in bridge design

 Design of structures for impact forces

 Design of flexible bridges for wind forces

 Design of offshore structures

 Design of buried pipelines for seismic action

 Fatigue design of tubular joints for wind and wave forces

 Design of chimneys for vertex shedding

 Design of transmission line towers for wind forces

 Design of very tall building of structures for fluctuating component of wind force
 Lecture 1-12
Future Vision

 Structural Dynamics is becoming more and more important as the slender and flexible
structures are made for economic design

 Seismic disaster mitigation is a topical subject requiring great deal of knowledge of

structure dynamics

 Similarly, protection of slender and flexible structures against cyclonic wind is an

expanding field requiring good knowledge of structural dynamics

 Reliability analysis of structures for dynamic loading such as wind, earthquake, blast etc.
is becoming important for insurance companies

 Structural control for environmental forces

 Deep water deployment of offshore structures

 Structural health monitoring

Lecture 2
 Lecture 2-1
Static Vs. Dynamics
 Static load is invariant of time and hence, the responses are also time invariant
 Static load means gradually applied load on the structure as opposed to suddenly applies load
 Force displacement relationship (defining static equilibrium) is governed by
K  F

 In which K = stiffness matrix corresponding to the kinematics d.o.f and F is the load vector
corresponding to the loads applied to the direction of d.o.f.
 Both dynamic force and displacement are time dependent; dynamic force may be of oscillatory
type or suddenly applied or moving, all producing oscillation in the system
 Because of the presence of mass in the structure and acceleration, a dynamic system is
characterized by the presence of inertia force; this force acts opposite to the motion of the body
 Because of the to & fro movement giving rise to velocity, a damping force (absorption of
imparted energy) is also present in the elastic system; the energy absorbed is transformed into
another form with the to & fro movement of the molecules
 Lecture 2-2
 Finally, as the elastic body undergoes to and fro movement, a restoring force (also to and fro)
is developed in the system

 Three types of forces apart from the externally applied force are present in a dynamic system
i.e. inertia forces, damping forces and restoring forces

 Velocity of the system is related to damping force; acceleration of the system is related to the
inertia force; displacement (time dependent) is related to the restoring force

 Kinematic d.o.f. are different than dynamic d.o.f in the sense that all dynamic d.o.f are
associated with mass; thus, dynamic d.o.f ≤ kinematic d.o.f.

 Dynamic equilibrium exists at every instant of time; dynamic equilibrium is the equilibrium of
forces between resisting (inertia, damping, restoring) and external forces at every instant of
time

 Dynamic equilibrium exists at each dynamic d.o.f.

 If inertia force (acceleration) << restoring force, it is called quasi static equilibrium and
governed by K (t )  p (t )

 In dynamic system, total energy at any instant of time is PE + KE; In static system, only PE
exists.
 Lecture 2-3
Idealization of structures
 A real life structure is idealized as 1D, 2D, and 3D discrete and continuous models

 In discrete method, the number of degrees of freedom decides the size of the problem

 Like the static analysis, the dynamic equilibrium equation of the modeled structure is called the
dynamics equilibrium equation of multi degree of freedom (MDOF)

 Single degree of freedom (SDOF) is a special case of MDOF; but they are treated separately for
the following reasons

 Analytical solutions of MDOF system can be comprehended easily if those of SDOF system, which
are simpler, are known

 Solution of SDOF system leads to the understanding of many fundamental concepts of dynamics

 Many structures may be suitably idealized as equivalent SDOF systems (two examples are shown)
 Lecture 2-4

 Restoring force represents the elastic action produced due to deformation of the
structure which tries to restore the structure to its original shape
 Elastic restoring force linearly varies with the displacement; spring force in the model
is represented by
FR  Ku
 For portal frame shown below, the stiffness corresponding to the sway displacement
is given by
3EI
K 3 EI, L
l
thus, for SDOF model, restoring stiffness is K

 For the beam behaving as a SDOF model as shown in figure below

m E
384EI
K
5l 3 u
l
 Lecture 2-5
Energy of the System
 At any instant of time t, the total energy in a freely vibrating system is made up of two
parts, kinetic energy EK and potential energy ES
1 1
mu t   k u t 
2 2
E  EK  ES 
2 2
2E k
 u t   u t   u t    2n u t 
2 2 2 2

m m
if u t   A sin  t   ;
2E
m
 
 A 2  2 cos 2 t     2n sin 2 t  
2E mA  n2 2

when ω = ωn; i.e., a freely vibrating system:  A2n2 ; E  Constant

m 2
 Thus the energy is fully conserved; and it is equal to input energy

 For systems with viscous damping, the total energy will decrease with time because
of the energy dissipation in viscous damping.

Ed   f d   cu u dt   cu 2 dt ;

t t t
t  ,
for Ed  input energy
0 0 0
 Linear and Nonlinear Dynamics Lecture 2-6

 Dynamic equation of motion which is a second order differential equation could be a linear or
nonlinear equation of motion depending upon the damping and restoring forces
 If any one of them is nonlinear i.e., restoring force varies nonlinearly with displacement or
damping force varies nonlinearly with velocity, the problem is considered as a nonlinear dynamic
problem
 Generally, inertia force remains linear i.e., it varies linearly with acceleration; however, there are
problems where inertia force is nonlinear e.g., variable added mass in offshore structure
 Solution of nonlinear dynamic equation is faced with many problems and is mostly solved by
numerical techniques
 Nonlinear dynamic problems do not have unique solutions; a number of solutions are possible
 It also gives rise to dynamic instability of the system consisting of certain undesirable response
phenomena such as divergence, period doubling, chaos, nonlinear jumps etc.
 Lecture 2-7

 When the structure goes into inelastic range, the force-displacement relationship no more
remains linear. It exhibits a hysteretic behavior as shown in the figure below

f f

f
y
fy
x x
xy x y

Variation of force with displacement under cyclic loading

f f
fy f
y

x x
xy x y

Idealized model of force Idealized model of force

displacement curve displacement curve
 Lecture 2-8
 Not only the stiffness varies with displacement but the direction changes (positive to negative or
negative to positive). Therefore, tracking the displacement is required to obtain the correct
stiffness at any state of deflection
 Representation of a nonlinear SDOF model is achieved by a nonlinear spring attached to the
SDOF, showing the nonlinear behavior of the structure
 Another kind of nonlinearity in restoring force is exhibited in which, force-displacement is
nonlinear but not of hysteretic type. This happens for geometrically nonlinear structure
(associated with large displacement)
 For loading and unloading the stiffness remains the same at any state of displacement as shown
in the figure f
Unloading
Loading
Dx
x
Loading Unloading
 Lecture 2-9
Nature of the Response of Structures

 Nature of response of the structure depends upon the state of the structure (Linear or Nonlinear)
and the nature of the dynamic force

 In certain cases, the nature of response may follow pattern different from that of dynamic force
(like blast loading)

 For long duration harmonic/periodic force, the response finally converges to the harmonic/periodic
nature of oscillation

 For the irregular dynamic force, response may be of irregular pattern different from that of force

 The nature of response under harmonic force, in the absence of damping, could be periodic, but
not harmonic, unlike the case when damping is present

 Under quasi-static action, the nature of response remains the same as that of the dynamic force in
elastic state; in the inelastic state, the two may differ
Lecture 2-10
Lecture 2-11
Lecture 2-12
Lecture 2-13
Lecture 2-14
Lecture 3
 Lecture 3-1
Vibration of Single Degree of Freedom

 For understanding the dynamic analysis of structures, study of the vibration of single Degree of
freedom (SDOF/ SDF) is of great importance because of two reasons

 Most of the analytical techniques used for the dynamic analysis of structure can be understood
properly, with the vibration analysis of SDOF system
 Many important results and conclusions drawn from them are extremely useful for
understanding dynamic behavior of structures

 The simplest dynamic system was shown before to describe different forces acing in the system;
this model is popularly known as SDOF model

 The mathematical model of the SDOF system is shown below. Along with it, is shown how is single
portal frame can be idealized as a SDOF system represented by the mathematical model
 Lecture 3-2

k u, p(t)

m
c

Spring-mass-dashpot system
Rigid beam
u, p(t)
Lumped mass
24 EI l, EI (m) All members
K 3 are inextensible
L

Idealized single frame

 Lecture 3-3
 The viscous dashpot in the figure represents the action of the damping force present in the
system

u(t)
k
M P(t)

u(t)
Ku
Mu M P(t)
Cu

 It was discussed before how the damping force is mathematically represented as an

equivalent viscous force resisting the motion of a piston within viscous fluid

 Dynamic equilibrium of the mass at any instant of time t gives rise to the equation of motion
of the SDOF system
 Dynamic Equation of Equilibrium Lecture 3-4

 It can be derived using three different concepts, all concluding that inertia force acts opposite to
the direction of motion ( or applied force)
 For oscillatory body, u is opposite to u(t) and p(t); so, Mu, Cu and Ku oppose P(t)

Mu  Cu  Ku  p(t)

 Rate of change of momentum is equal to the applied force
du
p(t) - cu ku  m  mu mu cu  ku  p(t)
dt
 D’Alambert reasoned that since the sum of the forces acting on a particle m results in its
 would produce a condition of equilibrium
acceleration u , the application of force -Mu

p(t)  f(t) - mu  0

mu cu  ku  p(t)
f(t)  -cu  ku
 The equation of motion is a second order differential equation. It can also be written as a set
of two coupled first order differential equations
 Lecture 3-5
Energy Formulation

An alternative statement is that the state of equilibrium corresponds to the state of minimum energy.
The above equation is an equation of motion for the un damped free vibration of the system. If the
damping and external dynamic force are included, then the equation of motion becomes the same as
that given by the damped equation of equilibrium. If the equation of motion is without an external force,
it is called an equation of motion for the free vibration of the system. On the other hand, if the external
force exists, it is called an equation of motion for the forced vibration of the system.
 Lecture 3-6
 State Space Equation

u  u (add an auxilliary equation)

mu  cu  ku  p (t )
c k p (t )
u   u  u 
m m m
 If a vector V is defined as

V  u 
T
u
then, the equation can be written as following matrix equation

V  AV  f
Where,

 0 1   0 
 
A k c  and f   p (t ) 
  
 m m  m 
 Lecture 3-7

 Equation is a first order matrix differential equation called state space equation of dynamic
equilibrium, since V at any instant of time t describes the complete state of the system.

 Equation is widely used in many dynamic application problems such as dynamic stability
analysis, structure control, state space analysis etc.

 Another reason for using the equation is that numerical methods for the solution of first order
differential equation are more exhaustive compared to the second order differential equation.

 Using MATLAB SIMULINK tool box, the solution of the equation is very popular.

 However, in conventional dynamic analysis of structures, the solution of the second order
differential equation is widely used.
 Lecture 3-8

 In the second order differential equation, if the damping term is removed, then the
equation of motion is called un damped equation of motion i.e.,
  ku  p (t )
mu
 Although undamped vibration is an ideal case, which does not exist in practice, the
concept of undamped vibration and the solution of the undamped equation of motion lead
to many important phenomena and definitions used in dynamics.
 Some of them include natural frequency of vibration, harmonic oscillation, periodicity,
steady state vibration, mode shapes for structures etc.
 When damping is present, the vibration is called damped vibration; damped vibration is the
practical case.
 The presence of damping in the equation of motion gives rise to a number of interesting
phenomena in dynamics, which are extremely useful in practice.
 Lecture 3-9
 They include bounded responses even at resonance, damped frequency of structure, transient
response, effect of initial condition, phase shift, impedance function etc.

 For analysis purposes, undamped and damped oscillations are separately treated, though their
solutions can be derived from one solution; similarly, both solutions are derived separately for
free and forced vibration.

 The reason for such treatment is that each type of solution gives some physically important
results which are important in explaining many concepts.

 Solutions of equations of motion consist of solving either a second order differential equation or
a set of coupled first order differential equations.

 Right hand side of the equation of motion has load, p(t) which could be of various forms as
described before.
 Lecture 3-10

Natural Frequency
 Solution of the undamped equation of motion provides the natural frequency and period of the
system
 Consider the equation

mu  ku  0

 Assuming the solution to be𝑢= 𝐴sin 𝜔𝑛 𝑡 and substituting for u and u



n 2 mA sin nt  kA sin nt  0

k
k  mn  0; n 
2
m
2
Tn 
n
Lecture 3-11
Lecture 3-12
Lecture 3-13
Lecture 3-14
Lecture 4
 Lecture 4-1
Free Vibration Problem
 The solution of equation of motion consists of two parts namely anduc (so
uh that
) final
u pdisplacement
is written as,
u  uc  u p
 uc is obtained by setting RHS as zero while, u p is obtained by choosing the solution to be of the
same nature as that of the RHS function, p(t)

 Solution for uc is called as a free vibration problem as described before since given an initial
condition denoted by u (0) and u (0) , the mass starts oscillating freely (without any applied load)

 As stated before, the free vibration problem is important in understanding a few fundamental
concepts in dynamics

 Assume u  ce t

 Substituting for u , u , and u in the equation of motion, the following algebraic equation is
obtained; the solution of the equation provides the expression for uc
 Lecture 4-2

( 2 cm   cc  kc)et  0

2 m  c  k  0

 c  c 2  4mk
 k
2m n 
m
c c2 k
  2

2m 4m m

c c2 c
   wn
2

2m 4m 2 2m

     2  n2
 Lecture 4-3
d    
2 2
n
2

u  c1e   t  id t  c2   t id t

 e   t [c1eid t  c2 e  id t ]
 e   t  A cos d t  B sin d t 
c c
 defining ccr  2mn 
2m ccr
c   ccr   2m n   n

d 2  n2   2n 2 d  n 1   2
u  e nt  A cos d t  B sin d t 
u  n e nt  A cos d t  B sin d t   e n t [d A sin d t  Bd cos d t ]

u (0)  n A  Bd t

u (0)  nu (0)
u (0)  A; B  u (0) 
d
 Lecture 4-4
 It is an exponentially decaying function. Hence, amplitude decreases with time. More
the value of  and n, more rapidly it decays
 Three conditions can arise namely, critically damped, over damped and under
damped
 When roots of  are the same, then;
 t
1, 2    ; 
n   ; d  0; u  Ae n

 Which shows that for a given initial displacement the displacement dies down
exponentially to zero without any oscillation as shown in Figure below,
 Lecture 4-5
 Further,   1 c  2mn c  this ccr damping is called critical damping,
when roots of  are real and distinct i.e. u  c e (   d ) t  c e (   d ) t
1 2
 For a given u(0) this also does not oscillate and decays down to zero as shown in
Figure, since   n ;  1 hence, it is called over damped
 When roots are complex, then an oscillation takes place as evident from the equation.
  its
This case is of importance in the free vibration of SDOF system. Since hence, n ;  is
1
called under damped. Practical values of <<1 of the order of 1-5%
 The under damped free vibration of SDOF is shown in the figure. The two constants
A, and B are evaluated from two specified initial conditions u(0) and ů(0)
 Lecture 4-6

 It can be easily shown that,

  u ( 0)

u (0)   n u (0)
B
d


n t u (0)  n u (0) 
ue u (0) cos d t  sin d t 
 d 
n t
  
 (0)  0; u  e
For, u u (0) cos d t  sin d t 
 1  2 
For small value of 

ue n t
u (0) cos d t   u (0) sin d t 
 Lecture 4-7

u (0)e nt
 For u(0) = 0 u sin d t
d
 When =0 i.e., un-damped free vibration of SDOF takes place,
u (0)
u  u (0) cos nt  sin nt
n
 The SDOF continues to vibrate harmonically without any decay of the amplification. It
can also be written as,
u  R sin(nt   )
where,
u (0) 2 n u (0)
R u (0) 
2
  tan 1

 2
n u (0)
=0 corresponds to u(0)=0; the SDOF vibrates as a sin function, as shown in the figure
 Lecture 4-8
(In lecture note, it is section 5.1)

 From the figure it is evident that,

2 2
Td  Tn 
d n

Td and Tn are called damped and undamped periods of oscillation. Energy input to free
vibration problem by giving initial conditions is,
1 1
E (0)  ku (0) 2  mu (0) 2
2 2
At any instant of time t, energy E(t)
1 1
E (t )  ku (t ) 2  mu (t ) 2
2 2
For free undamped vibration
2 2
1  u(0)  1 2  u (0) 
E (t )  k u (0) cos n t  sin n t   mn  u (0) sin n t  cos n t 
2  n  2  n 
 Lecture 4-9
(In lecture note, it is section 5.1)

 After simplification, it can be shown that

E (t )  E (0)

therefore, in free undamped vibration, input energy is conserved.

