Structural

Dynamics
& Earthquake
Engineering

Dr Alessandro
Palmeri
Structural Dynamics
Recap
& Earthquake Engineering
Fourier Series Lectures #3 and 4: Fourier Analysis + Frequency
Fourier Response Function for SDoF oscillators
Transform

Fast Fourier
Transform

Dr Alessandro Palmeri

Architecture, Building & Civil Engineering @ Loughborough University

Tuesday, 9th October 2018

 Intended Learning Outcomes

Structural
Dynamics
& Earthquake
Engineering

Dr Alessandro
Palmeri

Recap
At the end of this unit (which includes the tutorial next
Fourier Series
week), you should be able to:
Fourier Derive analytically the frequency response function
Transform

Fast Fourier
(FRF) for a SDoF system
Transform
Use the Fourier Analysis to study the dynamic
response of SDoF oscillators in the frequency domain
 Recap of Last-Week Key Learning Points

Structural
Dynamics
Unforced Undamped SDoF Oscillator
& Earthquake
Engineering Equation of motion (forces):
Dr Alessandro
Palmeri m ü(t) + k u(t) = 0 (1)
Recap Equation of motion (accelerations):
Fourier Series
ü(t) + ω02 u(t) = 0 (2)
Fourier
Transform
Natural circular frequency of vibration:
Fast Fourier r
Transform
k
ω0 = (3)
m
Time history of the dynamic response u(t) for given
initial displacement u(0) = u0 and initial velocity
u̇(0) = v0 :
v0
u(t) = u0 cos(ω0 t) + sin(ω0 t) (4)
ω0
 Recap of Last-Week Key Learning Points

Structural Unforced Damped SDoF Oscillator

Dynamics
& Earthquake
Engineering Equation of motion (forces):
Dr Alessandro
Palmeri
m ü(t) + c u̇(t) + k u(t) = 0 (5)

Recap
Equation of motion (accelerations):
Fourier Series ü(t) + 2 ζ0 ω0 u̇(t) + ω02 u(t) = 0 (6)
Fourier
Transform Viscous damping ratio and reduced (or damped) natural
Fast Fourier circular frequency:
Transform
c
ζ0 = <1 (7)
2 m ω0
q
ω 0 = 1 − ζ02 ω0 (8)
Time history for given initial conditions:
 
v0 + ζ0 ω0 u0
u(t) = e−ζ0 ω0 t u0 cos(ω 0 t) + sin(ω 0 t) (9)
ω0
 Recap of Last-Week Key Learning Points

Structural
Dynamics Harmonically Forced SDoF Oscillator (1/2)
& Earthquake
Engineering
Equation of motion (forces):
Dr Alessandro
Palmeri
m ü(t) + c u̇(t) + k u(t) = F0 sin(ωf t) (10)
Recap

Fourier Series
The dynamic response is the superposition of any particular
Fourier
Transform
integral for the forcing term (up (t)) and the general solution
Fast Fourier
of the related homogenous equation (uh (t)):
Transform
u(t) = uh (t) + up (t) (11)

General solution (which includes two integration constants

C 1 and C 2 ):
h i
uh (t) = e−ζ0 ω0 t C 1 cos(ω 0 t) + C 2 sin(ω 0 t) (12)
 Recap of Last-Week Key Learning Points

Structural
Dynamics Harmonically Forced SDoF Oscillator (2/2)
& Earthquake
Engineering
Particular integral:
Dr Alessandro
Palmeri up (t) = ust D(β) sin(ωf + ϕp ) (13)
Recap
Static displacement and frequency ratio:
Fourier Series

Fourier
F0
Transform
ust = (14)
k
Fast Fourier ωf
Transform β= (15)
ω0
Dynamic amplification factor and phase lag:
1
D(β) = q (16)
2 2
(1 − β2) + (2 ζ0 β)
2 ζ0 β
tan(ϕp ) = (17)
1 − β2
 Recap of Last-Week Key Learning Points

Structural
Dynamics
& Earthquake
Dynamic Amplification Factor
Engineering
50.0
Dr Alessandro
Palmeri Ζ0 = 0

