See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245389571

Development and experimental validation of real-time executable models of

primary fly-by-wire actuators

Article in Proceedings of the Institution of Mechanical Engineers Part I Journal of Systems and Control Engineering · June 2008
DOI: 10.1243/09596518JSCE546

CITATIONS READS

11 895

3 authors, including:

Gianpietro Di Rito Eugenio Denti

Università di Pisa Università di Pisa
47 PUBLICATIONS 257 CITATIONS 18 PUBLICATIONS 121 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Study of hybrid laminates (piezoelectric patches & composite leminates) View project

Development of Algorithms for the Health-Monitoring of Electro-Mechanical Actuators for Aircraft Applications View project

All content following this page was uploaded by Gianpietro Di Rito on 18 November 2015.

The user has requested enhancement of the downloaded file.

 Proceedings of the Institution of Mechanical
Engineers, Part I: Journal of Systems and
Control Engineering
http://pii.sagepub.com/

Development and experimental validation of real-time executable models of primary fly-by-wire

actuators
G Di Rito, E Denti and R Galatolo
Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering 2008 222: 523
DOI: 10.1243/09596518JSCE546

The online version of this article can be found at:

http://pii.sagepub.com/content/222/6/523

Published by:

http://www.sagepublications.com

On behalf of:

Institution of Mechanical Engineers

Additional services and information for Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control
Engineering can be found at:

Email Alerts: http://pii.sagepub.com/cgi/alerts

Subscriptions: http://pii.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://pii.sagepub.com/content/222/6/523.refs.html

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 523

Development and experimental validation of real-time

executable models of primary fly-by-wire actuators
G Di Rito*, E Denti, and R Galatolo
Department of Aerospace Engineering, University of Pisa, Pisa, Italy

The manuscript was received on 27 December 2007 and was accepted after revision for publication on 5 June 2008.

DOI: 10.1243/09596518JSCE546

Abstract: The availability of practical real-time models of primary flight actuators is a key
aspect for performing hardware-in-the-loop simulations of fly-by-wire flight control systems.
The solution could be offered by empirical models tuned on experimental data, but this
approach would imply that hardware-in-the-loop simulations could be performed only after
the actuators have been designed, constructed, and tested. The alternative approach, to
which this work refers, is to develop high-fidelity actuator models based on the component
physics and to reduce their complexity by trying to obtain a compromise between accuracy of
results and real-time execution requirements. In the paper, real-time models of a servo-
hydraulic actuator for primary flight controls are developed, taking into account the basic
features of the fly-by-wire implementation (the non-linear direct-drive servovalve dynamics,
the structural compliance, the oil compressibility, the saturation of commands, and the digital
controls) as well as other physical phenomena, which are often disregarded in hydraulic
actuator modelling (the hinge play, the flow forces on the servovalve spool, and the laminar
servovalve flow). The simulation results of models characterized by different levels of
complexity are compared with experimental data obtained by testing the aileron actuator of
a modern fly-by-wire aircraft, and the relative importance of the model characteristics is
highlighted and discussed, providing useful guidelines about actuator model reduction for real-
time applications.

Keywords: flight control systems, hydraulic actuators, modelling, real-time simulation,

hardware-in-the-loop testing

1 INTRODUCTION In the modern design approach, most of the

integration problems are determined and solved
Traditionally, the design of the aircraft systems well before the first flight, thanks to the large use of
(power plants, actuators, sensors, etc.) was carried simulation. As a consequence, there is a strong
out separately from that of the aircraft itself demand for accurate models able to predict the
(structure, propulsion, aerodynamics, control laws), dynamic behaviour of the various aircraft systems as
using simplified models for a rough estimation of the well as the interactions between them.
integrated system performances. The actual system The use of simulation for performing integration
interactions were evaluated and examined by studies on aircraft systems plays a key role in the
ground tests or even during flight tests. Of course, project of modern fly-by-wire (FBW) aircraft, whose
any problem or malfunction at this stage of the redundant flight control systems (FCSs) are char-
project inevitably causes a standstill of work with a acterized by a high level of sophistication [1]. In this
strong increase of time and cost. context, many efforts are made for developing
practical real-time executable models of the critical
FCS subsystems. The real-time models are used for
*Corresponding author: Department of Aerospace Engineering, in-flight condition monitoring [2] as well as for
University of Pisa, via Caruso, Pisa, Italy. email: g.dirito@ing. performing real-time/hardware-in-the-loop (RT/
unipi.it HIL) simulations [3], in which one or more FCS

