Sei sulla pagina 1di 14

Application of Singular Perturbation Theory

to Hydraulic Servo Drives - System Analysis and


Control Design
Bernhard Manhartsgruber

Department of Mechanics & Machine Design


Johannes Kepler University
Altenbergerstr. 69, 4040 Linz, Austria
manhartsgruber@mechatronik.uni-linz.ac.at

This paper describes the system analysis and control design of an electrohydraulic servo-drive using a
nonlinear composite feedback control approach based on singular perturbation theory (Kokotovic, Khalil
& O’Reilly 1986). A comparison with linear output feedback (P-control) and linear state feedback (LQR-
design) shows the bene¿ts of the proposed method. The nonlinear control design features feedback
linearization of the slow system (reduced system, corresponding to the incompressible model) and sim-
ple proportional feedback of the pressure difference (fast system). Experimental results are given for
step responses and sine tracking.

Key words: hydraulic drives, nonlinear control, composite feedback, singular perturbation

1 INTRODUCTION
Electro-hydraulic control systems are used for the exertion of large forces on their surroundings
or for the exact manipulation of heavy objects. Applications include pick and place robots
for die-cast part handling, oscillation drives for continuous casting machines, force and rolling
gap control in rolling mills, positioning of aircraft control surfaces and control of military gun
turrets.
Simple proportional output feedback as described in (Merritt 1967) suffers from the low damp-
ing of hydraulic servo-drives.
State feedback techniques have successfully been implemented to improve the dynamic behav-
ior of hydraulic position control systems. To keep track with variations of system and load
parameters, various adaptive control methods have been tested. Still, the behavior of hydraulic
drives is inherently nonlinear. Especially in applications where the dry friction force at the hy-
draulic cylinder is high, or where the pressure drop at the servo valve becomes low compared
to the load pressure – i.e. where the drive reaches its power limits – the nonlinear effects cannot
be neglected.
This paper deals with the comparison of nonlinear composite feedback design for hydraulic
servo drives (Manhartsgruber & Scheidl 1998) to LQR design and conventional P-control. The
three control designs are implemented on a test rig with a driven mass of 545 kg and a displace-
ment of 150 mm. The controllers are designed to handle a step input of 30 mm height without
saturation of the control input or the chamber pressures.
In additition to the investigation of the step inputs, the problem of harmonic oscillation is ana-
lyzed. Even in the case of equal piston areas and a fully symmetric design of the entire system,
a strong nonlinear effect is reported for the sine tracking problem due to the nonlinear pressure-
Àow relation of hydraulic valves (Feuser & Piechnick 1993).
Proportional feedback with a control gain near the stability margin results in oscillatory step
responses in the experiments. The sine tracking performance of this type of control is low due
to nonlinear effects. While the LQR design provides suf¿cient damping and fast step responses,
it does not improve the sine tracking behavior. The nonlinear composite feedback control solves
the sine tracking problem along with a further improvement of the step response.

2 MATHEMATICAL MODEL

m FP

p1 , V1 p2 , V2

xv

pS p0

Fig. 1: Hydraulic servo drive.

Figure 1 shows a servo-drive with a hydraulic cylinder featuring equal areas and a direction
control valve. The cylinder acts on the heavy mass 6. The process force 8 is an unknown
disturbance. The supply pressure R7 and the return line pressure Rf are assumed to be constant.
This can be accomplished by the use of hydraulic accumulators near the control valve.
2.1 Equation of Motion with Dry Friction
Due to the tight sealing, hydraulic cylinders feature a strong dry friction effect. The friction
force is denoted by 8s o . The behavior of this friction force is rather complex. The friction
force not only depends on the piston velocity, but also on the pressure in the cylinder chambers.
Additionally, the sealing system has internal dynamics: The deformation of the seal rings due to
the pressurization of the sealing system takes some time. In the case of very small displacements
the contact force of the seal rings is dominated by viscoelastic reaction forces caused by the
small deformation. Therefore, the dry friction model is not valid for small displacements. In this
case, a critical value of approximately 20-40 >m can be observed in experiments. Furthermore,
the friction behavior is subject to changes due to temperature variations, aging and wear.
To the authors knowledge, the friction behavior of hydraulic cylinders under dynamic con-
ditions, i. e. the aforementioned internal dynamics of the sealing system, has not yet been
investigated. Experimental studies under stationary conditions (Gessat 1997) show a typical
dry friction behavior: The friction force decreases with increasing velocity. Furthermore, the
friction force increases with increasing chamber pressure. In this paper a model of the form
3  _%  4
 