 For damped system, the total input energy gets dissipated. Expression for E(t) can be
obtained by using the corresponding values of A and B for damped free vibration. Since the
dissipation of energy is caused due to viscous damping, the dissipated energy E D is
t1 t1 t1

ED  
0
f D du  
0
(cu )u dt  
0
cu 2 dt

For t1 ED = input energy

 (In lecture note, it is section 5.2) Lecture 4-10

Determination of Damping Ratio ()

 Free damped vibration of SDOF is used to obtain the damping coefficient of the material of elastic
body.
 The stiffness of the elastic body is represented by the spring and the damping ratio corresponding
to the viscous damper represents the damping of the material.
 If amplitudes of the responses at the two successive cycles separated by the period T d are
considered, the ratio of the two responses are given by,
u (t ) e nt ( A cos d t  B sin d t )
  n ( t Td )
u (t  Td ) e ( A cos d (t  Td )  B sin d (t  Td ))

 enTd
2 2
r  log   n 
n 1   2 1  2
 Lecture 4-11
(In lecture note, it is section 5.2)

 log  is called logarithmic decrement which is measurable and is related to 

 <<1, an appropriate value of  can be determined as,

log 

2
 Otherwise, the following quadratic equation is solved to obtain .

(1   2 )r 2  4 2 2

(4 2  r 2 ) 2  r 2  0
r

4 2  r 2
 Once  is obtained, the damping coefficient of viscous damper can be calculated as
c
  ; c   ccr  2 mn
ccr
Lecture 5
Lecture 5-1
Lecture 5-2
Lecture 5-3
Lecture 5-4
Lecture 5-5
Lecture 5-6
Lecture 5-7
Lecture 6
 Lecture 6-1
Damping in Structural Systems
 As mentioned before, exact damping present in a structure is very difficult to define; only material
damping property can be evaluated experimentally; with the help of this information, damping of a
structural system is idealized using different types of model
 Under oscillation, an elastic mass under goes deformations resulting in alternate stretching and
compression of the molecular bonds; this absorbs some energy which is transformed into heat
energy; another mechanism also working at the molecular level is, friction; this also dissipates
energy
 Effects of both dissipation mechanisms are difficult to model; so experimental tests are conducted
to obtain damping property from which damping in the system is modeled
 There are different types of modeling to describe the damping force acting in structural vibration;
the type of modelling depends upon the problem at hand; apart from the linear modelling, there
exist nonlinear models of damping
 Any energy absorption device integrated into the structure can be modelled as a linear or nonlinear
damping
 Lecture 6-2
 Out of them the popular models are
 Viscous damping model
 Friction damping model
 Structural damping/hysteretic damping model

 During inelastic excursions under oscillatory motion, another kind of dissipation of energy takes
place due to the formation of hysteresis loops in force-displacement relationship; area within each
hysteresis loop represents energy dissipation; this dissipation of energy is directly included in the
analysis

 Except viscous damping model, other two types of damping and the damping due to inelasticity
can be also described by equivalent viscous damping

 The equivalent viscous damping is approximate but provides an easy solution to the problem
because of its linear characteristics
 Lecture 6-3
Viscous Damping Model
 In this model, damping force developed within the system is assumed to be proportional to the
velocity of the system like a force experienced by a piston moving within a viscous fluid

 Force developed per unit area is written as

F  Cu ; C is the coefficient of viscosity

 In structural models, this is represented by a dash pot shown before

 In one cycle of harmonic motion, the energy dissipated ED is

2 
ED   f D du  
0
cu u dt
u  u 0 sin t   
2  2 
u0 cos t    dt
2
 cu dt  c 
2
0 0

c  2mn  2   1 
  cu02  2 ku0  4 ES 0  ES 0  ku02 
n n  2 
 Lecture 6-4

 The dissipated energy varies

quadratically with the amplitude of
motion as shown in the following figure

 Graphical representation of the damping force fD

f D  cu t   cu0 cos t     c u02  u t 

2

 This can be re rewritten as 2 2

 u   fD 
    1
 u0   cuo 
 Lecture 6-5

 The equation of ellipse shown in following figure makes a loop in fD-u axes; the area of the loop
is;
  u 0  c u 0    c u 2
0

 Thus, area within the loop represents the energy dissipated in one cycle of motion
 Lecture 6-6

 Although the model is called viscous damping model, the name should have been equivalent
viscous damping model

 This is the case because the damping is obtained from experimental tests which will be
discussed later and then express the results of the test in terms of damping ratio by assuring
that damping force is proportional to velocity

 Important thing to note is that

 force-displacement curve for one cycle of oscillation provides a hysteresis loop
 area of the loop gives the energy dissipated in one cycle of oscillation

 This observation is useful in obtaining equivalent viscous damping for other types of damping
produced during oscillation both elastic and inelastic ranges
 Lecture 6-7
Rate Independent Damping
 It is seen that energy dissipated in one cycle of oscillation is dependent on the frequency of
oscillation for viscous damping model
 However, experiments on structural metals indicate that the energy dissipated in cyclic oscillation
is fairly independent of frequency
 This type of damping is called rate independent linear damping, structural damping, hysteretic
damping etc.; rate independent damping is exhibited in static hysteresis developed at
microscopic state in crystals at different spots but over all macroscopic behavior remains linear
 In spite of this observation, viscous damping model has been favored because of analytical
convenience
 Since rate independent damping model is independent of frequency, the damping force fD is
defined as k
fD  u


 in which k is the stiffness of the structure and  is damping coefficient

 Lecture 6-8
Plot of ED Vs. ω
 As a consequence, the
energy dissipated in one
cycle of oscillation is

ED   cu02
 k  2 k
 
 0u c
   
  ku02 with
 2 ESo

 The equivalent viscous damping for rate independent damping may be obtained by
matching damping energies at ω = ωn
E D  4ES 0 (for viscous damping) E D  2 ES 0 (for rate independent)
 Thus,

 ( viscous) 
2
 Lecture 6-9
Coulomb Damping
 Coulomb damping results from friction against sliding of two dry surfaces; this kind of forces
acts within the micro level of material during oscillation; however, the energy dissipation due
to this mechanism is small
 This type of damping becomes important when additional damping devices especially friction
damping is added
 Friction force is given by F  N

in which µ is the coefficient of static and kinetic friction taken to be equal and N is the normal
force; friction force is independent of velocity once the motion is initiated
 Direction of friction force opposes the direction of motion; sign changes as the direction of
motion changes as shown in figure
 Lecture 6-10
 Change of sign of friction force with change of the direction of motion introduces nonlinearity in
the system; however for harmonic excitation an exact solution is possible. Friction force-
displacement plot is shown in following figure

The area of the loop is E D  4 F 0

 The equivalent viscous damping is obtained by equating the energy dissipated in one
cycle of viscous damping to the area of friction damping is

  2 F 2 F
 The area of the loop is 4F 0  4 ES0  2 ku02   
n n   ku  u
n 0 n o
F
In which, F 
k
Lecture 6-11
Lecture 6-12
Lecture 6-13
Lecture 6-14
Lecture 6-15
Lecture 6-16
Lecture 7
 Lecture 7-1

Solution for Harmonic Excitation (Steady State)

 Solution of the equation of motion of SDOF subjected to harmonic excitation can be best
obtained by considering complex harmonic excitation; derivation is simpler and it provides the
solutions for both real harmonic and complex harmonic excitations.
 In subsequent lectures, this will be presented. Here, the response for real harmonic excitation
will be presented as it is discussed in all text books on dynamics.
 Equation of motion for SDOF subjected to harmonic excitation is given by:
mu  cu  ku  P0 sin t
 Solution consists of two parts i. e.
u  u c (u h )  u p
 uc is obtained by setting RHS as zero and is derived in previous lecture. uc is given by

uc  e nt  A cos d t  B sin d t 

 Lecture 7-2

 up is called the particular solution and is assumed to have the same form as that of
the RHS force term:

u p  C sin  t  D cos  t 

 Substituting for up in the equation motion, the following equation is obtained

m  C 2 sin t  D 2 cos t   c C cos t  D sin t   k C sin t  Dcos t   p0 sin t

Collecting ‘sin’ and ‘cos’ terms together following equation is obtained,

  mC 2  cD  kC  sin t    mD 2  cC  kD  cos t  P0 sin t

Equating ‘sin’ and ‘cos’ terms of the LHS and RHS give.

(k  m 2 )C  (c ) D P0

(k  m 2 ) D  (c )C  0
 Lecture 7-3

 Solution of the two equations give,

(k  m 2 ) Po
C
(k  m 2 ) 2  c 2 2 

c P0
D
(k  m 2 ) 2  c 2 2 

p0  (1   2 ) 
C  (1   2 ) 2  (2 ) 2 
k  

p0  2 
D  (1   2 ) 2  (2 ) 2 
k  

c 
 
2mn n
 Lecture 7-4

 Substituting for C and D in up and performing algebraic manipulation

p0
up  DAF sin(t   )
k
 
1
2 2 2 2
DAF  (1   )  (2 )

2
  tan 1
1  2
a is the phase lag i. e. response lags behind the excitation by angle 

The solution for up can also be written as,

u p  us DAFsin(t   )
p0
u
In which s 
k , called static displacement i.e., response of SDOF if the load were
applied statically.

Maximum displacement of up is thus

us DAF
 Lecture 7-5
 Plot of up and u are shown in figure below; u is given by
u  e nt  Acos d t  Bsin d t   us DAFsin(t   )
 Constants A and B are determined from initial condition u(0) and ů(0)

 It is seen that uc part gradually dies down with time leaving up part only after some time; the

uc part is thus called transient response and up part is called steady state response.

 For sustained excitation like P0sinwt it is the steady state response which is of interest and
is utilized for finding response of irregular excitation in frequency domain (discussed later).
 Lecture 7-6

Undamped Solution

 The undamped forced vibration under sinusoidal excitation is mainly of theoretical

interest; as discussed before, undamped vibration hardly exists in practice.

 The solution helps in understanding, the importance of damping in near resonating

condition.

 For  = 0 the solution, turns out to be,

p0
u sin t DAF  A cos n t  Bsin n t
k
1
DAF 
1  2

 With initial condition as u(0) and ů(0)

 u (0)   us sin t
u  u (0) cos nt    us 2
sin nt 
 n w 1    1   2
 Lecture 7-7
us sin t
 The steady state is the second part of the solution i.e.,
 The plot of u is shown in figure below, 1  2

 It is seen that the transient response does not die down with time; hence, total response
remains steady the with time and at any instant of time t, large or small, is not independent of
initial condition unlike the damped forced vibration
 Lecture 7-8

 The plots of DAF, and phase angle with frequency ratio are shown in figures below
 Lecture 7-9

Response at Resonance

For damped force vibration problem, at  = n and zero initial condition

ue n t
 Acos d t  B sin d t   us DAF sin(t   )
1
DAF  for   1,   90 
2

us
A
2

B
1  2
For  very small the equation simplifies to

us nt
u (e  1) cos nt (d  n )
2
 Lecture 7-10

 The plot of u for the above condition is shown in figure below

The response reaches to steady state with an envelop function (e nt  1)

 Lecture 7-11

For undamped vibration with zero initial condition.

us
u (sin t   sin nt )
1  2

When  = n and =1, u is indeterminate

In order to obtain the response at  = n L’Hospitals rule is employed,

dN
u 1  lim u s (  1) d (n t cos n t  sin n t )
 us
dD 2
d
us
  (nt cos nt  sin nt )
2
us 1
1
 1  (n t ) 2  2 cos(n t   );   tan 1
2 n t
us
As t  , u   nt cos(nt );   0
2
 Lecture 7-12

 The plot of u is shown in figure below

 It is seen that as t becomes large, the response also becomes large ie, it grows with time
signifying a divergent response
Lecture 7-13
Lecture 7-14
Lecture 7-15
Lecture 7-16
Lecture 7-17
Lecture 8
 Lecture 8-1
Interpretation of Steady State Response
 The plots of DAF vs  and phase angle vs  for a given value of  is shown in figure
below; the plot reveals a lot of useful information for dynamic analysis and design
of structure
 Lecture 8-2
 The DAF plot shows the following

 The variation of DAF vs  peaks near  = 1 (not exactly at 1)

 In the vicinity of  ≈ 1 the curve sharply rises; sharpness increases as the value of 
decreases

 After the peak is reached, there is a sharp fall in the value of DAF; for the  ˂˂ 1; de-
amplification occurs

 As 0, DAF1; further, the effect of  on the DAF is maximum near  = 1; the effect is
almost negligible for  >>1 and  <<1

 Based on these observations, the DAF curve is divided into 3 zones; quasi-static zone,
resonating zone and anti-resonance zone
 Lecture 8-3

 Thus in quasi-static zone a static analysis is sufficient with results increased by

about 20% to account for dynamic effect
 In anti-resonance zone, The  <1; the effect of  on DAF is negligible; it could be as
small as 0.01; dynamic analysis is important in this range in order to obtain the
correct value of DAF
 By decreasing the natural frequency much less compared to the excitation
frequency, the response can be controlled significantly
 In the near resonance zone, DAF increases significantly; the increase significantly
depends on ; at  = 1, it is 1/2 and is completely governed by ; for =0, DAF
which is shown in the previous lecture
 Note maximum DAF does not take place at = 1; it is obtained by
d ( DAF )
0 ; in the quasi-static zone, DAF varies within 1 to 1.2; effect of
d
damping on DAF is small; the value of ω at which the peak value of DAF occurs is
the valueofDAF
 p at 1
 p  n 1  2,2 is
2 1  2
 Lecture 8-4

 The variation of  vs  shows the following:

 For value of   0 ,   0 i. e. the response tends to be in the quasi-static zone. The

effect of damping on the phase angle is small

 As   1 (in the resonating zone),  tends to be 90 º i.e. response is in quadrature with

the excitation

 At   1 , the phase angle is 90 º irrespective of the damping; however, in the

resonance zone, the effect of damping is more on the DAF

 When >>1,  tends to become 180º; the response lags behind the excitation by 180̇º
i.e. the response takes place opposite to the direction of excitation; in anti resonance
zone, the effect of damping is small

 As damping becomes small (nearly to undamped case), the phase angle suddenly
changes from the in phase state to quadrature state, and then suddenly to out of phase
(opposite) state
 Lecture 8-5
Determination of Damping Ratio

 Figure bellow shows the DAF vs ; half-power band width is defined as,
b  a
n where b and a are the frequencies obtained by the
0.707 resonant amplitude.