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

524 G Di Rito, E Denti, and R Galatolo

components are integrated with real-time models of pressures). Concerning the simulation sampling
the whole aircraft (flight mechanics, propulsion, rate, the basic constraint has been the capability of
power plants, and all the other FCS subsystems that the models to be executed in real-time on common-
are not present in the test rig) for obtaining a off-the-shelf PCs. This choice, essentially driven by
preliminary performance characterization of the FCS the equipment that was available for the research,
as well as for assessing the FCS capability to detect, implies a wide applicability of the proposed results.
isolate, and compensate for the malfunctions of its Previous works of the authors [4–6] have been
critical subsystems. focused on the development of an RT/HIL simula-
The major problems connected with the real-time tion system for modern FBW/FCS*, as well as on the
simulation of FBW/FCS are related to the develop- development and the validation of high-fidelity
ment of primary flight actuator models. The inherent models of primary flight actuators [7–10], while in
complexity of a primary FBW actuator (i.e. the this paper reduced real-time models are developed,
redundant hydraulic and electrical sections), its pointing out the most important model features for
characteristic non-linearities, and its intrinsic high- the achievement of a good compromise between the
frequency dynamics typically lead to models char- accuracy of results and the limitation of computing
acterized by extremely small integration steps, which resources.
are not suitable for real-time execution. A solution
could be offered by semi-empirical actuator models
tuned on experimental data, but this approach 2 ACTUATION SYSTEM MODEL
would imply that RT/HIL simulations and condition
monitoring studies could be performed only after The actuation system model has been developed
the actuators have been designed, constructed, and with reference to the aileron servo-hydraulic actua-
tested. The alternative approach, to which this work tor of a modern FBW aircraft, which has also been
refers, is to develop high-fidelity actuator models used for the experimental validation of the model
itself (see section 3). As shown in Fig. 1, the
based on the component physics and to reduce their
reference actuator is a valve-controlled tandem
complexity by trying to obtain a compromise
cylinder, in which the hydraulic flows coming from
between accuracy of results and real-time execution
two constant supply pressure plants are regulated by
requirements. In fact, the description of some
two four-way proportional valves with rigidly con-
physical phenomena involved in the actuator dy-
nected spools. Figure 2 reports a more detailed
namics is not fundamental for the simulation of
scheme of the actuator internal architecture, which
most operating conditions, and a model reduction is
is characterized by a direct-drive servovalve (DDV)
convenient for obtaining real-time executable codes.
with quadruple electrical redundancy that provides
Concerning this aspect, it must be noted that
the component with the capability of tolerating up to
different sampling-rate requirements have to be
two electrical failures. Concerning this aspect, it is to
considered, according to the use for which the
be outlined that the models described in this paper
real-time models are intended. Actually, an actuator
refer to the normal operative condition (i.e. no
model for in-flight condition monitoring must be
failures), but parallel research activities have been
real-time executable at the sampling rate of the FCS
carried out [9, 10] and are ongoing about the
control computers (in the range of 100 Hz), while the development of multistate models for the simulation
sampling rate of a real-time actuator model for RT/ of failures.
HIL simulations is generally driven by the specific
needs of design and analysis.
In the work, attention is focused on the RT/HIL 2.1 Objectives of the real-time actuator models
applications, and real-time actuator models are
developed in order to reproduce the basic response The real-time actuator models proposed in the paper
of the actuator (bandwidth, rise-time, etc.), its major have been developed with the following basic
non-linearities (the variation of dynamic response objectives:
with the command amplitude), and its most sig- (a) capability to reproduce the actuator bandwidth
nificant high-frequency dynamics. In addition, for (4 Hz in the reference application; see section
use in condition-monitoring studies, the models are 3);
developed in order to provide the simulation of the
main dynamic states of the actuator (current to the *The FCS of the advanced jet trainer Aermacchi M346 was chosen
servovalve, servovalve position, rod position, fluid as the reference architecture.

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 525

Fig. 1 Primary flight control actuation

Fig. 2 Internal architecture of a tandem DDV actuator

(b) capability to simulate the most significant high- The accuracy and the time cost of the real-time
frequency dynamics of the actuator (in the models will be evaluated on the basis of the response
reference application, the servovalve has a of an actual FBW actuator in the no-load condition,
closed-loop bandwidth of 50 Hz, while the first with the purpose of pointing out the relevance of
resonance related to the actuator servoelastic some modelling details for the achievement of a
interactions is located at 75 Hz); good compromise between the accuracy of results
(c) capability to simulate the digital loops char- and the limitation of computing resources.
acterizing the actuator position control (in the
reference application, the higher sampling rate
is related to the servovalve position control, 2.2 DDV dynamics
which is performed at 1200 Hz*); As shown in Fig. 2, the DDV is composed of a
(d) capability to reproduce the variation of the quadruple linear force motor (LFM), of the hydraulic
actuator response with the command amplitude section (spool and sleeve) and of the position
(an experimental evidence for the reference transducer (linear variable differential transformer,
application, see section 3); LVDT). When no current circulates in the LFM coils,
(e) capability to be executed in real-time on a the armature (rigidly linked to the valve spool) is
common off-the-shelf PC (the maximum allow- centred with respect to its endstrokes, as in this
able sampling rate is in the range of 10 kHz). condition the centring spring is unloaded and the
*This requirement leads to the need to develop the models with magnetic fluxes induced by the permanent magnets
sampling rates that are multiples of 1200 Hz. provide equal and opposite forces. When the current

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

526 G Di Rito, E Denti, and R Galatolo

is not zero, the magnetic flux induced by the coils dimensional correction factor (fv) that depends on
causes an unbalanced armature polarization, so that the spool position.
a magnetic force is generated and the spool moves Concerning the definition of the DDV/LFM model
towards a direction that depends on the sign of the parameters, the values of the correction factor fv
current. (Fig. 3) have been obtained experimentally by
Modelling the dynamics of a DDV/LFM is a measuring (thanks to the internal LVDT shown in
complex topic, since several non-linear phenomena Fig. 2) the steady state spool position at different
are involved. First, the magnetic force is a non-linear current inputs, without flow through the servovalve.
function of both the current and the spool position, Successively, in the same test conditions, the
so the polarization efficiency is considerably in- parameters a1, a2, and a3 appearing in equation (1)
creased if the armature approaches the endstroke have been determined using a least-square approx-
[8]. Secondly, the current to the coils of the DDV/ imation of the hardware open-loop frequency
LFM is typically provided by dedicated servo- response at 5 per cent of the maximum current
amplifiers, whose dynamics (based on a closed-loop command (in this condition, the motor essentially
control of the coil current) can significantly influ- works with constant current-to-displacement gain,
ence that of the servovalve [11]. Moreover, the effect Fig. 3). Finally, the DDV/LFM model has been
of the friction between the spool and the sleeve of validated by comparing the model and the hardware
the servovalve can be important for the description in terms of both step responses and frequency
of a small displacement dynamics such as that of the responses, using different values of the input
DDV [9, 12]. amplitudes (up to 30 per cent of the maximum
The authors developed an accurate model of the current) in order to provide a good coverage of the
dynamics of a fault-tolerant DDV/LFM [9], providing valve working range. Some results of the comparison
a detailed description of the system parts (electrical between model and hardware responses are shown
redundancy, magnetic flux generation, dynamics of in Figs 4 and 5, and prove that the DDV/LFM model
currents, equilibrium of momentum of the spool, is capable of reproducing the hardware response
failure simulation), but this model has not been used over a quite wide range of operating conditions. The
for the real-time simulation of the whole actuator, DDV/LFM response can be represented by a first-
since its complexity would imply the use of high order dynamics characterized by a corner frequency
computing resources and a strong limitation to the of 5 Hz and a settling time to the step response of
accuracy of the models of the other actuator parts 130 ms. In order to point out the importance of the
would result. On the other hand, too simple an use of the correction factor fv for reproducing the
approach, such as the linear models provided by the DDV/LFM non-linear behaviour, Fig. 4 also reports
classical literature [13], is not applicable if the the response of the linear model obtained from
complete working range of the servovalve (i.e. small equation (1) if fv is set to 1: the linear model exhibits
as well as large openings) is to be described. For
these reasons, a compromise solution is proposed,
using a simplified model that provides an overall
characterization of the DDV/LFM open-loop dy-
namics from the servoamplifier current demand
(icom) to the valve spool displacement (xv)




a3 x v za2 x€v za1 x_ v zxv ~Kv fv ðxv Þicom ð1Þ

The structure of the model was selected in order to

maintain the physical coherence, so the DDV/LFM
dynamics is assumed to be characterized by three
dynamic states: two states are related to the
dynamics of the spool mass linked to the centring
spring, with one related to the closed-loop current
dynamics implemented by the servoamplifier [9, 11,
13]. In addition, in order to take into account
the increase in the polarization efficiency as the
armature approaches the endstrokes, the current-to- Fig. 3 Correction factor of the DDV/LFM current to
displacement gain (Kv) has been multiplied to a non- displacement gain

