HYDRAULIC ACTUATOR SYSTEM FOR ROTOR CONTROL

Heinz U l b r i c h * and Josef Althaus

I n s t i t u t e B o f Mechanics
Technical U n i v e r s i t y o f Munich
P.O. Box 202420
D-8000 Munich 2, Federal Republic o f Germany
In the last ten years several different types of actuators have been developed and fabricated
for active control of rotors. This paper deals with a special hydraulic actuator systenl ca.pablc of
generating high forces to rotating shafts via conventional bearings. The actively controlled hydraulic
force actuator features an electrohydraulic servo valve which can produce amplitudes a.nd forces at
high frequencies necessary for influencing rotor vibrations. The mathematical description will be
given in detail. The experimental results attained verify the theoretical model. Simulations already
indicate the usefulness of this compact device for application to a real rotor system.

1 Introduction
Rotating machinery can be found in many fields of the industrial world; many could be im-
proved with regard to speed of rotors/shafts, weight, noise, longevity and last .but not least.safety,
by applying active vibration control. The key to success for active measures in the improveme~~t of
the dynamics of rotating machinery lies in the availability of suitable actuators which have to satisfy
the following requirements: the actuator must be capable of amplitudes in the range of the vibr a t'Ion
amplitudes to be influenced on the one hand, and possess an appropriate frequency cha,racteristic on
the other hand. In the literature many papers deal with magnetic actuators, e.g. [3,$,5]. Magi~et~ic
actuators require a relatively large amount of space compared to the attainable forces. This deficiency
may be avoided by the use of hydraulically controlled chambers.
The main topic of this paper is the introduction of a newly developed hydraulic actuator ~ i ~ l ~ i c l i
is able to apply forces to rotating shafts via the bearings. For producing forces (pressure va.riations)
servo valves are used. The regulating distance is caused by elastic deformations of nlembranes (no
cylinder piston arrangements).
All influences which could be of practical relevance (oil compressibility, dynamic effects of
the servo valve and the fluid itself) are taken into account in the mathematical description of thc
system. Applying the theory of plates, the strain in the membranes can be calculated. A limitation
of the actuator system is given by the permitted strain which may not be exceeded in any area.
Because of the actuator dynamics an appropriate enlargement of the equations of motion for the
entire system is necessary. The equations of the actuator dynamics show that the chamber system
already applies damping effects to the rotor system without any control input. With thc aid of'
a well adapted controller the damping of the entire system can be essentially improved. This is
* P r e s e n t l y a t Technical U n i v e r s i t y o f Braunschwei g, Braunschweig, Federal
Republic o f Germany.

269
PRECEDING PAGE BLANK NOT FILMED
demonstrated by simulations. The input parameters, especially with regard to the actuator system,
have been determined by experiment.
For verifying the results achieved by computer simulations two different experimental test
facilities have been constructed (see Fig. 4 and Fig. 5). One test rig contains the active chamber
system. It allows the evaluation of its characteristic frequency. The experimental results which are
attained verify the theoretical model. The second test facility consists of anelastic rotor]s t r u c t i ~ r e
and two different types of magnetic actuators. In further investigations the two test facilities will be
combined in order to test the dynamic behavior of the elastic rotor controlled by the active chamber
system.

Nornenclat ure

system matrix
membrane area
coefficient of the actuator input
control matrix
bearing coefficients
lag coefficient of the control force
membrane stiffness
passive damping coefficient of the actuator
identity matrix
force vector
leakage coefficient
loop gain of the servo valve
constant of the oil compressibility and the membrane buckling
system matrices of the second order system
control force to the bearing housing
transfer function describing the fluid dynamics and the stiffness
transfer function of the chamber pressure
transfer function considering the servo valve dynamics, the
oil compressibility, and the membrane buckling
transfer function describing the servo valve dynamics
transfer function of the regulating distance
Jacobian matrix of translation of the bearing
pressure drop in the valve
supply pressure difference
chamber pressure difference
vector of generalized coordinates
oil flow through the loaded servo valve
oil flow through the unloaded servo valve
Laplace-operator
time
control vector
 control voltage to the servo valve
nominal voltage of the servo valve
state vector
acting directions
regulating distance (bearing deflection)
natural frequency of the servo valve
rotor frequency
servo valve damping coefficient