Ζ0 =0.05
Recap
10.0 Ζ0 =0.10
Fourier Series

Fourier Ζ0 =0.20
5.0
Transform
Ζ0 =0.50
Fast Fourier
D

Transform

1.0

0.5

0.0 0.5 1.0 1.5 2.0 2.5 3.0

Β
 Recap of Last-Week Key Learning Points

Structural
Dynamics
& Earthquake
Phase lag (= Phase of the steady-state response − Phase
Engineering of the forcing harmonic)
Dr Alessandro
Palmeri Π Π

Recap

Fourier Series 3Π 3Π
4 4
Fourier
Transform

Fast Fourier
Transform Π Π
jP

2 Ζ0 = 0 2

Ζ0 =0.05

Ζ0 =0.10
Π Π
4 4
Ζ0 =0.20

Ζ0 =0.50

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Β
 Fourier Series

Structural
Dynamics
& Earthquake
Engineering

Dr Alessandro
Palmeri
Fourier was a French
Recap
mathematician and
Fourier Series
physicis, born in Auxerre,
Fourier
Transform and he is best known for
Fast Fourier
initiating the investigation
Transform of Fourier series and their
applications to problems
of heat transfer and
vibrations

Jean Baptiste Joseph Fourier

(21 Mar 1768 – 16 May 1830)
 Fourier Series

Structural
Dynamics
& Earthquake We have obtained a closed-form solution for the dynamic
Engineering
response of SDoF oscillators subjected to harmonic
Dr Alessandro
Palmeri
excitation

Recap How can we extend such solution to a more general case?

Fourier Series

Fourier
Transform
Since the dynamic system is linear, the superposition
principle holds
Fast Fourier
Transform
The Fourier series allows us decomposing a periodic signal
into the sum of a (possibly infinite) set of simple harmonic
functions
We can therefore: i) decompose the forcing function in its
simple harmonic components; ii) calculate the dynamic
response for each of them; and then iii) superimpose all
these contributions to get the overall dynamic response
 Fourier Series

Structural
Dynamics If the forcing function f (t) is periodic with period Tp :
& Earthquake
Engineering n
X
Dr Alessandro f (t) = F0 + Fj sin(Ωj t + Φj ) = f (t + Tp ) (18)
Palmeri
j=1
Recap
where:
a0
Fourier Series F0 = (19)
2
Fourier q
Transform
Fj = aj2 + bj2 (for j ≥ 1) (20)
Fast Fourier
Transform aj
tan(Φj ) = (for j ≥ 1) (21)
bj
in which: Z Tp
2
aj = f (t) cos(Ωj t) dt (for j ≥ 0) (22)
Tp 0
Z Tp
2
bj = f (t) sin(Ωj t) dt (for j ≥ 1) (23)
Tp 0
2π
Ωj = j (24)
Tp
 Fourier Series

Structural Approximating a square wave of unitary amplitude and period

Dynamics
& Earthquake Tp = 2 s with an increasing number n of harmonic terms
Engineering
n=1 n=3
Dr Alessandro
Palmeri
1.0 1.0

Recap 0.5 0.5

force  kN

force  kN
Fourier Series 0.0 0.0

Fourier -0.5 -0.5

Transform -1.0 -1.0

Fast Fourier 0 1 2 3 4 0 1 2 3 4
Transform time  s time  s

n=5 n = 15

1.0 1.0

0.5 0.5
force  kN

force  kN
0.0 0.0

-0.5 -0.5

-1.0 -1.0
0 1 2 3 4 0 1 2 3 4
time  s time  s
 Fourier Series

Structural
Dynamics
& Earthquake
The same approach can be
Engineering adopted for a non-periodic
Dr Alessandro signal, e.g. the so-called
Palmeri
Friedlander waveform, which is
Recap often used to describe the time
Fourier Series history of overpressure due to
Fourier blast:
Transform
(
Fast Fourier p0 , if t < 0
Transform p(t) = −t/τ t

(25)
p0 + ∆p e 1 − τ , if t ≥ 0

where p0 is the atmospheric pressure, ∆p is the maximum

overpressure caused by the blast, and τ defines the timescale of
the waveform

Zero padding is however required, which consists of extending

the signal with zeros
 Fourier Series

Structural Approximating a Friedlander waveform (p0 = 0, ∆p = 100kPa,

Dynamics
& Earthquake τ = 0.01 s) with an increasing number n of harmonic terms
Engineering
n = 10 n = 20
Dr Alessandro
Palmeri 100 100