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 527

Fig. 4 DDV/LFM open-loop step responses: comparison between simulation and experiments
(no flow regulation)

good results at 20 per cent of the maximum current, tor models, whose objective has been the matching
but is unable to match the hardware response at of the basic actuator response.
lower or higher commands, while the maxima errors
of the non-linear model (not visible in Fig. 4) are less
than 4 per cent of the spool full stroke.
2.3 Reynolds-dependent orifice equations
Two basic limitations of the model can be empha-
sized. Firstly, the parameter tuning and the valida- An appropriate modelling of the flow through the
tion of the model have been performed without flow servovalve orifices is crucial for the development of
regulation, so the flow forces are not taken into hydraulic actuator models. The traditional approach
account. As a second issue, the model gives only an to the modelling of orifice flows is based on
overall characterization of the servovalve dynamics pffiffiffiffiffiffiffiffiffiffiffi
without describing it by first principles, so a rigorous Qo ~Ko jDPo jsgnðDPo Þ ð2Þ
cause-to-effect analysis is difficult to realize.
Concerning the first point, since the effect of the pffiffiffiffiffiffiffiffi
Ko ~Cd 2=rAo ð3Þ
flow forces counteracts the motor force, it is
expected that, when the fluid flows through the
where the discharge coefficient Cd is assumed to be a
valve, the steady state asymptote of the open-loop
constant. The relationship between the flowrate and
frequency response lowers (Fig. 5). Thus, in order to
the square root of the pressure drop is applicable
provide an approximated estimation of the flow
only if the orifice flow is turbulent. If the flow is
effect in the whole actuator model, a correction
laminar (e.g. at small valve openings or at small
current, proportional to the flow force calculated
pressure drops) this relationship tends to be linear,
later by equation (15) is added to the command
and many authors propose the assumption that the
current (icom in equation (1)).
discharge coefficient is a function of the orifice
As a final consideration, the DDV/LFM model
Reynolds number [14–17]. Following this recom-
does not take into account the valve hysteresis. This
mendation, the authors selected the relation be-
choice depends on two reasons: first, as a result of
tween the orifice discharge coefficient and the
the model structure (no equation of momentum is
Reynolds number provided by
defined), it is difficult to introduce the friction
effects, to which the hysteresis is mainly related pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(section 2.7); secondly, the effects of the valve Cd ~Cd t tanh Reo =Ret ð4Þ
hysteresis on the whole actuator dynamics can be
important for the simulation of limit cycles, but
these phenomena have been considered of minor r Qo Lo
Reo ~ ð5Þ
importance for the development of real-time actua- mAo

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

528 G Di Rito, E Denti, and R Galatolo

Fig. 5 DDV/LFM open-loop frequency responses: comparison between simulation and

experiments (no flow regulation)

The combination of equations (2) and (3) with 2Lo Ao Cd2 t

equations (4) and (5) leads to a relationship between lim Qo % DPo ð7Þ
Reo ?0 m Ret
the flowrate and the pressure drop that is approxi-
mately linear if small Reynolds numbers are con- Thus, once j 5 1, 2, the index related to the hydraulic
sidered and tends to be quadratic if the Reynolds units, and i 5 1,…,4, the index related to the valve
number is sufficiently high. Actually, if Reo ap- ports (Fig. 2), are defined, the flow through the eight
proaches zero in equation (4), the discharge coeffi- orifices of the actuator servovalve is given by
cient is given by equation (6) and the orifice
qffiffiffiffiffiffiffiffiffiffiffiffi
   
equation is reduced to equation (7). On the other Qij ~Kij DPij sgn DPij ð8Þ
hand, if very high orifice Reynolds numbers are
considered in equation (4), the discharge coefficient qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffi 
is reduced to a constant and equations (2) and (3) Kij ~ 2=rAij Cd t tanh Reij Ret ð9Þ
collapse into the traditional model of turbulent
orifice flow  
2rQij 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
Reij ~  ð10Þ
lim Cd %Cd t Reo =Ret ð6Þ mw 
Reo ?0

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 529

Once known* the basic geometry of the orifices the spool dynamics. The total axial flow force is
(i.e. the orifice width w, the valve underlap u, and the given by the following equation, where a force
spool radial clearance cr), all the orifice areas (Aij) contribution is positive if directed in the same
can be calculated, under the hypothesis that the direction as the spool displacement (Fig. 2)
servovalve is symmetrical
(
2 X
X 4 2 X
X 4 rQ2ij
wcr , xv v{u Ffa ~ Ffaij ~ ð{1Þi cos hij ð15Þ

1 ðxv Þ~ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Cc Aij
A1j ~A3j ~A ð11Þ j~1 i~1 j~1 i~1
w ðxv zuÞ2 zcr2 , xv ¢{u
The quantities Cc and hij appearing in equation (15)
( qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 represent the area contraction coefficient and the jet

2 ðxv Þ~ w ðxv {uÞ zcr2 ,
A2j ~A4j ~A xv ¡u ð12Þ angle related to the generic servovalve port [14]. The
wcr , xv wu
jet angle hij is assumed to be a function of the axial
1
orifice opening Xv ij
Finally, the symmetry of the servovalve ports allows
two leakage functions K̄1 and K̄2 to be defined, which
1 uzXv if i~1, 3
can be used to calculate the eight leakage coeffi- Xvij ~ ð16Þ
cients of the servovalve u{Xv if i~2, 4

K11 ~K31 ~K12 ~K32 ~K
1 ðxv ,Re Þ according to the Von Mises relationship
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   

