Documente Academic
Documente Profesional
Documente Cultură
CHRISTOPHER BALINO
CHRISTOPHER BRENNAN
SAM GREEN
ERIC HOLTZMAN
OLIVER QUADROS
Mechatronic Systems
Northeastern University
TABLE OF CONTENTS
INTRODUCTION AND LITERATURE REVIEW ............................................................ 3
THEORY .................................................................................................................. 4
RESULTS ................................................................................................................. 7
ANALYSIS ............................................................................................................. 15
CONCLUSION ....................................................................................................... 16
APPENDIX ............................................................................................................ 17
COMPARISON CHARTS FOR CONCLUSION ........................................................ 17
SYSTEM PARAMTERS FLOWCHARTS ................................................................. 19
EQUATIONS AND SAMPLE CALCULATIONS ....................................................... 20
SIMULINK AND MATLAB CODE ......................................................................... 21
SIMSCAPE MODEL ............................................................................................ 27
th
Cinquemani, S., Resta, F., and Monguzzi, M., Limits on the use of inertial actuators in active vibration control, 9
International Conference on Computing, Communications and Control Techniques, Orlando, Florida, USA.
2
Paulitsch, C., Gardonio, P., Elliott, S. J., Sas, P., and Boonen, R., Design of a Lightweight, Electrodynamic, Inertial
Actuator with Integrated Velocity Sensor for Active Vibration Control of a Thin Lightly-Damped Panel, ISMA2004,
239-254.
3
Winberg, M., Johansson, S., Claesson, I., Inertial Mass Actuators, Understanding and Tuning, Proceedings of the
International Congress on Sound and Vibration (ICSV) 11, St. Petersburg (2004).
4
Avitabile, P., Basics of Spectrum Analysis/Measurements and the FFT Analyzer, University of Massachusetts
Lowell (2001).
5
Irvine, T., An Introduction to Frequency Response Functions, www.vibrationdata.com, (2000)
PARAMETER
SYMBOL
UNITS
meACT
(Kg)
Inertial Mass
minertial
(Kg)
Sensor Mass
macc
(Kg)
Total Mass
(Kg)
Natural Frequency
f n o r
(Hz)
Spring Constant
ka
(N/m)
Damping Ratio
---
Damping Constant
ca
(Ns/m)
cc
(Ns/m)
THEORY
The active resonator absorber (ARA) becomes a passive absorber without feedback control due
to physical properties of the piezoelectric actuator. These properties can be determined
experimentally by recording the actuator vibrations upon excitation. As shown in Figure 1, a
stinger is responsible for excitation, and a laser vibrometer and accelerometer are responsible
for vibration characterization.
By employing the SDOF mathematical model depicted in Figure 2, the piezoelectric actuator can
be described in terms of stiffness ka and damping caby a linear, second order differential
equation given by
=
1
( )
where F represents the input force from the excitation stinger and M represents the combined
effect of the inertial mass, the accelerometer mass, and the effective mass of the piezo. From
the properties of second order systems, the natural frequency f n is readily given as
=
1
.
2
Note that the natural frequency of the actuator can be adjusted by the size of the inertial mass.
With a known resonance frequency of the structure, the stiffness ka becomes
= 2 2 .
F
Sensor
Inertial Mass
xa
ka
PZT Actuator
Rigid Structure
ca
Rigid Structure
Figure 2: The general primary structure depicted in Figure 1 (left), and the corresponding
mathematical model for a single degree of freedom (right).
Again, from the properties of second order systems, the damping constant ca can be expressed
as
= 2
or with appropriate substitutions from equation # above as
= 2
where is the damping ratio. The damping ratio and fundamental frequency can both be
experimentally determined by analyzing the frequency response of the structure (this process is
described in more detail below). Once the fundamental frequency and damping ratio are
determined and can be calculated using equations # and #. With and at hand, a
second order linear system can be simulated using Simulink. The Simulink model used is
included in Figure 3.