   
 
3k
R7   R7 

_% E 
 _|  F
8s o ' r}? C_ff n _f e D   n & R  n & R2 
_| 2 2

is used to cope with the inÀuence of the chamber pressures R E| and R2 E|. The equilibrium
state of the drive with 8 ' f is % ' _%*_| ' f and R ' R2 ' R7 *2. Thus, the dependence
of 8s o on R and R2 is linearized at R ' R2 ' R7 *2. A signi¿cant increase of the friction force
at higher velocities (up to 0.3 m/s) did not appear in the experiments. Thus, a linear term in the
friction force model as used in (Yun & Cho 1991) is missing. The equation of motion reads

_2 %
6 ' ER  R2    8s o  8 .
_|2
2.2 Pressure Dynamics
The bulk modulus of oil varies with temperature and pressure (Jinghong, Zhaoneng & Yuanzhang
1994, Manring 1997). In the case of small temperature variations and moderate supply pressure
values, a constant bulk modulus q can be assumed. The evolution of the chamber pressures is
then described by

Tf n % _R
q
'  _%
_|
n ^T ER7  R   ^T ER  Rf 
_|   t   t
%T
nr} %Tc4@ '? R{R7 3R
?
 r}  %T
%Tc4@
'? R{R
 3Rf
?

T2f 3 % _R
q
'  _%
_|
n ^T ER7  R2   ^T ER2  Rf 
_|   t   t
nr}  %Tc4@ '? {R?  r} %Tc4@ '? R{R
%T R7 3R2 %T 2 3Rf
?

where Tf and T2f are the initial chamber volumes at % ' f and '? denotes the nominal Àow
rate of the valve at a nominal pressure loss {R? per notch. The valve opening %T is masked by
the function
;
? f % f
r} E% G' % e,re .
=  %:
In addition to the square root behavior of the ori¿ce Àow, there are always small leakage Àows
at the control valve. The leakage Àow is proportional to the pressure drop with a coef¿cient ^T .
Since the leakage is inÀuenced by the valve spool position, this model is only an approximation
of the real behavior. Yet, the effect of leakage for the closed valve (%T ' f) is precisely incor-
porated in the mathematical model. Therefore the leakage model is a good approximation for
small valve openings. For large valve openings the inÀuence of the leakage Àow is negligible.
2.3 Valve Dynamics
If the effects of hysteresis and Àow force are neglected, the dynamics of the valve can be
described by an LTI system. The transfer function can be estimated from experimental in-
put/output data using Welch’s periodogram method. Figure 2 shows the results of this non-
parametric identi¿cation method together with a 6th order transfer function
10

magnitude [dB]
0
-10
-20
-30
-40
0
-100
phase [°]

-200
-300
-400
-500
10 100 1000
frequency [Hz]

Fig. 2: Estimated transfer function of the servo valve and least squares ¿t.

@f
C Er '
@f n @ r n @2 r2 n @ r n @e re n @D rD n rS
computed by a least squares ¿t. The coef¿cients @ are given in table 1.
Due to saturation effects, the linear behavior is only valid for very small amplitudes. The
frequency analysis was performaed with input/output data of a logarithmic sine sweep between
1 Hz and 10 kHz with a demand amplitude of 1 % of the full valve opening.
2.4 Dimensionless Equations
The application of perturbation techniques requires a properly scaled formulation of the prob-
lem. This is achieved by a dimensionless time coordinate

_  _
u
' /| ' E '
_ / _|
and by rescaling of the dependent variables in the form

R E| R2 E| % E| %T E|


  E  ' 2 E  ' 1 E| ' l E| ' ,
R7 R7 %4@ %Tc4@
with the supply pressure R7 , a rated displacement %4@ , and the maximum valve opening %Tc4@ .
With the new parameters

