Click here to load reader
Upload
fu-cheng
View
215
Download
3
Embed Size (px)
Citation preview
Vibration Control of a Three-Leg Optical Table by Mechatronic Inerter Networks
Yu-Chuan Chen, Sheng-Yao Wu and Fu-Cheng Wang
Department of Mechanical Engineering, National Taiwan University, Taipei, Taiwan
(Tel : +886-2-3366-2680; E-mail: [email protected])
Abstract: This paper develops a three-leg optical table, and applies a newly-developed mechatronic inerter network to
suppress vibrations of the table. Optical tables can insulate precision machines from two types of disturbances: ground
disturbances from the environment and load disturbances from the equipment. Using disturbance response decoupling
(DRD) techniques, we can effectively isolate the ground disturbances by soft passive suspensions and improve the load
responses by active control. This paper further applies mechatronic inerter networks to a three-leg optical table, and
optimizes the ground responses by connecting the networks to suitable electric circuits. We then apply DRD techniques
to improve the load responses without influencing the ground responses, and implement the optical table for
experimental verification. Based on the results, the proposed mechatronic inerter networks and DRD structures are
deemed effective in improving system responses.
Keywords: mechatronic inerter, disturbance response decoupling, optical table, vibration control, optimization.
1. INTRODUCTION
Optical tables are usually applied to insulate
vibrations for precision machinery. Traditional optical
tables, such as a pneumatic leg Newport I-2000 [1],
apply passive elements, e.g. dampers and springs, to
isolate machines from disturbances. Because passive
structures cannot fully satisfy the increasing precision
requirements, active optical tables are developed for
further performance improvement. For example, Kato et
al. [2] applied model-following control to pneumatic
actuators to improve table responses. Oh et al. [7]
combined passive rubbers and piezoelectric actuators to
suppress disturbance responses. Wang et al. [14, 16]
designed robust controllers to a two-layer isolation
platform with voice coil motors.
In general, an optical table needs to suppress two
main vibration sources: ground disturbances and load
disturbances. However, the suspension settings for
restraining these two disturbances are conflicting; i.e.,
the suspensions need to be soft for isolating ground
disturbances and stiff to suppressing load disturbances.
That is, the suspension design is a compromise between
these two settings. Therefore, Wang et al. [22,23]
proposed a double-layer optical table, as shown in Fig. 1,
and applied disturbance response decoupling (DRD)
techniques [11] to independently control the two
disturbances: the ground disturbance rz is controlled
passively by passive elements 1 ,
2 and 3 , while
the system responses to the load disturbance sF is
improved by the actuator A without influencing the
ground responses. Because the passive elements cannot
be easily adjusted according to the operating conditions,
this paper further extends these ideas by applying the
mechatronic inerter network and designing the connected
electrical circuits to suppress particular ground
disturbances.
The inerter have been applied to vehicle suspensions
[5–6, 13, 15, 18, 20–21] and building structures [17] to
suppress vibrations. Until now, there have been four
inerter realisations: the rack-pinion inerter [12], the
ball-screw inerter [13], the hydraulic inerter [19], and
the mechatronic inerter network [20]. The former three
are mechanical structures whose parameters cannot be
easily adjusted according to the operating conditions.
The mechatronic inerter network consists of two parts:
the ball-screw inerter and permanent magnet electric
machinery (PMEM). Therefore, we can design the
system impedance by connecting the PMEM to suitable
electrical circuits to optimize the system’s responses to
particular ground disturbances.
This paper is arranged as follows: Section 2
introduces the mechatronic inerter network. Section 3
derives the dynamics of a three-leg optical table and its
DRD structure. Section 4 identifies the transfer
functions of system components by experiments.
Section 5 improves the ground disturbances by
designing an optimal circuit that connects to the
mechatronic inerter networks. Section 6 applies robust
control to design active suspensions that improve the
load responses without affecting the ground responses.
The designed DRD control structure and active control
are then implemented for experimental verification.
Lastly, we draw conclusions in Section 7.
Fig. 1. The generalized DRD structure [21].
SICE Annual Conference 2014September 9-12, 2014, Hokkaido University, Sapporo, Japan
978-4-907764-45-6 PR0001/14 ¥400 © 2014 SICE 426
2. MECHATRONIC INERTER NETWORKS
The mechatronic inerter network, as shown in Fig.