2 =
1
2
The frequency response functions were used to determine the resonance frequency of the
structure. Both 1 and 2 were graphed and the most intense peak within the usable range of
frequency data was taken as the resonance frequency of the structure. The damping ratio was
then determined using the half-power bandwidth method around the fundamental frequency.
The damping ration is given by:
=
2 1
2
where2 and 1 were taken as the 3db frequency values on the positive and negative side of
the fundamental frequency ( ), respectively.
RESULTS
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
0
10
20
30
40
50
60
70
80
90
100
time(s)
Figure 4 displays the raw experimental data recorded by the accelerometer and laser
vibrometer with a sampling frequency of 4 kHz. The structure was subjected to 40 input linear
sine chirps lasting 1.25 seconds each. Accordingly, the recorded data was windowed into 40
blocks as seen by the dashed rectangles in Figure 1.Figure 5 displays a single block of recorded
data. During analysis, each block was analyzed individually. First the data was windowed using
the MATLAB flat top window algorithm and a fast fourier transform was performed. The
frequency data was then used to calculate the cross-power spectra for each block. The
calculated cross-power spectras were then averaged for all blocks and used to calculate the
Frequency Response Functions given by equations ##s.
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
0.6
0.8
1.2
1.4
1.6
1.8
time(s)
The coherence was calculated for the frequency response formulas and is displayed in Figure 6.
The coherence is near one within the range of 100 to 500 Hz showing that this data can be
accurately represented by a single input single output linear system with constant parameters.
For this reason all remaining analysis was restricted to this range.
Figure 7 displays the magnitude of the frequency response function calculated from the
experimental velocity data. Figure 8 displays the same information calculated from the
experimental acceleration data. The frequency response function phase data is displayed in
Figure 9 and Figure 10 for the experimental velocity and acceleration data respectively.
Experimental Data: Velocity Phase H
Phase (rads)
-1
-2
-3
-4
100
150
200
250
300
350
400
450
500
frequency(Hz)
Phase (rads)
-1
-2
-3
-4
100
150
200
250
300
350
400
450
500
frequency(Hz)
Figure 7 and 8 above were used to determine the resonance frequency and damping ratio of
the structure which were in turn used to calculate values of and as described in the
Theory section above. The resonance frequency was found to be = 178 and the damping
ratio was found to be = 0.111. The spring constant was calculated to be = 2.89 105
and the damping constant was found to be = 57.2
The simulation data was then analyzed using the same procedures in MATLAB as the
experimental data. Figure 10 displays the single block of simulated velocity and acceleration
data recorded in time.
1.5
0.5
-0.5
-1
-1.5
-2
0
0.5
1.5
2.5
time(s)
Figure 11 displays the simulated velocity data FRF magnitude and Figure 12 displays the
simulated acceleration data FRF magnitude. The simulated resonance frequency was found to
be = 179.6 for the acceleration data and = 177.6 for the velocity data. This
results in an average simulated resonance frequency of = 178.6 . Figure 13 displays the
simulated velocity FRF phase data. Figure 14 displays the simulated acceleration FRF phase
data.
0.018
H1
H2
0.016
0.014
0.01
0.008
0.006
0.004
0.002
0
100
150
200
250
300
350
400
450
500
frequency(Hz)
25
H1
H2
20
15
m/sec 2 /N
m/sec /N
0.012
10
0
100
150
200
250
300
350
400
frequency(Hz)
450
500
1.5
0.5
Phase (rads)
-0.5
-1
-1.5
-2
-2.5
100
150
200
250
300
350
400
450
500
frequency(Hz)
3.5
2.5
Phase (rads)
1.5
0.5
-0.5
100
150
200
250
300
350
400
frequency(Hz)
450
500
ANALYSIS
Shown in the Appendix, Figures 1, 3, 4, 5 and 6 represent side-by-side charts from comparing
experimental and theoretical data.Figure A1 plots acceleration and velocity in the time domain
for one block of response data. While the velocity plots have a similar shape, the experimental
data shows more noise and a much quicker decline in amplitude after the natural frequency
was reached. The acceleration data is similar in shape as well starting with a smaller constant
amplitude region, increasing in amplitude near the natural frequency, and then decreasing to a
smaller constant amplitude region. Looking at the end of each figure, the differences between
theoretical and real-world experimental systems can be seen. After the signal chirp stops (1.25
s) in the experimental response of Figure A1, the vibration response falls to almost to zero very
quickly. The theoretical vibrational response dissipates at a much slower rate. This clear
difference is due to external factors such as excess damping in other parts of the system
orunexpected flexibility.
