unit e.4 safety and security of buildings a matlab toolbox ... · a matlab toolbox for experimental...
TRANSCRIPT
1
The European Commission’s
science and knowledge service
Joint Research Centre
29 Maggio 2018
Milano
A MATLAB Toolbox
for Experimental
Modal Analysis of
Structures
Directorate E Space, Security and Migration
Unit E.4Safety and Security of Buildings
Daniel Tirelli
2
gaAccelerogram
dt...( ) ( ) ( ) gMa t Cv t R t MIa
Numerical Model
Servo-Hydraulic
Actuators
Displacement
Transducers
Reference
Frame
Measured
Restoring Force)(tR
Force
Transducers
Imposed
Displacement( )d t
The Pseudo Dynamic Method
ELSA Laboratory: The European Laboratory for Structural Assessment
Activity mainly for European Standards
in Civil Engineering
3
Dynamic Inputs
Wind RainSeism Traffic
Example of bi-directional test at real scale
Resonances
Storm on Volgograd Bridge (Russia 2010 ). Source: YouTube
For flexible structures, ex: Bridges, Cables etc…
Sometime resonances are visible to the naked eye.
4
s070108
10 12 14 16 18
-40
-30
-20
-10
0
10
20
30
40
s07: Cable 1 Fexi=100 N
X= 812
Y= 0Z= -150
time (sec)
Dis
plac
emen
t m
m
Measurement
on ELSA cable
Damping effect
ELSA Cable facility
Dangerous for structures
s094307
s094507
5 10 15 20 25
-15
-10
-5
0
5
10
s09:Snap back test:release of 700 N at 10 m
time (sec)
Dis
pla
ce
me
nt
[mm
]
Free Cable Dtime=18s
SSI Cable Dtime=4 s
Decay ratio at L/2 =0.22
Damping𝜻
Mode Shapes φi Amplitude at
different positions
What to understand from resonances?
From measurements at different positions of the structure the
toolbox calculates the three parameters which govern resonances
They depend from the structure only
(not from the input).
Frequency
5
Frequency~ (Ki/Mi)½
Damping
Mode Shapes φi
Amplitude at different
positions j
J positions
𝑀 𝑥′′ +
Inertial Forces
The three
parameters
calculated by the
toolbox from
experiments
Damping Forces
2. ζ. ω 𝑥′ +
Reaction Forces
[K]{x}=
- The mechanical state of the structure (damaged or not)
- Reference values to implement numerical models Could give:
Equation of motion of the structure
Dynamic Inputs
Fext
Source: Messaggero Veneto. Friuli Earthquake 1976
6
Practically
c030108
0 0.2 0.4 0.6 0.8 1-10
-8
-6
-4
-2
0
2
4
6
8
c03: Container with doors open:Cal:Impact no:1
X= 0
Y= 0Z= 130
time (sec)
Acc
ele
rati
on
m/s
/s
Real time history signals from a dynamic event
01
23
45
67
89
10
0
2
4
6
8
10
12
-30
-20
-10
0
10
20
30
1SE X042SE X05
3SE X061NE X10
2NE X113NE X12
1SE Y04
time [s]
2SE Y053SE Y06
1SW Y012SW Y02
3SW Y03
a09: Conf12: Snap-Back (run3): 150KN in X7
transducer
Acc
eler
atio
n m
/s/s
Fourier Transform
Aim of the Experimental Modal Analysis Toolbox:
extract parameters from “complex” signals
The “Genome” of the structural motion
7
Instrumentation(Input)
Hammer
Impact
Piezo-Patch
(Embedded sensor)
Accelerometer
Wireless Acc.Displacement
transducers
(Output)
Load Y
c080113
Load Y
c080213
Load Y
c080313
0 0.05 0.1 0.15 0.2-200
0
200
400
600
800
1000
1200
1400
c08: Container + doors open 60d.:R.Panel.Imp.Po:1
X= 0
Y= 0Z= 0
time (sec)
Forc
