Identification Methods for Structural Systems · 2016. 3. 2. · Introduction to the Frequency...
Transcript of Identification Methods for Structural Systems · 2016. 3. 2. · Introduction to the Frequency...
Identification Methods for Structural Systems
Prof. Dr. Eleni Chatzi
Lecture 3 - 2 March, 2016
Institute of Structural Engineering Identification Methods for Structural Systems 1
Fundamentals
Overview
General Harmonic Response
Frequency Response Functions
The Fourier and Laplace Transforms for SDOFs
Bode plots for SDOFs
Institute of Structural Engineering Identification Methods for Structural Systems 2
Introduction to the Frequency domain
General Harmonic Response of SDOF system
Let’s revisit the case of Forced Damped Vibration with aHarmonic Excitation F0cos(ωt)
mx + cx + kx = F0cosωt (1)
Then particular solution is of the type:
xp(t) = Ccosωt + Dsinωt
and its derivatives are:
xp(t) = ω(−Csinωt + Dcosωt)
xp(t) = ω2(Ccosωt − Dsinωt)
Institute of Structural Engineering Identification Methods for Structural Systems 3
Introduction to the Frequency domain
Let us plug these expressions in (??)[−mω2C + cωD + kC
]cosωt +
[−mω2D − cωC + kD
]sinωt = F0cosωt
If we equate the coefficients of similar terms:[−mω2C + cωD + kC
]= F0 &
[−mω2D − cωC + kD
]= 0
⇒ C =k −mω2
cωD & (k −mω2)
k −mω2
cωD + cωD = F0
which leads to the following solution for the two unknowncoefficients C , D.
D =(cω)
(k −mω2)2 + (cω)2F0 & C =
(k −mω2)
(k −mω2)2 + (cω)2F0
Institute of Structural Engineering Identification Methods for Structural Systems 4
Introduction to the Frequency domain
However, we know form trigonometry, that the expression
xp(t) = Ccosωt + Dsinωt
is equivalent to:
xp(t) = X0cos(ωt − φ) = X0cosωtcosφ+ X0sinωtsinφ
where X0 =√
C 2 + D2 & φ = arctan
[D
C
]Therefore,
X0 =F0√
(k −mω2)2 + (cω)2& φ = arctan
[cω
k −mω2
](2)
Institute of Structural Engineering Identification Methods for Structural Systems 5
Introduction to the Frequency domain
However we have previously defined that:
c = 2mωnζ, k = mω2n
Using the above relationships we can rewrite (??) as:
X0 =F0/k√(
1− ω2
ω2n
)2
+
(2ζ
ω
ωn
)2& φ = arctan
2ζω
ωn
1− ω2
ω2n
Institute of Structural Engineering Identification Methods for Structural Systems 6
General Harmonic Response of SDOF system
Summarizing, the particular solution then becomes:
xp(t) = X0cos(ωt − φ)⇒ xp(t) =F0
kH(ω)cos(ωt − φ)
whereF0
k= δst is also known as static deflection
and H(ω) is what is known as:
Frequency Response Function
H(ω) =1√(
1− ω2
ω2n
)2
+ (2ζ ωωn)2
Institute of Structural Engineering Identification Methods for Structural Systems 7
The half power Method
The half power Method for the estimation of Damping
The amplitude of the FRF at resonance is called the quality factor(Q) of the system
For ω ≈ ωn ⇒ H(ω) ≈ 1
2ζ= Q
Points R1, R2 where it holds thatH(ω) = Q/
√2 are called 1/2
power points. these are used forextracting the bandwidth of thesystem
t
( )F t
1τ
1F
2F
2τ
n
ωω
2R 1R
12
Qζ
=
2Q
( )H ω
1
n
ωω
2
n
ωω
1
m
k c ( )ty
( )tx t
( )F t
t∆
1t∆
0t∆ →
Institute of Structural Engineering Identification Methods for Structural Systems 8
The half power Method
The Bandwidth is defined as: ∆ω = ω2 − ω1 where
ω1, ω2 are calculated from:
H(ω) =Q√
2=
1
ζ√
2⇒ 1√
(1−ω2
1,2
ω2n
)2 + (2ζω1,2
ωn)2
=1
ζ√
2
ω21,2
ω2n
= (1− 2ζ2 ∓ 2ζ√
1 + ζ2)ω2nζ<<→ ω2
1,2 = (1∓ 2ζ)ω2n
and ω22 − ω2
1 = 4ζω2n ⇒ (ω2 − ω1)(ω2 + ω1) = 4ζω2
n
Institute of Structural Engineering Identification Methods for Structural Systems 9
The half power Method
Bandwidth - Identification of ζ