Tf n T2f Tf  T2f


Tf ' e'
2 Tf n T2f
u
'? R7  %4@ R7  R7
@/ ' %Tc4@ c K' S/ ' c 0' c
/Tf {R? Tf 6 %4@ /2 q
_ff _f ^T R7
_/ff ' c _/f ' c ^/ '
6 %4@ / 2 6 %4@ /2 /Tf
the model may be written as
d3 5=588:5<  4355
d4 5=567677  434<
d5 9=6763;<  4348
d6 4=3;9869  4345
d7 4=:8;4<7  43;
d8 4=43<;6:  437

Table 1: Coef¿cients of the transfer function C Er.

s s
0 E n e n K 1   ' K 1 n ^/ E  2  n r} El @/      r} El @/  (1a)
s s
0 E  e  K 1  2 ' K 1 n ^/ E  22  n r} El @/    2  r} El @/ 2 (1b)
1 ' S/ E  2   s  (1c)
  
 3km1 m
r}? 1 _/ff n _/f e  E n & E n 2  

3 EXPERIMENTAL SETUP
The test rig with a driven mass of 6 ' DeD kg is shown in ¿gure 3. The other parameters can
be found in table 2.
The experiments were performed with a MOOG D760-995A type servo-valve. The nominal
Àow rate at a pressure drop of D bar per notch is 2 l/min at full opening. Due to a signi¿cant
saturation of the Àow curve, the Àow gain is higher for small openings. Since the valve is
operated in the linear range during the experiments, the saturation is not modelled.
The dimensionless parameters related to the scaling frequency / '  rad/s are

@/ ' H2c K ' fbc S/ ' .HDc 0 ' fff.D


_/ff ' 2 Sc _/f ' Hc k ' 2f e ' fDc ^/ ' ffSe .

4 LINEAR CONTROL
In the following linear analysis the valve dynamics is assumed to be in¿nitely fast. Thus, the
control input  is equal to the valve opening l. The dry friction effect must be neglected in
order to make the system linearizable. With the state transformation
5 6 5 6
j  n 2 n e E  2   
9 { : 9   2 :
'9 :'9
7 1 8 7 1
:,
8
1 1
the linearization at the resting point 1 ' 1 ' j ' { ' f with input l ' f yields
5 ^/ ^/ 6 5 6
2 2e f f f
9
9
0 0 :
: 9 s :
9 e ^/  n e2 ^/  K : 9 2 @/ :
2 2 f 2 9 :
 '9
9   e2 0   e2 0
:
  e2 0 : n 9   e2 0 :l .
9 : 9 :
7 f f f  8 7 f 8
f S/ f f f
Fig. 3: Experimental Setup.

Classical feedback design (Merritt 1967) uses linearized models at midstroke (e ' f), because
the stability margins are worst at this position. With e ' f, the model is
5 ^/ 6 5 6
2 f f f f
9 0 : 9 s :
9 ^ K : 9 2@/:
9 f 2
/
f 2 : 9 :
 '9 0 0 : n9 0 :l .
9 : 9 :
7 f f f  8 7 f 8
f S/ f f f
The ¿rst state variable j becomes nonobservable and noncontrollable. The stability of this
subspace is guaranteed by ^/ : f. Other authors (Bobrow & Lum 1996, Plummer & Vaughan
1996) omit this analysis of the dynamics of the pressure sum and de¿ne the net force on the
piston or the pressure difference as a state variable and work with a state vector like
 A
3 ' { 1 1
resulting in a third order model. This reduction of the state space results in
5 6 5 s 6
^/ K 2@/
2 f 2
9 : 9 :
3 ' 7 f 0 f  0 8 3n 7 0 8 l .
f
S/ f f f

4.1 Proportional Feedback


Most industrial controllers for hydraulic drives use proportional feedback with a demand input
1 oes in the form
 
l ' & 1 oes  1 .
Symbol Value Unit Remarks
sV 433 bar supply pressure
 46633 bar effective bulk modulus
D 9> 74 cm5 effective piston area
{pd{ 483 mm rated displacement
Y43 758 cm6 volume of left chamber
Y53 8:8 cm6 volume of right chamber
g33 4<3 N friction parameters
g34 <8 N
 5 1

Table 2: Physical Parameters.