2, is composed of a ball-screw inerter and a PMEM with
the following system impedance [20]:
ˆ ( ),
ˆ( ) ( )
m
m m
a a e
KF sb s c
v s R sL Z s
(1)
where v is the relative velocity between the two
terminals and F is the corresponding force. Equation (1)
can be considered as two parts: the first part m mb s c
is the mechanical inerter and damper, while the second
part / ( ( ))m a a eK R sL Z s represents PMEM and the
connected electrical circuit Ze. The equivalent
impedance of the mechatronic inerter network is shown
in Fig. 3, where Km , Ra and La are the motor gain
constant, the armature resistance, and the armature
inductance, respectively. Therefore, we can adjust the
electric circuits Ze to change the impedance of the
mechatronic inerter network.
Fig. 2. The mechatronic inerter networks [20].
Fig. 3. An ideal network model [20].
3. SYSTEM DYNAMICS AND DRD
FILTERS
A three-leg optical table is shown in Fig. 4. The
dynamics of the table can be expressed as [22, 23]:
1 2 3 ,s s s p p pm z F u u u
1 2 3 ,f p f p r pI z T l u l u l u
1 2 ,f p f pI z T t u t u
1 1 1 ,u u p rm z u F
2 2 2 ,u u p rm z u F
3 3 3 ,u u p rm z u F
where
2 3 ( ),i i i i ip i a uu D z z
1 ( ),i i i ir u rF z z
for 1 3i .sm , I and I represent the mass, and
inertias of the table, respectively. sF , T and T
are
the force, and torque disturbances to the table. jA are
active actuators, while jrz are the ground disturbances
for 1 3j .ji are corresponding passive suspension
elements for 1 3i and 1 3j . The table
displacements 1 3~D D are derived as follows:
11 ,s f f uD z l z t z z
22 ,s f f uD z l z t z z
33 ,s r uD z l z z
Fig. 4. The full-optical table model with 3-legs.
We further assume1 2 3u u u um m m m ,
1 2 3k k k k for 1, 2, 3k , and define the
following symmetric transformation matrix ,1fL
and ,2fL :
,1
1 1 01
0 0 2 ,2
1 1 0
fL
,2
1
2fL
t ,
which decouple the full-table model into one half-table
and one quarter-table models, such that
,1 1 2 3( ) ( ) ( ) ,f r f
T T
b b fx x x L x x x
(2)
,2( ) ( ) ,froll fx L x (3)
in which x can be substituted by the following
SICE Annual Conference 2014September 9-12, 2014, Sapporo, Japan
427
parameter: strut displacement D , displacement rz
and uz , or control signals u , while the subscripts
,f rb b represent the front and rear bounce modes, ,f
represent the front roll modes.
We can represent the control structure of the table as
the linear fractional transformation (LFT) structure of
Fig. 5 with:
,f
z wP
y u
(4)
where
1 2 3 ,
T
s r r rw F T T z z z
1 2 3 ,
T
s u u uz z z z z z z
2 1 1 2 3 T
sy f f z z z D D D
4 3 1 2 3 T
u f f u u u ,
diag( , ),bp rollK K K
in which the 1 4~ f f are as follows:
1 ,1 diag(1, 1, 1, ),ff L
2 ,2 diag(1, -1, 1, 1, 1, ),ff L
1
3 ,2 diag(1, 1, ),ff L
1
4 ,1 .ff L
Fig. 5. The LFT structure.
Applying the DRD algorithms, we can design a suitable
filter 2U for the plant
fP P , such that the actuator
control signal u can only be activated by the load
disturbances T
sF T T , but not the ground
disturbances1 2 3
T
r r rz z z . The DRD filter
2U is
designed as follows:
2 3 2 3
2
2 3 2 3
2 3
02( )( ) 2( )( )
0 ,2( )( ) 2( )( )
0 02( ) ( )
s r
r f r f
s f
r f r f
f r f
m l I
l l l l
m l IU
l l l l
I
t t t
4. IDENTIFICATION OF SUSPENSION
ELEMENTS
In this section, we derive the transfer functions of
the suspension elements 1 , 2 , 3 , and the
actuator dynamics . For the mechatronic inerter
networks 1 , we consider its nonlinear factors, as
shown in Fig. 6(a), in which k is a parallel spring,
is the backlash , , s sk c are the elastic effects, and f
is the Stribeck friction model with the following
dynamics: /
( ) ( (1 ) ) sgn( ) ,sv v
static mf v F e v c v
, 0 1.c staticF F
where cF is the Coulomb friction and
staticF is the
static friction, v is the relative velocity between the
terminals, sv is the Stribeck velocity, and
mc is the
damping constant. We built the model in
Matlab/SimulinkTM
, and tuned the model parameters
by comparing the experimental and simulation data, as
illustrated in Fig. 6(b) and Table 1. Note that is
actually zero; this is because the backlash of a
ball-screw can normally be eliminated by preloading in
the manufacturing processes.