1 1
 Equating the expression for DAF with 
2 2
, the following equation is
1 1 1
obtained 1

2 2
1   2 2   2 
2 2

 

 Lecture 8-6

 Squaring both sides and showing the resulting equations

 1
 (1  2 ) 2
n
 Taking the first term of the Taylor series after expanding the RHS

 1 
n
 Subtracting the smaller root from the larger one gives,

b  a
 2
n
 thus b  a

2n

  can be obtained either from the DAF at   1 or from the half power band
width; the latter is called half power point method of determining the damping
ratio
 Relationship between Rd, Rv and Ra Lecture 8-7

u (t )  ust DAF sin(t   )

 ust Rd sin(t   )
u (t )  ust DAF cos(t   )

u(t )  ust 2 DAF sin(t   )

 Defining R   Rd andRa   2 Rd   R

u (t )  ust Rd sin(t   ); u (t )  ust R cos(t   ); u (t )  ust Ra sin(t   )

 Because of the relationship between Rd, Rv and Ra, they can be plotted in a
tripartite logarithmic plot

log Rd  log R  log w

log Ra  log R  log w

 Lecture 8-8
When R Rv and Ra are Plotted Separately
d
,
 Lecture 8-9
When Rd Rv and Ra are plotted together in TRIPARTITE plot
,

 The plot is shown in figure below,

1
Maximum value of Rv occurs at,   n ; ( Rv ) max 
2
n 1
 ; ( Ra ) max 
Maximum value of Ra occurs at, 1  2 2 2 1  2 2
 Lecture 8-10
PHASE PLOTS u

umax u / w

STABLE STEADY STATES

umax

umax u
Lecture 8-11
Lecture 8-12
Lecture 8-13
Lecture 9
 Lecture 9-1
Complex Harmonic Solution

 eiωt= cosωt + i sinωt

 If the results for harmonic excitations are known,

 For Fo cosωt excitation,
Fo
u (t )  us DAF cos t    ; us 
k
 For Fo sinωt excitation,
Fo
u( t )  us DAF sin ωt  α  ; u s 
k

 For Fo cosωt + i Fo sinωt

Fo
u (t )  us DAF cos t     i sin t    ; us 
k
i ωt α  Fo
 us DAFe ; us 
k
 Lecture 9-2
 The result as above can be obtained by solving the equation of motion in complex
domain; from the solution it is not only possible to obtain results of harmonic
excitation but also some important concepts in dynamics can be derived

 eiωt is called unit complex harmonic function; if solutions are desired in complex
domain eiωt is used in place of cosωt and/or sinωt

 For excitation, f(t)= Fo eiωt , particular solution is obtained by assuming u(t)= uo eiωt .

 Substituting in equation of motion

  m  ic  k  u
2
0  F0

u0  k  m 2   ic  F0
1

 For F0=1, u0 is called H(ω) and is inverse of the function

H( )  k  m 2  ic 
1
 Lecture 9-3
 H(ω) is called frequency response function (FRF) of the single degree of freedom system;
it represents a composite dynamic characteristics of the system for a particular excitation
frequency ω.

 By knowing the FRF of the system, the response of the system to any long term excitation
(leading to steady state response) can be obtained.

 FRF of a dynamic system can be determined experimentally; once it is known, the

response of the system to future excitations can be determined without explicitly knowing
‘k’, ‘c’ and ‘m’ of the system.

 In dynamic system identification and in dynamic structural health monitoring, determination

of H(ω) by experimental means is the key task.

• Since u0(ω) has to be of the form u0=a+ib.

F0 k  m 2 F0
 i
c F0
k  m 2   ic k  m 2  c2 2  k  m 2
 c 2 2 

a
k  m F 2
0
b
c F0
k  m  c  
2
k  m  c  
2 2 2 2 2

u o    a  ib  Re i R  a 2  b 2
 Lecture 9-4
b b c 2 mn 2 
  tan1 ;    
a a k  m 2  1   2 k  1   2  n

R
k  m   c 
2 2 2 2

F 
F0
2 0
k  m   c   k  m 2   c2 2
2
2 2 2 2
 

F0 1 F0
  DAF  us DAF
k 1      2 
2 2 2
k

Since, u 0    Re i

u t   u 0  e it  u s DAFe i t  

 Same result as it was obtained before; Since eiωt = cos ωt + i sinωt, when
excitation is cos ωt, then
u t   u s DAFcos t   
 Lecture 9-5
Concept of Dynamic Flexibility and Impedance Function
 Since u(t) = u(ω)eiωt and F(t)=F(ω)eiωt
u(ω) = H(ω) F(ω)

If F(ω)=1, then for unit harmonic load, displacement is H(ω)=a + ib


H    k  m  ic
2
1
; H ( ) is called dynamic flexibility matrix
 Similarly, if u(ω)=1, then for unit complex harmonic displacement, force required
is:

H    k  m 2  ic
1

 This is called impedance function for a system’s degree of freedom (for SDOF, it is
evident). Force has two parts: a real and an imaginary part

 Real part is associated with stiffness (represented by spring) and imaginary part is
associated with damping (represented by dash pots) for a given frequency

 Plots of real and imaginary parts with frequency provide the impedance function of
a dynamic system and are widely used in dynamic analysis of soil structure
systems
Lecture 9-6
Lecture 9-7
Lecture 9-8
Lecture 10
 Lecture 10-1
Transmissibility

 The results of harmonic support excitation are useful in solving problems of different
kinds including earthquake excitations.

 The problem is classified as transmissibility problem and therefore, the

transmissibility problem is discussed in detail.

 The problem becomes easier to handle if it is solved in complex domain.

 Lecture 10-2

 The first problem is shown in the pervious figure, where interest is to find what is the
fraction of the load applied to the mass is transmitted to the floor.

it
mu  cu  ku  P e

u  u eit (assumed as u p ); steady state same as u p
 k  m 2   ic  u  P
 
P
u 
 k  m 2
  ic
P i t  
u DAF e
k
 Lecture 10-3
1
DAF  1



 1  
2 2
 2 22


P
u  DAF i  eit  
k
R (t )  ku t   cu (t )
 c  i t 
 P DAF  1  i  e  
 k 
 c  i t 
 P DAF  1  i  e    P0 DAF 1  2i  e  
i  t 

 k 
1


 P DAF 1   2 
2
 2
e
i  t   
  tan 1  2 

R (t ) max 1   2   2 2

Tr    
1
P (t ) max
1   2 2
  (2 ) 2 
2
 
 Lecture 10-4

 Second problem is that of support excitation as shown in Fig.

If u
g  t   u go e it
then
mu  cu  ku  ( mu go )e it  Po e it

 Assuming steady state solution to be

u  u o e it like before

Po
DAFe  
i ωt-
u=
k

 m 2u
go  DAF e  i  t  
 u go eit
a  u  ug 

u
k
 i  t  
 u go e 
i  t  
u

go  2
DAF e ei
 Lecture 10-5

 u
go  2
DAF  u go  cos   i sin   
 e i t  

 u
go  2
DAF  u go cos  1  i tan  
 e
i t  

  2i   it  
go  DAF  u g o cos   1 
 u 2
2 
e
  1   

[1  tan 2   sec 2  ]
gives cos  1- 2  DAF
 2 2  2i   it  
 ugo DAF   1-  1  2 
e
  1   
 Lecture 10-6

 u go DAF 1  2i  e
i  t  

1

 ugo DAF 1   2i   e 

i  t    
2 2
  tan 1 2
 

1

ua max 
 
2 2
1  2
TR     T (ForceTransmissibility)
ugo  1   2 2   2 2  R

 
Similarly,it can be shown that if u g t   u go eit
ua max
then TR   TR (ug )
ugo
 Lecture 10-7

 The plot of TR vs γ is shown in the figure below.

The following things can be noted from the figure

√
 For γ ≤ 2 , damping effectively reduces the
transmissibility
 For TR to be less than 1, γ > √2 ; in this zone ξ
has an adverse effect.
 For TR to be effective in this zone ξ should be as
small as possible
 Since ξ has beneficial effect for γ  2 , judicious
selection of ξ is needed if frequency sweep is
expected (0 to operating frequency); for lower ω,
higher damping and for higher ω, lower damping
are required

 Higher damping ratio causes the maximum

amplification (resonance) to occur at ω≠ ωn ;
maximum TR is, therefore not at ω=ωn
 Lecture 10-8
Measurement Instruments
 Transducer is a simple device which measures dynamic motion; it consists of mass-
spring-dashpot attached to a rigid frame as shown in the figure.
 This transducer is used as seismograph.
 Lecture 10-9

 Principle of operation is that the equipment is mounted on the rigid base. The relative
displacement of the mass represents the ground acceleration. For a simple harmonic
motion, equation of motion:
 mugo
u t   Rd sin t  
k
R   
  d2 ugo sin  t  
 n   
R   
  d2 ug  t  
 n    
 Measured u(t) is a true independent modification of the ground acceleration record
with only a phase shift φ/ω; for a given frequency of excitation, the modification
amounts to a multiplication by a constant

 In order to make it applicable for the measurement of a realistic time history record,
both Rd and phase should be made frequency independent as far as possible
 Lecture 10-10

 This may be accomplished by keeping ωn around 50 Hz and ξ=0.7; for this condition Rd≈1
for a frequency band of 0 to 25 Hz which contains most of the frequencies; also, for this
frequency range φ linearly varies with ω, so that φ/ ω=1

 If ground displacement is to be measured, then either spring is made flexible or mass is

increased or both, so that natural frequency becomes much less compared to the most
frequency contents of the ground motion

 Under that condition,

u t   Ra u go sin t  
For   n , Ra  1 and   180 0 , therefore
u t   u go sint

 The measured displacement is negative of the ground displacement; damping constant of

the equipment is inconsequential since for >>1, DAF is almost independent of ξ
 Measured u(t) is a true independent modification of the ground acceleration record with
only a phase shift φ/ω; for a given frequency of excitation, the modification amounts to a
multiplication by a constant
Lecture 10-11
Lecture 10-12
Lecture 10-13
Lecture 10-14
Lecture 10-15
Lecture 11
 Lecture 11-1
Transient Dynamics

 Response of SDOF system to short and very short duration excitation is of considerable interest in
practice; for example, blast loading, impulse loading etc

 The loading is modeled as a summation of number of impulses and therefore, it is also named as
impact load analysis

 The name transient dynamics is given because of the reason that the initial condition of the system
is important and influences the response mostly unlike steady state response; over the short
duration, the effect of initial condition (which exponentially decays) does not die down significantly

 Since frequency domain analysis is strictly valid for finding the steady state response, it is not
applicable for transient dynamics

 Time domain solution is used for obtaining the response for impulse load. One of the classical
methods for determining the response for impulse loading is by performing Duhamel integration of
equation of motion; for integrable functions (loadings), the results can be obtained in closed form
 Lecture 11-2
 Where numerical integrations are to be performed, other time domain analysis techniques
may prove to be efficient, however, a variant of numerical Duhamel integration in
recursive form is also available

 The Duhamel integration of equation of motion is based on two premises:

(i) Idealization of load as a sum of series of impulses

(ii) Use of the known results of free damped oscillation of a SDOF system

 Solution of damped free vibration equation of motion gives;

 u 0   nu 0  
u t   e nt u 0 cos d t  sin d t 
 d 
 For u(0) and ξ=0, a special case, u(t) becomes
u 0 
u t   sinnt
n
If ξ  0, but u 0   0, then
u 0 
u t   e  n t
sin d t
d
 Lecture 11-3

 If the load p(t) as shown in figure is considered, then at time  , an impulse of p( )d
may be assumed to have been imparted on the SDOF

p(t)

t
dt
t
t -t
 Lecture 11-4

 This impulse will cause a response of SDOF at time t (or at the elapsed time t- τ after
the impulse is imparted) as;

u 0 
u t   sin n (t   )
n
 u (0) is the velocity imparted to the system at t=

p ( ) d
u 0  
m
p ( )
Then , u t   sin  n (t   ) d
m n
 h(t   ) p ( ) d

1
 In which h(t   )  sin  n (t   ) d is called the impulse response function ie.,
m n
response due to unit force.
 Lecture 11-5


 Using theory of superposition, total response at time t is the sum of the responses
produced due to all impulses imparted upto time t ie,
t t
u t    u t    ht   p d
0 0
t
1

mn  psin t  d
0
n

 Note that the initial velocity for free vibration solution is the velocity caused by the
impulse at time t =  ; velocities of the system at time t= caused by previous impulses
are not considered in the initial velocity as evident from the figure

u t 


t
+
u t 
t
 Lecture 11-6

 If damping is considered,

t
1
u t    e  n t   
p sin d t   d
md 0

 In transient dynamic analysis using Duhamel integration, it is customary to set ξ=0, for
two reasons:

(i) to obtain conservative estimate of the peak value (not much difference is obtained
if the peak occurs in early phase) for deriving impulse response function

(ii) to get a simpler function to integrate to obtain the closed form solution for most of
idealized impulses

 Solutions of a number of cases of short duration excitations (Impulse) are presented

using Duhmel integration in the following slides
 Lecture 11-7
Step Force

t
P0
 P(t)=Po, u t    sin n t   d
m n
0
P0 P0
 2
cos n t    t
 0  1  cos n t 
m n k
 2t 

 u st 1  cos 
 Tn 
 Lecture 11-8

u t  t
 Plot of with is also shown in figure below
u st Tn

 If damping is included then,

  
u t   u st 1  e  n t
(cos d t  sin d t 
 1  2

 It is seen that maximum value of u t  is slightly less than 2
u st
 Lecture 11-9
Pulse Forces

 Amongst different types of pulse forces, the three types of pulse forces namely, rectangular,
half sine and symmetric triangle received the attention of the practitioners. Most of the pulses
can be modeled by one of these types

 Response spectrums of these pulses are obtained in most text books and are briefly
presented here

 The method of finding the response spectrum follows the steps as below:

 Obtain the expression for u(t) in the forced vibration phase t<td

 Obtain the expression for u(t) in the free vibration phase t>td

 Find the maximum value of u(t) in the two phases and retain the one which has greater value

 Express the maximum value as a fraction of

 Lecture 11-10
Rectangular Pulse
 Forced vibration phase is same as obtained before for constant force.
u t  2 t
 1  cos t  td
u st Tn
 Free vibration phase is characterized by free oscillation with u t d and u t d 
u t d 
u t   u t d cos n t  t d   sin n t  t d 
n
 Substituting for u t d  and u t d  and simplifying

u t   t    t t 
  2sin d sin 2   d  t  td
u st  Tn  
   Tn 2Tn 
 Lecture 11-11
Half Cycle Sine Pulse
 The half cycle sine Pulse is given as
t
p t   po sin
td

 For ≠1/2, forced vibration phase has u(t) as:

ut 1   t  Tn  t 
  
sin     
sin  2  t  t d
2  
u st  Tn    t d  2t d  Tn 
1   
 2t d 
 Lecture 11-12
Triangular Pulse

 The final results for the responses are:

  t T 2t  t
2  n sin  0t  d
  t d 2t d Tn  2

ut   t Tn  2  t d  2   t d
 2 1   2sin  t    sin   t  td
u st   t d 2t d  Tn  2 Tn   2

2 Tn 2sin 2  t  t d   sin 2 t  t   sin 2t   t  t
  2t d  T  2  T
d
T

n 
d
  n n
 Lecture 11-13
Concept of Response Spectrum

 The plot of with for an impulse force of duration td is called the response spectrum of impulse.

