BOND GRAPH MODELING OF A HYDRAULIC VIBRATION SYSTEM: SIMULATION

AND CONTROL

Farid Arvani, Geoff Rideout, Nick Krouglicof, Steve Butt

Faculty of Engineering, Memorial University of Newfoundland, St. John’s, Canada

(farvani, g.rideout, nickk, sdbutt) @mun.ca

ABSTRACT A typical hydraulic control system consists of a

power supply, a flow control valve, an actuator, a
Electro hydraulic drives are often used where the transducer and a servo controller for closed-loop
dynamic response requirements of the driven load in control. The control valve adjusts the flow of hydraulic
terms of speed, cycling, accuracy and stability are fluid into actuator chambers to move the actuator to the
particularly severe. One such application is the control desired position based on the control signal generated
and drive system associated with a high frequency by the servo controller. A force control scheme operates
hydraulic vibrator. A hydraulic shaker controls a valve similarly by using proper load transducer(s).
which directs high pressure hydraulic fluid to generate
controlled vibration. Hydraulic shakers typically have
long stroke and large force and are useful up to many
hundreds of Hertz vibration frequency.
In this paper, bond graph modeling is used to study and
simulate the effect of various components in a hydraulic
circuit on a shaker performance. The intention is that
the electrohydraulic system vibrates at a specified force
or displacement amplitude and frequency for the
controlled vibration of a load attached to the rod. This
paper describes the simulation and control of such
system. Feedforward PID control scheme is utilized in
the simulations to control the system and to track a
Figure 1: Schematics of the modeled hydraulic system
variable reference signal.
In this paper, the hydraulic dynamics of a valve-
Keywords: Hydraulics Actuator, Vibrator, Control
controlled hydraulic cylinder is developed and
System, Bond graph modeling
simulated, and a simplified model of a high frequency
hydraulic system is presented. The models are derived
1. INTRODUCTION
from first principles and implemented using bond graph
Hydraulic control systems are used to control the
methodologies. Finally, a control system is proposed to
position, speed or force of resisting loads. A linear
perform a tracking control scheme for a hydraulic
hydraulic cylinder or a rotary-motion hydraulic motor
shaker. Figure 1 depicts the schematics of the modeled
usually provides the final drive. A positive
system. Solid lines show the bidirectional hydraulic
displacement pump at a relatively high pressure
signals while dashed lines designate the electrical
between 7 and 35 MPa delivers the hydraulic fluid to
control signals. Servovalves and actuators are the most
the system to power the actuators.
important parts of this hydraulic system and therefore
Hydraulic control systems are used where
has been the main focus of this work. However other
relatively large forces and torques accompanied with
components have been modeled in detail as well.
fast, stiff response of resisting loads are required such as
in industrial presses, machine tools and flight
2. MODELING METHODOLOGY
simulators, to name a few. They are used for closed-
The simulated hydraulic circuit consists of the bond
loop control of a response variable(s) such as in aircraft
graph model of a closed–center four-way directional
and industrial robot control; and in manual control over
valve, hydraulic actuator, two pressure relief valves,
a substantial powered motion such as mobile equipment
three accumulators and four hydraulic hoses. 20-sim
and road vehicle steering (Merritt 1967).
package is used to create and simulate the bond graphs.
Understanding the hydraulic system dynamics
A tracking PID controller is implemented to perform
plays an important role in designing controllers for
position control on a linear actuator.
ensuring a good system response or in finding out why
an existing system is not performing satisfactorily
(Dransfield 1981).