1 ðxv ÞCd t tanh 1
~ 2=rA Re=Ret ð13Þ Xvij 1zp=2sinhij {ln tan p=2{hij 2 coshij
~    ð17Þ
Cr 1zp=2coshij zln tan p=4{hij 2 sinhij
K21 ~K41 ~K22 ~K42 ~K
2 ðxv ,Re Þ
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi It must be mentioned that no invasive experiments
~ 2=rA
2 ðxv ÞCd t tanh Re=Ret ð14Þ have been performed on the servovalve, so the flow
force model has not been experimentally validated.
The quantities Cd t and Ret were used as tuning However, this fact does not imply a significant
parameters of the model, and their values were
limitation to the study: considering that the max-
determined experimentally by comparing simulation
imum flow force predicted by equation (15) is in the
and experimental results of actuator open-loop
order of 5 per cent of the maximum DDV/LFM force,
responses at different servovalve openings with no
load applied. In particular, the tests with large the flow force is not expected to be fundamental in
openings (i.e. high actuator velocities) were used to the servovalve dynamics (at least for normal opera-
calculate Cd t, while Ret was determined by tests with tive conditions). In addition, since the DDV/LFM
small valve openings (i.e. low actuator velocities). model provides only an overall dynamic character-
The results of the parameter tuning is shown in ization from the command current to the spool
Fig. 6, where the leakage functions K̄1 and K̄2 are displacement, without defining a moment equation,
plotted as functions of the spool position for the flow force cannot be directly introduced in the
different Reynolds numbers. model. Thus, the effect of the flow force was
reproduced by adding to the DDV/LFM command
a correction current that is proportional to the flow
2.4 Flow forces force calculated by equation (15). For these reasons,
The flow forces acting on the servovalve spool have the study of the flow force effects on the actuator
been modelled using a classical literature approach dynamics proposed in section 3.2 must be consid-
[14]. The flow force due to the generic servovalve ered as a rough sensitivity analysis.
port Ff ij is given by two contributions: Ffr ij (radial to
the spool axis) and Ffa ij (parallel to it). The latter
contribution is always applied in a direction that 2.5 Tandem cylinder dynamics and system
tends to diminish the orifice area, so that it counter- equations of motion
acts the DDV/LFM magnetic force, and it can affect The flowrates entering in the actuator chambers
(Qaj, Qbj) and the actuator pressures (Paj, Pbj) are
*The parameters can be determined using basic servovalve calculated (Fig. 2) by using the mass continuity
characteristics such as the no-load flow (for the orifice width) equations applied on both servovalve ports (equa-
and the valve pressure gain (for the underlap and the clearance). tions (18) and (19)) and actuator chambers (equa-

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

530 G Di Rito, E Denti, and R Galatolo

DDV servovalve leakage functions K̄1 and K̄2

Fig. 6

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 531

tions (20) and (21)), while the chamber volumes are the proposed models do not take into account the
calculated by equations (22) and (23). aerodynamic load simulation, and the term Hs only
provides structural reaction (see equation (30)).
Qaj ~Q1j {Q4j ð18Þ
2.6 Hinge play model
Qbj ~Q2j {Q3j ð19Þ
The calculation of the two structural reactions
  appearing in equations (24) to (26) was done taking
P_ aj ~b Qaj {Saj x_ Pr Vaj ð20Þ
into account the play on the hinges. This hinges
  effect was simulated using a model developed by the
P_ bj ~b Qbj zSbj x_ Pr Vbj ð21Þ authors, and the basic mechanical system depicted
in Fig. 7 is helpful to describe it. The mass m moves
Vaj ~Vaj0 zSaj xPr ð22Þ under the action of the external force Fext and of the
structural reaction Fstr according to
Vbj ~Vbj0 {Sbj xPr ð23Þ
€free ~Fext zFstr
mx ð27Þ
The actuator pressures are finally used in the
system equations of motion If there is play on the hinge, two distinct states can
be defined: if there is contact between the pivot and
Is €
ds ~Hs ð24Þ the sleeve of the hinge, the structural deformation
(xstr) is imposed by the mass motion (xfree) and the
X
2   structure (on condition that the force Fstr acts in
€ h zx
m P ðx €Pr Þ~ Paj Saj {Pbj Sbj zHs =hr ð25Þ such a direction that the pivot is pushed on the
j~1
sleeve) reacts with damping and stiffness (28a). If
there is no contact, the structural reaction Fstr is set
X
2  
mh x€h ~{ Paj Saj {Pbj Sbj zFh ð26Þ to zero and the deformation xstr is regulated by a
j~1 first-order linear dynamics (28b)
which were obtained by neglecting friction and by
if jxfree {xstr j¢xPL
representing the actuation system as a three-degree-
of-freedom mechanical system whose dynamic x_ str ~x_ free
states are the absolute displacement of the actuator xstr ~xfree {xPL sgnðxfree Þ
housing (xh), the relative displacement of the piston (
{Cstr x_ str {Kstr xstr if sgnðCstr x_ str zKstr xstr Þ~sgnðxstr Þ
with respect to the housing (xPr), and the control Fstr ~
0 if sgnðCstr x_ str zKstr xstr Þ=sgnðxstr Þ
surface deflection (ds)*.
ð28aÞ
28(a)
The lever arm hr, which can generally vary with the
actuator stroke, has been assumed constant in the else
model, since this is the typical result of the kine-
(
matics design (in the reference application, the x_ str ~{ðKstr =Cstr Þxstr
maximum variation of the lever arm, corresponding 28(b)
ð28bÞ
Fstr ~0
to the actuator full stroke, is less than 5 per cent).
Thus, once known the structural reaction on the
actuator housing Fh and the hinge moment applied
on the control surface Hs (a term that includes both
the external torque due to the aerodynamic loads
and the structural reaction of the link between the
actuator rod and the control surface), the actuation
system dynamics is completely determined.
Concerning the hinge moment Hs, since the experi-
ments for the model validation (section 3.2) were
performed with no load applied on the flight actuator,

*The formulation is valid if the inertial loads induced by the

aircraft motion and by the oscillatory motion of the actuator
housing about its hinge line are negligible. Fig. 7 Basic dynamic system with hinge play

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

532 G Di Rito, E Denti, and R Galatolo

Thus, once eh and eP, the plays on the actuator important for the simulation of disturbance dy-
hinges, are defined, the structural reactions appear- namics such as limit cycles, hysteresis, and stick–slip
ing in equations (24) to (26) are given by motions. Therefore, specific research activities are
ongoing for the optimization of the DDV/LFM model
Fh ~{Ch x_ h {Kh ½xh {eh sgnðxh Þ ð29Þ developed in reference [9], in order to obtain a real-
time model of the whole actuator that also takes into
  account the friction effects.
Hs ~{Cs d_ s {d_ s d {Ks ðds {ds d Þ ð30Þ