(a) The nonlinear network.
(b) The responses.
Fig. 6. Identification for1 .
Table 1. Model parameters for Fig. 6.
Parameters value Units
sF 30.5468 N
0.9577 -
sv 9.1195 mm/s
mc 589.1063 Ns/m
mb 47.0146 kg
sk 142.4507 kN/m
sc 3.6794 kNs/m
k 29.3750 kN/m
For the spring-damper sets 2 and
3 , we
SICE Annual Conference 2014September 9-12, 2014, Sapporo, Japan
428
considered the structure shown in Fig. 7, where can
be replaced by 2 or
3 . We generated a swept
sinusoidal input signal rz and measured the output
signalsz , and derived the transfer function from
rz to
sz by applying the subspace system identification
method [8]. We repeated the experiments six times to
account for system variation and uncertainties, and
denoted the transfer functions as ( ), 1, ,6iG s i . On
the other hand, we assumed 2 cs k and derived
the theoretical transfer function r sz zT as:
2.
r sz z
s
cs kT
m s cs k
Thus, we can calculate the values of c and k that
minimize the gap between r sz zT and ( )iG s , as in the
following:
,
, arg min max ( , ), ,r sz z i
c k ic k T G i
in which ( , )r sz z iT G
represents the gap [23] between
iG andr sz zT . The result gives:
2 96 21200.s
3 57289.
Fig. 7. Identification for
2 and 3 .
Lastly, we gave the piezoelectric transducer (PZT) a
swept sinusoidal signal input voltage inv and measured
the corresponding output displacement outd , and
derived the PZT dynamics as:
0.002616( ) .
628.7
out
in
ds
v s
5. OPTIMIZATION OF CIRCUIT Ze
In this section we apply the mechatronic inerter
network to 1i of Fig. 4, and design a circuit Ze to
optimize the ground responses. We simplified the design
procedures by regarding the friction and elastic as
model uncertainties and consider 1i as follows:
2
1 1 ,( )i
m
m m
a a e
Kb s c s k
R sL Z s
in which 21.0975aR , 1.5maL ,
7011VNs / A / mmmK , and mb , mc , k are given in
Table 1.
Regarding table responses, we considered the
following performance index:
0 ( ) ( ) ( ) ,2 2
s
l tJ z t z t z t (5)
to derive the optimal electrical circuits Ze,0 that
minimizes 0J as follows:
,0 0arg min .eZ J
We limited Ze as the following second-order impedance: 2
2 1 0
,0 2
1 0
,e
b s b s bZ
s a s a
(6)
because realization of higher-order impedance is costly
and difficult, and only slightly improves the system
performance that can be achieved by a second-order one
[9]. We used a full optical table with the following
parameters: ( ) 0.002616 ( 628.7)s s , 140.16 kgsm ,
15.5 kgum 217.27kg mI
29.92kg m ,I
0.6 mr fl l l , 0.45 mr ft t t , 3 57289,
2 96 21200s , , and optimized the networks of [4]
to obtain the optimal Ze,0 as follows:
1 6,0
5 2
1 2
2.484 10 0.004813 1
2.616 10 0.0092 2.107 10e
s s
s sZ
.
Jiang and Smith [4] showed that regular bi-quadratic
transfer function of (6) can be realized as a five-element
network, as shown Fig. 8 with 5
1 5.2308 10R ,
2 0.5232R ,5
3 9.4940 10R , 0.0092FC ,
0.0027L . The comparison in Fig. 9 and Table 2
indicate that Ze,0 can effectively reduce the
time-domain responses, and frequency responses
( )sZ at the frequency between 0.2Hz and 10Hz.
Table 2 illustrates an improvement of 3.62% for the
time responses, and an improvement of 32.96% for the
frequency responses at the concerned frequency range.
Fig. 8. The electric circuit structure[4].
(a) The time-domain load responses.
0 1 2 3 4 5 6 7 8 9 10-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
Am
plit
ude (
m/s
2)
time (sec)
without Ze
with Ze
SICE Annual Conference 2014September 9-12, 2014, Sapporo, Japan
429
(b) The frequency-domain load responses.
Fig. 9. Experimental responses.