The closed loop model is


5 s 6 5 s 6
^/ 2@/ K 2@/
9 2 & 2 : &
3 ' 7 0 0 0 8 3n 9
7 f 0 :8 1 oes
f f 
S/ f f f
with a stability criterion

s K ^/
f & 2 2  2 .
0 @/
A gain of & ' f is used in the experiments. With this gain the controlled system handles
input steps of f mm without saturation (cavitation or pressures above the supply pressure) of
the chamber pressures. The LQR controller and the nonlinear controller are designed to meet
the same restrictions for a step height of f mm.
The proportional gain & is limited by the amount of leakage at the control valve. Alternatively,
the damping can be modelled by internal leakage at the cylinder or by viscous friction forces.
Yet, the modelling of leakage at the valve is more accurate and realistic, since the sealing of the
cylinder chambers is almost ideal and viscous friction forces can be neglected at low velocities.
4.2 LQR Control
In order to improve the dynamical behavior of the controlled drive, a state feedback with an
extra demand input
 
l ' & 1 oes  1  &( 1  &{ {
can be designed by either pole placement (Finney, de Pennington, Bloor & Gill 1985, Vaughan
& Plummer 1990, Plummer & Vaughan 1996) or by optimization of a performance index
(Bobrow & Lum 1996, Jacobs & Roth 1982). The performance index used is

]"
 A 
3 " 3 n o l2 .
f
This is the well known LQR-problem (Athans & Falb 1966). The model
5 6 5 6
f. f Df. .2f
3 ' 7 f f  8 3n 7 f 8 l
.HD f f f
with the weight matrices
5 6
ff f f
" '7 f f f 8 o'
f f f
yields the controller gains

& ' Sc &( ' ffSc &{ ' fD .

5 NONLINEAR COMPOSITE FEEDBACK CONTROL


The linearization of the system for the design of linear controllers results in the paradox situation
that the leakage of the valve limits the controller gain. Most of the dissipation in the system
is due to dry friction effects. Thus, a nonlinear control design should yield better results. The
method of composite feedback control (Kokotovic et al. 1986) uses a two time-scale approach
for the design of slow and fast feedback functions for the slow and fast dynamics of the system.
The application of this model for the position control of hydraulic servo systems has been
proposed in (Manhartsgruber & Scheidl 1998).
The model (1a,1b,1c) is rewritten in the new state
5 6 5 6
P  n 2
9 { : 9   2 :
9 : 9 :
3'9 :'9 :
7 1 8 7 1 8
1 1
yielding

0 E n e n K 1 Pn2 { ' K 1 n ^/ E  P  {
 
(2a)
t t
n r} El @/   2  r} El @/ Pn{
Pn{
2

0 E  e  K 1 P32 { ' K 1 n ^/ E  P n { n
 
(2b)
t t
n r} El @/   2  r} El @/ P3{
P3{
2
 
1 ' S/ {  s  r}? 1  (2c)
 
_/ff n _/f e3km1m E n & EP  


The dynamic behavior of the servo-valve is described in the frequency domain by1

l ' C Er  .

5.1 Exact Feedback Linearization of the Reduced Model


 Therefore,
The servo-valve is very fast compared to the dynamics of the ”slow” states 1 and 1.
the valve dynamics is neglected (l ' ) in the reduced model. Furthermore, the unknown
disturbance force s E  is neglected in the following control design.
Setting 0 ' f in the differential equation system (2a,2b,2c) results in a differential algebraic
4 a x
The signals  and x are denoted without hats (> a) in the frequencey domain, since the hat denotes fast time
in this paper.
system where the algebraic subsystem has the unique solution

Pf ' ,
v
@/ @2/   2K   2 
{f '  2 n r}?  1 n 2  r}? 1 2 @/  K1 .
1 (3)
2^/ e^/2 ^/ e^/ ^/
Substitution of these solutions into eq. (2b) yields
  