2.8 Multirate position control

where the quantity ds d is expressed as a function of
the actuator piston displacement Modern DDV actuators for FBW/FCS are character-
ized by two closed-loop position controls: one acting
xPr zxh {eP sgnðxPr zxh Þ on the DDV spool and another on the actuator rod.
ds d ~ds 0 { ð31Þ
hr These two control loops are typically performed by
in which ds0 is the control surface deflection at means of digital actuator interface module (AIM)
xPr 5 0 and xh 5 0. cards, which are part of the flight control computer
Concerning the tuning of the parameters of the (FCC) and, since the DDV needs a bandwidth larger
hinge play model, some of them, such as the than that of the actuator, are usually executed at
stiffness of the actuator junctions (Ks and Kh) or different sample rates.
the play on actuator hinges (eP and eh), have been set In order to reproduce this feature, digital closed-
to the nominal characteristics of the actuator test rig loop position controls have been developed in the
used for the model validation, while the damping of Matlab/Simulink environment by using discrete trans-
the actuator junctions have been calculated theore- fer functions [19]. As shown in Fig. 8, the DDV
tically, by assuming that the mass of the actuator controller is proportional, with a lag–lead filter for
housing and the surface inertia oscillate on the the attenuation of the high-frequency components of
related structural spring with a damping factor of the spool dynamic response, while the actuator
0.05 (the value is typically assumed in the vibration controller has a similar structure, but there is also a
analysis of metallic structures [18]). notch filter specifically designed for the attenuation of
the high-frequency disturbances due to servoelastic
interactions [20]. The sampling rates used for the
digital controls have been set to 600 Hz for the ram
2.7 Sliding friction and 1200 Hz for the DDV, with reference to the digital
The simulation of the sliding friction is often achieved controllers of the primary flight actuators of a modern
by means of complicated models characterized by FBW jet trainer. Moreover, in order to reproduce the
event-driven dynamics [12], which allow three differ- typical command filtering implemented by the AIM
ent phases of motion to be described: the sticking, the cards, a rate-limiter function and a voltage saturation
slipping, and the transition between them. When this were respectively inserted in the actuator and in the
approach is used for friction simulation on systems DDV command paths (Fig. 8).
characterized by low mass, high-frequency dynamics The multirate control was used for the control of
(and simulation problems) arise. Actually, in refer- the simulated actuator as well as for the actual
ences [9] and [10] the authors used a single-state actuator, during the tests for the experimental
elastoplastic model for the simulation of the sliding validation of the model described in section 3.2.
friction acting on the DDV spool, obtaining a stable
and accurate numerical solution at a sampling rate of
2.9 Real-time models with different levels of
48 kHz. This implies that the model of the whole flight
complexity
actuator is not real-time executable on a common off-
the-shelf PC, while the DDV/LFM model is real-time In order to evaluate the relative importance of the
executable, but one PC must be exclusively dedicated actuator model features for a practical real-time
to the servovalve simulation. For these reasons, the simulation, five models characterized by different
sliding friction simulation has not been considered in levels of complexity have been developed (Table 1).
this work. All the models were derived from a basic version
However, as described in reference [10] by means (Model 1) that includes the main features of the
of a non-real-time simulation, the friction effects are actual implementation (the DDV/LFM dynamics,

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 533

the structural flexibility, the oil compressibility, the

saturation of commands, and the digital controls).
Starting from the Model 1, more complex versions
were obtained by adding the hinge play, the flow
forces, and the laminar servovalve flow, which are
often disregarded in the modelling of hydraulic
actuators.

2.10 Performances of the models

All the actuator models reported in Table 1 have
been developed in the Matlab/Simulink environ-
ment, the related real-time executable codes have
been obtained by means of the xPC Target tools, and
a common off-the-shelf PC (AMD Athlon 2.1 GHz)
was used for the real-time execution.
The selection of numerical parameters such as the
solver algorithm and the integration step has been
obtained by performing an extensive simulation
Multirate actuator position control

campaign, using different solver configurations. The

results of the models have been compared in terms of
accuracy (using the actual actuator response; see
section 3) and time cost. It must be mentioned that, in
order to evaluate the time cost correctly for the use of
the models in RT/HIL condition monitoring studies,
eight signals have been acquired at 1200 Hz during the
real-time simulations: the actuator demand, the
actuator position, the servovalve position, the servo-
valve current, and four actuator pressures.
Table 2 synthetically reports, for each of the five
models and for different solver algorithms, the average
total elapsed time (ATET) obtained with a reference
Fig. 8

integration sampling rate of 12 kHz (step of 83.3 ms)

and a minimum stability sampling rate (SSRmin). It
can be seen that the ATET increases if the solver
passes from the Bogacki–Shampine method to the
Dormand–Prince method, while the SSRmin decreases.
This is the consequence of increasing solver
complexity, which is detrimental for the time cost as
well as beneficial for numerical stability.
Concerning the effects of the detailed model
features on the overall numerical performances,
Table 2 points out that the time cost of the model

Table 1 Actuator models with different levels of

complexity
Reynolds-dependent
Hinge play Flow forces orifice equations
Model 1 No No No
Model 2 Yes No No
Model 3 No Yes No
Model 4 No No Yes
Model 5 Yes Yes Yes

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

534 G Di Rito, E Denti, and R Galatolo

Table 2 Performances of the real-time actuator models

Bogacki–Shampine Runge–Kutta Dormand–Prince
ATET at 12 kHz (ms) SSRmin (kHz) ATET at 12 kHz (ms) SSRmin (kHz) ATET at 12 kHz (ms) SSRmin (kHz)
Model 1 13.8 4.8 16.3 3.6 23.5 3.6
Model 2 14.3 6 16.6 4.8 24.0 4.8
Model 3 19.3 4.8 24.2 3.6 33.1 3.6
Model 4 16.6 4.8 19.6 3.6 25.8 3.6
Model 5 22.0 6 26.0 4.8 34.8 6

