Professional Documents
Culture Documents
electrical signal processing allows the There is no common basis for benchmarking the
implementation of a full range of possible linear and different control strategies for direct comparison. It
non-linear control schemes. can be said though, that all mentioned works on non-
linear approaches, except for non-linear feed-forward
In the mid eighties, the use of different linear single
compensation, have not reportedly been transferred
input single output (SISO) feedback laws for pressure
into widespread applications. Still today, linear
control in hydraulics was studied. A driving factor
controllers on the basis of PID are almost exclusively
was the availability of digital simulation and
applied. Often they are used in combination with
programmable control hardware. However, some
non-linear extensions like limiters, switched integral
early works also used analogue devices to test
characteristics and feed-forward compensation.
different feedback types.
Forster summarizes in his work on valve controlled Parametrization problem
electro-hydraulic load simulation that simple linear
controllers (P or PID) are well suited for pressure The cited works offer a large variety of possible
control, Forster (1984) and Forster (1988). The P solutions for pressure control but the parametrization
controller cannot achieve zero steady state error in of the proposed controllers remains an issue for the
the presence of disturbance. The PID controller can user. This is also true for position control even
be optimized towards reference tracking or though much more research is dedicated to it
disturbance rejection but not both at the same time. compared to pressure control. It seems that most
Forster proposes a disturbance feed-forward control research papers focus on specific applications and are
to compensate the pressure loss due to the outflow of tied to experimental set-ups. An interesting and
pressurized fluid during cylinder motion or leakage. practical paper which gives guidelines to the choice
Y. Liu (1985) compares in theoretical and of control structure and for the parametrization of
experimental investigations P, PD and PI controller position control is presented by Noskievič (1996 and
performance for a supply pressure control using a 2002) on the basis of linear analysis.
variable displacement pump. Ulrich (1989) discusses In pressure control, a straight-forward method to
the compensation of line dynamics using linear parametrize the PID or any of the other controllers
control schemes. Yang (1999) et al. develop a two for a general case is not described by any of the
degree of freedom type I-PDD2 controller for a load above listed works. Mostly, parameters are found by
simulator and compare results with a PID controller. root locus method or pole placement, sometimes in
Also, adaptive and non-linear controllers were combination with numerical optimization. However,
developed to compensate non-linear effects and the the strategies are not explained and therefore cannot
changing plant dynamics at different operating points. be repeated for a different setup without modelling
and simulation.
Guo & Hovestäd (1989) present an adaptive PI
controller for pressure control. The parameters of the To address the issue to parametrize a PID controller
PI controller are adapted to the changing capacity of for a pressure control application, Boes et al. (2003)
an accumulator which is connected to the outlet of published three simple formulas to calculate the
the pressure controlled pump. A sliding mode control parameters for a common pressure control
controller for pressure control is developed by Park problem as shown in Figure 0.1. The proposed rules
& Kim (2009). yield unstable response as is easy to show with Routh
Hurwitz criterion. Obviously the publication has a
Force control is closely related to pressure control. typographical mistake in the sign of the integrator
However, as Alleyne & R. Liu (2000) stress, the coefficient . The (correct) PID feedback gains are:
dynamics, especially inertia, of the mechanical
system has a strong influence on the closed loop (√ )
dynamics and must be taken into account. They (1)
develop and implement a Lyapunov based non-linear
controller which takes a lightly damped load into ( √ )
account. They also propose a parameter estimation (2)
and friction compensation scheme. Kennedy & Fales
(2010) design different force controllers (P, PID and [ √ ( ) ]
H∞) on the basis of an experimentally identified open (3)
loop plant dynamics and find parameters for
nominal/robust stability using uncertainty and
performance measures. Plummer (2007) proposes a The advantage of these gain formulas is that the
robust force control scheme, which does not require parameters required to calculate , and are
an exact model of the mechanical system. This is usually known to the commissioning engineer:
possible by introducing a flexible element such as a pressure chamber capacity , valve natural
spring between the actuator and the mechanical frequency , valve damping ratio and the flow
structure on which the controlled force is applied. gain value . For details on how to calculate these
3 Review # 287-10
̈ ̇ (8)
The orifice equation describes the valve flow as a
function of the valve opening and pressure difference.