1 ' S/ {f  s  r}? 1 _/ff n _/f e3km1 m . (4)
Thus, the reduced system is of second order and has the control input . The goal of the exact
feedback linearization is to ¿nd a slow feedback function
 
 
 ' Kr 1c 1c
that results in the standard linear behavior

1 n 2 B /S 1 n /2 1 ' /2  (5)
S S
of the drive position 1 due to the new control input . A comparison of the nonlinear reduced
model (4) with the linear equation (5) results in the requirement
     
{f 1c 1c  '  / 2 E  1  2B/ S 1 n r}? 1 _/ff n _/f e3km1 m . (6)
S
S/
This can be achieved by the inversion of equation (3) resulting in the slow feedback function
 
   y
 K1 n ^/ {f 1c 1c x
  ' x 2
Kr 1c 1c w    
@/   r}? 1 {f 1c 1c
 
 

with {f 1c 1c  according to eq. (6).

50
magnitude [dB]

-50

-100
10-3 10-2 10-1 100 101 102 103

-180
phase [°]

-360

-540

-720
10-3 10-2 10-1 100 101 102 103
ω [1]

Fig. 4: Bode Plot of the Open Loop Transfer Function.


sωc2
s2ωc2 + 2δωc s +1
.
Reference Model Friction Force Ffr ( ξ )
v
.
ξ . ζs
z −1
z
ζs (ξ,ξ,v ) ζ

ξ
Numerical
Differentiation Slow Feedback
ζf

ψ1
- - k∆
ψ2

Hydraulic Drive

Fig. 5: Composite Feedback Control with Reference Model for Friction Compensation.

5.2 The linearized dynamics of the fast variables


The fast dynamics is investigated in the fast time scale

Y Y
' c E  ' '0 .
0 Y
Y
The transfer function of the valve dynamics may be rewritten by

@ f
T Er ' 2
@
f n @  0 r n @ 2 E0 r n @
E0 r n @ e E0 re n @
D E0 rD n @ S E0 rS
 ' @ *0 . The Laplace variable r is transformed into the new time by
where @

r ' 0 r .
The valve transfer function in the transformed frequency domain is

@ f
T E
r ' .
@
f n @  r n @
2 n @ r n @ r 2
e r e n @
D r D n @ S r S
In the fast time scale the linearized dynamics of P and { at 1 ' 1 ' f is
5 ^/ ^/ e 6 5 s e 6
  2 2    2@/
P 9   e2   e2 : P 9 e :2
 '7 ^ e ^/ 8  n 7 s 8l .
{ / { 
2 2 2@/
  e2   e2   e2
The transfer function from the input l to the output { is
s
E 2@/ r n 2^/
C r '   .
e 2 2^/ 2^/
rn rn
ne e
The inÀuence of the reference position e is negligible. The nominal model at midstroke (e ' f)
is
s
f E 2@/
C r ' .
r n 2^/
5.3 Proportional Control of the Fast Dynamics
The slow feedback function Kr copes with the dominant nonlinearities. The deviations of the
perturbed system behavior from the reduced model – i.e. the compressibility effects – have to be
eliminated by a second feedback, the fast feedback function Ks . Since the deviations between
the perturbed equations and the reduced system are  E0 small (Fenichel 1979) a simple linear
control design can be used for this purpose.
In this paper, a proportional feedback
     
 
Ks 1c 1c c { ' &{ {f 1c 1c   {
with the desired pressure difference according to eq. (6) and a constant gain &{ is designed in
the frequency domain. Figure 4 shows the magnitude and phase plots of the open loop transfer
function &{  T E f E
r  C r for &{ ' . The stability margin is reached at &{ ' e2. A range
of &{ ' f.D to &{ ' D has been successfully used in the epxeriments to get a well damped
response of the pressure control.
5.4 Composite Feedback
The control loop is closed by simply adding the slow and fast feedback functions to a composite
feedback (Kokotovic et al. 1986)
   
  n Ks 1c 1c
 ' Kr 1c 1c  c { .
The structure of the control system is shown in ¿gure 5. The velocity is not measured but
computed by numerical differentiation of the position signal. The friction force compensation
can be improved by using the velocity of a reference model according to eq. (5) instead of the
signal from the numeric differentiation.
The parameters used in the following experiments are

/ S ' 2Z  eDc B ' fHc &{ ' f.D .