2 Description of the Actuator Device

The device of this compact system is shown in Fig. 1 by a schematic. It consists of t\\~o
components: a commercially-available servo valve, which is able to transform electrical signals into
flow- and pressure variations of the hydraulic fluid, and an elastic chamber system, which tra.nsduces
, the pressure variations into displacements and into forces acting on the outer bearing housing (in
Fig. 1 only one control direction is displayed).
The servo valve is supplied by the system pressure which is to be kept constant,ancl the retur~l
pressure R z 0. If the spool of the valveis in theineutralipositionl(Fig.ljthe~pressureintlie chambers is
half of the system pressure. After a shift of the spool by an input voltage Uv # 0.a pressure difference
appears between the two output orifices. Because of the elasticity of the membranes this enahlcs a
shifting of the rotor in the radial direction. The chamber system itself consists of four cylintlricill
chambers which are equally spaced in a circle around the bearing housing. Each chamber is sealccl on
the top and bottom by an elastic membrane. In order to decouple forces into both acting directions,
the bearing housing is supported against the membrane system by linear roller guides. The influence
of friction is thereby reduced as well. Alternative support solutions are given by design variations. To
eliminate the linear roller guides the bearing can be mounted via elastic rods to the outer housing.
The two opposing chambers are both controlled by one servo valve (one valve per force direction).
This compact actuator system is capable of generating very large forces and can thereby in flu-
ence even large turbines weighing several tons. Actuator pistons as an alternative to tlie dcfor~i~al~lc
chambers have the drawback that they possess relatively large moving masses. In adclition, sealing
problems and friction forces on the contact surfaces would occur. All these influences have negative
impacts on the frequency characteristic. An appropriate design of the membranes enables the desirecl
amplitudes of motion without exceeding the stress limits of the material.

3 Theory
\lit11 the aid of the theory of plates the calculation of the stress in the membranes causccl by
a given load can be performed, e.g. [8]. Experiments have shown that the real stresses are always
lower than the calculated stresses. This is caused by thelelastic fixing of the membrallie in the real
system compared to the theoretically stiff fixing. For calculating the stresses the load must be l<no\\in.
Therefore it is necessary to describe the dynamics of this active machine element.

3.1 Chamber System Dynamics

3.1.1 Servo Valve

The dynamics of the servo valves are given by the manufacturer (MOOG, see [2]) as a PT2-
element. The valve flow Qv for the unloaded valve can be given as a function of the input voltage
Uv in the Lnplnce-domain (with s as a complex variable)
1

where II'~ is the loop gain of the valve, which depends on the supply pressure APs = P - R (sec
Fig. 1) and the geometry of the valve itself, w, is the natural frequency of the valve, and 5, is the
damping ratio of the valve. In the case where the valve is loaded, the fluid flow depends on the output
pressure APv (load differential pressure). For this reason the equation for the fluid flow passing a
pressure-regulating valve (here between the spool of the valve, inlet- and outlet orifice, see Fig. 1) is
applied. In general the fluid flow Q is proportional to the square root of the differential pressure A P

Applying the relation that the sum of the pressure drops in the valve A P and the load pressurc
APv is equal t o the drop of the supply pressure APs,

AP+APv =APs ,
and considering eq. 2 and the fluid flow properties in the valve we obtain

After a linearization in the neighborhood of APv = 0,eq. 4 becomes

Q = Qv - I(,, . APv , lip,leakage coefficient . (5)

The linearization applied in eq. 5 is valid for APv 5 $APs where I&, is theoretically a.
function of Qv. For low input voltage Uv 5 &UN (UN = nominal voltage) this dependance can be
neglected. Fig. 2 shows the course of the fluid flow Q as a function of Qv and the differential pressure
of the valve APv as given by the manufacturer. The slope of these curves at the operating point.
(APv = 0) is given by I(,, appearing in eq. 5. In contrast to the nonlinear eq. 4 in practice one ca,i~
establish that even at Uv = 0 (which results in Qv = 0, eq. 1) the fluid flow is non-zero for A P v # 0.
This is caused by leakage where the spool is in the neutral position. The linear eq. 5 takes this inlo
account. Fig. 2 shows this behavior of the valve by the line through the origin with a slope which is
non-zero. For derivation of the equation of motion and the transfer characteristic of the entire SYS~CIII
(see later), the linear relation by eq. 5 is used. The nonlinearity of the relation between the fluid flow
Q and the pressure drop APv can be taken into account by I(,, as a function of Qv.
3.1.2 Influence of Fluid Flow, Membrane Stiffness and Oil Compressibility