Normally we assume the controlled pressure to be
4 Review # 287-10
With ( ) (22)
( ) ( )
̇ (17) (23)
( ) ( )
Linearizing gives: Mathematically, at least three conditions must exist
for a unique calculation of parameters , and .
̇ (18) These conditions can be stated as:
The capacity is the ratio of volume over effective a) Damping ratio of zeros of PID same as
fluid bulk modulus : valve damping ratio
5 Review # 287-10
b) phase margin (large phase margin settings of Eqs. (1)-(3) and Eqs. (4)-(7) in the
provides good robustness against parameter following section.
uncertainty)
c) Set gain cross-over frequency a factor Control performance
1/3 below valve natural frequency The control design is implemented on a MTS 100 t
(determines expected closed loop system servo-hydraulic load frame in the civil engineering
bandwidth) structural laboratory of the American University of
Beirut, Figure 0.1.
The control design with these conditions is quite
conservative but can be tailored for special The valve of the hydraulic axis is a high-response
applications, if needed, see appendix. servo-valve of type MTS 252.25 with 56 l/min
nominal flow at 35 bar pressure drop and a cut-off
Condition a) states frequency of 170 Hz. It is mounted very close to the
cylinder on a special manifold and supplied with
(24)
200 bar pressure. The double rod cylinder has a
Condition b) states that the phase ( ) at the gain piston diameter of 292.1 mm, a rod diameter of
cross-over frequency of the open loop Eq.(23) 152.4 mm and a stroke of 254 mm. Therefore the
should be . This means that maximum volume of the cylinder in extended piston
position is 12.39 l. For the experimental results the
( ) piston is fully extended and the pressure is controlled
( | ) (25)
( ) in the larger chamber. Since the piston is blocked, it
is appropriate to calculate the stiffness of the oil
Inserting and (condition c) yields volume using an average bulk modulus of 14000 bar.
The pressure sensor attached to the cylinder is a
DMP 333 industrial pressure transmitter (BD|sensors)
( )| (26) with accuracy of 0.1 % of full scale output of
0-200 bar. This high level of accuracy is optional for
This equation can be solved for and we get these types of sensors and is achieved by internal
signal conditioning. The response time is therefore
(√ ) (27) unusually slow around 200 ms according to the data
sheet. We assume that we can model the pressure
Damping and natural frequency of the controller have sensor dynamics as a second order element with
now been determined. Finally, the integral gain is natural undamped frequency of 5 Hz and damping
calculated from the magnitude criterion. At the gain ratio of 0.7.
cross-over frequency, the open loop magnitude
equals 1 = 0 dB. Therefore the integral gain can be
determined from
( )
| | (28)
( )
to be
(29)
(30)
(31)
0.1867 0.0047
1 0.1191
170 0.9241
10 0.0024
0% 0.1291
12.39 l 0.5
70 bar 0.0043
56 0.0432
35 bar 0.2007
7 Review # 287-10
Figure 0.2: Theoretical closed loop step response Figure 0.3: Experimental closed loop step response
(linear systems)
The transient responses have similar shapes in
One can see that the ITAE optimized pressure control comparison with Figure 0.2. The ITAE optimized
(Eqs. (4)-(7)) has almost no overshoot and yet a swift control has almost no overshoot due to the
rise time. The cancellation of an open loop zero by prefiltering of the reference signal. The other PID
the pole of the reference input prefilter proves to be response curves exhibit a characteristic bend 25 ms
very effective. Since the parameterization is based on after the reference step change. The sharp rise in the
optimization, the plant dynamics is well utilized. beginning suggests that the open loop bandwidth is
higher than expected from the response time stated in
The step responses of the frequency response design
the pressure sensor data sheet. Probably the signal
and the one proposed by Boes et al. (2003) are
filtering in the transducer is adaptive to the rate of
characterized by a large overshoot of around 36% of
change of signal. It goes beyond the scope of this
the step height. This overshoot, caused by the integral
paper to further investigate this issue. What the
action, is not the result of bad tuning. The PID
results show, is that the control feedback has a degree
control developed by Forster (1988) has a similar
of robustness to deal with this modeling uncertainty.