Table 2. Statistic data from Fig. 9
Without
,0eZ With
,0eZ Improvement
(0, 10)tJ 1.7541 1.6906 3.6159%
(0.2, 10)J
0.0679 0.0455 32.9622%
6. ACTIVE CONTROLLER DESIGN
Consider the LFT structure of Fig. 5; all stabilizing
controllers can be parameterized as [11]:
1
1 1 2 22 1( ) .K I QU P Q
Therefore, we can design a proper controller K1 for the
filtered plant 2 22U P . We applied the standard H
loop shaping procedures [23–25] with the following
weighting functions:
1 2
200( 5000), 1,
100
sW W
s
(7)
and designed the controllers as follows:
1 diag( , , ),bp warp rollK K K K
where 2 6 10
2 4
779.3 7.096 10 1.6 10
216.3 1.163 1,
0roll
s sK
s s
0,warpK
0,
0bp
c
c
K
KK
2 6 10
2 4
779.3 7.096 10 1.6 10
216.3 1 10.
.163c
s s
s sK
Note that we regarded the table as a rigid body, and
ignored table’s warp mode. The experimental results of
the active control are shown in Fig. 11 and Table 3. First,
we note that the ground disturbance did not activate the
control signals iu ; i.e., the DRD structure is effective.
Second, the time-domain and frequency-domain load
responses are effectively suppressed by the designed
robust controller. Table 3 illustrates an improvement of
3.97% for the time responses, and an improvement of
16.89% for the frequency responses at the concerned
frequency range. In the practical application, we can
adjust the concerned frequency range by tuning the
weighting functions of (7).
Fig. 10. The three lags system
(a) The DRD effects
(b) The time-domain load responses.
(c) The frequency-domain load responses.
Fig. 11. Experimental responses of the active control.
10-1
100
101
102
103
10-6
10-5
10-4
10-3
10-2
Frequency (Hz)
Am
plit
ude (
m/s
2)
without Ze
with Ze
0 0.5 1 1.5 2 2.5 3 3.5 4-5
0
5
Fs
0 0.5 1 1.5 2 2.5 3 3.5 4-1
01
zr
0 0.5 1 1.5 2 2.5 3 3.5 4-2
0
2x 10
-5
ub
0 0.5 1 1.5 2 2.5 3 3.5 4-2
02
x 10-5
uth
0 0.5 1 1.5 2 2.5 3 3.5 4-5
0
5x 10
-5
uphi
time (sec)
0 1 2 3 4 5 6 7 8-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
time (sec)
ddot(
zs)
(m/s
2)
open-loop
active control
10-1
100
101
102
103
10-6
10-5
10-4
10-3
frequency (Hz)
ddot(
zs)
(m/s
2)
open-loop
active control
SICE Annual Conference 2014September 9-12, 2014, Sapporo, Japan
430
Table 3. Statistic data from Fig. 11.
Open
loop
Active
control
Improve
ment 1/2
26
0( )sz t dt
0.8799 0.8450 3.968%
1/2216
2
0.33( )sz d
0.0112 0.0093 16.89%
7. CONCLUSION
This paper has proposed a three-leg optical table, and
applied mechatronic inerter networks and DRD
techniques to control vibrations of the table. First, we
introduced the mechatronic inerter network, whose
impedance can be adjusted by connecting suitable
electrical circuits. For example, we designed a
second-order circuit to optimize the ground responses.
Second, we applied the DRD structure and designed
robust loop-shaping controllers to improve the system’s
load responses without influencing the ground responses.
Lastly, the designed mechatronic inerter network and
active controllers were implemented for experimental
verification. Based on the results, the proposed design
was deemed effective.
REFERENCES
[1] Newport 2011 Vibration Isolators, “http://www.newport.com/”.
[2] Kato T, Kawashima K, Funaki T, Tadano K and Kagawa T, “A new, high precision, quick response pressure regulator for active control of pneumatic vibraiotn isolation tables”. Precision Engineering Vol. 34, No. 1,43–48, 2010
[3] Chen MZQ, Hu Y, Huang L, Chen G , “Influence of inerter on natural frequencies of vibration systems”. Journal of Sound and Vibration, Vol. 333, No. 7, pp. 1874–1887, 2014.
[4] Jiang JZ and Smith MC , “Regular positive-real function and five-element network synthesis for electrical and mechanical networks”. IEEE Transactions on Automatic Control , Vol. 56, No. 6, pp. 1275–1290, 2011.