The application of the continuity equation results in an additional relation for the fluid flow

Q = A* . SL + Ii*. A& .
A* is the characteristic membrane area, and K* represents the oil compressibility and the deforms.-
tion of the membrane into a direction that doesn't lead to a movement of the bearing (rnei~lbra.lie
buckling), see Fig. 3. K* is dependent on the oil volume VL, the bulk modulus of the fluid PI,-, and
the membrane geometry characterized by I&,

Simulations have shown that the fluid losses caused by oil inertia, friction and losses at the inlct
and outlet orifices are negligible. So the pressure difference at the outlet orifices of the valve is
approximately equal to the chamber pressure difference, constant A Pv.

,3.1.3 Recapitulation of Equations

The equilibrium of forces applied with respect to the membranes (entire asrangelllent of two
membranes acting in one direction) leads to

F* = A * . APv - CL - XL. , (8)

F' the control force to the bearing

where CL is the entirelmembrane~stiffness~inthelforce'direction~and,
housing, see Fig. 3. Including this equation we obtain F*as a function of the control voltage Uv ant1
the membrane deflection, respectively; the bearing displacement XL in the,direc tioil of the force

Fv is the transfer function describing the dependence on the control voltage

and FLis the transfer function considering the influences of the fluid dynamics and the stiffness of
the arrangement
A * ~
-s
FL= ICpq
1+Ch"S
+
CL . (11)

The constant CK is represented by the leakage coefficient I(,, and by the oil compressibility a.11tl
membrane buckling Ii'* eq. 7,
I<*
CIC= - . (12)
Iip*
3.1.4 Representation in Time Domain

For including the actuator system into the complete system. the representation of eq. 9 in
the time-domain is more convenient. With the equations above we get a set of differential equations
which describe the control force F* as a function of the input voltage Uv and bearing displacenlent
XL,

The coefficient CL describes the additional stiffness effects with regard to the entire system (see later).

3.1.5 Simplification and Interpretation of Equations

For designing control concepts and to explain the fundamental operation. the equa,tions will
be simplified. In order to obtain a simple relationship between the control voltage Uv, the regulating
distance XL, and the control force Fq the PT2-behavior of the servo valve will be approximated by a.
linear relationship (FVaI,, = 1). This is valid at frequencies under the cut-off frequency of the valve.
This simplification leads to the following form

$2
where b = - and dL = A*2
+ CKcL.
TEQconstant CIi representing the oil compressibility and the membrane buckling in the eclua-
tion above has an important influence. If CIc is negligible the actuator works as a passive spring and
damper unit characterized by the coefficients CL and dL. Additionally a control force Fc acts that
is proportional to the input voltage Uv (see Fig. 9). In the other case, if the frequency or; = 1
li
is below the frequency range of application, the actuator possesses an integral behavior. As a result,
the feedback of velocities and accelerations leads to stiffness and damping effects. In this caae or
if w ~ cis inside the range of application (PTI behavior),a proportional behavior (see Fig. 9) can be
achieved by an internal phase shifting controller (e.g. feedback of pressure APv) or an additional
passive throttle between the two output orifices.
By an appropriate design of the membrane (constants c ~ A*, , of the oil volume IfL, oS
the supply pressure APs, and by the choice of the suitable valve size (constant IiT, and Ii;,,), the
actuator can be tuned in an optimal way. In the case of a sudden pressure drop (APs t 0) the rotor
bearing is always sufficiently supported by an adequate membrane stiffness c ~ .

3.1.6 Transfer Characteristic of Actuator

In the following the transfer characteristic of the complete actuator system is investiga tecl
applying a sinusoidal input voltage Uv to the valve. The1 purpose oil this i s to get a comparison t,o
the experiments which are described in chapter 3.1.6 and 5. The resulting outputs are the regulating
distance X L and the pressure difference APv between the opposing chambers. Introducing a ma.ss
which simulates the mass of the rotor, including the support system,the control force appearing in
eq. 9 has to be equal t o
F* = r i ~X L - s 2 . (15)
The transfer functions for the regulating distance Fzand the chamber pressure Fp can be expressed

The coefficients b and d L are given in eq. 14.