Analyzing coherence plots can give better insight to the relationship and similarities of data sets
or functions. In this experiment, the transfer function representing the systems is calculated
two different ways, using the power spectra and FFTs. Shown in Figure A2 the coherence
between the functions for both velocity and acceleration data, is approximately 1 for the range
of 100-500 Hz. This means that the FRF functions found for the experimental data, best
represent the system in this range. Any data taken in frequencies less than 100 Hz can be
skewed based on low frequency external factors. Also, data taken in frequencies higher than
500 Hz can be impacted by noise and any other factors. This region correlates to the
experimental set-up because the signal chirp varies from 0-500 Hz. Any frequency data at
greater than 500 Hz is not due to the excitation of the system making the coherence very poor
and that data range should not considered as noise.
Creating FFT plots will allow for identification of natural frequencies of a system. By looking at
Figures A3 and A4 in the Appendix, the locations and intensities of the natural frequencies
between experimental and theoretical data can be compared. All FFTs are limited to a
frequency range of 100-500 Hz which was determined to be the range of most coherent data.
Graphically, both the experimental and theoretical data sets have a natural frequency of 178
Hz. The intensities of each peak are much different due to the plethora of noise measured in
the experimental data. Not seen in the FFTs for experimental data are the large peaks in the
noise frequency range of 500-2000 Hz. Because of all noise, the counts of natural frequencies
is much lower than in the theoretical data set leading to a less intense peak. However, when
the noise is ignored and the graph is scaled to the natural frequency peak, the graphs look very
similar. The other peak at the beginning of the experimental velocity graph, Figure A3, at (~125
Hz), can be assumed to be noise data because it does not correlate to the acceleration data.
This gives only one natural frequency, 178 Hz, of the system in the frequency range of 100-500
Hz.
Phase plots can also be an important tool in comparing theoretical and experimental system
data. The two phase plot comparisons are labeled as Figures A4 and A5 in the Appendix. A
noticeable difference between the two system types are the orientation of the phase plots. The
experimental values given an increase both velocity and acceleration phase across the natural
frequency of 178 Hz. The theoretical simulation shows a decrease in phase change in the similar
region. This change could be due to a sign difference in the initial force vector which would
simply flip the slope of the phase plot. However, the transient region of phase change does
have the same location at the natural frequency of 178 Hz, regardless of the measurement
(velocity and acceleration), or the system type (experimental or theoretical).
CONCLUSION
Often times, oversimplification in modelling can cause variations between experimental and
theoretical values. Assuming a linear system, the experiment can more simply be modelled to
produce theoretical data. By calculating the experiment damping coefficient , the values for
spring constant k and damping constant c can be calculated for both the acceleration and
velocity data sets. These values can be implemented into the MATLAB script and Simulink
model,Table A2and Figure A7 respectively, to create a theoretical system. Overall, the
theoretical model, assuming linear and one input, was an acceptable simulation for this
experiment. There were many differences mainly lack of noise, orientations, and intensities
between the two experiments. However, for the purpose of this experiment, the model proved
similar finding natural frequency and general shapes of some of the graphs. The coherence of
the experimental data also proved to be acceptable in the region of 100-500 Hz. Because any
further analysis of data and plots were done in this range, the experimental results can be
taken as valid.