shape with the same amount of overshoot. The
Furthermore, it is evident that the bandwidth of the
frequency response design based controller is slightly
frequency response design based controller is slightly
faster than the one proposed by Boes et al. (2003)
higher than the controller according to Boes et al.
with a little bit more overshoot. The control
(2003) This matches with the ideally expected
performances are summarized in Table 2.
response shown in Figure 0.2. The rise time and
Table 2: Performance comparison of theoretical settling time of the controller proposed in this paper
response are shorter. Also the damping is slightly better. The
performance of all three PID control schemes is
Criterion Freq.R. Boes ITAE summarized in Table 3.
Overshoot 37.1 % 35.2 % 1.7 % Table 3: Performance comparison of experimental
Rise time 0.22 s 0.25 s 0.22 s response
overshoot compared to the other PID controls in this Bakirdogen & Liermann (2010) would suggest that
study. This statement can be backed up by the parameters proposed by Boes et al. (2003) are
comparison of the phase margins of the open loop also found on basis of an optimization, certainly a
transfer functions (without prefilter), compared in more conservative one.
Table 4. The phase margin of the frequency response
design is , as expected from the control design Conclusion
condition. The phase margin of the control proposed
PID controllers have long been used for pressure
by Boes et al. (2003) is a little bit higher, at ,
control applications but parametrization still poses a
whereas the phase margin of the ITAE optimization
problem for the commissioning engineer.
based control is only . Therefore, if good
Conventional tuning rules take time in the
robustness against noise and model uncertainty is
commissioning process and do not always lead to
desired, the frequency response design based
sufficient results. The feedback formulas proposed by
controller presented in this paper or the parameters
Boes et al. (2003) (with corrected sign in equation for
proposed by Boes et al. (2003) are the better choices
integral gain) are a great help because they base on
over the ITAE optimization based design.
physical plant parameters which can be read from
Table 4: Open loop phase margins of PID controls component data-sheets and from plant documentation.
Also, control tuning is possible by varying a single
Freq.R. Boes ITAE parameter, the capacity of the system. However, it
Phase is not clear how these tuning rules were derived.
margin The tuning rules proposed in this paper Eqs. (29)-(31)
Gain cross 1.67 Hz 2.24 Hz 3.6 Hz are derived from straight-forward requirement
over freq. specifications with a transparent frequency response
design technique. The derivation is conservative but
can be tailored for special applications, if needed, see
The rules proposed by Boes et al. (2003) seem to appendix. The final product, the parametrization by
yield very similar results to the controller presented known plant parameters and the single parameter
in this paper. In this example, the dominant second tuning capability, is effective and very relevant for
order system (the pressure sensor dynamics) is industrial practice.
assumed to have a critically damped behavior with a
The performance characteristics of the controller
damping ratio of 0.7. For other damping ratios the
based on frequency response are similar to the one
result of the two controllers is not so similar. Figure
proposed by Boes et al. (2003). If excellent reference
0.4 shows the phase margin of the considered
tracking is prioritized over disturbance rejection, the
controllers over varying damping ratios of the
ITAE optimized control parameters presented in
dominant second order system. It is evident that the
Bakirdogen & Liermann (2010) Eqs. (4)-(7) should
frequency response design based controller maintains
be considered.
a constant phase margin of , while with the other
controllers the phase margin decreases with Future studies should focus on relevant practical
increasing damping of the dominant second order issues such as noise, non-ideal differentiation, and
system. added phase lag due to sampling time and sensor
dynamics. Also a relationship between capacity,
expected pressure steps and valve size could be found
on basis of this study, which may help the designer to
choose the right valve size for a pressure control
application.