only including the simulation of the hinge plays for the experimental validation of the actuator
(Model 2) is almost the same of that of the basic models developed during the research. The RT/HIL
model (Model 1), but the hinge plays have a negative simulations are performed using a system of PCs
impact on the numerical stability, since event-driven networked with a broadcast memory. The system of
dynamics are introduced (section 2.6). With respect to PCs is made of a host PC, which is used for the
the basic model, Model 3 shows a notable increase in model prototyping in the Matlab/Simulink environ-
the ATET, due to the significant number of calcula- ment as well as for the generation of the executable
tions related to the flow force model (section 2.4). On code by means of the xPC Target tools, and seven
the other hand, the addition of the Reynolds-depen- Target PCs, which are used for real-time code
dent orifice equations, although implying the solution execution [6].
of eight algebraic loops (equations (8) to (10)), leads to The plant used for the experimental validation of
a negligible increase in computing resources. the real-time models includes some of the elements
The choice of the solver for real-time models was of the whole RT/HIL system, and is composed of a
done in the perspective of RT/HIL applications, host PC, a TCP/IP hub, one of the Target PCs (used
aiming to reduce the number of Target PCs to be for the flight actuator control), and the hydraulic rig
used for the simulation of a complete FBW/FCS. For (Fig. 9).
this reason, the Runge–Kutta method was considered The experiments for the model validation have been
to be the best solution, since it provides simulations performed in the no-load condition (note, in Fig. 9,
that are more robust than the Bogacki–Shampine that the flight actuator is not connected to the load
method and is less time-consuming than the Dor- actuator). This choice depends on the fact that the
mand–Prince method. characterization of the actuator response has been
Thus, once the Runge–Kutta method is selected, a done with respect to the relatively small displace-
real-time simulation of two actuators including ments, in which the effects of the aerodynamic load
hinge play, flow forces, and the Reynolds-dependent can be considered negligible. In addition, the no-load
orifice equations (Model 5) can be executed on the condition is a typical configuration used for the
same PC. The number of actuator models running experimental characterization of a position-controlled
on one PC rises to three if hinge play and flow forces actuator, since the system exhibits the minimum
are neglected (Model 4), and to four if Reynolds- damping [14]. However, before using the models in
dependent orifice equations are also eliminated the RT/HIL simulations, the validation will be com-
(Model 1), but this choice, as discussed in section pleted by also including the loaded conditions.
3.2, would lead to unreliable results.

3.2 Comparison between simulation and

3 EXPERIMENTAL VALIDATION OF THE experiments
MODELS
Two test campaigns have been performed, using a
tandem servo-hydraulic actuator employed for the
3.1 Test equipment
aileron and the rudder of a modern FBW jet trainer.
In the laboratories of the Department of Aerospace The first test session has been dedicated to the DDV/
Engineering of the University of Pisa (DAEUP), a RT/ LFM (section 2.2), the second one to the whole
HIL simulation system for FBW/FCS has been actuator.
recently set up [4–6]. The equipment includes a Frequency response and step response tests were
hydraulic test rig for primary flight control actuators, performed on the flight actuator without external
which can be used for RT/HIL simulations as well as load, using different input amplitudes. In particular,

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 535

Fig. 9 Test equipment for the experimental validation

the frequency responses (obtained by applying the introduced (Model 2), so this feature has been
Fourier transform to the actuator output signal) were considered negligible at this stage of the research*
performed by commanding the actuator at 0.5 and 5 (Fig. 11). The same consideration can be made if
per cent of the maximum demand, with reference to only the flow force model is added (Model 3)
FCS requirements related to different aircraft control (Fig. 12).
laws functions (e.g. augmented stability and autopi- The quality of results is strongly enhanced if the
lot functions). Input amplitudes of 1 and 10 per cent Reynolds-dependent orifice equations are intro-
of the maximum demand were used for the step duced (Model 4), obtaining a model that is capable
responses, in order to obtain the model validation of satisfactorily reproducing the hardware response
also in ‘off-design’ conditions. for both large and small commands (Fig. 13). Model
Experiments highlighted a significant variation of 4 succeeds in simulating the bandwidth dependence
the actuator response with respect to the input from the input amplitude essentially because the
amplitude: the bandwidth ranges from 2 to 4 Hz if servovalve leakage coefficients vary with respect to
the input amplitude ranges from 0.5 to 5 per cent of the orifice Reynolds numbers. Actually, when the
the maximum command (Figs 10 to 14). This non- actuator is commanded with large input amplitudes
linear behaviour can negatively affect the robustness (more than 2 per cent of the maximum command)
of the aircraft control laws, since the actuator response the servovalve openings and the flowrates entering
changes with respect to the flight manoeuvres. and coming out from the actuator are sufficiently
As shown in Fig. 10, the results of Model 1 match high, so that a turbulent flow is established in the
very well the experimental data if the response to large orifices, resulting in a high servovalve flow gain
commands is considered. On the other hand, the (Fig. 6). On the other hand, when the actuator is
simulation results at small commands are completely commanded with small input amplitudes (less than
different from the experimental data. This highlights 1 per cent of the maximum command), the Reynolds
the fact that, for reproducing the actuator response to number in the orifices is strongly reduced, resulting
large commands, the simulation of hinge plays, flow
forces, and laminar servovalve flow are not important,
while some of them can be fundamental if the *The simulation of the hinge plays can be relevant for the study
actuator moves with small displacements. of aeroservoelastic interactions, which will be considered in
No significant differences in the actuator response future development of the research, when the validation of the
can be observed if the hinge play model is actuator models in a loaded condition will be carried out.

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

536 G Di Rito, E Denti, and R Galatolo

Actuator frequency response: comparison between experiments and simulation (Model 1)

Actuator frequency response: comparison between experiments and simulation (Model 2)

Fig. 10

Fig. 11

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 537

Fig. 12 Actuator frequency response: comparison between experiments and simulation (Model 3)

Actuator frequency response: comparison between experiments and simulation (Model 4)

Fig. 13

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

538 G Di Rito, E Denti, and R Galatolo

in a significant decrease of the servovalve flow gain

and, as a consequence, of the actuator bandwidth.
Therefore, as far as the basic actuator response is
concerned, the Reynolds-dependent orifice equa-
tions must be considered an important model
feature, while the hinge plays and the flow force
can be regarded as minor effects. Figure 14 provides
a further proof of this conclusion, since the
frequency response of Model 5 provides results very
similar to those obtained by Model 4.

Actuator frequency response: comparison between experiments and simulation (Model 5)

Finally, Fig. 15 shows a comparison between the
step responses of the actual flight actuator and those
predicted by Model 4. It can be seen that the model
is capable of reproducing satisfactorily the transient
dynamics at the piston as well as at the servovalve
level.