[5] Jiang JZ, Matamoros-Sanchez AZ, Goodall RM and Smith MC, “Passive Suspensions Incorporating Inerters for Railway Vehicles” Vehicle System Dynamics, Vol. 50, No. 1, pp. 263–276, 2012.
[6] Jiang JZ, Matamoros-Sanchez AZ, Zolotas A, Goodall R and Smith MC , “Passive suspensions for ride quality improvement of two-axle railway vehicles”, Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, DOI: 10.1177/0954409713511592, 2013.
[7] Oh JS, Han YM, Choi SB, Nguyen VQ and Moon SJ, “Design of a one-chip board microcontrol unit for active vibration control of a naval ship mounting system”, Smart Mater Struct, Vol. 21, No. 8, pp. 1-15, 2012 .
[8] Overschee PV and Moor BD , “N4SID-subspace algorithms for the identification of combined deterministic-stochastic system”, Automatica Volume,
Vol. 30, No. 1, pp. 75–93, 1994 [9] Papageorgiou C and Smith MC , “Positive real
synthesis using matrix inequalities for mechanical networks: application to vehicle suspension”. IEEE Transactions on Control Systems Technology , Vol. 14, No. 3, pp. 423–435, 2006.
[10] Smith MC , “Synthesis of mechanical networks: The inerter”. IEEE Transactions on Automatic Control ,Vol. 47, No. 10, pp. 1648–1662, 2002.
[11] Smith MC and Wang FC, “Controller parametrisation for disturbance response decoupling: application to vehicle active suspension control”, IEEE Transactions on Control Systems Technology, Vol. 10, No. 3, pp. 393–407, 2002.
[12] Wang FC and Smith MC, “Performance benefits in passive vehicle suspensions employing inerters”, Vehicle System Dynamics, Vol. 42, No. 4, pp. 235–257, 2004.
[13] Wang FC and Su WJ , “The impact of inerter nonlinearities on vehicle suspension control”. Vehicle System Dynamics, Vol. 47, No. 7, pp. 575–595, 2008.
[14] Wang FC, Tsao YC and Yen JY , “The application of disturbance response decoupling to the vibration control of an electron beam lithography system”, Japanese Journal of Applied Physics, Vol. 48, No. 6, pp. 1–5, 2009.
[15] Wang FC, Liao MK, Liao BH, Su WJ and Chan HA, “The performance improvements of train suspension systems with mechanical networks employing inerters”, Vehicle System Dynamics Vol. 47, No. 7, pp. 805–830, 2009.
[16] Wang FC, Hong MF and Yen JY , “Robust Control Design for Vibration Isolation of an Electron Beam Projection Lithography System”, Japanese Journal of Applied Physics, Vol. 49, No. 6, pp. 1–7, 2010.
[17] Wang FC, Hong MF and Chen CW, “Building suspensions with inerters”, Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, Vol. 224, No. 8, pp. 1605–1616, 2010.
[18] Wang FC and Liao MK, “The lateral stability of train suspension systems employing inerters”, Vehicle System Dynamics, Vol. 48, No. 5, pp. 619–643, 2010.
[19] Wang FC, Hong MF and Lin TC , “Designing and testing a hydraulic inerter”, Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, Vol. 225, No. 1, pp. 66–72, 2011.
[20] Wang FC and Chan HA, “Vehicle suspensions with a mechatronic network strut” Vehicle System Dynamics, Vol. 49, No. 5, pp. 811–830, 2011.
[21] Wang FC, Hsieh MR and Chen HJ, “Stability and performance analysis of a full-train system with inerters”, Vehicle System Dynamics, Vol. 50, No. 4, pp. 545–571, 2012.
[22] Wang FC, Yu CH, Tsai Jeff TH and Yang SH ,”Decoupled robust vibration control of an optical table”, Journal of Vibration and Control , Vol. 20, No. 1, pp. 38–50, 2014.
[23] Zhou K, Doyle JC and Glover K, Robust and Optimal Control. Prentice Hall, New Jersey, 1996.
[24] Wang FC, Chen LS, Tsai YC, Hsieh CH and Yen JY, “Robust Loop Shaping Control for a Nano-Positioning Stage”, Journal of Vibration and Control, Vol. 20, No.6, pp. 885–900, 2014.
[25] Wang FC, Yu CH, Tsai JTH, and Yang SH, “Decoupled Robust Vibration Control of an Optical Table”, Journal of Vibration and Control, Vol. 20, no.1, pp. 38–50, 2014.
SICE Annual Conference 2014September 9-12, 2014, Sapporo, Japan
431