4 Test Facilities
Fig. 4 shows the test rig which was designed to investigate the chamber system experimentally.
The investigation is conducted only in the vertical direction. The facility mainly consists of the
housing 1 where the four oil chambers 2 are integrated. The distance rod 3 performs the function of
the bearing housing and assures that the opposing chambers both move the same absolute distance.
For investigating the influence of masses additional masses can be mounted on the distance rod 3.
The influence of friction caused by the linear roller guides (see Fig. 1) can be studied by loading
the chambers for the horizontal direction with an appropriate pressure in order to press the linear
roller guides via distance sleeves 4 on the distance rod. The entire arrangement is niountecl on
the foundation including the servo valve 5 . Servo valve and chambers are connected via pipes 6.
Between actuator and hydraulic control unit (not shown in Fig. 4) are the connections to the first
stage pressure unit, main control pressure unit (pressure P shown in Fig. I ) , return pressure unit
(indicated by pressure R in Fig. l),and an additional pressure supply for the linear guiding bearings.
To enable a broad range of experimental investigations, variations of the supply pressure APs,
membrane thickness h, contact pressure for the guiding bearings, additional masses m (see eq. 15)
and of course the control voltage Uv (function generator) can be conducted.
The measurement equipment installed allows one to gather the following information: regu1a.-
ting distance with the aid of an inductive displacement sensor 7, and oil pressure in the cha.rnbess
by piezo electric pressure sensors 8. Additional information is given by acceleration pick ups (not
shown) mounted on the distance rod and by strain gauges t o determine the strain of the membranes.
A special hydraulic unit was designed to assure that the supply pressure is constant. This is
required because the fluid flow changes with frequencies up to 200 Hz and higher. These requirenlents
can be met with the aid of a controllable radial piston pump and an. appropriate reservoir.
The test rig shown in Fig. 5 is used to investigate different control concepts. It consists of an
elastic rotor structure 9, a magnetic bearing 10 for simulating different excitation forces, the bearing
r s A furthcl.
unit 11, and actuator system which is presently realized by electromagnetic a c t ~ a ~ t o12.
description of the test rig and the test results attained by this facilitya.~.e:given
e.g. in [:3].
 The active chamber system will be integrated in the test facility Fig. 5 after conclusion of the
experiments with the test rig Fig. 4. For this purpose the chambers for the horizontal direction liave
to be activated by a second servo valve,and the distance rod has to be replaced by the bearing unit.
Finally the electromagnetic actuators 12 are substituted by the active chamber system.

5 Transfer Characteristic

In the following the transfer functions 3;, and Fp. eq. 16 and 17,for the regulating distance X L
and the chamber pressure APv will be investigated. The goal is to compare theoretical and experi-
mental behavior of the actuator and to determine the system parameters. These are the l~lernbra.ne
stiffness CL, the leakage coefficient I(,,, the loop gain of the servo valve I{vand the coefficient It'*
which describes the oil compressibility and the membrane buckling. The simulation is based on the
data shown in Fig. 6 which are equivalent to the data of the test facility. The results can be seen in
Fig. 7.

5.1 Theoretical Results

In the plot, of the amplitude of the regulating distance in Fig. 7,one can recognize a PT1
behavior up to a frequency of about 170 Hz.Its cut-off frequency is very low (about 2.2 fIz) and is
given by z, see eq. 14. Beyond this frequency an extremum occurs in the amplitude at approximately
300 Hz.It corresponds to a pole in the transfer function of XL and APv. This extremum results fsoni
a function a ( s ) which can be obtained by the denominator of eq. 16. This function has the value
a(s) = 0 at the frequency of w, = 27r - 287 Hz (s = iw),

Eq. 19 shows that this frequency is mainly dependent on the membrane area A*, on the mass 172
(simulating the rotor mass including the bearing unit) and on the oil compressibility and membrane
buckling I{*. The servo valve has no influence. The expression can be interpreted as a spring
coefficient of the oil support and the!membranebuckling. This natural frequency causes a phase shift
of 180'. The rest of the curve is characterized by a drop in the amplitude due to the PT2-behavior
of ;the servo valve, see chapter 3.1.5. In the plot of the chamber pressure a further zero point
can ,be seen at 110 Hz. It corresponds to the natural frequency of the one-DOF-oscillator, function
P ( s ) from eq. 17