3.3 Limitations of the models and future

developments of the research
The basic limitation of the proposed real-time
actuator models is that the validation has been
performed only with reference to no-load condi-
tions. Before the use in RT/HIL simulations, the
validation should also be completed including the
loaded conditions, since some model features that
appear negligible without external load can be
important when the load is applied. As an example,
the hinge plays are essential for the study of
aeroservoelastic interactions, since they imply the
presence of a ‘dead-zone’ in the structural reaction
that can induce a diminution of the flutter velocity
[21]. For these reasons, parallel research activities
have been carried out on the experimental char-
acterization of the dynamic stiffness of the actual
actuator [22] as well as on the design and imple-
mentation of a force control system on the DAEUP
hydraulic test rig [23]. Further developments of the
work will be focused on the validation of the stiffness
response of the real-time actuator models and to a
final validation by means of flight manoeuvre
Fig. 14

simulations during RT/HIL tests.

In addition, it must be mentioned that the
actuator models have been validated with reference
to the normal operative condition, so further studies
will be carried out for the development of real-time
models for fail-operative conditions. For instance, it
is expected that the flow forces can be important if
one or more electrical failures occur to the DDV/
LFM coils [9]. In these conditions, the DDV/LFM
force capability is lower and the effect of the flow
forces on the servovalve dynamics is stronger,
especially when the actuator moves at high velocity.

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 539

Fig. 15 Actuator step response (all demands given at 0.167 s): comparison between experiments and simulation (Model 4)

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

540 G Di Rito, E Denti, and R Galatolo

Concerning the proposed models, the most re- tic interactions can arise. In fact, the most remark-
markable limitation is related to the DDV/LFM able limitation of the proposed real-time actuator
model: it is semi-empirical and provides only an models is that the validation was performed only
overall dynamic characterization from the command with reference to no-load conditions. Before using
current to the valve spool displacement, but it does the models in the RT/HIL simulations, the validation
not allow the flow forces or the friction to be should be completed by also including the loaded
introduced appropriately (for this reason, the models conditions, and further developments of the re-
are not suitable for predicting limit cycles or search are ongoing in this direction.
hysteresis).
Finally, further simplifications of the actuator
model can be attempted by neglecting the oscillatory
REFERENCES
dynamics related to the actuator case and to the
actuator piston, which are respectively characterized 1 Collinson, R. P. G. Fly-by-wire flight control.
by pulsations of 450 and 1000 Hz, well outside the Computing and Control Engng J., 1999, 141–152.
frequency range of interest. 2 Bonnice, W. F. and Baker, W. Intelligent fault
diagnosis and failure management of flight control
actuation systems. NASA Contractor Report
4 CONCLUSION 177481, 1988.
3 Webster, F. and Smith, T. D. Flying qualities flight
Real-time models of a servo-hydraulic actuator for testing of digital flight control systems. RTO-AG-
primary flight controls are developed, taking into 300, Flight Test Techniques Series, Vol. 21, Canada
Communication Group Inc., 2001.
account the basic features of the fly-by-wire im-
4 Bernardini, M., Cecchelli, G., Denti, E., and
plementation (the structural compliance, the oil
Galatolo, R. A parallel computing network for the
compressibility, the non-linear direct-drive servo- real-time simulation of fly-by-wire flight control
valve dynamics, the saturation of commands, and systems (in Italian). In Proceedings of the 17th
the digital controls) as well as other physical AIDAA National Congress, Rome, Italy, 2003, vol. 1,
phenomena, which are often disregarded in hydrau- pp. 549–558.
lic actuator modelling (the hinge play, the flow 5 Denti, E. and Galatolo, R. A research laboratory for
forces, and the laminar servovalve flow). the study of fly-by-wire flight control systems (in
In order to evaluate the relative importance of Italian). In Proceedings of the 17th AIDAA National
Congress, Rome, Italy, 2003, vol. 2, pp. 997–1004.
the actuator model features for a practical real-
6 Denti, E., Di Rito, G., and Galatolo, R. Real-time
time simulation, models characterized by different
hardware-in-the-loop simulation of fly-by-wire
levels of complexity were developed in the Matlab/ flight control systems. In 25th International Con-
Simulink/xPC Target environment, and their perfor- gress of the Aeronautical Sciences (ICAS), Ham-
mances were evaluated on a common off-the-shelf burg, Germany, 2006.
PC in terms of time cost and accuracy (by comparing 7 Denti, E., Di Rito, G., and Galatolo, R. Fly-by-wire
the simulation results with the response of an actual flight control systems: modelling primary hydraulic
FBW actuator tested in no-load conditions). actuators with direct-drive servovalve (in Italian).
The study points out that the use of Reynolds- In Proceedings of the 17th AIDAA National Con-
gress, Rome, Italy, 2003, vol. 2, pp. 1005–1015.
dependent orifice equations for modelling the
8 Di Rito, G. Experimental validation of theoretical
servovalve flow (usually regarded as a minor effect)
and numerical models of a DDV linear force motor.
can be essential for reproducing the variation of the In Proceedings of the 3rd FPNI-PhD Symposium,
actuator response with respect to the command Terrassa, Spain, 2004, pp. 105–114.
amplitude. Other detailed model features, such as 9 Di Rito, G. and Galatolo, R. Modelling the
the hinge plays and the flow forces, were shown to dynamics of a direct-drive servovalve for fail-
be less important for the simulation of the basic operative primary flight control actuators. In
actuator response. Proceedings of the 3rd International Conference
This does not mean that hinge play and flow forces on Recent advances in aerospace actuation systems
and components, Toulouse, France, 2004, pp.
are not important at all. Actually, it is expected that
67–72.
the flow forces can be important if a fail-operative
10 Di Rito, G., Denti, E., and Galatolo, R. Experi-
condition is studied, when one or more electrical mental validation of the model of a fault-tolerant
failures occur to the DDV/LFM coils. On the other servo-hydraulic flight actuator for FBW/FCS (in
hand, the hinge plays can be important when a Italian). In Proceedings of the 18th AIDAA National
loaded condition is concerned, since aeroservoelas- Congress, Volterra, Italy, 2005.