APPENDIX
COMPARISON CHARTS FOR CONCLUSION
Experimental Data: Velocity and Acceleration vs. Time
2
4
0.8
1.5
0.6
0.4
0.2
-0.2
-0.4
0.5
-0.5
-1
-0.6
-1.5
-0.8
0.6
0.8
1.2
1.4
time(s)
1.6
1.8
-2
0
0.5
1.5
time(s)
Figure A1: Response over time comparison of experimental (left) and simulation (right) data.
2.5
0.018
H1
H2
0.016
0.014
m/sec /N
0.012
0.01
0.008
0.006
0.004
0.002
0
100
150
200
250
300
350
400
frequency(Hz)
Figure A3: Velocity FFT comparison of experimental (left) and simulation (right) data.
25
H1
H2
20
m/sec 2 /N
15
10
0
100
150
200
250
300
350
400
450
500
frequency(Hz)
Figure A4: Acceleration FFT comparison of experimental (left) and simulation (right) data.
450
500
0.5
-0.5
-1
-1
-2
-1.5
-2
-3
-4
100
1.5
Phase (rads)
Phase (rads)
150
200
250
300
350
400
450
-2.5
100
500
150
200
250
300
350
400
450
500
frequency(Hz)
frequency(Hz)
Figure A5: Velocity phase plot comparison of experimental (left) and simulation (right) data.
Experimental Data: Acceleration Phase H
3.5
2.5
Phase (rads)
Phase (rads)
-1
1.5
1
-2
0.5
-3
0
-4
100
150
200
250
300
350
400
450
frequency(Hz)
500
-0.5
100
150
200
250
300
350
400
450
frequency(Hz)
Figure A6: Acceleration phase plot comparison of experimental (left) and simulation (right) data.
Model
PCB 288D01
P o l yt e c C L V - 2 5 3 4
P C B 7 1 2 A0 2
P C B 7 1 2 A0 1
S/N
772
552
554
Sensitivity
98.2 mV/g
98.57 mV/lbf
2 mm/s/V
-
Mass
19.2 g
12.14 g (effective)
7.199 g (effective)
500
= =
=
=
= 2
1
= 2
2
(A1)
(A2)
(A3)
= = 2
2 2 = 4
(A4)
For Velocity
= 177.6355 []
21 = 200.6882, 22 = 200.7284
21 + 22
2 =
= 200.7083 []
2
11 = 161.2840, 12 = 161.2505
11 + 12
1 =
= 161.2673 []
2
2 1
=
= .111017
2
= 200 + 19.2 + 12.14 103 = .231354 []
= 2 2 = 2.882 105
= 4 = 57.32
Figure A7: Simulink model used to run simulation data with MATLAB code from table A2
Table A2: MATLAB code for analyzing experimental data, calculating parameters, and running theoretical
simulation.
%Scr ipt to An aly z e ex per ime nta l D a ta a nd Run Si mul a tion to ve rif y
%Exp eri men tal Da t a. I ncl ude s A D -H O C An aly sis an d C o mmen ted No tes
%%
clea r;
clos e a ll ;
A=im por tda ta( 'pr o j1_d ata set _ED IT. c sv' ) ;
Time =A. dat a(: ,1) ;
Chir p=A .da ta( :,2 ) ;
Acce ler ome ter =A. d ata( :,3 );
Tran sdu cer =A. dat a (:,4 );
SimT ime =Si mul ink V eloc ity Dat a.t ime ( :,1) ;
SimV elo c=S imu lin k Velo cit yDa ta. sig n als. val ues (:, 1);
SimF orc e=S imu lin k Forc eDa ta. sig nal s .val ues (:, 1);
SimA cce l=S imu lin k Acce ler ati onD ata . sign als .va lue s(: , 1);
SimF Len gth =si ze( S imFo rce (:, 1), 1); %Tak e l e ng th of F orce Da ta