281
2.1. Servovalve modeling
This section describes the methodologies to model a
servovalve which is generally comprised of a torque
motor and a valve spool assembly (Jelali and Kroll
2003; Dransfield 1981; Poley 2005). The servovalve
consists of a torque motor; application of electrical
signal to its coil will generate a torque. The flapper of
the torque meter is integrated with the spool assembly.
The flapper acts as a mechanical feedback for the
control of spool position (Maskrey 1978; Jones 1997;
DeRose 2003).
The servovalves can be modeled using two
approaches. The first method is to model the
subcomponents of the valve. This requires the design
parameters for the mechanical and electrical
components of the valve. It is possible to derive
meaningful transfer functions for electrohydraulic
Figure 2: Physical and schematic configuration of the
servovalves, and several papers have reported such
modeled valve (Johnson 2008)
work (Jelali and Kroll 2003). Unfortunately,
servovalves are complex devices and have many
The total valve coefficient is defined to be the flow
nonlinear characteristics which are significant in their
coefficient at the flow rating conditions:
operation. These nonlinearities include: electrical
hysteresis of the torque motor, change in torque-motor QN
output with displacement, change in orifice fluid (2)
PN
impedance with flow and with fluid characteristics,
change in orifice discharge coefficient with pressure It is necessary to know the coefficient of each of
ratio, sliding friction of the spool, and others (DeRose the four lands. The valve ratio is defined as
2003, Maskrey 1978).
Experience has shown that these nonlinear and (3)
non-ideal characteristics limit the usefulness of
theoretical analysis of servovalve dynamics in systems
design. Instead, the more meaningful approach is to where K VPL and K VRL have units of Holes and represent
approximate measured servovalve response with the valve coefficients for the pressure and return lines,
suitable transfer functions (Thayer 1965, Clarke 1969). respectively. It is safe to assume that valve ratio is
Moog has performed sophisticated analyses of unity in absence of evidence to the contrary as there is
servovalve dynamic response including computer no reason to deliberately design a non-symmetrical
simulations of various nonlinear effects, and up to valve. The relation between total valve coefficient and
eight-order dynamics (excluding any load dynamics). coefficients for valve lands is
The results indicate that these complex analyses have
not contributed significantly to servovalve design due to 1 (4)
uncertainties and inaccuracies associated with the
higher-order effects.
(5)
2.1.1. Servovalve Hydraulic Modeling
The actual flow through the valve is dependent upon The above equations are derived based on the fact
electrical command signal and valve pressure drop. The that two orifice resistances are in series and are related
flow for a given valve pressure drop can be calculated by
using the square root function for sharp edge orifices:
(6)
P
Q QN PN
(1)
It is noteworthy that these parameters take into
where Q [gpm] = calculated flow, QN [gpm] = rated account the spool geometry, discharge coefficient and
flow, P [psi] = actual valve pressure drop and PN oil density (Johnson 2008).
[psi] = rated valve pressure drop. The servovalve used Servovalve delivers a control flow proportional to
in these simulations is high performance Moog- D765 the spool position which is also proportional to the
series valves and the values for QN and PN are square root of the pressure drop across the valve.
provided in the valve datasheet. The modeled valve’s Combining equations (1) and (2) and comparing with
schematics (Johnson 2008) is shown in figure 2. the well known formula for flow–pressure relation for
turbulent flow,

282
 A typical performance graph for a high-response servo-
valve Moog D765 is shown in figure 4 where the
Q C A √ΔP (7) frequency response of the valve is plotted for different
command percentages, taken from the manufacturer’s
it is clear that provides us with the coefficient in the datasheet.
above formula easily which contains parameters for the
spool geometry, discharge coefficient and oil density.
Torque motor can be modeled as a series L-R
circuit, neglecting any back-EMF effects generated by
the load (Poley 2005). The transfer function is

I
(8)
V L R

where R is the combined resistance of the motor coil

and the current sensor resistor of the servo amplifier
and L is the inductance of the motor coil. Values of
these parameters for series and parallel wiring
configurations of the motor are published in the
manufacturer’s data sheet.
Control flow, input current and valve pressure drop
are related by
Figure 4: Frequency response of Moog D765
servovalve (Courtesy of Moog Inc.)
Q KV iV √ΔP (9)

where iV is the input command corresponding to a

A second order model yields an approximation to
specific valve opening and ΔP, pressure drop across the
actual behavior. Standard second order system
valve, is given by ΔP P PT PL where
identification of the frequency response can be
P , PT and PL are system pressure, return line pressure
performed where suitable values for natural frequency
and load pressure, respectively. The change in flowrate
and damping ratio will need to be determined from the
with current and load pressure is shown in figure 3.
graph. Natural frequency ωV can be read fairly
accurately from the -3dB or 90 degree phase point of
the 5% curve in figure 4. Damping can be determined
from an estimate of the magnitude of the peaking
present. For an under-damped second order system, the
damping factor SV can be shown to be related to peak
amplitude ratio (MV ) by the formula (Poley 2005).