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

 Development and experimental validation of real-time executable models 541

11 Lin, J. S. and Chen, C. L. Buck/boost servo- a2 coefficient of the second-order time

amplifier for direct-drive-valve actuation. IEEE derivative in the DDV spool dy-
Trans. Aerospace and Electronic System, 1995,
namics equation (s2/rad2)
31(3), 960–967.
12 Olsson, H., Åström, K. J., Canudas de Wit, C., a3 coefficient of the third-order time
Gäfvert, M., and Lischinsky, P. Friction models derivative in the DDV spool dy-
and friction compensation. Eur. J. Control, 1998, namics equation (s3/rad3)
4(4) Ac ij contracted orifice area for the ith
13 Fitzgerald, A. E., Kingsley, C. Jr, and Kusko, A. port of the jth hydraulic section (m2)
Electric machinery: the processes, devices and Aij orifice area for the ith port of the jth
systems of electromechanical energy conversion,
hydraulic section (m2)
3rd edition, 1971, (McGraw-Hill, New York).
14 Merritt, H. E. Hydraulic control systems, 1967 (John Ao orifice area (m2)
Wiley & Sons, Inc., New York). Ā1 orifice area (m2)
15 Borutzky, W., Barnard, B., and Thoma, J. An Ā2 orifice area (m2)
orifice flow model for laminar and turbulent cr valve radial clearance (m)
conditions. Simulation Modelling Practice and Cc orifice contraction coefficient
Theory, 2002, 10(3), 141–152.
Cd orifice discharge coefficient
16 Baudry, X. and Mare, G. C. Linking CFD and
lumped parameters analysis for the design of flow Cdt orifice discharge coefficient for tur-
compensated spool valves. In Proceedings of the bulent flow
1st FPNI-PhD Symposium, Hamburg, Germany, Ch structural damping of the junction
2000, pp. 249–258. between the actuator housing and
17 Di Rito, G. Experiments and CFD simulations for the aircraft structure (kg/s)
the characterisation of the orifice flow in a four-
Cs structural damping of the junction
way servovalve. Int. J. Fluid Power, 2007, 8(2),
37–46. between the actuator piston and the
18 Ashley, H. and Bispligkoff, R. L. Principles of control surface (kg m2/s rad)
aeroelasticity, 1962 (Dover Publications Inc., New Cstr structural damping (kg/s)
York). fv correction factor on the DDV/LFM
19 Franklin, G. F., Powell, D. J., and Workman, M. L. current-to-displacement gain
Digital control of dynamic systems, 2nd edition, Fext external force (kg m/s2)
1990 (Addison-Wesley Publishing Company, Bos-
ton, Massachusetts). Ff ij flow force related to the ith port of
20 Flight control design – best practices, RTO Tech- the jth hydraulic section (kg m/s2)
nical Report 29, Canada Communication Group Ffa ij axial component of Ff ij (kg m/s2)
Inc., 2000. Ffr ij radial component of Ff ij (kg m/s2)
21 Brase, L. O. and Eversman, W. Application of Fh structural reaction of the junction
transient aerodynamics to the structural nonlinear between the actuator housing and
flutter problem. J. Aircraft, 1988, 25(11), 1060–1068.
the aircraft structure (kg m/s2)
22 Di Rito, G. and Galatolo, R. Experimental char-
acterisation of the dynamic stiffness of a fly-by- Fstr structural reaction (kg m/s2)
wire actuator (in Italian). In Proceedings of the hr horn radius (m)
18th AIDAA National Congress, Volterra, Italy, Hs structural reaction of the junction
2007. between the actuator piston and the
23 Di Rito, G. Experimental implementation of a control surface (kg m2/s2)
motion-compensated force control in a hydraulic
i index related to the servovalve port
workbench for flight actuators. In Proceedings of
the Third International Conference on Recent icom command current (A)
advances in aerospace actuation systems and Is moment of inertia of the control
components, Toulouse, France, 2007, pp. 51–57. surface (kg m2)
j index related to the hydraulic section
Kh structural stiffness of the junction
APPENDIX between the actuator housing and
the aircraft structure (kg/s2)
Notation Kij leakage coefficient related to the ith
a1 coefficient of the first-order time port of the jth hydraulic section (m7/
derivative in the DDV spool dy- kg)1/2)
namics equation (s/rad) Ko leakage coefficient ((m7/kg)1/2)

JSCE546 F IMechE 2008 Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

542 G Di Rito, E Denti, and R Galatolo

Ks structural stiffness of the junction Vaj fluid volume in chamber a of the jth
between the actuator piston and the hydraulic section (m3)
control surface (kg m2/s2 rad) Vaj 0 initial value of Vaj (m3)
Kstr structural stiffness (kg/s2) Vbj fluid volume in chamber b of the jth
Kv current to displacement gain of the hydraulic section (m3)
DDV (m/A) Vbj 0 initial value of Vbj (m3)
K̄1 leakage function ((m7/kg)1/2) w orifice width (m)
K̄2 leakage function ((m7/kg)1/2) xh actuator housing displacement (m)
m mass (kg) xPL radial hinge play (m)
mh housing mass (kg) xP r piston displacement with respect to
mP piston mass (kg) the actuator housing (m)
Paj pressure in chamber a of the jth xstr structural deformation (m)
hydraulic section (kg/s2 m) xv valve spool displacement (m)
1
Pbj pressure in chamber b of the jth xv ij axial orifice opening in the ith port of
hydraulic section (kg/s2 m) the jth hydraulic section (m)
Qij flowrate entering ith port of the jth
hydraulic section (m3/s) b bulk modulus (kg/s2 m)
Qaj flowrate entering chamber a of the ds actual control surface deflection
jth hydraulic section (m3/s) (rad)
Qbj flowrate entering chamber b of the dsd demanded control surface deflection
jth hydraulic section (m3/s) (rad)
Qo flowrate through an orifice (m3/s) DPo orifice pressure drop (kg/s2 m)
Reij Reynolds number referred to the flow eh radial play in the hinge that links the
in the ith port of the jth hydraulic actuator housing to the aircraft struc-
section ture (m)
Reo orifice Reynolds number eP radial play in the hinge that links the
Ret transition Reynolds number actuator piston to the control surface
Saj piston area in chamber a of the jth (m)
hydraulic section (m2) hij jet angle of the flow in the ith port of
Sbj piston area in chamber b of the jth the jth hydraulic section (rad)
hydraulic section (m2) m oil absolute viscosity (kg/s m)
u valve underlap (m) r oil density (kg/m3)

Proc. IMechE Vol. 222 Part I: J. Systems and Control Engineering JSCE546 F IMechE 2008

Downloaded from pii.sagepub.com at Unipi / Celdes on July 15, 2011

View publication stats