List of Notations
propoprtional control valve for pressure Zehner, F., 1987. Vorgesteuerte Druckventile mit
controlled variable displacement pumps]. O direkter hydraulisch - mechanischer und
+ P : Zeitschrift für Fluidtechnik, 29(11), elektrischer Druckmessung [Pilot operated
792-796. pressure valves with direct hydro-mechanic
and electric pressure sensing]. RWTH
Noskievič, Petr 1996. Auswahlkriterium der Aachen University.
Reglerstruktur eines lagegeregelten
elektrohydraulischen Antriebes [Criterion
for selection of control law for hydraulic Appendix: Gain formulas with variable
position servo drives]. O + P : Zeitschrift phase margin
für Fluidtechnik, 39(1), 49-51.
The control design conditions a)-c) in section 0 are
Noskievič, Petr 2002. Closed loop control of the proposed based on experience and practical
system with the modes of different considerations. The choice of phase margin is a
dynamics and damping. International conservative choice with the advantage that it makes
Carpathian Control Conference ICCC' 2002, the gain formula for more compact, since
235-240 becomes infinite. Choosing a smaller phase margin
yields a shorter rise time but increases the overshoot
Murrenhoff, H., 2008. Servohydraulik - geregelte comparatively. The ratio b between natural frequency
hydraulische Antriebe [Servo-hydraulics - of dominant second order system and desired gain
closed loop controlled hydraulic drives], cross-over frequency seems to be ideal around
Aachen: Shaker. . Choosing it significantly higher or lower both
leads to slower system response. Mathematically it
Park, S. & Kim, J.L.J., 2009. Robust control of the cannot be larger than 1.
pressure in a control-cylinder with direct
drive valve for the variable displacement For variable phase margin, the criterion for the phase
axial piston pump. Proceedings of the angle of the open loop is:
Institution of Mechanical Engineers, Part I:
Journal of Systems and Control Engineering, ( )| ̂ (32)
223(4), 455-465.
which yields
Plummer, A.R., 2007. Robust electrohydraulic force
control. IMechE, 221, pp.717-731. ( ( ) ( ))
( )(
̂. (33)
)
Ulrich, H., 1993. Elekro-hydraulische Druckregelung
mit Verstellpumpe für unterschiedliche Solving for with and ̂ gives
Verbraucher und Leitungsnetze [Electro-
hydraulic pressure control with variable √
displacement pump for different actuators [ ( )
and pipe networks]. RWTH Aachen (34)
University.
]
Ulrich, H., 1989. Kompensation der The PID gains are calculated according to Eq. (29)-
Leitungsdynamik bei Druckregelungen mit (31).
Verstellpumpen [Line effect compensation
with pressure control using variable Author
displacement pumps]. O + P : Zeitschrift für
Fluidtechnik, 33(12), 930-936. Matthias Liermann
Dr.-Ing Matthias Liermann is Assistant Professor in
the Department of Mechanical Engineering at the
Watton, J., 2009. Fundamentals of fluid power American University of Beirut. He received his
control, Cambridge University Press. doctoral degree 2008 from RWTH Aachen
University, Germany and joined AUB in 2009. His
current research interests are in the field of control
Yang, K., Oh, I. & Lee, I., 1999. Pressure control of and fluid-mechatronics with emphasis on the
hydraulic servo system using proportional analysis, simulation and design of smart fluid power
control valve. Journal of Mechanical systems. For more info, please visit:
Science and Technology, 13(3), 229-239. http://staff.aub.edu.lb/~ml14/Homepage/index.html .