Since,ω1 + ω2
2= ωn, we have that
Bandwidth = ∆ω = ω2n = ω2 − ω1 ≈ 2ζωn
Therefore, from experimental response we can evaluate the dampingcoefficient:
ζ =ω2 − ω1
2ωn
Institute of Structural Engineering Identification Methods for Structural Systems 10
Base Excited Systems
Structural Systems excited at the base - ground motion(earthquake)Two alternatives exist for formulating the Equation of Motion
Absolute Motion x(t):
mx + c(x − y) + k(x − y) = 0⇒mx + cx + kx = cy + ky
Relative Motion z(t) = x(t)− y(t)
m(z + y) + cz + kz = 0⇒mz + cz + kz = −my
n
ωω
2R 1R
12
Qζ
=
2Q
( )H iω
1
n
ωω
2
n
ωω
1
m
k c ( )ty
( )tx
Institute of Structural Engineering Identification Methods for Structural Systems 11
Base Excited Systems-Absolute Motion
Base Excited Systems-Absolute Motion
mx + cx + kx = cy + ky
Total solution xtot = xh + xp.
The homogeneous solution is already explored.Assuming that the base excitation is of harmonic type, i.e.,y(t) = Y0cosωt, the particular solution will also be harmonic.
In accordance with the harmonic force excited system shown earlier,we now obtain:
xp(t) = Y0
√(k2 + (cω)2)
(k −mω2)2 + (cω)2cos(ωt − φ)⇒
xp(t) = Y0H(ω)cos(ωt − φ)
Institute of Structural Engineering Identification Methods for Structural Systems 12
Base Excited Systems-Absolute Motion
Base Excited Systems-Absolute Motion
We therefore can define the following terms.
Gain Function or Transmissibility of Displacement:
Td = H(ω) =
√1 + (2ζr)2
(1− r2)2 + (2ζr)2
and Phase:
φ = tg−1
(2ζr3
1 + (4ζ2 − 1)r2
), r =
ω
ωn
Institute of Structural Engineering Identification Methods for Structural Systems 13
Base Excited Systems-Absolute Motion
Transmissibility of Displacement
Institute of Structural Engineering Identification Methods for Structural Systems 14
Base Excited Systems-Absolute Motion
Transmissibility of Displacement
This signifies how larger the maximum displacement of the system X0 is with
respect to the maximum amplitude of the input harmonic force Y0.Base excited systems: absolute motionBase excited systems: absolute motion
h l f i i 0 d l i f ll l f• The value of Td is unity at r=0 and close to unity for small values of r.
• For an undamped system ζ=0, Td ∞ at resonance (r=1).
• The value of Td is less than unity (Td <1) for values of r >√2 (for any amount of damping ζ)
• The value of Td is equal to unity (Td=1) for all values of ζ at r=√2
222
2
)2()1()2(1
rrr
YX
o
o
ξξ+−
+=
Institute of Structural Engineering Identification Methods for Structural Systems 15
Base Excited Systems-Relative Motion
Relative Motion: z(t) = x(t)− y(t)⇒ mz + cz + kx = −myBase excited system: Relative motionBase‐excited system: Relative motion
d l f2rZ• In nondimensionless form,
• The gain function for the relative motion for the base‐excited system is
222 )2()1( rrr
YZ
o
o
ξ+−=
The gain function for the relative motion for the base excited system is shown in the figure:
Institute of Structural Engineering Identification Methods for Structural Systems 16
Arbitrary Excitation
Response of SDOF System to Arbitrary ExcitationLet’s consider the impulse response:
Newton:∫ t+∆tt Fdt = (change in momentum) =∫ t+∆t
t mxdt =∫ t+∆tt m
x
dxdt ⇒ Fdt = mxt+∆t −mxt
Assuming mxt = 0⇒ xt+∆t =Fdt
m.
n
ωω
2R 1R
12
Qζ
=
2Q
( )H iω
1
n
ωω
2
n
ωω
1
m
k c ( )ty
( )tx t
( )F t
t t t+ ∆
Assume the unit impulse at t = 0, where
Fdt = F = 1⇒ F =1
∆t. F behaves like the Dirac δ
function∫ t+∆tt δ(t)dt = 1⇒ mx0 = 1⇒ x0 = 1 =
1
mand x0 = 0 (no move yet). Hence, the solution is the free
response with I.C. x0 = 0, x0 =1
m.