MV (10)
V V

A reasonable estimate of peaking based on the 5%

response curve would be about 1.5 dB, which
corresponds to an amplitude ratio of about 1.189.
Solving the above equation, a value of damping is
determined to be about ζV 0.48. Using these values, a
model of the servo-valve spool dynamics may be
constructed. The input to the model will be the torque
motor current normalized to the saturation current
obtained from the datasheet, and the output will be the
Figure 3: Relationship between flowrate and pressure normalized spool position. SIMULNK modeling is used
drop across a servovalve for different input command to obtain the simulated frequency response of the
percentages (Johnson 2008) transfer function for 5% opening (shown in figure 5).
The servovalve is not the primary dynamic element
in a typical hydraulic servo valve. The servo valves are
sized in a way that its natural frequency is at least three
times the natural frequency of the hydraulic actuator
and therefore it is only necessary to model the valve for
a low range of frequencies without major accuracy loss.

283
 The model of the torque motor and the servovalve
dynamics is described in the previous section. The
values for the parameters are taken from the
manufacturer’s data sheet. Modulated resistors are used
to model the sharp edge orifices. As it is shown in
figure 7, the modulated resistors are located between
pressure drops between four valve ports. Ps, Pa, Pb and
Pt represent supply pressure, port A and B pressures and
return line pressure, respectively. The spool position is
provided by the valve dynamics computed by the
transfer functions and the R-elements are modulated by
the value of the spool position. In this model, the
pressures of port A and B of the valve are respectively
connected to the actuator chambers A and B. Port S is
Figure 5: Simulated frequency response of Moog D765 connected to an effort source which models an ideal
valve source of power capable of supplying constant pressure
at any flow required (Muvengei and Kihiu 2010). Port T
The step response of the servo valve is shown in figure is the pressure for the return line. The infinitesimal
6. The open loop settling time is 0.04s. leakage between port S and T is also modeled as a
Resistor.

2.2. Actuator Modeling

I_saturation The model of the actuator consists of two C-elements
for each chamber. These capacitors model the oil
1 k 2 compressibility in the chambers. It is notable that while
0.59s + 100 s2 + 2  s +  2 the piston rod is retracting or extending, the chamber
Constant TorqueMotor Servovalve
volume is changing and therefore two modulated
capacitors are used to model this effect. The leakage is
modeled using Hagen-Poisseille equation (Muvengei
and Kihiu 2010). The position of the piston is controlled
by the modulation of the I–element. The model is
shown in figure 8.

Figure 6: Transfer function for valve spool positioning

system and its step response

2.1.2. Servovalve Bond Graph Modeling

The nonlinear pressure-flow relation is modeled using
the modulated R-elements for each valve land utilizing
the equations and formulation in section 2.1.1.

Figure 8: Bond graph model of the linear actuator

2.3. Pressure Relief Valve Model

The pressure relief valve is modeled (shown in figure 9)
so that if the pressure is above a threshold, it will bleed
the hydraulic fluid to generate balanced pressure.
Figure 7: Bond graph model of the servo valve

284
 First a constant reference position is used to carry out
position control on the cylinder. The stroke length is
30cm and the initial position of the piston is 15cm. The
controller will extend the piston 5 cm and stop (shown
in figure 11).

Figure 9: Bond graph model of the pressure relief valve

2.4. Hoses Model

There are four hoses that run between the servovalve,
pump and the actuator. Compliance elements are
included to take the expansion of the tubes radially, and
oil compressibility into account. An R-element for the
pressure loss due to the wall friction in the hose is also
considered (Hölcke 2002). An I-element is added to
account for the fluid inertia (Karnopp, Margolis, and
Rosenberg 2006).

2.5. Bond Graph Model of Hydraulic Circuit

The full bond graph model using the subcomponents of
the previous subsections is shown in figure 10. This
model includes the model of the servovalve, two
pressure relief valves, three accumulators and four
corresponding hydraulic hoses. This model is simplified
by removing the models for the accumulators and the
hoses for simplification in analysis and using a PID Figure 11: Step response of the system for 5cm
controller to perform position control. movement – Desired and actual piston location (above)
and corresponding control signal (below)
3. CONTROLLER DESIGN AND SIMULATION
In this section, a feedforward PID is implemented to In the second scenario, a sinusoidal profile is used
simulate the desired motion on the actuator (Caputo to perform target tracking to simulate the behavior of
1996). The controller utilizes the piston position of the the system for oscillation generation. In this case, a
actuator as a feedback signal which is obtained by the sinusoidal signal, is
integration at the I-element associated with the piston. used to perform position tracking and as it is shown in
The generated model is used to perform PID control on the figure 12, it can perform the tracking with a good
two scenarios. The following figure shows the complete performance. The control signal to the valve is shown in
bond graph model with the implemented control. figure 12.