 The method of finding the response spectrum follows the steps below:

 Obtain the expression for u(t) in the forced vibration phase of oscillation

 Obtain the responses for t=0, and in the free vibration phase of oscillation

 Find the maximum value of u(t) in the two phases of oscillation and retain the one which has
greater value

 Express the maximum value as a function of

 The plot of the maximum value of the response as function of is called the response spectrum
for the blast load
 Lecture 11-14

 The comparison of the response spectrums for the three pulses is shown in figure.
Lecture 12
Lecture 12-1
Lecture 12-2
Lecture 12-3
Lecture 12-4
Lecture 12-5
Lecture 12-6
Lecture 12-7
Lecture 12-8
Lecture 12-9
Lecture 12-10
Lecture 12-11
Lecture 13
 Lecture 13-1

Time domain Analysis

 In time domain analysis of SDOF, equation of motion is integrated in time

 For integrable forcing function u (t )  u h  u p discussed earlier

A variant of time domain analysis is Duhamel integration widely used for short duration/pulse

loading seen before

In principle, it can be used for an irregular p(t)

 There are other methods (popular) for time domain analysis of SDOF for irregular p(t)

 Two such methods will be discussed here, one for solving 2 nd order differential equation and the

other for first order differential equation

 They are outlined in subsequent slides

 Lecture 13-2

Explicit and Implicit Integration Schemes

 All schemes are time marching schemes; integration is carried out over discrete intervals of time

 If at a current time, responses of an SDOF is determined with the help of known responses at the
previous time steps and the dynamic load at the immediately previous time step, then it is called an
explicit integration scheme

 On the other hand, if the scheme requires the additional information on the responses of the
current time step (which are not known) and the dynamic load at the current time step, then it is
called implicit integration scheme

 There are several explicit and implicit integration schemes in the literature

 Out of them, one explicit scheme and one implicit scheme widely used in structural dynamics will
be discussed here; the central difference scheme and Newmark’s beta method
 Lecture 13-3
Central Difference Integration Scheme
 The central difference scheme is based on the Taylor series expansion at a point t
 The derivatives of the function are explained as
ui 1  ui 1
u i 
2 t
ui 1  2ui  ui 1
ui 
t 2
 Applying the derivatives in the equation of motion

kuk 1  Pk Pk
uk 1 
k
1 1
k  a0 m  a1c; a0  ; a1 
t 2 2 t
 2m  c m  1
PK  PK  b0uk  b1uk 1; b0    k  2  ; b1    
 t   2 t  t

 is determining after finding .

 Lecture 13-4

Newmark’s  - method
 With known displacement, velocity & acceleration at kth time, it calculates the corresponding
quantities at k+1th time; Fk+1 is known.

 Two relationships are used for this purpose; they mean that within time interval t , the
displacement is assumed to vary quadratically

k+1 = uk + 1- δ  uk Δt+uk+1δΔt

u
1 
  k   uk+1
2 2
uk+1 = uk +u
k Δt+  -β Δt u +β Δt
2 
Making use of the two equations, substituting for derivatives at K+1th time in the equilibrium
equation

Pk 1
kuk 1  Pk 1; uk 1 
k
 Lecture 13-5
d
u
k 1  a (u
1 k 1  u k )  t  (1  d )u k  (1  )uk  t
2

k  a0 m  a1c  k

Pk 1  Pk 1  (a0 m  a1c )uk  (c0 m  c1c )uk  (d 0 m  d1c )uk

1 1 1    1    
a0  ; b0  ; c0  ; c 
1   1 ; d
     1 ; d
 1    1   t ; d  a1 t
 t 2
 t 2
 t    2   2 

Pk 1
k 1  Pk 1 ; uk 1 
mu ; m  (m  a2 c  a3 k )
m

Pk 1  Pk 1  c2uk  d 2uk  d3uk

1  2
a2    t ; a3    t 2 ; c2  k ; d 2  (c  k  t ); d3  c(1   ) t  k      t
2 
 Lecture 13-6

 Substituting the two basic relationships in the equation of motion and performing algebraic
manipulation, following recursive relationship can also be obtained

qk+1 = FNqk +HNFk+1

ui  β  Δt 2 
   1  
in which qi = u i HN =    δΔt 
u   mα  
i  1 
 
 1 2
          
2 2 3 2
 α -ω 2
nβ Δt αΔt- 2ξω nβ Δt - ω 2
nβ Δt α Δt -β α+  Δt 
2
1 
α -2ξωnδΔt- ωnδ Δt  αΔt- δ α+   Δt
2
FN =  -ωnδΔt 2 2

α 
 -ω 2
n -2ξω n -ω 2
n Δt -  
 

  1  2nt     t  2 2
n

1
  2n (1   )t  n2 (   )  t 
2

2
 Lecture 13-7
Solution for Earthquake Excitation
 If the support of the SDOF is subjected to time history of excitation , the RHS of the equation of the
motion is
P (t )   mug

 By dividing the equation of motion by m, the following equation is derived

 For the support excitations problem, the above two formulations can be now easily used to find the
responses at k+1 th. step by setting m=1 and p(t)=
 If absolute values of the responses are required, then g(t) should be also specified

𝑢
¨ 𝑎 ( 𝑘+1 ) = 𝑢
¨ 𝑘 +1 + 𝑢
¨ 𝑔( 𝑘+1 )
Lecture 13-8
Lecture 13-9
Lecture 13-10
Lecture 13-11
Lecture 13-12
Lecture 13-13
Lecture 14
 Lecture 14-1

Frequency Domain Analysis

 Unlike the time domain analysis, the frequency domain analysis varies the frequency to obtain
responses for the arbitrary time history of the force

 For this purpose, is transformed to the summation of a series of harmonic forces

 This is accomplished by applying the classical Fourier series to decompose as a summation of

several harmonic functions of ordered frequency

 Frequencies are in ascending order having a fixed frequency interval

 Responses of the SDOF system are essentially obtained by sweeping the frequency from lowest to
the highest, the key variable is the frequency. Hence, the solution procedure is known as frequency
domain solution

 Note that the steady state response of the SDOF system for a harmonic excitations is known in the
closed form and is used as the formation block for a frequency domain analysis
 Use of Fourier Analysis Lecture 14-2

 Fourier decomposition of dynamic loading/excitation provides one of the most important tools
in solving dynamic problems in frequency domain

 Also it provides a physical understanding of dynamic composition of excitation especially in

knowing its frequency contents which are again helpful in estimating the dynamic effects being
produced by the excitation

 Fourier decomposition along with the use of theorem of superposition make the frequency
domain analysis is easy and elegant

 With the availability of FFT and IFFT in most mathematical programs like MATLAB, the
frequency domain analysis can be performed routinely for any structure

 Data input to FFT and IFFT will be explained and the steps involved in carrying out the
analysis will be explained in the subsequent lecture

 Use of FFT for frequency domain analysis requires the knowledge of FRF of the system and
working knowledge in complex frequency domain

 For the purpose of understanding how Fourier synthesis is used in solving problems the
following example is explained
 Lecture 14-3
Fourier Decomposition
 Fourier series expansion of an arbitrary formation of time is given by
a
x(t )  a0   an cos nt  bn sin nt
n 1
T /2
1
a0  
T -T /2
x(t )dt

T /2
2
an  
T -T /2
x(t ) cos nt dt

T /2
2
T -T/2
bn  x(t ) sin nt dt

In which ; The amplitude of the harmonic at is given by

2 2
 2 T /2   2 T /2 
An  an  bn    x (t )cos ntdt     x (t ) sin nt dt
2 2 2

 T -T /2   T -T /2 
 Lecture 14-4

Eqn. 13.1 can also be expressed as


x(t )  c0   cn sin(nt  n )
n 1

cn  An ; co  ao
bn  -1 
n  tan  
 an 

p (t )  p0   pcn cos nt  psn sin nt
n 1

T T T
2 2 2
1 2 2
p0 
T T p (t )dt , pcn 
T T p(t )cos nt dt , psn 
T T p(t )sin ntdt
- - -
2 2 2

p (t )  p0  pc1 cos 1t  ps1 sin 1t  pc 2 cos 2t  ps 2 sin 2t
 Lecture 14-5
pc1
uc1 (t )  DAF1 cos(1t - 1 )
k
pc 2
uc 2 (t )  DAF2 cos(2 t - 2 )
k
In a similar way,
ps1 ps 2
us1 (t )  DAF1 sin(1t - 1 ) us 2 (t )  DAF2 sin(2 t - 2 )
k k
In which
1 2 i i
DAFi  ; i  tan ; i 
-1
(i  1, 2)
[(1-  i )  (2 i ) ]
2 2 2
1
1-  i 2 2
T
Finally,

u (t )  uc1 (t )  uc 2 (t )  u s1 (t )  u s 2 (t )

For an SDOF system, the equation of motion under the support excitation is given by
 2nu   2 nu  u g
u
 Lecture 14-6

 The plot of Vs. is called the frequency amplitude spectrum

 Conventionally this plot is shown as a continuous function of frequencies and for this, the discrete
values of obtained at each frequency are manipulated for an equivalence in the rms sense
Lecture 14-7
Lecture 14-8
Lecture 14-9
Lecture 14-10
Lecture 14-11
Lecture 14-12
Lecture 15
 Lecture 15-1
Fourier Series and Fourier Integral

 Fourier series and Fourier integral are extensively used in structural dynamics for the

formulation of problem in frequency domain

 The solutions are known as solutions using Fourier transforms; similarly, solutions

using Laplace transform are also used in vibration problem

 It is possible to change from Laplace transform to Fourier transform because of the

relationship that exists between the two

 The fundamentals of Fourier series and Fourier integral are briefly described here
 Lecture 15-2
Frequency contents of time history
 Fourier synthesis of time history record of (load/ ground motion) provides frequency
contents of it

 It provides useful information about the dynamic excitation & also forms the input
for frequency domain analysis of structure

 Fourier series expansion of x(t) can be given as

a
x(t ) = a0 +  a cosω
n=1
n n t + bnsinωn t
T /2
1
a0 = 
T -T /2
x(t )dt

T /2
2
an = 
T -T /2
x(t ) cosωn t dt

T /2
2
bn = 
T -T /2
x(t ) sinωn t dt

ωn = 2πn / T
 Lecture 15-3

 The amplitude of the harmonic at ω n is given by

2 2
2 T /2
 2 T /2

A = a
2
n
2
n + b 2
n =   x(t ) cosωntdt  +   x(t ) sinωntdt 
 T -T /2   T -T /2 

 Fourier series equation can also be represented in the form


x(t ) = c0 +  c sin(ω t
n=1
n n + φn )

cn = An
b 
φn = tan-1  n 
 an 
 Lecture 15-4

 Plot of cn with n is called Fourier Amplitude Spectrum

 The integration in fourier series is now efficiently performed by FFT algorithm
which treats fourier synthesis problem as a pair of fourier integrals in
α
complex domain 1

-iω t
x( iω ) = x( t ) e dt
2π -α
α


iω t
x( t ) = x( iω ) e dω
-α

 Standard input for FFT is N sampled ordinates of time history at an interval of

∆t

 Output is N complex numbers; first N/2+1 complex quantities provide

frequency contents of time history; other half is complex conjugate of the first
half
 Lecture 15-5
Fourier series & integral
 Fourier series decomposes any arbitrary function x(t) into Fourier components

a0 α  2πkt 2πkt 
x(t ) = +   ak cos +bk sin 
2 k=1  T T 
T T

2 2
2πkt 2 2
ak =
T  x(t )cos
T T
dt a0 =
T  x t  dt
T
- -
2 2
T

2 2
2πkt
bk =
T  x(t )sin
T T
dt
-
2

 T
  T

a    2     2 
x(t )  o     x (t )cos( k t ) dt  cos( k t )     x ( t )sin(k t ) dt sin(k t )
2 k 1   -T  k 1   -T 
 2   2 
 Lecture 15-6

2
 T  ,     d  It can be shown that
T
 
x(t )  2  A ( ) cos (t ) d   2  B ( )sin (t ) d 
 0  0

 
x(t )   A   cos t d   B   sin t d
- -

 The complex harmonic function is introduced to define the pair of

Fourier integral


1

-it
x( )  A( ) - iB ( )  x (t ) e dt
2 -

x(t )   x( )eit d 
-
 Lecture 15-7

For the Fourier integral to be strictly valid



 x(t ) dt  
-
 Discrete form of Fourier integral is given by
 2 kr 
1 N -1 -i 
xk ( )   xr e N 
N r 0
N -1  2 kr 
i 
xr (t )   xk e  N 

k 0
 FFT & IFFT are based on DFT.
 From x k , Fourier amplitude Ak is obtained

Ak  2  ck2  d k2  k  1......
N
2

xk  ck  id k ; A0  c0
 Lecture 15-8

 In MATLAB, xr, is divided by N/2 (not N), then

Ak =  c k2 +d k2 
c0
A0 
2
 Parsavel’s theorem is useful for finding mean square value
T
1 a0 2 1
T0 x (t ) 2
dt 
4 2
  ( a 2
k  b 2
k)

T
1 1 N -1 2 N -1
x(t ) dt   xr   xk
2

2
T0 N r 0 K 0

(X(t) is divided by N not N/2 as in MATLAB)

 Lecture 15-9

Use of Fourier Series & Integral in Dynamic Analysis

 An SDOF subjected to harmonic (both complex and real) excitation was discussed in
detail.
 The steady-state response is shown to be harmonic with a phase shift; the
expressions for the responses are taken as standard responses for the harmonic
excitation which is characterized by a frequency and an amplitude
 The response analysis using this standard response is popularly known as frequency
domain solution; this means that the variable used in the solution is the frequency of
excitation
 The fundamental basis of the development of the frequency domain analysis is the
Fourier series expansion of the arbitrary function f(t) in terms of frequency contents;
this has been discussed before
 Lecture 15-10

 In the development of the frequency domain analysis of an SDOF system the

following steps are used:

 The arbitrary loading f(t) is Fourier synthesized into harmonic components

 From these harmonic components, the Fourier spectrum can be constructed

 Solution of the SDOF system for each harmonic component is obtained using
the standard results

 For linear systems, theorem of superposition is used to combine the standard

results for all harmonic components of the loading
 Lecture 15-11

 With the advent of FFT algorithm this analysis is performed in a more compact way;
the steps used are:
 The time history of arbitrary loading f(t) is discretized in N sampled values at Δt
intervals
 They are fed into FFT which provides N discrete complex values at Δω interval.
 First (N/2)+1 discrete values are considered
 H(ω) for the system is obtained at discrete values of frequency of 0 to (N/2)Δω at an
interval of Δω
 Using (N/2)+1 values of FFT of f(t) and (N/2)+1 values of H(ω), a set of (N/2)+1 values
of x(ω) is constructed; after introducing complex conjugates, N values of x(ω) are
obtained (this part is explained in the next slide)
 IFFT of x(ω) provides the response x(t)
Lecture 15-12
Lecture 15-13
Lecture 15-14
Lecture 15-15
Lecture 15-16
Lecture 16
Lecture 16-1
Lecture 16-2
 Lecture 16-3

Use of FFT and IFFT in Frequency Domain Solution

 A generalized complex harmonic load is expressed as

𝑖 𝜔𝑡
𝑝 (𝑡 )= 𝑝 ( 𝜔 ) ⅇ
 Steady state response of an SDOF may be written as

 Substituting in the equation of motion and equating bot sides of the equation

 Note that is the same as that described in Lecture 9 as FRF.