n
ωω
2R 1R
12
Qζ
=
2Q
( )H iω
1
n
ωω
2
n
ωω
1
m
k c ( )ty
( )tx t
( )F t
t∆
1t∆
0t∆ →
Institute of Structural Engineering Identification Methods for Structural Systems 17
Arbitrary Excitation
Unit Impulse Response for impulse at t = 0:
x(t) =
(x0cosωd t + (
x0
ωd+ x0ζωn)sinωd t
)e−ζωnt
with x0 = 0, x0 =1
m⇒ x(t) =
1
mωdsinωd te
−ζωnt , t > 0
For a unit impulse of magnitude Fdτ , applied at t = τ :
x(t) =Fdτ
mωdsinωd(t − τ)e−ζωn(t−τ), t > τ
Institute of Structural Engineering Identification Methods for Structural Systems 18
Arbitrary Excitation
For 2 unit impulses of magnitude F1dτ , F2dτapplied at t = τ1, t = τ2:
x(t) =F1dτ
mωdsinωd(t − τ1)e−ζωn(t−τ1)
+F2dτ
mωdsinωd(t − τ2)e−ζωn(t−τ2)
t
( )F t
1τ
1F
2F
2τ
n
ωω
2R 1R
12
Qζ
=
2Q
( )H iω
1
n
ωω
2
n
ωω
1
m
k c ( )ty
( )tx t
( )F t
t∆
1t∆
0t∆ →
Institute of Structural Engineering Identification Methods for Structural Systems 19
Arbitrary Excitation
Duhamels’s (or convolution) Integral Similarly, for n finiteimpulses:
x(t) =n∑
i=1
Fidτ
mωdsinωd(t − τi )e−ζωn(t−τi )
Adding up, for dτ → 0, we obtain the continuous expression:
x(t) =
∫ t
0
F(τ)
mωdsinωd(t − τ)e−ζωn(t−τ)dτ ⇒
x(t) =
∫ t
0F(τ)h(t − τ)dτ
Institute of Structural Engineering Identification Methods for Structural Systems 20
From Time to Frequency Domain
SDOF systems: From Time to Frequency Domain
The Fourier Transform
Definition: F (ω) = F{f (t)} =∫∞−∞ f (t)e−iωtdt
Inverse: f (t) = F−1{F (ω)} =1
2π
∫∞−∞ F (ω)e iωtdω
Fourier Transform of a cosine
Institute of Structural Engineering Identification Methods for Structural Systems 21
From Time to Frequency Domain
SDOF systems: From Time to Frequency Domain
Fourier Transform of a cosine - Note
The Fourier Transform (FT) of a cosine simply consists in twosymmetric spikes at values corresponding to the specific frequency ofthat cosine (symmetric transformation).
Mathematically this can be written in the form of two deltafunctions δ(ω ± ω0) or δ(p ± p0) in Hz.
This reflects the fact that the frequency content of a perfect cosine(or sine) function contains a single frequency component.
see Laplace and Fourier transform pdfs uploaded on the website
Institute of Structural Engineering Identification Methods for Structural Systems 22
From Time to Frequency Domain
The Laplace Transform
Definition: F (s) = L{f (t)} =∫∞
0 f (t)e−stdt, s = σ + iω
Inverse: f (t) = L−1{F (s)} =1
2πilimT→∞
∫ γ+iTγ−iT F (s)estds
Basic Laplace Transform Property
L[df
dt
]=
∫ ∞0
e−stdf (t)
dtdt
Applying integration by parts:∫udv = uv −
∫vdu, for u = e−st ,
v = f , we obtain:
L[df
dt
]= e−st f (t)
∣∣∞0−∫ ∞
0(−s)e−st f (t)dt
= −f (0) + s
∫ ∞0
f (t)e−stdt ⇒ L[df
dt
]= sF (s)− f (0)
Institute of Structural Engineering Identification Methods for Structural Systems 23
The Laplace Transform
Similarly,
L[d2f
dt2
]= s2F (s)− df
dt(0)− sf (0)
L[dnf
dtn
]= snF (s)− dn−1f
dtn−1(0)− ...− sn − 2
df
dt(0)− sn−1f (0)
Example - Obtaining the Transfer Function (TF):Apply the Laplace Transform on a 2nd order ODE:
d2y
dt2+ 2
dy
dt+ 3y = 4u
L→ s2Y (s)− dy
dt(0)− sY (0) + 2sY (s)− 2Y (0) + 3Y (s) = 4U(s)
0 I .C .−→ Y (s) =4
s2 + 2s + 3U(s) : Transfer Function
Institute of Structural Engineering Identification Methods for Structural Systems 24
The Laplace Transform
Transfer FunctionA transfer function (also known as the system function) is a mathematicalrepresentation, in terms of the system frequency, of the relation betweenthe input and output of a linear time-invariant system with zero initialconditions and zero-point equilibrium.A linear time -invariant (LTI) system is characterized by two properties:
Linearity which means that the relationship between the input and theoutput of the system is a linear map. If input x1(t) produces response y1(t)and input x2(t) produces response y1(t) then the scaled and summed inputα1x1(t) + α2x2(t) produces the scaled and summed responseα1y1(t) + α2x2(t) where α1, α2 are real scalars.