6 EXPERIMENTAL RESULTS, CONCLUSIONS, AND OUTLOOK

6.1 Step responses


The step responses for the P-controlled system are given in ¿gure 6a. Step heights of 30, 15 and
6 mm have been tested according to a step from fc ffDc ff2 to nfc nffDc nff2 and
back in the normalized scale. Since the scaling frequency is / ' , the time scale coincides
with the physical time | measured in seconds.
The step response behavior for LQR control in ¿gure 6b is much better, with higher usage of the
available valve opening. The feedback of the pressure difference { results in a small increase
of the stationary position error due to the stiction effect. The usual workaround for this problem
is an addtional feedback of the integrated error. In the presence of dry friction such additional
integrators must be disabled for small position errors to avoid limit cycles.
a) b) c)
0.1 0.1 0.1
0.05 0.05 0.05
ref

ref

ref
0 0 0
ξ, ξ

ξ, ξ

ξ, ξ
-0.05 -0.05 -0.05
-0.1 -0.1 -0.1
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2

2
2 2
1

d ξ/d τ

d ξ/d τ
d ξ/d τ

0 0 0
-1
-2 -2
-2
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2
1 1 1

0.5 0.5 0.5

0 0 0


-0.5 -0.5 -0.5

-1 -1 -1
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2
0.2 0.5 1

0.1 0.5

0 0 0
ζ

ζ
ζ

-0.1 -0.5

-0.2 -0.5 -1
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2
τ τ τ

Fig. 6: Step responses: (a) P-control, (b) LQR-control, (c) nonlinear composite feedback control

The nonlinear composite feedback control (Figure 6c) shows a further improvement of the step
response. The behavior of the nonlinear control for small position errors will be improved
in future work by adding the aforementioned error integrator together with a small region of
inactivity around zero. Furthermore, the behavior of friction of hydraulic cylinders should be
investigated in more detail.
6.2 Sine oscillation
The bene¿ts of nonlinear control can be shown by a comparison of the sine response of the
hydraulic drive. The sine oscillation is best viewed in a phase space plot with position 1, velocity
1 and acceleration 1. The acceleration is replaced by a measurement of the force in the piston
rod. The rod force 8 was normalized in the way

8
s' .
R7 
In addition to the three-dimensional 1  1  s trajectories, a projection onto the 1  s -plane is
shown in ¿gure 7.
The systems with proportional feedback (Figure 7a) and LQR control (Figure 7b) show a non-
linear behavior of the trajectories during harmonic oscillations. The nonlinearities can be com-
pensated by nonlinear control (Figure 7c) to give a precise harmonic motion of the mass. The
sine oscillations were performed at a frequency of  Hz with Amplitudes of 150 and 105 mm
(1 and 0.7 in the dimensionless scale). The trajectory with 150 mm amplitude is slightly af-
fected by the nonlinearity of the Àow curve of the servo-valve. The problem could be solved by
incorporating this nonlinearity in the model.
a) b) c)
0.6 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2


f

0 0 0

f
-0.2 -0.2 -0.2

-0.4 -5 -0.4 -5 -0.4 -5

0 0 0
-1.5
-1 d ξ/d τ -1.5
-1 d ξ/d τ -1.5
-1 d ξ/d τ
-0.5 5 -0.5 -0.5
ξ 0 ξ 0 5 0 5
0.5
1 0.5 ξ 0.5
1 1

Fig. 7: Sine oscillation: (a) P-control, (b) LQR-control, (c) nonlinear composite feedback control