 Since and are complex quantities, is also a complex quantity and should be better written as

 For brevity is written as simply

 Lecture 16-4

 If the is an arbitrary function of time, the FFT of provides a discrete set of N values of spaced a

 Each value is a complex number ; using DFT, the becomes a sum of for frequencies at an interval
of

 Thus, for each frequency, is obtained as

 The basis of frequency domain analysis in the complex domain is based on the above equation

 Ideally maybe obtained as a continuous function of however, because Fourier pair of integrals are
performed using DFT, the responses are obtained at an interval of and ; because of the presence of
the pairs of complex conjugates, the is obtained up to the cut-off frequency
 Lecture 16-5

Steps for frequency domain analysis using FFT and IFFT

 The following steps are used in obtaining a steady-state response of an SDOF system under an arbitrary
dynamic force :

 is discretized into N sampled values at an interval of time

 These values are fed into FFT, which provides N discrete complex values at an interval of frequency
(i=1…N)

 Take the first N/2+1 discrete values; is named as the Nyquist or cut-off frequency

 Find for the SDOF for to (N/2+1 values) at an interval of

 Since IFFT requires N discrete values of out of which complex conjugates shall appear after N/2+1
values, the complex conjugates are added after the computed values of ; how the complex
conjugates appear has been described in lecture 13 and example problems

 Feed N values of

 IFFT outputs N discrete values of

Lecture 16-6
Lecture 16-7
Lecture 16-8
Lecture 16-9
Lecture 17
 Introduction to MDOF System Lecture 17-1

 All realistic structures are modeled as either discrete MODF system or continuum system

 A continuum system has infinite number of degrees of freedom (d.o.f.) leading to infinitesimally small
mass attached to each d.o.f.

 This led to writing the equation of motion in a different form using the restoring action to be written as
that written for continuous elastic curve/continuum

 However, continuum system is often idealized as an assemblage of finite number of elements

transforming it to be a MDOF system

 Before writing equation of motion for MDOF system, following points deserve attention:
 All kinematic d.o.f. of a structure are not the dynamic d.o.f.

 Those kinematic d.o.f. which have masses attached to them are declared as dynamic d.o.f.

 Dynamic d.o.f. are mostly the condensed kinematic d.o.f. of a system; those d.o.f. which do not have
corresponding masses are condensed out so that d.o.f. to be handled in the dynamic problem are
reduced
 Lecture 17-2
 A common example is the condensation of rotational d.o.f. in the framed structure as shown in
this figure; rotations are condensed out when point mass lumping is used (mr2 = 0 as r = 0 for a
point).

At all joints

At all joints

Kinematic d.o.f. = 12 Kinematic d.o.f. = 9

Dynamic d.o.f. =3 (Inextensible condition) Dynamic d.o.f. = 2 (Inextensible condition)

Rotational d.o.f. are condensed out

 Lecture 17-3
Rigid slab

Dynamic d.o.f. = 6
Extensible members 6 d.o.f. at
each joint

Flat slab II

Kinematic d.o.f. = 72
Dynamic d.o.f. = 9
Any three independent d.o.f. can be selected
System I - non-diagonal mass matrix (discussed in lecture 18)
System II – diagonal mass matrix
 Lecture 17-4
 Matrix condensation may be carried out by any one of the following procedure:

• 
K  K   K Δθ K θθ1 K θΔ 
• Finding flexibility matrix corresponding to the dynamic d.o.f. by applying unit loads to
those d.o.f. successively and analyzing the structure with full kinematic d.o.f.
 Depending upon problem to be solved, the method of condensation is selected; both methods
have specific advantage
 Condensation of stiffness matrix to some specific selected dynamic d.o.f. is also required for
certain problems ( for example 3D building)
 Consistent mass matrix is quite popular in FEM modeling of 2D and 3D continuum; in this
formulation, acceleration at any instant of time within the element is assumed to be proportional
to the displacement field
 Thus, the shape function as that of the displacement is used to obtain inertia force distribution
within the element; the same energy principle used for deriving the stiffness matrix, is used to
obtain the mass matrix for the element
 The name consistent is used because the mass matrix generated is consistent with the
displacement field assumed within the element
 Lecture 17-5
 Generally, for framed structure point mass lumping is used; in consistent mass matrix,
rotational d.o.f. may be retained

 Torsional rotation about the vertical axis of 3D frames with rigid slab is included in the dynamic
d.o.f.; mass moment of inertia of the slab about the vertical axis is attached to torsional d.o.f.

 Here, point mass lumping and mass moment of inertia of the slab in 3D frames will be
discussed only

 The generation of stiffness matrix corresponding to the dynamic d.o.f. forms the major task in
writing the equation of motion

 M martix is normally a diagonal matrix for point mass lumping and can be easily formed;
however, for some types of structures it may not remain diagonal and for them, it may have to
be generated from first principles

 When the rotations are involved in the kinematic d.o.f., they are condensed out as they do not
form the dynamic d.o.f.; without condensation the problem can be formulated by unnecessarily
increasing the size of the problem; masses corresponding to rotational degrees of freedom are
set to zero
Lecture 17-6
Lecture 17-7
Lecture 17-8
Lecture 17-9
Lecture 18
 Lecture 18-1
Equations of Motion for MDOF
 Equations of motion for MDOF system are developed for two degrees-of-freedom (d.o.f.)
first which will be then generalized for n d.o.f.
 Consider the two d.o.f. as shown below; sway movement of a 2-D shear frame may be
idealized as the two d.o.f. spring dashpot as shown in the figure

 At any instant of time t, the free body diagram of the masses are shown in the figure

Two d.o.f. system

2-D shear frame

Free body diagram

 Lecture 18-2

 Dynamic equilibrium of forces at time t leads two equations

m1u1  c1u 1  k1u1  k 2 u 2  u1   c 2 u 2  u 1   p1 t 

m2 u2  c 2 u 2  u 1   k 2 u 2  u1   p 2 t 
 In matrix form the set of equations becomes
m1 0  u1  c1  c 2  c 2  u 1  k1  k 2  k 2  u1   p1 t 
0            
 m2  u2    c 2 c 2  u 2    k 2 k 2  u 2   p 2 t 
+ cu + ku = p  t 
mu
m, c and k are 2 × 2 matrices; u and p are vectors of size 2 × 1
.
 Thus, for a two d.o.f. system, the equations of motion can be written in a matrix form
with 2×2 matrices and 2×1 dynamic displacements (velocities and acceleration
vectors).

 Displacement vector is associated with K matrix; velocity vector is associated with C

matrix; and acceleration vector is associated with M matrix.
 Lecture 18-3

 It can be easily shown that k matrix is the stiffness matrix of the 2D frame/spring system
corresponding to the two d.o.f. (u1 and u2)

 c is called damping matrix which is not explicitly known but is defined by certain assumptions;
however, it has the same form as that of K matrix (not always true)

 m is called mass matrix which is found to be diagonal and therefore, is easy to generate;
however m matrix is not necessarily a diagonal matrix for point lumped mass system- it depends
upon the type of structure and the d.o.f. chosen (it will be explained later with the help of
examples)

 Now, it is possible to extend the equation of motion for two d.o.f. to n d.o.f. which follows as:

 The equations of motion can be written in a matrix form for the unknown vector of displacement
at the d.o.f. and a specified vector of excitations at the d.o.f.

 The sizes of vectors will be n × 1

 The K matrix will be n × n stiffness matrix of the structure corresponding to the dynamic d.o.f.
 Lecture 18-4

 The c matrix will be of the same form (not always true) and size as that of stiffness matrix; the
elements of the matrix are written in terms of coefficients c1, c2 etc. which are not explicitly known

 The m matrix will be n × n diagonal matrix (not always true) whose elements are the masses
attached to the d.o.f.

 The equations of motion remains the same as the previous matrix equation except the sizes of
matrices and vectors are changed to n × n and n × 1 respectively

 The load vector consists of the dynamic forces corresponding to d.o.f.; thus its size is also n × 1

 Note that if m(t), i.e., moment excitation is included in the load vector, then matrix condensations
are not needed; rotational d.o.f. are included in the dynamic d.o.f.

 Mass matrix contains mass moment of inertias corresponding to the rotational d.o.f ; in many soft
wares condensation procedure is avoided and rotational d.o.f are included in the analysis to
make the program a generalized one
 Lecture 18-5
Support Excitations
 When the supports of MDOF system undergo excitation, then the equations of motion can be
written extending the equation of motion of SDOF under support motion i.e.

mu  cu  ku  mu g

is extended to
MU CU  KU  Mrug

where r is the influence coefficient matrix or vector usually consisting of 1 and zero elements; some
examples are shown

 If support excitations are different at different supports, then r is a matrix of size n × q , where q is
the number of supports

 r is a matrix whose kth column elements contain the displacements corresponding to non support
d.o.f. (included in the dynamic d.o.f.) when an unit displacement is applied at the kth support
keeping other support d.o.f. locked

 The derivation of the equation of motion with single or multi-support excitation is provided in the
subsequent lectures
 Lecture 18-6
Example problems
Write the equation of motion for the MDOF systems corresponding to the dynamic degrees of
freedom shown in the figures. Assume the damping matrix to be mass and stiffness proportional.

Fig. 1
Lecture 18-7
Lecture 18-8
Lecture 18-9
Lecture 18-10
Lecture 18-11
Lecture 18-12
Lecture 18-13
Lecture 19
 Lecture 19-1
Examples for Support Exciations
 For single support single component excitation

mu + cu + ku = -mIug
For two component ground motion
1 0 1 0 - - - - - -
IT  
0 1 0 1 - - - - - - 
g T  u
u 

g1 u g2 
 For three component ground motion
1 0 0 1 0 0 - - - -
I T   0 1 0 0 1 0 - - - - 
 0 0 1 0 0 1 - - - - 
u 
g T  u
g1 ug2 u g3 
 Lecture 19-2
Example 3.1: Determine I for the following structures .

Solution :
T u1 v1 u 2 u 3  T u1 u 2 u 3 
I =   I =  
 1 0 1 1   1 1 1 

u2
u3
u3
u2
v1
u1
u1

xg
xg
Shear building frame
Bracket frame
 Lecture 19-3

Case 1 I T
 1 0 0 1 0 0 Single component
1 0 0 1 0 0 
Case 2 I  
T
 Two component
 0 1 0 0 1 0  u1 and v 1

2 v2
u2

1 v1
u1

3-D model of a shear building frame

Fig3.4c
 Lecture 19-4
Non-Diagonal Mass Matrix

 For lumped mass system/ point lumped mass system the mass matrix could be a no

diagonal mass matrix depending upon the problem

 Generally it is thought that mass matrix for lumped mass system is diagonal

 There are number of examples where mass matrix turns out to be non-diagonal

 Such mass matrices are derived from the first principle using the definition that mass is

defined as inertia force per unit acceleration

 The derivation follows the same procedure as adopted for deriving the stiffness matrix

of the structure

 One example is shown and the derivation of the mass matrix is obtained using virtual

work principle
 Lecture 19-5
Example 3.2 : Find the mass and stiffness matrices
for the two models of 3D frame shown in Fig 3.5.

Solution :
 4 2 2   4 1 3  0 
  m   
k  k  2 3 2  ; m   1 4 3 ; I  0 
6 1 
 2 2 3   3 3 6   
u3

u2
m
P (model1) = - 3 -3 6 
y T T
x L eff xg
C.M. 6
u1
L

k k
&&
xg

Model-1
 Lecture 19-6

For Model -2

1 0 0 
 3 0 0.5 L 
  0 
   
k k 0 3 0.5 L  ; m  m 0 1 0  ; I  0 
  1 
0.5 L 0.5 L 0.5 L2  0 0 L   
6
q

P (model2) = -m 0 1 0 
T T
C.M. u eff xg
C.R.
v
L
&&
xg
k k

Model-2
 Lecture 19-7
For multi support excitation, equation of motion
 m ss m sg  
 

u t
  css csg   
u t
  kss ksg   
u t
 
0  
m   + c   + k   =  
 gs m gg  
ug 
  gs c gg  
 ug   gs k gg  
 ug   pg 
  
u t = u + rug
t

m ss u + m sg ug + c ss u+ c sg u g + k ss u g + k sg u g = 0
t
+ c ss u t + k ss u t = - m sg u g - c sg u g -k sg u g

or m ss u
t + c ss u t + k ss u t = -k sg u g
m ss u
c ss u+ k ss u= - (m sg + rm ss )u g - ( c sg + rc ss )u g - (k sg + rk ss )u g
m ss u+
k ss us + k sg ug = 0
us = - k ss -1 k sg ug = rug
rk ss + k sg = 0
c ss u+ k ss u= - rm ss u g
m ss u+
Lecture 19-8
Lecture 19-9
Lecture 19-10
Lecture 19-11
Lecture 19-12
Lecture 20
 Lecture 20-1
Frequencies and Mode Shapes for MDOF
 Free vibration analysis of MDOF system is similar to that for SDOF system i.e. solution is obtained by
setting RHS of the equation as zero
 Two cases are considered separately, undamped and damped vibrations

 The solution provides natural frequencies and mode shapes of the MDOF system

 Natural frequencies are defined as the frequencies of oscillation of the MDOF system when it is set to
vibrate freely (without any exciting force) starting with an initial condition; Unlike, the SDOF system it
has not only one frequency of oscillation but has n frequencies of oscillation
where n is the degree of freedom (d.o.f.) of MDOF system

 Each natural frequency of oscillation is associated with a particular nature of vibration, called the
mode shape

 Natural frequencies and mode shapes are the outcomes of the free vibration analysis of the MDOF
system; the problem which is solved to obtain them is known as eigenvalue problem
 Lecture 20-2
 Physically, when a MDOF system is set to vibrate freely, it vibrates with the lowest frequency
of oscillation known as fundamental frequency of oscillation and associated mode shape is
known as fundamental mode
 Fundamental mode shape is the simplest possible mode of oscillation that the MDOF can
undergo; it requires least amount of energy

 Undamped free vibration of MDOF system is represented by

 + ku = 0
mu
 From the solution for SDOF system, the response may be assumed as
u  u0 sin  t 
 Substituting for u and u

 m 2

 k u0 sin  t   0
or
 k - m 2   0
or
 
k - m 2  0 k - m  0
  2
 Lecture 20-3

 The determinant leads to a polynomial equation in  ; if n is the d.o.f. then the order of the
polynomial is also n
 There are many ways to solve the above problem numerically; up to n = 3, the polynomial
equation may be easily solved to find the values of  ; it can also be obtained by plotting the
value of determinant against ω for smaller order of n

 The successive values of ω can be obtained from the curve where it cuts the ω axis as
shown below

Det
ω3

ω1 ω2 ω4 ω
 Lecture 20-4

 The problem can be posed as a classical eigenvalue problem for which many
standard solutions are available; the problem is cast as an eigenvalue problem as
below
k - m  u  0
ku  m u
 m 1 2 m1 2  u
m -1 2 km -1 2 u  I  u
Au   u
in which A  m -1 2 km -1 2 of size n × n

 Eigenvalues of matrix A provide n values of  and hence n values of ω.