Time invariance which means that whether we apply an input to the
system now or T seconds from now, the output will be identical except for
a time delay of the T seconds. That is, if the output due to input x(t) is
y(t), then the output due to input x(t − T ) is y(t − T ). Hence, the
system is time invariant because the output does not depend on the
particular time the input is applied.
Institute of Structural Engineering Identification Methods for Structural Systems 25
The Laplace Transform
More Laplace Transform Properties
L{δ(t)} = 1, δ(t):Dirac
L{f (αt)} =1
αF( sα
)L{eαt f (t)} =
1
αF (s − α) (Frequency Shift)
L{f (t − α)H(t − α)} = e−αsF (s) (Time Shift) where H is the
Heaviside step function H(n) =
{0, n < 01, n ≥ 0
L{f ∗ g (t)} =∫ t
0 f (τ)g(t − τ)dτ = F (s)G (s) (Convolution)
Institute of Structural Engineering Identification Methods for Structural Systems 26
The Laplace Transform
Example: Equation of Motion
mx + cx + kx = f (t)L,0I .C .−→ ms2X (s) + csX (s) + kX (s) = F (s)⇒
X (s) =1
ms2 + cs + kF (s)⇒ H(s) =
1
ms2 + cs + kTransfer Function
Assuming s = iω we obtain the complex Frequency Response
Function (FRF):
H(iω) =1
mω2 + ciω + k
We can then use what we define as the Inverse Laplace Transformin order to determine x(t)
Institute of Structural Engineering Identification Methods for Structural Systems 27
The Laplace Transform
Basic Inverse Transform Properties (also look at given tables)
L−1
{1
s + α
}= e−αt
L−1
{1
(s + α)2
}= te−αt
L−1
{1
(s + α)(s + β)
}=
1
β − α[e−αt − e−βt
]Hence, x(t)L−1{X (s)} = L−1{ 1
ms2 + cs + kF (s)} =
L−1{ 1
m(s + α)(s + β)F (s)}
where,α, β =c ∓ i
√4mk − c2
2massuming an underdamped system
Institute of Structural Engineering Identification Methods for Structural Systems 28
The Laplace Transform
Also for f (t) = δ(t) (unit impulse) ⇒ F (s) = 1
Then, x(t) =1
m(β − α)
[e−αt − e−βt
]
Using k = mω2n, c = 2mωnζ, ωd = ωn
√1− ζ2
⇒ x(t) =1
mωde−ζωnt [sinωd t]
Obviously this agrees with the time domain derived SDOF impulseresponse
Institute of Structural Engineering Identification Methods for Structural Systems 29
Linear Systems
The Frequency Response Function (FRF)Assume a linear system characterized by its Transfer Function:
Y(s) = G(s)U(s), where s ∈ C (Laplace Domain)
Evaluated at the imaginary axis s = iω, the TF yields the FRF, G(iω):
Case of a SISO System
d2y
dt2+ 4
dy
dt+ 3y =
du
dt+ 2u
L, 0 ICs−→
(s2 + 4s + 3)Y (s) = (st2)U(s)⇒ G (s) =s + 2
(s + 3)(s + 1)
Hence the FRF is, G (iω) =iω + 2
(iω + 3)(iω + 1)
Institute of Structural Engineering Identification Methods for Structural Systems 30
The FRF
As shown FRF describes the response of the system for sinusoidalinputs. It is also the FT of the system’s impulse response.