80 80
Recap
pressure  kPa

pressure  kPa
60 60

Fourier Series 40 40

20 20
Fourier
Transform 0 0
-20 -20
Fast Fourier -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Transform time  ds time  ds

n = 40 n = 80
100 100

80 80
pressure  kPa

pressure  kPa
60 60

40 40

20 20

0 0

-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
time  ds time  ds
 Fourier Series

Structural
Dynamics Once the forcing signal is expressed as:
& Earthquake
Engineering n
X
Dr Alessandro f (t) = F0 + Fj sin(Ωj t + Φj ) (18)
Palmeri
j=1
Recap
The dynamic response can be evaluated as:
Fourier Series
n
Fourier F0 X
Transform u(t) = uh (t) + + uj (t) (26)
k
Fast Fourier j=1
Transform
where:
Fj
uj = D(βj ) sin(Ωj t + Φj + ϕj ) (27)
k
in which:
Ωj
βj = (28)
ω0
2 ζ 0 βj
tan(ϕj ) = (29)
1 − βj2
 Fourier Transform

Structural
Dynamics
& Earthquake
The Fourier Transform (FT) can be thought as an
Engineering extension of the Fourier series, that results when the
Dr Alessandro
Palmeri
period of the represented function approaches infinity

Recap

Fourier Series
The FT is a linear operator, often denoted with the
Fourier
symbol F, which transforms a mathematical function of
Transform time, f (t), into a new function, denoted by
Fast Fourier
Transform
F (ω) = Fhf (t)i, whose argument is the circular
frequency ω (with units of radians per second)

The FT can be inverted, in the sense that, given the

frequency-domain function F (ω), one can determine
the frequency-domanin counterpart, f (t) = F −1 hF (ω)i,
and the operator F −1 is called Inverse FT (IFT)
 Fourier Transform

Structural
Dynamics
& Earthquake
Engineering In Structural Dynamics, the time-domain signal f (t) is
Dr Alessandro often a real-valued function of the time t, while its
Palmeri
Fourier transform is a complex-valued function of the
Recap circular frequency ω, that is:
Fourier Series

Fourier F (ω) = FR (ω) + ı FI (ω) (30)

Transform

Fast Fourier
Transform where:
√
ı = −1 is the imaginary unit
FR (ω) = <hF (ω)i is the real part of F (ω)
FI (ω) = =hF (ω)i is the imaginary part of F (ω)
q
|F (ω)| = FR2 (ω) + FI2 (ω) is the absolute value (or
modulus) of F (ω)
 Fourier Transform

Structural
Dynamics
There are several ways of defining the FT and the IFT
& Earthquake
Engineering
(depending on the applications)
Dr Alessandro
Palmeri In this module, we will always use the following
Recap mathematical definitions:
Z +∞
Fourier Series

Fourier
F (ω) = Fhf (t)i = f (t) e−ı ω t dt (31)
Transform −∞
Z +∞
Fast Fourier 1
Transform
f (t) = F −1 hF (ω)i = F (ω) eı ω t dω (32)
2π −∞

Note that, according to the Euler’s formula, the following

relationship exists between the complex exponential function and
the trigonometric functions:
eı θ = cos(θ) + ı sin(θ) (33)
 Fourier Transform

Structural
Dynamics
& Earthquake
Engineering

Dr Alessandro
Palmeri The main reason why the FT is widely used in Structural
Dynamics, is because it allows highlighting the distribution
Recap
of the energy of a given signal f (t) in the frequency domain
Fourier Series

Fourier
Transform The energy E is always proportional to the square of the
Fast Fourier
Transform
signal, e.g.:
1
Potential energy in a SDoF oscillator: V (t) = 2 k u 2 (t)
1
Kinetic energy in a SDoF oscillator: T (t) = 2 m u̇ 2 (t)
 Fourier Transform

