Eurotev 22-06-051 Feedback Loop for the mechanical Stabilisation Jacques Lottin*...
-
Upload
nathan-johnson -
Category
Documents
-
view
218 -
download
0
Transcript of Eurotev 22-06-051 Feedback Loop for the mechanical Stabilisation Jacques Lottin*...
Eurotev 22-06-05 1
Feedback Loop for the mechanical Stabilisation
Jacques Lottin*([email protected])
Laurent Brunetti*([email protected])
Mihai Corduneanu**Vlad Cozma **
*LISTIC-ESIA, Université de Savoie, Annecy, France**Universitatea Politehnica, Bucuresti, Romania
Eurotev 22-06-05 2
Overview :
- Presentation of active vibration reduction- Principle of rejection- Control scheme- Programs- Experiments- Results- Conclusion
Collaboration with and
Eurotev 22-06-05 3
Mechanical part
control
Measurement :(continuous + resonances)
Disturbance : -ground: cultural noise…-equipment: motors, flows…
Excitation : (strength applied with actuators…)
Presentation of Active Vibration Reduction
Eurotev 22-06-05 4
Principle of rejection
Assumptions:
- There are a few resonances which are independentsmall amplitudes, linearity, …
- amplitude and phase of each resonance are constant or slowly varying with respect to the signal period
- frequency of each resonance is knowncomputed by means of Fourier transform
- there is no accurate model available
Eurotev 22-06-05 5
Principle of rejection
Current objectives: Independently reduce the main resonances
For example, an identification of one mock-up…
Eurotev 22-06-05 6
Principle of rejection
- rule 2: make the components of excitation converge to values such that the global effect of disturbance and excitation is null at sensor location
lumped / distributed
- rule 1: decompose each resonance as a weighted combination of sine and cosine measurement, disturbance, excitation
with:
( ) sin( ) sin( ) cos( )s cf t t f t f t cos( )
sin( )s
c
f
f
Eurotev 22-06-05 7
Principle of rejection
An exact direct compensation would require the knowledge of eight elementary transfer functions:
at least at one frequency
Mechanical part
disturbance
excitation
measurementpc
ps
fc
fs yc
ys
Disturbance is not well definedMeasurement/Excitation transfers are badly known
Eurotev 22-06-05 8
Model (State space representation for one frequency):
Control scheme
Disturbance
With the different components are decomposed in sine and cosine:
Input(Strength)
State
(1)
( ) ( ) ( ) ( )
( ) ( )n n n
n
x t A x t B u t G p t
y t C x t
( ) ( ), ( ) , ( )
( )
)
( ) ))
(
(( s
c
s s
c c
f t p tu t p t
f
yx t
t p t
t
y t
Eurotev 22-06-05 9
An : represents the dynamic of the system.
Control scheme
Dynamic response
open loop : first order-> Setting time = 3 * time-constant
: time-constant
10
,1
0nA
Eurotev 22-06-05 10
Control scheme
Bn ,Gn : Transfer’s matrix of the strength and the disturbance.
(1) Steady state then :
1st case: disturbance = 0
y = kyf ( fs sin (t + yf ) + fc cos (t + yf ) )
(2)
(2) & (4):
(3)
also:
(4)
(5)
(6)
( ) ( )
( )
cos sin
sin cos
cos sin
s
( ) ( )
( ) ( )
(
n os
)
i cyp yp
ypyp
yf yf
yfy
y
f yn n
n n
f
p
B A
G A
k
k
1 1s s s sn n n n
c c c c
x y f pA B A G
x y f p
cos sin
s
( ) ( )
( ) o ( )in c syf yf
yfyf y
s s
c cf
y f
y fk
Eurotev 22-06-05 11
Estimation of the disturbance using a state observer :
Control scheme
Requirement: the new model should include the disturbance in the components of the state:
The new state vector:
The model:(7)
(8)
e e e e e
e e
x A x B u
y C x
s
ce
s
c
y
yx
p
p
Eurotev 22-06-05 12
State observer:
, , 00 0 0
n n ne e e n
A G BA B C C
Control scheme
Where:
Final relation:
(9)
(10)
(Luenberger)
ˆ ˆ ˆ( )
ˆ ˆe e e e
e e
x A x B u L y y
y C x
ˆ ˆ
ˆ ˆ
ee e e e e
e
e e
ux A LC x B L
y
y C x
Eurotev 22-06-05 13
Control law:
Control schemeState feedback control :
(1) =>
To reject the disturbance:
(11)
(12)
(13)1p n nK B G
( ) 0n p nB K G
( ) ( ) ( )
ˆ( ) ( )
x p
x p
u t K x t K p t
K x t K p t
( ) ( )n n x n p nx A B K x B K G p
Eurotev 22-06-05 14
Control scheme
Mechanicalpart
spectralanalysis
signalprocessing
observer
statefeedback
signalrebuilding
actu
ator
sens
or
yc(wi)ys(wi)
fc(wi)
fs(wi)
y(wi)f
)(ˆ is wp)(ˆ ic wp
Eurotev 22-06-05 15
ProgramMain program:(Matlab / Simulink / XPC Target toolboxes)
Analog Input board
Analog Output board
Control of disturbances
Summation of each command
Algorithm forone frequency
Eurotev 22-06-05 16
ProgramProgram for one frequency :(Matlab / Simulink / XPC Target toolboxes)
State feedback
State observerSignal
processingBand-pass filter
for the disturbanceSensor
command rebuilding
Actuator
Algorithm for one frequency :
Eurotev 22-06-05 17
« a steel beam »
2 loudspeakers2 opposite PZT
Accelerometer(only for
monitoring)
ExperimentsMock-up:
Eurotev 22-06-05 18
ExperimentsMock-up:
Sensor PZT(bottom)
Actuator PZT (on top)
Eurotev 22-06-05 19
Experiments
Layout :
Development PC (host)
Matlab + Simulink+ XPC Target
Toolboxes
Dedicated PC (target)
XPC Target
Ethernet network
Input / Output analog boardTexas InstrumentsPlant
Sample time : 0.0007 s
Eurotev 22-06-05 20
ResultsRejection of 6 resonances : (without and with rejection)
Resonances of : -beam-support
Eurotev 22-06-05 21
Results
Some zooms…
Without rejectionWith rejection
Eurotev 22-06-05 22
ResultsRejection of 6 resonances : (disturbances and control)
Resonances of :-beam-support
Disturbances
Control
Eurotev 22-06-05 23
ResultsPhase robustness of yf (matrix Bn) : 1st resonance : margin of /2
Robustness OK :Identified phase
Stability limit :Identified phase +2/20Identified phase -8/20
Without rejection
Eurotev 22-06-05 24
ResultsGain robustness of kyf (matrix Bn) : 1st resonance : factor 10.
Identified gainIdentified gain * 0.1Identified gain * 10
Beginningof control
Worse in setting time
Time
Right convergence in
steady time(stability limit…)
Eurotev 22-06-05 25
Conclusion
Status:- Validation is OK- Robustness is ~OK
Current and future works:-Spectral analysis of the disturbance in real time-Translation of the last studies on a new model with a parallelepiped beam (2.5 meters long).(testing the algorithm and choosing the appropriate actuators….)
Jacques LOTTIN : [email protected] BRUNETTI : [email protected]