The FRF can be visualized using Bode Plots
Assume G (iω) =1
iω + 1⇒ G (iω) =
1
1 + ω2− iω
1 + ω2
Bode Magnitude Plot
For ω = 1⇒ |G (iω)| =
∣∣∣∣1− i
2
∣∣∣∣ =
1√2⇒ −20log |G | = −3.02dB
( )ϕ ω
45o−
90o−
1 100.1
5.7o
5.7o
( ) logscaleω
45 / decadeo−
3dB
( ) logscaleω
( ) ( )20log G i dBω
20 / decadedB−
30−
10−
0
0.1 1 10
20−
Institute of Structural Engineering Identification Methods for Structural Systems 31
The FRF
Bode Magnitude Plot - Asymptotes
For ω << 1⇒ 1
iω + 1≈ 1 and −20log |G | = −20log(1) = 0
For ω >> 1⇒ 1
iω + 1≈ 1
iωand
∣∣∣∣ 1
iω
∣∣∣∣ =1
ω
Then assuming ω2 = 10ω1 ⇒
20log
(1
ω2
)= 20log
(1
10ω1
)= 20log
(1
10
)+ 20log
(1
ω1
)Hence we can approximate the decline per “decade”
20log
∣∣∣∣ 1
iω2 + 1
∣∣∣∣− 20log
∣∣∣∣ 1
iω1 + 1
∣∣∣∣ ≈ 20log
(1
ω2
)− 20log
(1
ω1
)⇒
−20(dB/decade)
Institute of Structural Engineering Identification Methods for Structural Systems 32
The FRF
G (iω) =1
1 + ω2− iω
1 + ω2⇒ φ(ω) = tan−1(−ω)
Bode Phase Plot
For ω = 1⇒ φ(ω) = tan−1(−1) = −45o
For ω =1
10⇒ φ(ω) = tan−1(
1
10) = −5o
For ω = 10⇒ φ(ω) = tan−1(−10) = −90o
( )ϕ ω
45o−
90o−
1 100.1
5.7o
5.7o
( ) logscaleω
45 / decadeo−
3dB
( ) logscaleω
( ) ( )20log G i dBω
20 / decadedB−
30−
10−
0
0.1 1 10
Institute of Structural Engineering Identification Methods for Structural Systems 33
The FRF
Generalizing:
Suppose we had G (iω) =α
α + iω=
1
i(ωα
)+ 1
Then, the same plots apply for ω =ω
α
3dB
( ) logscaleω
( ) ( )20log G i dBω
20 / decadedB−
30−
10−
0
0.1α α 10α
20−
( )ϕ ω
90o−
α 10α 0.1α
5.7o
5.7o
( ) logscaleω
45 / decadeo−
Institute of Structural Engineering Identification Methods for Structural Systems 34
The FRF
Returning to the SISO Example
We had G (iω) =iω + 2
(iω + 3)(iω + 1)⇒
20log |G (iω)| = 20log |iω + 2|+ 20log
∣∣∣∣ 1
(iω + 3)
∣∣∣∣+ 20log
∣∣∣∣ 1
(iω + 1)
∣∣∣∣= 20log
∣∣∣2i ω2
+ 1∣∣∣+ 20log
∣∣∣∣∣∣ 1/3
(iω
3+ 1)
∣∣∣∣∣∣+ 20log
∣∣∣∣ 1
(iω + 1)
∣∣∣∣⇒Superposition of Plots
20log |G (iω)| =
20log
(2
3
)+ 20log
∣∣∣i ω2
+ 1∣∣∣+ 20log
∣∣∣∣∣∣ 1
(iω
3+ 1)
∣∣∣∣∣∣+ 20log
∣∣∣∣ 1
(iω + 1)
∣∣∣∣Institute of Structural Engineering Identification Methods for Structural Systems 35
The FRF
Plot Superposition
3dB
( ) logscaleω
20 / decadedB 10
20
0.2 2 20
0
( )20log 1 2 dBω +
+ + 220log 3
3dB
( ) logscaleω
( )120log 13
dBω +
20 / decadedB−
30−
10−
0
0.3 3
30
20−
+
3dB
( ) logscaleω
( )120log 1
dBiω +
20 / decadedB−
30−
10−
0
0.1 1 10
20−
+
Institute of Structural Engineering Identification Methods for Structural Systems 36