Structural
Dynamics
& Earthquake
Engineering
According to the Parseval’s theorem, the cumulative energy
Dr Alessandro
Palmeri E contained in a waveform f (t) summed across all of time t
Recap
is equal to the cumulative energy of the waveform’s FT F (ω)
Fourier Series
summed across all of its frequency components ω:
Fourier
Transform
Z +∞ Z +∞
Fast Fourier 1 2 1
Transform E= α f (t) dt = α |F (ω)|2 dω (34)
2 −∞ 2π 0

where α is the constant appearing in the definition of the

energy (e.g. α = k for the potential energy and α = m for
the kinetic energy)
 Fourier Transform

Structural
Dynamics
& Earthquake
Engineering

Dr Alessandro
Example: For illustration purposes, let us consider the
Palmeri following signal in the time domain:
Recap 2
Fourier Series
f (t) = F0 e−(t/τ ) cos(Ω t) (35)
Fourier
Transform consisting of an exponentially modulated (with time scale τ )
Fast Fourier
Transform
cosine wave (with amplitude F0 and circular frequency Ω),
whose FT in the frequency domain is known in closed form:
√
 
2 2 2 1
F (ω) = Fhf (t)i = π τ F0 e−τ (ω +Ω )/4 cosh Ω τ2 ω
2
(36)
 Fourier Transform

Structural
Dynamics
Effects of changing the time scale τ = 1, 3, 5 s (while Ω = 1 rad/s)
& Earthquake
Engineering f (t) |F (ω)|
Dr Alessandro 1.0
Palmeri Τ= 1 s 4
Τ= 1 s
Τ= 3 s
0.5 3 Τ= 3 s

ÈFÈ  F0
Τ= 5 s
Recap
f  F0

Τ= 5 s
2
0.0
Fourier Series
1
Fourier -0.5
Transform 0
-10 -5 0 5 10 0 2 4 6 8 10 12
ts Ω s  rad
Fast Fourier
Transform
f (t)2 |F (ω)|2 /π

1.0 6
Τ= 1 s
5 Τ= 1 s
0.8 Τ= 3 s

ÈF 2  HΠ F02L
4 Τ= 3 s
0.6 Τ= 5 s
f 2  F02

3 Τ= 5 s
0.4
2
0.2 1
0.0 0
-10 -5 0 5 10 0 2 4 6 8 10 12
ts Ω s  rad
 Fourier Transform

Structural Effects of changing the circular frequency Ω = 1, 3 rad/s (while

Dynamics
& Earthquake τ = 1 s)
Engineering
f (t) |F (ω)|
Dr Alessandro
Palmeri 1.0
W= 1 rads 4
W= 1 rads
W= 5 rad.s
0.5
Recap 3 W= 5 rad.s

ÈFÈ  F0
f  F0

Fourier Series 0.0 2

Fourier -0.5 1
Transform
-1.0 0
-10 -5 0 5 10 0 2 4 6 8 10 12
Fast Fourier ts Ω s  rad
Transform

f (t)2 |F (ω)|2 /π

1.0 6
W= 1 rads
W= 1 rads
0.8 W= 5 rad.s 5
W= 5 rad.s

ÈF 2  HΠ F02L
4
0.6
f 2  F02

3
0.4
2
0.2 1
0.0 0
-10 -5 0 5 10 0 2 4 6 8 10 12
ts Ω s  rad
 Fourier Transform

Structural
Dynamics
& Earthquake
The FT enjoys a number of important properties, including:
Engineering

Dr Alessandro
Palmeri Linearity:
Recap

Fourier Series
Fha1 f1 (t) + a2 f2 (t)i = a1 F1 (ω) + a2 F2 (ω) (37)
Fourier
Transform

Fast Fourier
Transform Time shift:
Fhf (t − τ )i = e−ı ω τ F (ω) (38)

Time scaling:

1 ω 
Fhf (α t)i = F (39)
|α| α
 Fourier Transform

Structural
Dynamics
& Earthquake
Moreover (very importantly):
Engineering

Dr Alessandro
Palmeri Derivation rule:
dn
Recap
 
Fourier Series F f (t) = (ı ω)n F (ω) (40)
dt n
Fourier
Transform

Fast Fourier
Transform Convolution rule:
Z +∞
Fh{f ∗ g}(t)i = f (t) g(t − τ ) dτ
−∞
Z +∞
(41)
= f (t − τ ) g(t) dτ
−∞
= F (ω) G(ω)
 Fast Fourier Transform

