Recent Researches in Automatic Control

Modeling of Hydraulic Control Valves

PETR CHALUPA, JAKUB NOVAK, VLADIMIR BOBAL
Department of Process Control, Faculty of Applied Informatics
Tomas Bata University in Zlin
nam. T.G. Masaryka 5555, 760 01 Zlin, Czech Republic
CZECH REPUBLIC
chalupa@fai.utb.cz

Abstract: The paper deals with development of a mathematical model of valves of a hydraulic system. A three
tank laboratory model (Amira DTS200) was investigated and characteristics of its valves were measured and a
process of creating a mathematical model of the valves is described in detail. Although the three tank system is
a classical modeling task this paper focuses on nonlinearities which are present in real system and other
differences between ideal mathematical model and real-time system. The valves contain hysteresis and other
nonlinearities. Despite the fact that all valves the system is equipped with are of the same type, big differences
were observed between their characteristics. The approach to modeling of the system is not restricted to the
particular system but can be used for many real-time hydraulic systems.

Key-Words: Control valve, Modeling, Identification

1 Introduction model of the valves of DTS200 Three-Tank System

Almost all current control algorithms are based on a [5] and to design the models in MATLAB-Simulink
model of a controlled plant [1]. Some information environment. The DTS200 laboratory equipment
about controlled plant is necessary for design of a was developed by Amira Gmbh, Duisburg,
controller with satisfactory performance. A plant Germany and serves as a real-time model of
model can be also used to investigate properties and different industrial systems concerning liquid
behaviour of the modelled plant without a risk of transport.
damage of violating technological constraints of the The models of valves serve as a part in process
real plant. There are two basic approaches of of creating a model of whole DTS200 system. The
obtaining plant model: the black box approach and major reason for creating the model of this
the first principles modelling. laboratory equipment are big time constants of the
The black box approach [2], [3] is based on plant and thus time consuming experiments. A
analysis of input and output signals of the plant. model, which represents the plant well, can
Usage the same identification algorithm for wide set considerably reduce testing time of different control
of different controlled plants is the main advantages approaches. Then only promising control strategies
of this approach. The knowledge of physical are applied to the real plant and verified.
principle of controlled plant and solution of set of The paper is organized as follows. Section 2
mathematical equation is not required. Main presents the modelled system – Amira DTS200.
drawback of a black box model persists in fact that Derivation of initial ideal using first principles
it is generally valid only for signals it was calculated modelling is carried out in Section 3. Section 4 and
from. 5 presents characteristics and calibration water level
The first principle modelling provides general sensors and pumps respectively. Section 6 consists
models valid for wider range of plant inputs and or results of measurements of valves.
states. The model is created by analyzing the
modelled plant and combining physical laws [4].
But there is usually a lot of unknown constants and 2. The DTS200 System
relations when performing analysis of a plant. The photo of main part of Amira DTS200 system is
The paper uses combination of both methods. shown in Fig. 1. The system consists of three
Basic relations are derived using mathematical interconnected cylindrical tanks, two pumps, six
physical analysis. Values of model parameters are valves, pipes, water reservoir in the bottom,
identified on the basis of real-time measurements. measurement of liquid levels and other elements.
The goal of the work was to obtain a mathematical Both pumps pump water from the bottom reservoir

ISBN: 978-1-61804-004-6 195

 Recent Researches in Automatic Control

to the top of the left and right tanks Valve positions signals which states that the valve is fully opened or
are controlled and measured by electrical signals, fully closed respectively.
which allow precision setting of their position. The overall number of inputs to the modelled
plant DTS200 is 14:
• 2 analogues signals controlling the pumps,
• 12 digital signals (2 for each of the 6
valves) for opening / closing of the valves.
The plant provides 21 measurable outputs which
can be used as a control feedback or for
measurements of plant characteristics:
• 3 analogue signals representing level
heights in the three tanks,
• 6 analogues signals representing position of
the valves,
• 12 logical signals (2 for each of the 6
valves) stating that corresponding valve is
fully opened / closed.
Fig. 1. Amira DTS200 – three tank system