 For each value of  , solution of the matrix equation  k  m  u  0 can be

determined by assuming any value of an element ui of vector u as 1; generally u1 is
set to unity.
 Lecture 20-5

 Then other vales of the elements of the vector u can be determined by solving the matrix
equation; the solution provides relative value of ui (i = 1……….. n) with respect to u1 = 1; an
example problem shows the procedure

 Thus, the values only provide a shape for displacements of different d.o.f; this shape is called
the mode shape ( i ) corresponding to frequency ω i

 An important property of mode shapes is that they are orthogonal w.r.t. to the mass and
stiffness matrices i.e.
iT m j  0 if i  j

iT m j  0 if i  j

the same condition holds good with respect to the stiffness matrix; the reason for this is that
eigenvectors of matrix A are orthogonal w.r.t. to the A matrix; it is mathematically established
 Lecture 20-6

 A formal proof for the orthogonality properties of the mode shapes is given below

 Pre multiplying ith mode equation by

 Similarly,

 Deducting two equations

 Tj ki -  iT k j  i2 Tj mi -  2j iT m j

 Taking symmetric property of k, m i.e.,  Tj ki  iT k j

(i2   2j )iT m j  0

 If
i   j ,iT m j  0
 From the second equation it follows for
Lecture 20-7
Lecture 20-8
Lecture 20-9
Lecture 20-10
Lecture 20-11
Lecture 20-12
Lecture 20-13
Lecture 20-14
Lecture 20-15
Lecture 20-16
Lecture 20-17
Lecture 20-18
Lecture 21
 Mode Shape Normalisation Lecture 21-1
 Mode shapes can be normalized in different ways since mode shapes represent only
the relative displacement of the d.o.f.
 Consider the mode shape in the following example:
1   0.036 0.036  0.125
T

 20 0 0 
 2  0.158 0.158 0  m   0 20 0  m
T

 0 0 60 
 3   0.154 0.154 0.03
T

 The normalization is done with the first value taken as unity and with maximum value
taken as unity; the corresponding mode shapes are,
 1  1 3.47  1   0.288 1
T T
1  0.288
 2  1 1 0  2  1 1 0
T T

 3  1 1 0.195  3    1 1 0.195
T T

 Another way of normalization is with respect to the mass matrix i.e.

1
1   0.0144  0.0144 0.05
T
1T m1  m1 ;
m1
1
 2   0.05 0.05 0
T
 T2 m 2  m2 ;
m2
1
 3    0.0167 0.00 32 5
T
 3T m 3  m3 ; 0.0167
m3
 Lecture 21-2
Properties of Mode Shapes and Frequencies
 If mode shapes are normalized as above then
 T m  I
 Undamped mode shapes have the following properties apart from the orthogonality:

 All points pass through their maximum and minimum values at the same instant
 All points pass through zero at the same instant in time
 The mode shapes can be described by a sign valued real number
 All points are either totally in phase or out of phase with any other point on the structure
 The mode shapes from the undamped case are same as the proportionally damped
case

 Since C matrix is not explicitly known, it is obtained by making certain assumptions

c  m  k
 One of the common assumption is that C matrix is mass and stiffness proportional i.e.
values of a and b are determined with the help of the undamped natural frequencies of the
system; more information on the topic will be discussed in a subsequent lecture
 Lecture 21-3
 Complex mode shapes have the following characteristics

 All the points do not pass through their maxima at the same in time- points
appear to have time log

 All the points do not pass through zero at the same instant of time

 Mode shapes can not be described by real valued numbers- the shapes are
complex valued

 The different d.o.f will have some general phase relationship that will not
necessarily be in phase or 180 degrees out-of phase with other d.o.f.

Undamped mode shapes Damped mode shapes

 Lecture 21-4
Construction of C Matrix from Natural Frequencies
 As discussed before damping matrix is classified as classical and non-classical damping
matrix; for non-classical damping advantage of modal analysis can not be used
 If damping matrix is classical, then the advantage of modal analysis can be taken and the
knowledge of complete C matrix is not required (modal ratio/damping ratios are sufficient
for analysis)
 However there are quite a few situations for which knowledge of C matrix of the structure
is required even if it is a classical one; they include :

 Consideration of soil structure/water structure interaction

 Nonlinear dynamic analysis of structure

 Dynamic analysis of hybrid structure

 Structural control of dynamic system

 Where large number of modes are to be considered or determination of modal initial

conditions from structural initial conditions are required
 Lecture 21-5
 For above cases, c matrix is derived from the damping ratio using certain assumptions
 The common assumptions made is that the c matrix is proportional to either m matrix or
k matrix or both; this kind of damping is known as proportional damping or Rayleigh
damping (after the name of Rayleigh)
 The constants of proportionality are equated to the damping ratio since it represents the
total dissipation of energy during vibrations in a comprehensive manner and is obtained
from experimental test

 If it is assumed that c matrix is proportional to m, then

c  a0 m
Say, cn  nT cn ; then cn  a0 mn
mn  nT mn
Since for SDOF,
c  2mn
For the nth mode,
2mn nn  a0 mn
a0
n  ; a0  2 nn
2n
 Lecture 21-6

 If c is proportional to k, then
c  a1k
Say, c n  a1nT kn ;  a1n2 mn
2mn nn  a1n2 mn
a1n 2 n
n  ; a1 
2 n
 If C is proportional to both, then
c  a0 m  a1k
nT cn  a0nT mn  a1nT kn ;
cn  a0 mn  a1n2 mn
2mn nn  a0 mn  a1n2 mn
a0 a1
n   n
2n 2
 If i and j are any two modes, then a0 and a1 can be solved in terms of i ,  j , i and  j
as (withi   j   )
2i j 2
a0   ; a1  
i   j i   j
 Lecture 21-7

 with a0 and a1 as known c is constructed

 Figures show the variations of  n with nfor all the three cases
 Lecture 21-8

 For determining a0 and a1 , any two natural frequencies can be adopted; they depend
upon the problem being solved and the number of modes which are considered

 There is no fixed criteria for the selection of natural frequencies i and j ; following
examples give some ideas for the selection

 Say the first 5 modes are being consider in the analysis i.e., contributions from modes
higher than 5 modes are neglected (assuming they are small)

 The values of 1 and 4are selected for this purpose; from the figure it is seen that
1 and  4 are set to  , the specified damping ratio; this has two implications
  2 and 3 will be less than  ; thus contributions of second and third mode will be over
estimated

 6 and damping ratios for higher modes will be more and hence, their contributions
will be less (which is consistent with the assumption made)
 Lecture 21-9
 It may be argued that over estimation of the contributions of second and third on the
response will compensate for neglecting the contributions of the higher modes
 Note that selection of first two frequencies is not always right
Lecture 21-10
Lecture 21-11
Lecture 22
 Solution of Equations of Motion Lecture 22-1

 Direct solution of MDOF solves the n x n coupled matrix differential equation

 cu  ku  p(t)
mu
 Solutions can be performed in frequency or time domain; in time domain the solution is an
extension of that for SDOF:

 If P(t) is a vector of irregular time histories of excitations, then it is solved using numerical
integration schemes
 The Newmark’s –b - method described for SDOF system can be easily extended to the
MDOF system

 The modifications consists of

 Converting u to vector u and the corresponding derivatives

 Replacing m, c, k by matrices M, C, K
 Replacing the equations by ku = p(t) or kuk+1 = pk+1
 In which k is a matrix and p(t) is a vector
 Solution requires inversion of the matrix k. once for all
 uk 1 vector is obtained by satisfying the equation of motion at k+1 the time station
 Lecture 22-2

 The equations of motions can be solved by a number of other numerical integrations

schemes like Wilson –  – method, Houbolt’s method, Alpha method etc

 Duhamel integrations, in principle can be extended to MDOF system; however, it is

cumbersome and therefore, not generally used
 Difficulty with direct integrations is fixing the value of t; general thumb of rule for good
T
accuracy is that t  where T is the time period of the system; for SDOF
10
it is known, but for MDOF system it is difficult to determine

 Without performing the frequency analysis, it is difficult to fix the cut off mode and
consequently the cut off T

 Other problem of direct integration is that errors are accumulated easily because of the size of
the problem; so very small t is required

 Trial values of t are adopted to obtain solutions of the MDOF system requiring more
computational time to fix the required value of t
 Lecture 22-3
 The alternative form of the equation (see the SDOF solution)

q  FN qk  H N FK 1
qi  [ui ui ui ]T

H n   1m -1 [  ( t ) 2   t ]T

 1 2
n   n   n      t  -   α + γ   t  
2 2 3 2
  - ω 2
  t   t - 2 ω   t - ω 2
  t
2
 
 - 2 ωn  t - ωn   t  t -  α + γ  t
1 2
FN    -ωn   t
2 2

 
 -ωn2 -2 ωn - ωn2  t -γ 
 

α  I  2 ωn  t  ωn2    t 
2

2 1
  ) t 
2
γ  2 ωn (1   ) t  ωn (
2
 In which I is an identity matrix of size n x n ; and are diagonal matrices; β, are scalars
 Lecture 22-4

Determination of other Responses

 The numerical integration scheme provides displacement vector

 For obtaining other responses, member end displacements and rotations are required

 For this, the member ends are identified and accordingly, the d.o.f. are consistent with global co-
ordinate system

 The following procedure is adopted to find the response quantities of interest.

 From the condensation relationships, find rotations at the joints corresponding to the ends of the
member
-1
θ = kθΔ - kθθ k

 Once the rotations are known, then the displacements and rotations at the ends of the members are
transformed from global to local co-ordinate system using transformation relations

; T is the transformation matrix

 Lecture 22-5

 Member end forces are determined by multiplying the vector of displacements in local co-ordinate
by the member stiffness matrix in the own co-ordinate system;
f  = k  

 In whichk, f’, k’ and are consistent with the member coordinates; k’ is the member stiffness matrix,
is member and displacement vector, f’ is the member end force vector

 Note that although the rotational d.o.f. are not included in the dynamic d.o.f., the rotations do occur
at the joints at every instant of time

 Even though the mass moment of inertia is not present at the joints, these rotations produced due to
the translational inertia force contribute to the member end response

 In addition to the moments, shears are also developed at the joint

 The member end responses, other than translations corresponding to the dynamic d.o.f., are
significantly influenced by the idealizations adopted in generating the mass matrix

 The simple lumped mass and consistent mass matrices include rotational inertia force which result
in increased rotations at the joints leading to increased moment and shear forces at the joints
 Lecture 22-6

 Determination of other response quantities other than displacement response require

rotational d.o.f, which are normally condensed out. This in turn requires some
additional computational effort
Examples
For the same 3 strorey frame, find an expression for the displacement responses at t =
2.52 s , given the following:

t  2.5s; t  0.02 s; u T  0.02 0.035 0.06  cm

u T  0.1 0.02 0.25 cm / s; PT  2.52 s   0.5 0.2 0.4 
1 1
 ;  ;
2 4
uT  0.6 0.12 0.15 cm / s 2 ;
_
M C
K  K
 t 2  t
4M 2C
  K
 0.02   0.02 
2
 Lecture 22-7
2 0 0  6.548 2.354 0   2 1 0 
4m 
2 0   100m  2.354 6.548 2.354   1177  m  1 2 1
_
K 2 
0
 0.02  0 0 1   0 2.354 3.274   0 1 1 

 43663.6 1412.4 0 
 m  1412.4 43663.6 1412.4 
 0 1412.4 11504.4 
 0.02   0.01 
_  4M 2C    2  4M   2
P t t     0.035   10   0.02  2C  C    0.02   10
  0.02   0.02    0.06 
2
 0.25 
   
 2 0 0   0.6   0.5 
   
 m  0 2 0  0.12  102  0.2 
 0 0 1   0.15  
 0.4 

_ _
K ut t  P t t
1 _
  _
ut  t   K  P t t
 
 Lecture 22-8

 0.01    0.01   0.02   0.01 

  2 2  102      
u t t  0.02  10    0.018    0.035   0.02  0.02 
 0.25   0.02    0.06   0.06   0.025 
   
 0.02 
 
  0.0348 
0.0537 
 
Mut t  Cut t  Kut t  Pt t
0.0092 
 
ut t   0.19 
 0.49 
 
Lecture 22-9
Lecture 22-10
Lecture 22-11
Lecture 22-12
Lecture 22-13
Lecture 23
 Lecture 23-1
Frequency Domain Solution
 Solution of MDOF system in frequency domain is just an extension of that of SDOF
system,; however, the solution does not lead to the same elegant expression and
understanding of the responses (the classical DAF vs  plots)
 Frequency domain analysis is based on the following:

 Fourier synthesis of excitation, so that excitation can be decomposed in to a

number of harmonic excitations in the form of A Sinit i 
i
 Theorem of superposition

 Solution of the dynamic system for harmonic excitation

 The solution provides a steady state solution which is independent of initial condition;
hence its use requires careful consideration
 A more compact form of the solution technique exists using FFT and IFFT algorithms
(solution technique is known as frequency domain analysis using FFT)

 Since FFT and IFFT algorithms are readily available now frequency domain solution
using the two algorithms are preferred now
 Lecture 23-2

 Equation of motion for harmonic excitation takes the following form

mu+ cu + ku = p sin  t
mu+ cu + ku = p(t)
p  t  =  p1 sinω1t p2 sinω2 t ...... pn sinωn t 
T

 The first one is taken up for illustration of frequency domain solution by direct
method; The second one is not amenable to frequency domain solution by direct
method

 Assuming solution of the form;

u  uc cos  t  us sin  t
m  -uc 2 cos t - us 2 sin  t   c  -uc sin  t  us cos  t   k  uc cos  t  us sin  t   p sin  t

 Equating like terms

 mus 2  c uc  kus  sin  t  p sin  t
 
 mus 2  c uc  kus  cos  t  0
 
 Lecture 23-3
 Equating like terms
 k - m 2

 c 2  
c  us
k - m  uc

p
0

-1
 k - m 2
 
us
uc

 c
c 
2
k  c 
 
p
0

 Solution of the above equation provides us and uc for a particular value of 

 Unlike SDOF system, the above solution can not be manipulated algebraically to an
elegant form

 When the responses are bending moments and shear forces for a member then
rotations at the joints are recovered using the condensation relationship

 Once the rotations are known at the joints, member end displacements are
completely known; member stiffness matrix multiplied by the member end
displacement provides the member end forces
 Lecture 23-4

Frequency domain analysis for non integrable periodic loading

 For solving the problem in frequency domain analysis, the irregular time histories are
Fourier synthesized and the equation of motion is solved in frequency domain as it is done
for SDOF; note that each time history has the same form of periodic function with variation
of amplitude only

 The solution procedure takes the following steps:

 Fourier synthesize each time history as

N
pi (t )   f sin( t  )
ij j ij
j 0
b 
2 j   tan -1  j fij  a 2j b2j
j  ij aj 
T  

 Consider j=1 to N where N is the number of terms included in the analysis; for
each value of j, find Fij, ωj and φij
 With these, determine the jth force vector; the ith element of the vector is

f ij sin  j t  ij 
 Lecture 23-5


 Since sin  jt  j  remains the same for elements of any vector, any element may
be written as

{ ftj }sin  j  j 
 Assuming displacement vector to be of the same form i.e.,

 
u j  ucj cos  jt  j  usj sin  jt  j  

u usj  can be determined from equation by

T
 For a particular frequency ωj,
cj
replacing {P} by {Fij}

 The ith element of the    

u displacement
 u cos vector
ij t  ujisudenoted
cij  j sin byt 
ij  sij  j ij 
 
 Lecture 23-6

 Which can be written as


uij  uij sin  j t  ij -  ij 
 In which

2 2
uij  ucij  usij

 ucij 
 ij  tan -1

 us 
 ij 
N

u t    u j sin  j t  ij -  ij 
j 1
Lecture 23-7
Lecture 23-8
Lecture 23-9
Lecture 23-10
Lecture 23-11
Lecture 23-12
Lecture 23-13
Lecture 24
 Solutions Using FFT and IFFT Lecture 24-1