Structural
Dynamics
& Earthquake The FT is a very powerful tool, but we can use it mainly if we have a
Engineering
simple mathematical expression of the signal f (t) in the time domain
Dr Alessandro
Palmeri
Very often the signal f (t) is known at a number n discrete time instants
Recap within the time interval [0, tf ]
Fourier Series

Fourier In other words, we usually have an array of the values fr = f (tr ), where:
Transform
tr = (r − 1) ∆t is the r th time instant
Fast Fourier
Transform r = 1, 2 · · · , n is the index in the time domain
∆t = tf /(n − 1) is the sampling time (or time step)
νs = ∆t −1 is the sampling frequency (i.e. the number of points
available per each second of the record)

Can we still use the frequency domain for the dynamic analysis of linear
structures?
 Fast Fourier Transform

Structural
Dynamics
& Earthquake The answer is yes...And the Fast Fourier Transform (FFT) can be used to
Engineering
transform in the frequency domain the discrete signal fr
Dr Alessandro
Palmeri
The FFT (which is implemented in any numerical computing language,
Recap including MATLAB and Mathematica) is indeed an efficient algorithm to
Fourier Series compute the Discrete Fourier Transform (DFT), of great importance to a
Fourier
wide variety of applications (including Structural Dynamics)
Transform

Fast Fourier The DFT is defined as follows:

Transform
n
X
Fs = DFThfr i = fr e2 π ı(r −1)(s−1)/n (42)
r =1

where n is the size of both the real-valued arrays fr in the time domain
and of the complex-valued array Fs in the frequency √ domain (i.e.
r = 1, 2, · · · , n and s = 1, 2, · · · , n), while ı = −1 is once again the
imaginary unit
 Fast Fourier Transform

Structural It can be proved mathematically that, for ω < 2 π νN = π/∆t, the

Dynamics
& Earthquake array Fs , computed as the DFT of the discrete signal fr , gives a
Engineering
numerical approximation of the analytical FT of the continuous
Dr Alessandro
Palmeri
signal f (t).

Recap In other words:

Fourier Series

Fourier Fs ≈ F (ωs ) (43)

Transform

Fast Fourier where:

Transform
ωs = (s − 1) ∆ω is the sth circular frequency where the DFT
is computed
∆ω = 2 π/(n ∆t) is the discretisation step on the frequency
axis
νN = νs /2 is the Nyquist’s frequency, and only signals with
the frequency content below the Nyquist’s frequency can be
represented
 Fast Fourier Transform

Structural Comparing FT (red solid lines) with FFT (blue dots)

Dynamics
& Earthquake (∆t = 0.7 s; n = 58; ∆ω = 0.155 rad/s; νN = 0.714 Hz)
Engineering
f (t) |F (ω)|
Dr Alessandro
Palmeri 1.0 æ æ
æ
æ
2.5 æ æ
0.8
2.0
Recap 0.6 æ
æ

ÈFÈ  fmax
æ
f  fmax

0.4 æ
1.5 æ

Fourier Series 0.2

æ

1.0 æ
æ

Fourier 0.0 æææææææææææææææææææææ

æ
æ æ
ææææ
ææææææææææææææææææ
æ
æ
æ
0.5 æ
æ

Transform
æ æ
-0.2 æ
æ
æ æ
-0.4 æ æ
0.0 æ æ æ æ æ æ æ æ æ æ æ æ æ æ

Fast Fourier 0 10 20 30 40 0 1 2 3 4
Transform ts Ω s  rad

FR (ω) FI (ω)
æ

2 æ
æ
2
æ
1 æ 1 æ
æ
FR  fmax

FI  fmax
æ
æ
æ æ
æ

0 æ
æ æ æ æ æ æ æ æ æ æ æ æ æ æ 0 æ æ
æ
æ æ
æ
æ æ æ æ æ æ æ æ æ æ æ æ æ æ
æ
æ
æ
æ æ
æ æ
-1 -1
æ
æ