e N
Ac.LP.Cen
c080101
Ac.LP.Cen
c080201
Ac.LP.Cen
c080301
0.05 0.1 0.15 0.2-6
-4
-2
0
2
4
6
8
c08: Container + doors open 60d.:R.Panel.Imp.Po:1
X= 300
Y= 234Z= 132
time (sec)
Acc
ele
rati
on
m/s
/s
Output signals Y(t) in position jInput signals X(t) in position i
Truck Impact
8
The toolbox is divided in two parts: in each of them
MATLAB eases and shortens the signal processing
1Process the data during the
acquisition
methods
Slide 9controls
Slide …interpolations
Slide 13
2After the acquisition extracts the
experimental modal parameters
displaying results in different forms
Extraction
Slide 14Mode shapes
Slide …Videos
Slide 20
9
Equality Limitations
( )
( )
i k
j k
A
F
Use of reciprocity property of the FRF (Transfer Function)
൯= 𝐻𝑗𝑖(𝜔𝑘𝐻𝑖𝑗(𝜔𝑘)
൯𝐴𝑗(𝜔𝑘
)𝐹𝑖(𝜔𝑘
The toolbox runs
with two types of
tests
Acquisition Methods
j Accelerometer positions
‘Classic’ Test or monitoring
Ref.
j limited by the number
of Output transducers
j is not limited ! More experimental data more accurate
results. It is possible to test additional positions etc…
Possible only with a fast check of the
coherence function (values near 1)MATLAB 1
Accelerometer Position
j Hammer Impact positions
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Ham
Fast Impact Hammer Test (FIHT)
Ref.Accer Position
10
Control of the measurements
MATLAB: 1A first “Time Reduction” converting number
of curves in few histograms
Correct measurement
Represented by histogram of the coherence
with a
positive exponential fitting
FRF curves
After Noise filtering
Erroneous measurements ?(on displacement transducer)
Tenths or hundreds of coherence
curves to check
Coherence curves
11
Spatial control of the measurements MATLAB: 2
A Typical Example Hammer tests on the first
composite bridge of a
motorway in Europe
(Asturia -Spain)
0
200
400
600
800
1000
1200
-100-50050100
0
50
100
150
200
250
0.94
Ac.Po.Z5 CarDi.8 Po.Z289
Di.11 Po.Z42
0.91
0.93
0.95
0.940.67
0.94
0.94
0.94
0.95
X
0.93
0.88
0.95
0.95
0.96
Map of the Filtered Mean Coherence for each position of impact:
FRF Output=Ac.Po.Z4
0.91
0.94
0.94
0.96
0.91
Ac.Po.Z5
0.96
Di.2 Po.Z0
0.93
0.95
Ac.Po.Z4 Ac.Po.Y4
Y
0.91
Ham.po.-Z1
Z
Output Transducer
Map of the coherence
Position to be
controlled !
Position and mean value of the Impact Load
Map of the Impacts
12
Use of Macro for a fast 3D design of the structures MATLAB: 3
beam([-28 1385],[-126 126],[112 150],nf, [.6 .6 .6] );
Deck designed in grey color :
irsbeam([-28 1385],[60 -60 -90 90 60],[0 0 112 112 0],12,‘g‘)
Carbon Beam with irregular section designed in green
( [X], [Y], [Z], fig.Nb., color)
Design of a Building in ELSA (Duarem project)Examples:
Bridge element in ELSA (Prometeo project)
13
Low frequencies distribution example : 0< f <10 Hz
Color grid in correspondence with the vibration amplitude levels
Ex: Prometeo Bridge Project
Interpolations of irregular mesh of measurement positions MATLAB: 4
1- Create a regular cartesian mesh:
[xi,yi] = meshgrid(mi1:step1:ma1,mi2:step2:ma2)
2-Interpolate the grid by the irregular distribution of
measurements.