 In complex domain, the solution of equation motion in frequency domain is performed

in more compact form using FFT and IFFT algorithm

 For SDOF system, it has already been described; For MDOF system, the concept can
be easily extended in matrix form

 The equation of motion is excited by eiωt at each DOF one at a time, then
+ cu + ku  I eit
mu
u(t )  u( )eit [k  m 2  i c ]u( )  I

u(i )  [k  m 2  i c ]1  h(i )

 For eiωt positioned at the second place
u(i )  h(i )
 u(iω) vectors thus obtained, if arranged as vertical column in sequence for n DOF,
then the matrix becomes h(iω) matrix itself where
h(i )  [k  m 2  i c ]1
 Lecture 24-2
h(i )  [k  m 2  i c ]1
 This matrix is called the frequency response function popularly known as FRF matrix
and is similar to flexibility matrix (better called dynamic flexibility matrix)
 Following this definition dynamic stiffness matrix is
[k  m 2  i c ]

 Note that these matrices are complex matrices

 Using FRF of a MDOF system, the MDOF system can be analyzed for any set of
excitation periodic and irregular; the responses are steady state solutions
 FRF of a MDOF system preserves the entire dynamic characteristics of the system in
a complete form; if FRF of a structure can be determined from experimental test,
then the system can be analyzed for any excitation without the knowledge of any
other thing
 System identification consists of finding FRF using prototype testing (especially used
for damaged structure)
 Analysis of MDOF system in frequency domain using FRF for any arbitrary loading
can be performed by frequency domain analysis using FFT as developed for SDOF
 Lecture 24-3

 The procedure is an extension for MDOF system and consists of the following steps

 Obtain condensed stiffness matrix form the full stiffness matrix

k  k - k k -1 k
   

  - k -1 k 
 

 Find C Matrix using Rayleigh Damping

c  m  k
 α and β are determined from two selected frequencies of the system (discussed
before)
 Fourier transform P(t) using FFT; it requires FFT of time histories; p1(t) …. pn(t)

 Take first N/2 +1 terms of the FFT of the time histories

 Form a matrix of p(ω) of size nx(N/2+1); first column corresponds to 0; second
column corresponds to Δω and so on, ith element of a column j corresponds to the ith
frequency content of the load pi(t)

 Obtain h(iω) for ω=0, (Δω),N/2 Δω

 Lecture 24-4

 Multiply h(iω) with each column of p(ω) matrix to obtain u i  of size nx(N/2+1)

 Obtain u i  from u i  by adding complex conjugates to each row of u i 

 Size of u i  is nxN matrix

 Make IFFT of each row of u i  ; note that both FFT and IFFT are to be performed for
n number of discrete series

 This provides discrete values of response u1(t) to un(t) at an interval of Δt; u(t) matrix
is of size nxN

 Once u(t) is obtained, θ(t) can be obtained

 (t )  - k-1 k  u(t )

 For any other response, the member end displacements and rotations at the two
ends of the member in global coordinate are selected from the u(t) and θ(t) vectors at
each time t.

 They are transformed to the local co-ordinate using the transformation matrix of the
member.
Lecture 24-5
Lecture 24-6
Lecture 24-7
Lecture 24-8
Lecture 24-9
Lecture 25
 Normal Mode Theory Lecture 25-1

 Normal mode theory is one of the elegant methods that has made dynamic analysis of
many structural vibration problems simple and easy to understand

 Another important feature of the method is that it enables one to physically understand the
dynamic problem in terms of its natural frequency of vibration

 Normal mode theory stipulates that the response of a MDOF system resembling a structure
subjected to dynamic excitation is weighted summation of its mode shapes (normal modes)
expressed mathematically as
u z

u is the displacement vector in structural co-ordinates  is the mode shape matrix,

Z is the vector generalized co-ordinates
 Both u and z are functions of t ; z is also called modal co-ordinates; z(t) is the time
dependent weighting functions; size of the vector depends upon the number of modes
considered in the analysis; accordingly, the size of  varies
 If z is Nx1, u is Nx1 and  is NxN, then the normal mode theory gives the exact results;
however, if size of z is less than u, then the result becomes inaccurate; number of modes or
generalized co-ordinates to be considered for obtaining good results depends upon the
problem and the response quantity of interest
 Lecture 25-2

 If it is considered that there are n d.o.f. and m is the number of modes considered,
then
unx1  nxm zmx1
 1 z1  2 z2  3 z3  ......  n zn
in which i is of size nx1, i is the mode number

 The above equation clearly illustrates that the response u is a weighted summation
of mode shapes, weighting functions being zi

Substituting for u in equation of motion and pre-multiplying by  , both LHS and

T

RHS of the equation of motion,
z   T c z   T k z   T p  t 
 T m 

if c is assumed to be Rayleigh damping then, c   m   k

 Substituting for c
 
z    T m    T k z   T k z   T p  t 
 T m 
mz ( m   k ) z  kz  p(t)
 Lecture 25-3

in which, because of the orthogonal property of mode shapes, m and k becomes

diagonal matrices

 As a result, the set of equations are uncoupled; each equation becomes an

equation of motion for SDOF:
zi  ci zi  k i zi  pi (t )
mi  i  1.......n
if m modes are only considered, then i =1…m

 The SDOF equations are solved for zi by using any technique in time/ frequency
domain described before; the technique to be adopted depends upon the duration of
loading and response quantity of interest.

 Once zi ( i = 1…..m) are obtained, u may be obtained as

u  nm zm1

 Equations of motion in generalized co-ordinates are SDOF equations; they may be

also written in the following form:
 pi (t )
Z i  2 ii Z i  i Z i 
2

mi
 Lecture 25-4

 in which ki
mi  iT mi  i2
mi
     2 
i
2
i i

 
i   i
2i 2
 As described before, in Rayleigh damping, the constants  and  may be

determined by two frequencies of the structures; generally, first frequency may be
considered

 The equations of motion as obtained above is preferred because of two reasons:

 It requires only undamped mode shapes and frequencies of the system to be

known; finding mi is trivial

 Damping does not become a big problem for the analysis provided (i)
proportional damping is assumed and (ii) damping ratio  is assumed to be the
same in all modes
 Lecture 25-5

 Thus, normal mode theory is widely used in structural dynamics for solving MDOF
systems in the linear range

 How many modes provide a good estimate of the response attracted the attention of
researcher

 The studies revealed that consideration of only first few modes is sufficient to get good
results since excitations generally do not excite higher modes( i.e. resonance takes
place)

 When excitations have high frequencies or when responses other than displacements
like BMs are to be obtained, consideration of large number of modes is required

 There are other ways to obtain the responses, other than the displacements, using less
number of modes in the modal analysis technique

 One such method will be discussed in the subsequent lecture

 The method requires the quasi static solution for the dynamic load
Lecture 25-6
Lecture 25-7
Lecture 25-8
Lecture 25-9
Lecture 25-10
Lecture 25-11
Lecture 25-12
Lecture 25-13
Lecture 25-14
Lecture 25-15
Lecture 25-16
Lecture 26
 Lecture 26-1

 Use of normal mode theory for support excitation is important for

 it allows to understand contribution of each mode of vibrations and

 provides a background to the development of response spectrum method

 For single support excitation, equation of motion

 cu  ku  -mI ug

mu

 Using normal mode theory, the developed SDOF system;

zi  2ii zi  i2 zi  i ug


N W i
  r 1 r r ; W is the weight of the rth floor
 
i 2 r
 N W  i
r 1 r r

 Each SDOF is associated with a thus,

 The above SDOF can be solved by any method described before

 Lecture 26-2

 After finding the displacements u(t), the member end forces are obtained in two ways

 Find the rotations θ from condensation relationship and then find the displacements

and rotations at the ends of a member in local co-ordinates using transformation;

multiply the member stiffness matrix by the member end displacement vector

(described before)

 Find the mode shape coefficient for the response quantity of interest and use the mode

summation equation to find u, replacing u by the response quantity of interest;

response quantity of interest could be base shear, member end forces, drift etc

 The mode shape coefficient for the response quantity of interest is obtained by a

separate static analysis of the structure

 Lecture 26-3

 In order to find the mode shape coefficients for the response quantity of interest, the
following procedure is adopted:
 Analyze the MDOF system for the static load vector p   2 m and find the
i i i
response quantity of interest R ( it could be one response quantity or a number of
i
response quantities represented by vector; accordingly R could be a single quantity
i
or a vector)
 Repeat the procedure for p ( i = 1….m)
i
 Arrange R in the form of a matrix; size of R will be s x m, where s is the number of
response quantities; note that if R contains member end forces, then global to local
transformation is required
 R matrix is mode shape matrix for the response quantities of interest
 This mode shape matrix replaces the displacement mode shape matrix in that the
relationship between the actual response and the modal response
Lecture 26-4
Lecture 26-5
Lecture 26-6
Lecture 26-7
Lecture 26-8
Lecture 26-9
Lecture 26-10
Lecture 27
 Lecture 27-1
 For most of the problems, consideration of first few modes is sufficient to get good
estimate of the responses; how many modes are to be considered is generally decided by
the mass participation factor 
given
i as
n
i  miir
i  r 1

in which n is the number of d.o.f. and i is the mode shape number, i is the mode
participation factor defined later, and m is the total mass of the system; m is decided
based on
m


i 1
i 1

 Even if the number of modes to be considered is selected based on the above concept,
the response quantities like bending moment and shear force may not be obtained
accurately; for that more number of modes are to be considered

 One approach called mode acceleration approach is developed to find good estimate of
response using less number of modes
 Mode Acceleration Approach Lecture 27-2

 The uncoupled SDOF equation of motion in generalized co-ordinate zi can be re-written as

1  pi t   1
zi  2    2  
zi  2i zi 
i  mi  i
 i be the mode shape coefficient for a response quantity of interest R(t), then
i i 
m  p t m
1
R t    2   2  
zi  2ii zi i
i 1 i mi i 1 i

 The first term is computed with m=n ie. by considering all modes and the second term is
calculated using only m number of modes

 The first term in that case provide quasi static response of the system for total load p(t) which
can be proved as below:
ku(t )  p(t )
T k  z  T p
z and z terms; therefore
The equation is equivalent to the modal equation without 
zi obtained from above equation is the quasi static generalized displacements for the ith
mode ie,
 Lecture 27-3
i T p
zi 
mii2
 Quasi static part R t  of response quantity of interest R t  may be written as
n
R t    i zi m  n 
i 1
n iT pi n
p
  2   i2 i
i 1 i mi i 1 i mi

 Thus response R t  may be written as

m
1
R t   R t     
zi  2ii zi i
i 1 
i
2

 Since first part of the response considers contribution from all modes, the error is introduced
only due to the second term

 Generally first part dominates the response R t  ,the response obtained by using the above
equation provides a better estimate of R t  with less number of modes than mode summation
approach
Lecture 27-4
Lecture 27-5
Lecture 27-6
Lecture 27-7
Lecture 27-8
Lecture 27-9
Lecture 27-10
Lecture 27-11
Lecture 27-12
 Modal Response Spectrum Analysis Lecture 28-1

 Use of normal mode theory has given rise to the popular Response Spectrum Analysis of Structures for
Earthquake

 A multi degree of freedom system subjected to single point support excitation due to ground motion provides the
following equation of motion: mu cu  ku  -mI u g

 Using normal mode theory the above equation motion can be written as a set of uncoupled equation of motion in
generalized co-ordinate
zi  2ii zi  i2 zi  i ug (i  1.....m)


i T mI  rN1 Wrri
in which i   T m  N (for buildings)
 
2
i  r 1 Wr r
i
i

 Is called mode participation factor denoting the contribution of the mode to the overall response, as is evident
from the SDOF equation of motion in generalized co-ordinate
 Mass participation factor i described before is used to find number of modes to be considered

 Solution of the modal equation by any method (time domain and frequency domain) provides the maximum value
of
 Lecture 28-2

 For different values of for a specified value of maybe obtained for the given time history of
 The plot of with provides the displacement spectrum of the ground acceleration ; this is designated by vs.
plot
 Once this is known for the ground acceleration two more spectrums are constructed from the
displacement response spectrum
 They are and ; they are plotted against ; vs. is called response spectrum of spectral acceleration

 If this acceleration spectrum is known for a given time history of acceleration, then maximum response of
the MDOF system can be obtained from static analysis of the system for an equivalent static load, if it is
assumed that that the MDOF is vibrating in only a single mode i.e. in mode
 Thus the equivalent static load for a particular mode of vibration is determined for the specific dynamic
excitation in the form of ground acceleration

 Static analysis of the structure is performed for the equivalent static load
 Lecture 28-3

 The equivalent static load is obtained in the following manner:

 It has been shown that

Fi  ki S di i

The maximum displacement for the mode:

S ai
zi  S di 
max
i2
The corresponding lateral force providing the maximum displacement in the mode maybe written as

Sai Sai
Fi  k ii  mi2 ii
i
2
i
2

Fi  i mi Sai
 Thus, if for the mode, is known, then can be determined provided, and spectral acceleration (for) are
known
 Lecture28-4

 If the structural system is analyzed for lateral load, then the responses provide the maximum response that
the structure would have provided for if it vibrates only in the mode

 The equivalent lateral load analysis is known as the modal response spectrum analysis of the structure

 Use of normal mode theory requires that contributions from all modes (m-modes) considered are added to
get the final responses

 In time history analysis, this is done by using the well-known relationship

u z
 In response spectrum method of analysis, the maximum response in each mode of vibration is obtained
 Since maximum response in each mode of vibration can occur at different times, the simple additions of the
maximum response obtained in the response spectrum method analysis will not give the final response

 Therefore, the responses in each mode of vibration are combined using some combination rules; one such
combination rule is called SRSS rule
 This introduces some approximations in the response spectrum method of analysis; the effect of these
approximations have been investigated and it has been found that in most cases, it doesn’t introduce much
error in the prediction of maximum responses
 Lecture 28-5

 The total response is obtained by summing up the product of and

m
u   i zi
i 1

 If the general case is considered, then R is

m
R   iR zi
i 1

is the ith mode shape coefficient of R

 This simple summation is not possible for the response spectrum method for two reasons:

 Response spectrum method uses absolute maximum value of ignoring the sign of

the response

 Maximum does not take place at the same time for all modes; thus, simple

summation to obtain maximum response becomes inadmissible

 Lecture 28-6
 In earthquake literature, three summation rules are prescribed

 These summation rules do not have strong theoretical base; however, they are formed to provide
either a conservative or a good estimate of R

Rules are:

 ABSSUM (Absolute sum of the responses in all modes)

 SRSS (Square root of the sum of the squares of the responses in all modes)

 CQC (Complete quadratic combination of the responses in all modes)

m
 ABSSUM rule: R   Ri
i 1

m
 SRSS rule: R  Ri2
i 1

m
 CQC rule: R  Ri2  ij Ri R j
i 1

 m is the number of modes considered and R is the response quantity of interest; is correlation
between mode responses
 Lecture 28-7

 ABSSUM provides a conservative estimate of the response as it is evident from the nature of the
summation

 SRSS rule provides a reasonably good estimate of the response for widely spaced frequencies of
structure

 In this approach, it is assumed that there is no correlation between the modal responses. This
assumption is not truly valid as there exists some correlation between modal responses