7 NOMENCLATURE
6 mass of moving parts [kg] ^T leakage param. [me s/kg]
 effective piston area [m2 , cm2 ] / time scaling factor [1/s]
q effective bulk modulus [N/m2 , bar] | time [s]
%4@ rated displacement [m, mm] % E| position [m]
Tf c T2f initial chamber volumes [m , cm ] R E| c R2 E| pressures [N/m2 , bar]
R7 supply pressure [N/m2 , bar] %T E| valve opening [V]
Rf return line pressure [N/m2 , bar] @/ Àow gain [1]
{R? nominal pressure drop [N/m2 , bar] K displacement ratio [1]
'? nominal Àow rate[m /s, l/min] S/ inertia force ratio [1]
_ff kinetic Coulomb friction [N] ^/ leakage constant [1]
_ff n _f static Coulomb friction [N] dimensionless time [1]
k friction force decay exponent [s/m]  E  c  2 E  dimensionless pressure [1]
& sensitivity: friction/pressure [m2 /N] 1 E  dimensionless position [1]
8s o friction force [N] l E  valve opening [1]
%Tc4@ maximum valve opening [V]

REFERENCES
Athans, M. & Falb, P. (1966). Optimal Control, McGraw-Hill.
Bobrow, J. E. & Lum, K. (1996). Adaptive, high bandwith control of a hydraulic actuator,
Transactions of the ASME. Journal of Dynamic Systems, Measurement, and Control
118: 714–720.
Fenichel, N. (1979). Geometric singular perturbation theory for ordinary differential equations,
Journal of Differential Equations 31: 53–98.
Feuser, A. & Piechnick, M. (1993). Antriebsdimensionierung unter berücksichtigung von
zylinderÀüchen- und steuerkantenverhältnissen, O+P Ölhydraulik u. Pneumatik 37: 509–
516.
Finney, J. M., de Pennington, A., Bloor, M. S. & Gill, G. S. (1985). A pole assignment
controller for an electrohydraulic cylinder drive, Transactions of the ASME. Journal of
Dynamic Systems, Measurement, and Control 107: 145–150.
Gessat, J. (1997). Reibungsverhalten von Hydraulikdichtungen und Führungselementen, O+P
Ölhydraulik und Pneumatik 41: 743–746.
Jacobs, M. & Roth, J. (1982). Rechnerunterstützte Auslegung und Realisierung von Regelungs-
konzepten für elektrohydraulische Antriebe, 5. Aachener Fluidtechnisches Kolloquium,
Fachgebiet Hydraulik, Band 2, pp. 239–262.
Jinghong, Y., Zhaoneng, C. & Yuanzhang, L. (1994). The variation of oil effective bulk modu-
lus with pressure in hydraulic systems, Journal of Dynamic Systems, Measurement, and
Control 116: 146–150.
Kokotovic, P. V., Khalil, H. K. & O’Reilly, J. (1986). Singular Perturbation Methods in Con-
trol: Analysis and Design., Academic Press, London etc.
Manhartsgruber, B. & Scheidl, R. (1998). Non-linear control of hydraulic servo-drives based
on a singular perturbation approach, in C. A. Burrows & K. A. Edge (eds), Proceedings
of Bath Workshop on Power Transmission and Motion Control, Professional Engineering
Publishing, London, pp. 301–313.
Manring, N. D. (1997). The effective Àuid bulk-modulus within a hydrostatic transmission,
Journal of Dynamic Systems, Measurement and Control 119: 462–466.
Merritt, H. E. (1967). Hydraulic Control Systems, John Wiley, New York.
Plummer, A. R. & Vaughan, N. D. (1996). Robust adaptive control for hydraulic servosys-
tems, Transactions of the ASME. Journal of Dynamic Systems, Measurement, and Con-
trol 118: 237–244.
Vaughan, N. D. & Plummer, A. R. (1990). Some issues in pole placement control of electro-
hydraulic servos, Proceedings of Third Bath International Fluid Power Workshop, Bath,
UK, pp. 6–28.
Yun, J. S. & Cho, H. S. (1991). Application of an adaptive model following control tech-
nique to a hydraulic servo systeme subjected to unknown disturbances, ASME Journal of
Dynamic Systems, Measurement and Control 113: 479–486.

Potrebbero piacerti anche