High frequencies distribution example: 10 Hz < f
Produce an iso-surface with: ci = griddata(Xcap1,Ycap2,Vcou,xi,yi);
14
Hammer Impacts (Input)
Allows to measure the
global mechanical properties
of the container or one of its
components.
Signals in time domain
3
x 25
x 5
x 3
= 1125
3 impacts at each position
25 positions for each panel
5 panels (each container)
3 transducers (Force+ 2 Acc)
Total: minimum time signals
Example of structure to illustrate the process of modal parameters extraction:
a smart composite container prototype
Vibrations
Induced by rain, wind,
transportation, damaging
etc..
Ac.EP.Cen
c080106
0.1 0.2 0.3 0.4
-3
-2
-1
0
1
2
c08: Container + doors open 60d.:R.Panel.Imp.Po:1
X= 590
Y= 120Z= 132.5
time (sec)
Acce
lerati
on m
/s/s
15
Signal processing to obtain modal parameters
250 Coherences functions
(“control” of the measurements)
Ac.LP.Cen
c051871
Piez.L.Cen.P
c051872
20 40 60 800
0.2
0.4
0.6
0.8
1
c05: Floor: Imp.Po:18
X= 300
Y= 234Z= 132
Frequency [Hz]
Coh
er.A
cce
[]
(125 positions*2 Outputs) =250 FRFs files
From Time domain (1125 signals) toFrequency domain (Fourier Transform)
Control of the data Physical interpretation of the data
Algorithm of automatic
identification of 3 parameters
of each peak in the
experimental and noisy 250
curves, and display of the
results in a few seconds
Matlab, step:5
250 vectors
of different
dimensions3 ×
Frequency
Amplitude
Damping
16
250
frequency
vectors0 50 100 150 200 250 300
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Point Nb.
|FR
F o
r F
FT|
cstruct:Position no:8 94.1 Hz
1.07 %
50.74 Hz
2.58 %
78.49 Hz
2.55 %
84.8 Hz
1.67 %
36.96 Hz
2.74 % 61.61 Hz
2.71 % 70.6 Hz
2.69 %
31.16 Hz
2.49 %28.24 Hz
2.92 %19.77 Hz
2.09 %
Sig.Processing: Danbox3
0 50 100 150 200 250 3000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Point Nb.
|FR
F o
r F
FT|
cstruct:Position no:17
90.83 Hz
1.09 %
51.05 Hz
2.31 %
78.99 Hz
2.1 %59.51 Hz
5.41 %96.7 Hz
0.617 %
83.51 Hz
1.37 %
39.69 Hz
2.38 %70.37 Hz
2.42 %29.4 Hz
2.39 %33.25 Hz
4.45 %25.81 Hz
4.26 %10.9 Hz
5.5 %
Sig.Processing: Danbox3
No:1
No:250
MATLAB, step:6 ( fastening with matrix)
Optimisation to obtain the final modal parameters
Which is the best clustering?
Solution in the set of frequency matrix
= Max(P/S)
Calculate for each Δfi , a Statistic S, and a Physical parameters P:
S=Mean Std. Deviation for all columns (modes)
P= Stability of the cluster (modes) when Δfi increases.
: Mf
1 column for each
frequencies cluster found
Δfi
Δfi
Clustering in
mi modes
For each window dispersion Δfi : 1 Frequency Matrix Mfi
250linesMf1
m1 col.
… Mfn
mn col.