This characteristic frequency has no effect on the regulating distance and on the controllability of
the rotor system.
5.2 Comparison of Theory and Experiment

Fig. 8 shows the measurements of the regulating distance XL and chamber pressure APv. The
results are obtained by noise excitation for the input voltage Uv.
A comparison with the results attained by theory shows good agreement. The characteristic
frequencies at 110 Hz and 293 Hz are due to the effects described above (functions a ( s ) and P ( s ) ,
eq. 19 and 21). This shows that all important influences on the transfer characteristic are incluclecl
in the transfer functions eq. 16 and 17. The step shaped curve at lower frequencies is only caused
by the discrete evaluation of the output signals by the measurement electronics.
Measuremellts of the transfer characteristic of the servo valve (described by a PTz-system)
showed a better amplitude behavior at frequencies beyond the cut-off frequency than predicted by
the PT2-behavior.
The experiments have shown that the simplified equations for describing the performa.nce of'
the active chamber system, eq. 14, are valid up to the cut-off frequency of the servo valve (here
about 200 Hz), see chapter 3.1.5. For a higher frequency range it is necessary to use a servo va.lve
with a higher cut-off frequency and to minimize the oil compressibility and the membrane bucltling
by minimizing the oil volume and by an appropriate design of the membrane.

6 Example of Application of the Actuator

6.1 Formulation of the Complete System

In designing a controller, a complete mathematical formulation of the system is

assumeld. A very efficient method of describing the rotor system;to be controlled is modeling it as a,
hybrid multibody system (see [6]). This leads to the second order differential equation

with the vector q t IRf of the generalized minimal coordinates (f is the number of degrees of freecloln),
Mt is the mass matrix, P t EXftf is the matrix of velocity-proportional forces, Q t lTCfljis the
matrix taking into account the displacement-proportional forces, hi t EXf is the vector of the i-t,h
external force (e.g. control force), R is the rotor frequency and t indicates the time.
The control forces acting indirectly (via the rigid bearing unit) on the rotor can be expressc-:d
by

where J L t IR2'f is the Jacobian matrix of translation belonging to the subsystem bearing unit alld
I?,* and F,* the control forces given by differential eq. 13,actingin the x- and y-directions. The radial
movement of the bearing described by XL and y~ (perpendicular to xL) can be formulatecl with the
aid of the Jacobian-matrix as

Introducing an enlarged state space vector

with Qvx,Qvy,asthevalve~flow.serving,the x- and:y-directionis ,.respectively (eq. l),and APvx7APvy a.s

the valve output pressure supplying the x- and y-directions, respectively, welohtain the sta.t,eequa.tion

where A represents the system matrix, B the control matrix and

the control vector consisting of the control voltages Uvx proper for x- and Uv, proper for the y-
direction.
The matrices used in eq. 26 can be expressed as follows

where the submatrix Q is

-
Q =Q+ J;J~-CL .
CLis the membrane stiffness appearing in eq. 13 and M , P , Q are the system matrices from eq. 22.
The matrix E j E m?.?j represents the identity matrix.

6.2 Amplitude Characteristic of the Rotor

A simple rotor running in journal bearings is used. The rotor is supported by the active
chamber system on its left side (Fig. 9). In the modeling of the rotor system.mass-, stiffness-, ancl
gyroscopic effects and additional damping by the journal bearings are considered. As admissible
shape functions only two mode shapes are taken into account. In most cases higher modes of real
rotor systems are damped very well because of bearings or material damping.
 In Fig. 10 are plotted frequency response functions versus rotor frequency. The exciting force
is caused by an unbalance of the rotor. In Fig. 10a the scaled amplitudes & at the runner position are
displayed without active forces (solid line) and with active forces (dotted line). Fig. lob
shows the scaled nutation angles 5 (slope of the elastic line) at the same location. The displaceuncnt,
amplitude shows a significant peak at the first natural frequency (bending vibration) and at the
second natural frequency (nutation, indicated by the nutation angle). The feedbacli is realized by
displacement, velocity, and pressure signals taken only at the actuator location. The controller was
designed as a "constant controller" for the rotor frequency 100 Hz. Using this very simple controller,
the results already demonstrate that a considerable reduction of the resonance amplitudes can be
achieved.