 However, with increased spacing between two frequencies of the structure, this correlation dies down

 CQC rule takes into account this correlation ; hence , provides a better estimate of R

 For ordinary frames, SRSS rule is widely used in practice

 For 3D building models with high eccentricity the frequencies are closely spaced and therefore CQC
rule is to be used for finding the responses (Eqns.5.14 and 5.15 of Fig. in the next slide refer to those
by Der Kiewrigian and Rosen blueth respectively)
Lecture28-8
Lecture 28-9
Lecture 28-10
Lecture 28-11
Lecture 28-12
Lecture 28-13
Lecture 28-14
Lecture 28-15
Lecture 28-16
Lecture 28-17
Lecture 29
 Lecture 29-1
Continuum system
 Infinite degrees of freedom exist for the continuum system; thus, dynamic equilibrium equation is
written similar to that for beam and plate
 For static equilibrium, beam equilibrium equation

d 4u
EI 4  p
dx

 For quasi-static case,

d 4u (t )
EI  p (t )
dx 4

 If is not quasi-static but dynamic, two more terms, inertia and damping would be present in the
above equation
 2u u  4u
m 2 c  EI 4  p (t )
t t x

 m is the mass per unit length; c is damping constant per unit length
 Above equation is written using the knowledge of beam theory, inertia force and damping force
 Lecture 29-2

 In subsequent section, the formal derivation of the equation of motion will be presented

 Direct solution of equation of motion is highly cumbersome. Therefore, the equation is solved using
normal mode theory

 The background of how the normal mode theory is used for the continuum system is presented
subsequently

 It also provides an understanding of the meaning of mode shapes of continuum system

 Normal mode theory as applied to MDOF system can be extended to develop the normal mode
theory as applied to the continuum system

 The mode shapes and frequencies of the continuum system is presented here for the beams and
axial member

 Unlike discrete system, mode shapes and frequencies of continuum system is derived by solving
the partial differential equation
 Continuum System Lecture 29-3

 Normal mode theory shows that the response of a humped mass system can be represented by
u  i zi (i  1..........m)
 For vibration only in one mode
u z
 The above equation leads to ;
 If mode shape coefficients for particular mode is known, then the responses of different d.o.f are known by
knowing the response of one quantity
 For large number of d.o.f, the curve passing through the curve passing through the points representing the
mode shape co efficient can be approximated by a function
 If this function is known or can be guessed (or assumed) then the responses at different d.o.f can be obtained
by knowing only
 Above observations can be extended for a continuum in which d.o.f are infinite; so a smooth function will
represent mode shape of the continuum
 If these functions are known or can be assumed, then responses of the system can be determined by knowing
i.e. solution of a SDOF system

 Dynamic analysis of continuum system for arbitrary /periodic /support excitation problem is popularly performed in
this fashion i.e., using normal modes
 Development of the equation of motion of the axial vibration of a rod Lecture 29-4

u P

x AE

 2u P
AE 2 
x x

 Since the inertia force balances the external force per unit length

P  2u
dx  Adx 2
x t
P
 Substituting for
x
 2u E  2u

t 2
 x 2
 2u 1  2u E
 2 2 c
x 2
c t 
 Lecture 29-5
 Solution of the equation of motion of the axial vibration of a rod
 The general solution of the equation of motion is
u  F1  ct  x   F2  ct  x 
Regardless of the type of function F, the ct  x  argument upon differentiation leads
to 2 F 1 2 F
2

x c 2 t 2
Hence, the differential equation is satisfied

 For component u  F1  ct  x  , it can be easily shown that the wave propagates in

positive x direction with speed c and the opposite phenomena is experienced for the
component u  F2  ct  x 

 So, the general solution consists of wave propagations in two opposite directions for
one dimensional wave propagation problem
 Solution of the equation of motion can be obtained by the method of separation of
variable is, u ( x, t )  u ( x)q (t )
Substituting in equation of motion,
1 2 y 1 1 2q

u x 2 c 2 q t 2
 Lecture 29-6

 Since both the sides are independent of each other, the ratios of the terms on the two
sides should be the same and hence,
2
 2u  w 
2
  u 0
x  c 
 2q
 2
q0
t 2

The general solution of the two equations are:

 
u  A sin x  B cos x
c c
q  C sin t  D cos t
 Constants of the two equations are obtained from the initial and boundary conditions

 Consider a stretched string between two fixed points separated by l, u (0, t )  u (l , t )  0


 Then B = 0 and u  (C sin t  D cos t )sin x
c
 Note A could be a function of t (not x)
ω
 For u (l , t )  0 , gives sin l=0
c
 Development of the equation of motion for the flexural vibration of the beam Lecture 29-7

From equilibrium consideration, consider the figure bellow.

v  2u M d 2u
 p  m 2 ;v  ; M  EI 2
x t x dx

Substituting from and,

 2u  2  d 2u 
M ( x) 2  2  EI ( x) 2   p ( x, t )
t t  dx 

 The above two derivations are without damping; since damping force negates the excitation force, it can be simply
added to the left hand side of the equation of motion i.e., equation of motion takes the form
d 4u
mu cu  EI 4  p ( x, t )
dx
in which is the mass per unit length ; is the moment of inertia per unit length and is the excitation per unit length
of the continuum
 Lecture 29-8

 For support excitation only, the above equation maybe easily modified by bringing in total acceleration in place of
relative acceleration ; equation of motion takes the form
d 4u
mut  cu  EI 4  0
dx
t
 u  u g  x  the equation becomes
Since, u
d 4u
 cu  EI ( x ) 4  m( x )u g
m( x)u
dx

 The equation of motion can be written also in terms of total displacement i.e.,

d 4ut d 4u g
t  cut  EI
m( x)u 4
 cu g  EI
dx dx 4
Since does not vary with for single point excitation, equation becomes
d 4ut
t  cut  EI
m( x)u 4
 cu g
dx
 For multi support excitation again, writing the equation of motion for the continuum becomes somewhat complex
although in principle it can be done
 Therefore, the continuum system is discretized in such cases
Lecture 29-9
Lecture 29-10
Lecture 29-11
Lecture 29-12
Lecture 29-13
Lecture 29-14
Lecture 29-15
Lecture 29-16
Lecture 30
 Lecture 30-1
Natural Frequencies and Mode Shapes of Continuum System
 Natural frequencies and mode shapes of the continuum system is obtained by solving the undamped equation of
motion without any excitation i.e.,
 2u d 4u
m( x) 2  cu  EI ( x) 4  0
t dx

The solution is obtained by assuming of the form

u ( x, t )   ( x) z (t )

which means that the vibration of the beam is distinguished by a single time history ; displacement along the
beam at any time obeys a function  ( x)

 This assumption leads to:

 2u  4u
2
  ( x) z (t )   iv
( x) z (t )
t x 4

Substitution in the equation of motion gives

 z (t ) EI ( x ) iv ( x )

z (t ) m( x) ( x)
 Lecture 30-2
 Thus the above two ratios are equal and say is equal to ; this leads to two equations
z (t )   2 z (t )  0
EI ( x) iv ( x)   2 m( x) ( x)  0
 The first equation shows that it is nothing but the frequency of a SDOF

 The second equation shows that there is an infinite number of frequencies and corresponding mode shapes
for the eigen value problem represented by the second equation

 If , then the second equation can be written as

 iv  x    4  x   0

Where,  4  2m
EI
The general solution of the equation is

  x   c1 sin  x  c2 cos  x  c3 sinh  x  c4 cosh  x

 Four unknown constants are evaluated from four displacement boundary conditions of the continuum; upon
finding to , the trigonometric function equated to zero gives the value of
 Lecture 30-3

 The solutions are presented for

 Simply supported beam
 Cantilever
 Fixed-fixed beam
 Propped cantilever

 For simply supported beam:

  0   0; EI   0   0
  L   0; EI    L   0
First set of conditions provide Leading to
  x   c1 sin  x  c3 sinh  x

Second set of conditions provide

c1 sin  L  c3 sinh  L  0
 2  c1 sin  L  c3 sinh  L   0
From these equations,
c3 sinh  L  0
 Lecture 30-4
Since cannot be zero (then),
which leads to
c1 sin  L  0
is a trivial solution
gives

n 2 2 EI
n  2 n  1, 2,3, 4....
L m
n x
and n  x   c1 sin L

 Cantilever beam:  0   0   0   0
EI    L   0 EI    L   0

Substitution of these conditions lead to

sin  L  sinh  L cos  L  cosh  L  c1 
cos  L  cosh  L  0
  sin  L  sinh  L  c2 
Setting determinate of the matrix as zero gives
1  cos  L  cosh  L  0
 Lecture 30-5
Numerical solution (not closed form) provides
 n L  1.875, 4.694,7.855,11 for n  1, 2,3, 4
3.516 EI 22.03 EI 61.7 EI 121 EI
1  2
;  2  2
; 3  2
; 4  2
;
L m L m L m L m
And
 cos  n L  cosh  n L 
n  x   c1   cos  n x  cosh  n x    sin  n x  sinh  n 
x
 sin  n L  sinh  n L 

 Fixed-fixed beam:
 0     L   0
  0      L   0

Application of the boundary condition gives

cos  n L.cosh  n L  1
1 L  4.73;  2 L  7.85; 3 L  11;  4 L  14.13
 Lecture 30-6
 Propped Cantilever
 0    L   0
  0     L   0
B.C provides
tan  n L  tanh  n L  0
n  x   c1 sin  n x  sinh  n x   n cosh  n x  cos  n x 
sin  n L  sinh  n L
n 
cos  n L  cosh  n L

 Approximate evaluation of natural frequency in a mode of vibration can be obtained if mode shape is
approximated by a function of x

 This approximation mode shape can also be termed as shape function; if a shape function is assumed, then
forced vibration (undamped) in a mode can be expressed as
u ( x, t )  z sin t. ( x)
u ( x, t )  z  cos t. ( x)
 Lecture 30-7
 Equation leading to determination of natural r frequency provides,
th

EI ( x)r ( x)  2r m( x)r ( x)

 Pre-multiplying by  n and integrating

 n ( x)EI ( x)r ( x) dx  
L L
  m( x) n ( x) r ( x)dx
2
r
0 0
L L
 EI ( x) n ( x)r ( x)dx    m( x) n ( x) r ( x)dx
2
r
0 0

 Similarly, one gets L L

 EI ( x) n ( x)r ( x)dx    m( x) n ( x) r ( x)dx
2
n
0 0

 Subtracting one equation from the other

L
(   )  m( x) n ( x) r ( x)dx  0
2
n
2
r
0
L
 Thus, orthogonality with respect to mass is obtained as 
0
m( x) n ( x) r ( x)dx  0 rn

 n ( x)EI ( x)r ( x) dx  0
L
 Similarly orthogonality with respect to stiffness can be derived as 0
 Lecture 30-8
Determination of Approximate Natural Frequency by Rayleigh’s Method
 Maximum potential energy of the system over a vibration cycle
L
1
Es   EI u0  dx
2

 2u 0
2
In which u0  2
x at max disp L
z02
Es   EI   x  dx
2

2 0

 Maximum kinetic energy is associated with maximum velocity

L
z02 2
Ek    m   x  dx
2

2 0

L L
 Equating and z02 z02 2
       
2 2

2 0  EI   x dx 
2
  m  x dx
0
L

 EI    x  dx
2

2  0
L

 m   x  dx
2

 This is known as rayleigh’s quotient for a system.

Lecture 30-9
Lecture 30-10
Lecture 30-11
Lecture 30-12
Lecture 30-13
Lecture 31
 Lecture 31-1
Normal Mode Theory for Continuum System
 For continuum system, the responses are obtained mostly using normal mode theory; however;
there are techniques to solve the continuum equation by direct integration of equation of motion
 For the use of normal mode theory, u(x, t) is written as

u ( x, t )   n ( x) zn (t )
n 1

in which n ( x) is the nth mode shape and zn is the nth generalized coordinates
 Obeys the orthogonality condition, which was proved previously; orthogonality conditions are given
by n ( x)
l

 m n ( x) m ( x) dx  0 for m  n
0

n
 ( x )  EI  
m ( x )  dx  0 for m  n

 Substituting for u(x, t) in the equation function

 

m ( x ) z  
 n n  m zn (t )  p( x, t )
 EI   ( x ) 
m 1 n 1
 Lecture 31-2
multiplying both sides by m and integrating over length l,
L L L

 zn  mmn dx   zn   m  EI n dx   m p ( x, t ) dx
0 0 0
 Using the orthogonality condition and introducing damping
zn  2n zn  n2 zn  pn (t )
 n  1......
L
mn   m n  dx
2
in which
0
l

kn   n  EI n dx
0
L
k cn
n2  n ; cn   c n  dx;
2
2n 
mn 0
mn
L

 n ( x) p( x, t )dx
pn (t )  0
L

 m n 
2

0
 Lecture 31-3
 Thus, the continuum system is converted to n number of SDOF system (n  ) like the discrete
MDOF system; by considering truncated mode it is possible to solve m number of equation to
find the response quantities of interest
 Once a number of SDOF systems are obtained, the solution procedures for them remain the
same as those of discrete MDOF systems
 For continuum system, n ( x) and n are to be obtained for various boundary conditions; the
standard cases have been discussed before
Support Motion
 When a continuum is subjected to support motion, the equation of motion takes the form similar
to the discrete system
 Consider the cases shown in figure below
 Lecture 31-4
 The equation of motion has been derived before
 If the supports undergo different motions, then writing the equation of motion for continuum
becomes difficult. The problem involves finding an equation for the elastic curve of the continuum
when a unit support motion is given keeping the other supports locked

 The solutions is difficult for the general case for obtaining a closed form solution; however, for
certain cases, it may be possible
 In such situation, the continuum may be converted to discrete lumped mass model and use the
problem solution for multi support excitation

 Whenever the response spectrum method of analysis is used for support excitation equivalent
static load in each mode of vibration can be determined in the same way as that was done for
discrete system

 The equivalent static load in each mode is continuous over the structure, since mode shapes are
continuous functions

 The mean peak response can be determined using SRSS rule

 Lecture 31-5
 Responses other than the displacement is obtained either by direct method or by using response
mode shape coefficient method

 In the continuum approach, the rotations are not needed, unlike discrete approach, to obtain any
response quantity of interest. The appropriate derivative of the displacement provides the response

 For example, if bending moment at only section is required

 2u
m(t )  EI 2
x

 A better approach to find the response is to use the following expression

m
R (t , x  l )   zi (t )i ( x  l )
i 1

 The mode shape coefficient for the response is obtained by solving the static problem with the
distributed load
 Lecture 31-6
Mode Acceleration Method
 Like the discrete method, the mode acceleration method can be used to find more accurate
responses using the limited number of modes

 The relevant equation remains the same as that for the discrete method

m
1 m  pi t  
R  R    2 ( 
zi  2i zi )i  R    2  Z i  i
i 1 i i 1  i mi 

 is the quasi-static response and denotes the mode shape co-efficient for the response

 The quasi-static response is obtained by performing the static analysis of the continuum for the load

 Note that, the quasi-static response of each type requires the derivatives of In many cases, these
derivatives may have to be obtained by numerical method
Lecture 31-7
Lecture 31-8
Lecture 31-9
Lecture 31-10
Lecture 31-11
Lecture 31-12
Lecture 31-13
Lecture 31-14
Lecture 31-15
Lecture 31-16
Lecture 31-17
Lecture 31-18
Lecture 31-19
Lecture 31-20