17
0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.050
0.05
0.1
0.15
0.2Conf11:1st Snap-Back:150KN in X7
Sum
of
Norm
alised S
TD
s
Max. Freq.dispersion [Hz]
0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.050
20
40
60
Max. Freq.dispersion [Hz]
Nb.o
f m
odes
*Stat.Solutions
P= cluster stability
S= std.[ Mfi ]
Δfi
Solution of the modal parameters extraction
Sig.Processing: Danbox3 [email protected]
CSTRUCT25 Highest modes/FRF.Minimum of. 8 values/modeFreq.Dispersion limits=0.3 to 1.3(Hz)Results from Output in: Acc4,
Solution Number:1Df<1.0470 Hz
Mode Freq. Damp. Fre.Disp.Nb. (Hz) (%) (%)--------------------------------------
01 8.3104 8.052 7.27402 22.764 1.089 1.42203 26.791 3.728 3.13904 29.280 1.766 3.14405 31.301 2.537 2.17906 35.294 2.548 2.25007 39.343 2.580 1.65508 42.393 1.480 1.31509 44.491 2.301 1.80310 63.185 1.554 1.62911 66.269 1.445 1.04712 90.515 1.329 0.713
Solution Number:2Df<0.7830 Hz
Mode Freq. Damp. Fre.Disp.Nb. (Hz) (%) (%)--------------------------------------
01 8.3104 8.052 7.27402 22.764 1.089 1.42203 26.738 3.537 2.14704 29.297 1.717 1.32505 31.301 2.537 2.17906 42.393 1.480 1.31507 44.390 2.096 1.66408 66.257 1.433 0.801
Solution Number:3Df<0.4830 Hz
Mode Freq. Damp. Fre.Disp.Nb. (Hz) (%) (%)--------------------------------------
01 22.764 1.089 1.42202 26.666 3.648 0.80103 29.297 1.717 1.32504 29.357 1.622 1.18505 31.268 2.470 1.12506 42.400 1.473 1.17507 66.397 1.447 0.63308 90.575 1.367 0.418
Mode shapes
Other Example: Duarem Frame in ELSA
2nd In-plane bending X2=7.42 Hz
18
0 50 100 150 200 250 3000
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Point Nb.
|FR
F or
FFT
|
cstruct:Position no:19
80.25 Hz
1.46 %
40.49 Hz
1.87 %77.67 Hz
1.06 %22.97 Hz
1.44 %
30.17 Hz
0.846 %
74.95 Hz
1.2 % 85.18 Hz
2.06 %32.33 Hz
3.2 %42.85 Hz
1.51 %65.65 Hz
1.74 % 93.02 Hz
2.15 %
52.81 Hz
1.3 %60.21 Hz
1.56 %57.09 Hz
0.925 %5.866 Hz
4.52 %
Sig.Processing: Danbox3
Example of the main mode
shapes of the roof panel in
correspondence of the peaks
~23 Hz
Spline interpolation for a better shape interpretation
~64.7 Hz
~30 Hz
MATLAB, step:7
[xi,zi] =ndgrid([…],…
yi = interpn(X',Z',Y',xi',zi','spline');
Example of a mode shapes of
one of the vertical panel
19
Results recorded in video format without interpolation
20
Results recorded in video format with interpolation MATLAB, step:8
21
Use and Interpretation of the Results: examples
Right panel fr ~ 72,3 HzLeft Panel fl~77,2 Hz
Stiffness comparaison:
Left panel stiffer than the right
panel fl > fr
Mode ratio
Freq.ratioExp.
Freq.ratioTheo.
Y2/Z2 1.0661.077
Y3/Z3 1.08𝒇𝟑,𝟎~133 Hz
𝒇𝟎,𝟑~144 Hz
Theory of thin plates 𝒇𝒎,𝒏 =𝑐
2𝜋
𝑚𝜋
𝑎
2
+𝑛𝜋
𝑏
2Comparison with theory of thin plates:
in agreement with the theory
a=236
b=219
22
Conclusions
The methodology allows the experimental modal analysis based
only on measurements processing, with structured automated steps.
Automatisation of:
• Near Real-time Signal Processing during the tests for measurements control.
• Possibility to adopt Fast Hammer Testing method (faster method)
• Spatial control of the measurement to immediately detect the errors locations.
• Powerful Interpolation to control and for a great help in the interpretation of the
results.
• 3D Visual animated representation of the results.
23
Thank [email protected]