SimF blk _si ze= (fl o or(S imF Len gth )); %Div ide Fo rce da t a in to blo cks
SimF blk _si ze= Sim F blk_ siz e -m od( Sim F blk_ siz e -1 ,2) ; % M ake blo cks od d
tsf= 400 0; %sa mpl i ng f req uen cy
g=li nsp ace ( -t sf/ 2 ,tsf /2, Sim Fbl k_s i ze)' ; % est abl ish freq uen cy
vect or
% De ter min e t he i ndex of th e z ero freq uen cy
zero ind ex= fin d(g == m edi an( g)) ;
% Ap ply fl at top wind ow to eac h b l ock, ca lcu lat e F F T an d p owe r
spec tra
win= fla tto pwi n(S i mFbl k_s ize ); %fl a ttop win
a=1;
fori =1: a;
% FF T o f i npu t
Sim X(: ,i) =ff t shif t(f ft( win .*S i mFor ce( Sim Fbl k_s i ze*( i 1)+1 :Si mFb lk_ siz e *i,1 ))) ;
% FF T o f o utp ut f or V ELO CIT Y
Sim YVe loc (:, i )=ff tsh ift (ff t(w i n.*S imV elo c(S imF b lk_s ize *(i 1)+1 :Si mFb lk_ siz e *i,1 ))) ;
% In put -ou tpu t c r ossp owe r s pec tra for eac h d ata bl o ck
SimG XYV elo c(: ,i) = SimX (:, i). *co nj( S imYV elo c(: ,i) );
% Ou tpu t -i npu t c r ossp owe r s pec tra for eac h d ata bl o ck
SimG YXV elo c(: ,i) = SimY Vel oc( :,i ).* c onj( Sim X(: ,i) );
% In put au top owe r spe ctr a f or eac h dat a b loc k
SimG XXV elo c(: ,i) = SimX (:, i). *co nj( S imX( :,i ));
% Ou tpu t a uto pow e r sp ect ra for ea c h da ta blo ck
SimG YYV elo c(: ,i) = SimY Vel oc( :,i ).* c onj( Sim YVe loc (:, i ));
% FF T o f o utp ut f or A CCE LER ATI ON
Sim YAc cel (:, i )=ff tsh ift (ff t(w i n.*S imA cce l(S imF b lk_s ize *(i 1)+1 :Si mFb lk_ siz e *i,1 ))) ;
% In put -ou tpu t c r ossp owe r s pec tra for eac h d ata bl o ck
SimG XYA cce l(: ,i) = SimX (:, i). *co nj( S imYA cce l(: ,i) );
% Ou tpu t -i npu t c r ossp owe r s pec tra for eac h d ata bl o ck
SimG YXA cce l(: ,i) = SimY Acc el( :,i ).* c onj( Sim X(: ,i) );
% In put au top owe r spe ctr a f or eac h dat a b loc k
SimG XXA cce l(: ,i) = Si mX (:, i). *co nj( S imX( :,i ));
% Ou tpu t a uto pow e r sp ect ra for ea c h da ta blo ck
SimG YYA cce l(: ,i) = SimY Acc el( :,i ).* c onj( Sim YAc cel (:, i ));
end
% Av era ge cro ssp o wer and au top owe r spe ctr a f or VEL O CITY
SimG XYV elo c=m ean ( SimG XYV elo c,2 );
SimG YXV elo c=m ean ( SimG YXV elo c,2 );
SimG XXV elo c=m ean ( SimG XXV elo c,2 );
SimG YYV elo c=m ean ( SimG YYV elo c,2 );
% Av era ge cro ssp o wer and au top owe r spe ctr a f or ACC E LERA TIO N
SimG XYA cce l=m ean ( SimG XYA cce l,2 );
SimG YXA cce l=m ean ( SimG YXA cce l,2 );
SimG XXA cce l=m ean ( SimG XXA cce l,2 );
SimG YYA cce l=m ean ( SimG YYA cce l,2 );
% Ca lcu lat e F RF f or V ELO CIT Y
SimH 1Ve loc =Si mGY X Velo c./ Sim GXX Vel o c;
SimH 2Ve loc =Si mGY Y Velo c./ Sim GXY Vel o c;
SIMSCAPE MODEL