-2 -2
æ
æ

0 1 2 3 4 0 1 2 3 4
Ω s  rad Ω s  rad
 Fast Fourier Transform

Structural
Dynamics Working with discrete signal can be tricky... A typical example is the phenomenon
& Earthquake of aliasing
Engineering

Dr Alessandro In signal processing, it refers to: i) different signals becoming indistinguishable

Palmeri
when sampled; ii) the distortion that results when the signal reconstructed from
samples is different from the original continuous signal
Recap

Fourier Series

Fourier
Transform

Fast Fourier
Transform

In the figure above, the red harmonic function of frequency νred = 0.9 Hz is
completely overlooked as the sampling rate is νs = 1 Hz (black dots), and
therefore the Nyquist’s frequency is νN = 0.5 Hz < νred

The reconstruction will then identify (incorrectly) the blue harmonic function of
frequency νblue = 0.1 Hz < νN
 Frequency Response Function

Structural
Dynamics
& Earthquake The equation of motion for a SDoF oscillator in the time domain
Engineering
reads:
Dr Alessandro 1
Palmeri ü(t) + 2 ζ0 ω0 u̇(t) + ω02 u(t) = f (t) (44)
m
Recap

Fourier Series By applying the FT operator to both sides of Eq. (44), one obtains:
Fourier  
Transform 1
F ü(t) + 2 ζ0 ω0 u̇(t) + ω02 u(t) = F


f (t)
Fast Fourier m
Transform
1
∴ Fhü(t)i + 2 ζ0 ω0 Fhu̇(t)i + ω02 Fhu(t)i = Fhf (t)i
m (45)
1
∴ (ı ω)2 U(ω) + 2 ζ0 ω0 (ı ω) U(ω) + ω02 U(ω) = F (ω)
m
1
∴ −ω 2 + 2 ı ζ0 ω0 ω + ω02 U(ω) =


F (ω)
m
 Frequency Response Function

Structural
Dynamics
The equation of motion in the frequency domain (the last of Eqs.
& Earthquake (45)) has been posed in the form:
Engineering

Dr Alessandro 1
−ω 2 + 2 ı ζ0 ω0 ω + ω02 U(ω) =


Palmeri F (ω) (46)
m
Recap

Fourier Series
where F (ω) = Fhf (t)i and U(ω) = Fhu(t)i are the FTs of
dynamic load and dynamic response, respectively
Fourier
Transform

Fast Fourier We can rewrite the above equation as:

Transform

F (ω)
U(ω) = H(ω) (47)
m
in which the complex-valued function H(ω) is called Frequency
Response Function (FRF) (or Transfer Function), and is defined
as:
−1
H(ω) = ω02 − ω 2 + 2 ı ζ0 ω0 ω (48)
 Frequency Response Function

Structural
Dynamics
& Earthquake
FRF for ζ0 = 0.05 – Note that: |H(ω)| = D(ω/ω0 )
Engineering

Dr Alessandro
10
Palmeri ÈHÈ
Recap RXH\
5
Fourier Series
IXH\
H  Ω20

Fourier
Transform

Fast Fourier
0
Transform

-5

-10
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Ω  Ω0
 Frequency Response Function

Structural Procedure for the dynamic analysis of SDoF oscillators in the frequency domain:
Dynamics
& Earthquake
Engineering 1 Compute the DFT of the dynamic force at discrete frequencies ωs (see Eqs.
Dr Alessandro
(42) and (43)):
Palmeri Xn
F (ωs ) ≈ DFThfr i = fr e2 π ı(r −1)(s−1)/n (49)
r =1
Recap

Fourier Series
2 Define analytically the complex-valued FRF of the oscillator:
Fourier  −1
Transform
H(ω) = ω02 − ω 2 + 2 ı ζ0 ω0 ω (48)
Fast Fourier
Transform
3 Compute the dynamic response in the frequency domain (see Eq. (47)):
F (ωs )
U(ωs ) = H(ωs ) (50)
m

4 Compute the dynamic response in the time domain at discrete time instants
tr through the Inverse DFT (IDFT):
1 Xn
u(tr ) ≈ IDFThUs i = Us e2 π ı(r −1)(s−1)/n (51)
n s=1