p
MV
PID
SP
Controller Sine

Servovalve Actuator
Ca
Constant 0
C

R
1 MR MR 1 
R
I_saturation
MR_SA MR_AT
0 TF Integrate
1 k 2 Aa

s2 + 2  s +  2
MR
0.59s + 100
MR_ST
TorqueMotor Servovalve
1 R 1 1
Se 0 0 1 0 Rl
Pp

0 TF
I 1 I
1 MR MR 1 Ab
Im IM
MR_SB MR_BT
1 R
Rr1
0
C
Cb

Se 0 1
Pe

Figure 10: Bond graph model of the hydraulic system

285
 Actuator parameters
Description Value
Rod Diameter 1 ¾”
Piston Diameter 2 ½”
Length of cylinder 0.3m
Length of piston 0.01m

REFERENCES
Merritt, H. E., 1967. Hydraulic Control Systems. New
York: John Wiley and Sons.
Jelali, M. and Kroll, A., 2003. Hydraulic Servo
Systems- Modelling, Identification and Control,
Springer Verlag.
Dransfield, P., 1981. Hydraulic Control systems-
Design and Analysis of Their Dynamics. Springer
Verlag.
Muvengei, M. and Kihiu, J., 2010. Bond Graph
Modeling of Inter-Actuator Interactions in a Multi-
Cylinder Hydraulic System. International Journal
of Mechanical, Industrial and Aerospace
Engineering. World Academy of Science,
Figure 12: Response of the system for sinusoidal piston Engineering and Technology, 32-41.
movement – Desired and actual piston location (above) Johnson, J.L, 2008. Designer`s Handbook for
and corresponding control signal (below) Electrohydraulic Servo and Proportional Systems.
4th edition. IDAS Engineering Inc.
Feedforward PID controller enables increased Karnopp, D. C., Margolis, D. L, and Rosenberg, R. C.,
dynamic system responsiveness while decreasing 2006. System Dynamics: Modelling and
positioning and velocity error. Feedforward gains are Simulation of Mechatronic Systems. 4th edition.
predictive gains that contribute to the control signal in New York:John Wiley and Sons.
the form of a linear combination of the target velocity Maskrey, R.H. and Thayer, W.J., 1978. A Brief History
and acceleration terms. The result is higher performance of Electrohydraulic Servomechanisms. ASME
and durable machine control due to smoother motion. Journal of Dynamic Systems, Measurement and
The control signal is applied to the valve and in both Control Vol 100, No 2, pp 110-116.
cases is appropriate in the sense that they do not excite Thayer, W.J., 1965. Transfer Functions for Moog
the valve with jerk that usually causes immature failure Servovalves. Technical Bulletin 103. Moog Inc.
of the components. Jones, J.C., 1997. Developments in Design of
Electrohydraulic Control Valves. Technical Paper.
APPENDIX A - PARAMETERS AND SYMBOLS Moog Inc.
Clarke, D.C., 1969. Selection and Performance Criteria
Symbol Description Value for Electrohydraulic Servovalves. Technical
Oil Density 867 kg/ Bulletin 103. Moog Inc.
Hose 1 Area 0.000791 ( 1 ¼”) DeRose, D., 2003. Proportional and Servo Valve
Hose 2 Area 0.000506 ( 1”) Technology. Fluid Power Journal. March/April,
Hose 1 length 20 cm 8-15.
Hose 2 length 20 cm Caputo, D., 1996. Digital Motion Control for Position
and Force Loops. Proc. National Conference
Kinematic Fluid 27cSt
Confreence. Fluid Power. Vol: Issue 47th:263 Vol
Viscosity
1-268.
Absolute Fluid 0.023409 Pa.s
Hölcke, J., 2002. Frequency Response of Hydraulic
Viscosity
Hoses. Licentiate Thesis. Royal Institute of
B Bulk modulus 1.5 GPa
Technology, KTH, Stockholm.
of Oil
Poley, R., 2005. DSP Control of Electro-Hydraulic
E Elastic Modulus 30 MPa Servo Actuators. Application Report SPRAA76.
of hydraulic Texas: Texas Instruments.
hoses D765 Series Servovalves. Product Datasheet. Moog
m Rod and piston 2.5 kg Inc.
mass 20sim version 4.1. Controllab Products B.V.. Enchede,
M Load mass 500kg Netherlands. http://www.rt.el.utwente.nl/20sim/.

286