Conclusion Remarks
The key to attaining the desired influence on the dynamics of rotor systems lies in the availa-
bility of suitable actuators. The development of the active chamber systems introduced in this papel.
will be an essential step towards industrial applications. This actuator works as a spring (n~ernbsane
stiffness) and damper (hydraulic effects in the servo valve),and in addition to these, passive.forcesa
control force is acting proportional to the input voltage at the servo valve. One of the main goals oC
the paper is to confirm the theoretical model of this new actuator concept by experiment. The theose-
tical and experimental results achieved by these investigations show a very good conformity. Further
experimental investigations of the actuator system included in the rotor system will demonstsate the
usefulness of applying this active machine element for active vibration control of rotors. The usc of
a more sophisticated controller (additional measurement information, adaptive controller realized by
special electronic equipment, hybrid controller) will show a wide range of possible applications for
such an actuator device.

References
[I] ANDERSON, W.R.: Controlling Electrohydraulic Systems. Marcel Dekker, Inc. New Yosl.; ancl
Basel, 1988.
[2] THAYER, W.J.: Transfer Functions for Moog Servo Valves. Technical bulletin 103, Moog Inc.
Controls Devision, East Aurora, N.Y. 14052, 1965.
[3] FURST, S.; ULBRICH, H.: An Active Support System for Rotors with Oil- Film Bearing. 4th
International Conference on Vibration in Rotation Machinery of the Institution of Mechanical
Engineers, Edinburgh, Scotland. Paper C 261188, 1988, pp. 61-68.
[4] INNERHOFER, G; HAMMER, J.: Low-Cost Magnetic Bearing Reaction Wheel. Proc. of the
IFAC Symp., Oxford, England, 1980.
[5] LIARD,G.: Aktive Magnetlager - ein Schliissel zu hoheren Spindeldrehzahlen. I<ugella.gcr-
Zeitschrift 56, 213, 1982, pp. 8 - 13.

[GI ULBRICH, H.: Dynamik und Regelung von Rotorsystemen. Fortschr.-Ber. VDI-Z, Reihe 11,
NR. 86, 1986.
[7] ULBRICH, H; ALTHAUS, J.: Actuator Design for Rotor Control. Twelfth biennial ASME Con-
ference on Mechanical Vibration and Noise , Montreal, Canada, Sept. 17-20, 1989.
[8] TIMOSHENKO, S.; WOINOWSKY-KRIEGER, S.: Theorie of Plates and Shells. New Yorlc,
Toronto, London, McGraw-Hill Book Company, 1959.

Figure 1: Active chamber system with servo valve

Figure 2: Fluid flow through the valve under load

280
 F.1 1 XL

Figure 3: Model of the chamber and elastic membrane

Figure 4: Test rig for active chamber system

28 1
OWEGOFSAL PAGE
BLACK AND WH1TE PHOTOGRAPH
 6RIGBNAE PAGE
BLACK AND WHITE PWOlOGRAPh

Figure 5: Test rig for rotor control

damping ratio of the servo valve

natural frequency of the servo valve
leakage coefficient
loop gain of the servo valve
characteristic membrane area
compressibility and membrane buckling
membrane stiffness
mass
amplitude of the input voltage

Figure 6: Data for the transfer characteristic of the actuator

282
 pltasc r,, [degree]

-100

-200

-300-
3.16 10.0 31.6 100.0 316.2
frequency [Iiz]
phase APII [degree]

loo--

r t : , ~ ~ ~ : e c . ' : ' G ' " ' ~ " :

3.16 10.0 31.6 100.0 316.2

frcquency [Hz]

Figure 7: Simulated transfer charateristic

dB

10 - 293 Hz

regulating distance XL

-10

-30 - 1
frequency [Hz] I

0 -
chamber pressure 4PL

-40 - 1

Figure 8: Measurement of the amplitude characteristic

283
 L- - -A- -'-l
Active Chamber System

Figure 9: Geometry and mechanical model of the investigated active rotor system

0 200 400 600 800 1000 1200

-without control rotor speed [radls]
--- with control
Figure 10: Response of the unbalanced rotor running through the critical speeds
a) runner displacement
b) nutation angle