A simplified scheme of the system is shown in

3 Initial ideal model
Fig. 2. The pump P1 controls the inflow to tank T1 This chapter is focused to derivation of
while the pump P2 controls the liquid inflow to tank mathematical model of a valve. This derivation is
T2. There is no pump connected to the middle tank based on ideal properties of individual components.
Ts. The characteristic of the flow between tank T1 The ideal flow of a liquid through a pipe can be
and tank Ts can be affected by valve V1, flow derived from Bernoulli and continuity equations for
between tanks Ts and T2 can be affected by the valve ideal liquid:
V2 and the outflow of the tank T2 can be affected by 1 ⎫
valve V3. The system also provides the capability of Δhg = v 2 ⎪
2 ⎬ q = S v 2 g Δh (1)
simulating leakage from individual tanks by opening
q = SV v ⎪⎭
the valves V4, V5 and V6.
where ∆h is a difference between liquid levels on
P1 both sides of the pipe (e.g. difference between levels
P2 of tanks that are interconnected by the pipe), g is the
T1 T3 T2 standard gravity, v is the liquid velocity and SV is the
flow space of the pipe. The flow space SV is
controlled by the valve position p.
h1
V1 h3 V2 h2 V3 SV = p ⋅ SV max 0 ≤ v ≤1 (2)
where SVmax is the maximal flow area of the valve.
leakage (V4) leakage (V5) leakage (V6) Since the flow through a valve depends only
Fig. 2. Scheme of three tank system Amira DTS200 on the level difference, the valve position and
constants representing pipes and cylindrical tanks,
Pumps are controlled by analogue signal in range the change of water level in tank T1 can be written
from -10V to 10V. Heights of water level are as follows:
measured by pressure sensors. Each valve is dh1
operated by two digital signals which control motor = k1 h1 − hs ⋅ sign ( h1 − hs ) − k4 h1 (3)
dt
of particular valve. First signal orders to start
closing of the valve while the second signal is used The area of all three tanks is the same and is
for opening of the valve. If none of the signals is symbolized by ST. The k is a parameter representing
activated the valve remains in its current position. valve position
Each valve also provides three output signals:
analogue voltage signal correspond to the current SV max 2 g
ki = pi i = 1, 2,...,6 (4)
position of the valve and two informative logical ST

ISBN: 978-1-61804-004-6 196

 Recent Researches in Automatic Control

and q represents inflow as change of water level in “opened” signals of individual valves. Before these
time: signals drop down the valves are said to be opened.
The vertical lines in the right part of Fig. 3 represent
qi′
qi = i = 1, 2 (5) the changes of “closed”. From these lines onward,
ST the valves are said to be closed.
It can be observed that the initial and final
Similar equations can be derived for the other
positions of the valve as well as the positions
two tanks. The model obtained by using ideal
corresponding to changes of “opened” and “closed”
properties and behaviour of plant parts if further
signal differ. But all the valves are moving at almost
referred as “ideal model”. This model of whole
the same speed vvalve.
three tank system is successfully used in many
control system studies as a demonstration example vvalve = −0.175 MU s (6)
[6], [7], [8].
Valve positions corresponding to hard
constraints and validity of “opened” and “closed”
state are summarized in Table 1.
4. Characteristics of the valves
As stated in Section 2, each of plant’s 6 valves is
Table 1
driven by two dedicated logical signals. These
Valve positions for important states
signals are used for starting valve’s motor in closing
“closed” “opened”
or opening direction respectively. If none signal is full
Valv signal signal full open
activated the valve remains in its current position. closed
e no. change change [MU]
Activation of both signals at in a particular time [MU]
[MU] [MU]
represents an invalid state and valve motor is
1 0.5199 0.4066 -0.6065 -0.6945
stopped.
Each valve provides three output signals. The 2 0.4550 0.3880 -0.5719 -0.7493
current valve position is determined by analogue 3 0.5462 0.3996 -0.5423 -0.6501
signal. Higher values of signal represent closed 4 0.4926 0.3594 -0.5751 -0.7151
valve and lower values represent opened valve. The 5 0.4868 0.3296 -0.5157 -0.6718
other two signals are logical and state that valve is 6 0.5242 0.4416 -0.5698 -0.6525
opened or closed respectively.

4.1 Valve limits and speed 4.2 Valve flow parameter for outflow valves
Process of opening all valves at once from fully Valve flow parameters ki as appear in (3) were
closed state to fully opened is presented in Fig. 3. computed from measurements of draining through
full close => full open
individual valves which are connected to outflow
pipes (V3, V4, V5 and V6). The draining of a tank to
the reservoir situated below the tanks is described
0.4
by differential equation based on (3):
0.2 dh ( t )
= −k h ( t ) (7)
dt
valve pos [MU]

Integrating in an appropriate time range leads to

-0.2 the equation of time course of water level:
valve 1 k2 2
h (t ) = ⋅ t − k h ( 0) ⋅ t + h ( 0)
-0.4
valve 2 (8)
valve 3 4
-0.6 valve 4
valve 5 where h(0) is initial water level. An example of
valve 6
-0.8 draining is presented in Fig. 4. At the beginning of
0 1 2 3 4 5 6 7 8 9 10
time [s] the experiment, the tank was full and all valves were
closed, then valve V4 was partially opened, its
Fig. 3. Closing all valves in full range position was recorded and time course of water level
height was measured.
This process represents moving of valve position
in full range of its hard constraints. The vertical
lines in left part of Fig. 3 represent changes the

ISBN: 978-1-61804-004-6 197

 Recent Researches in Automatic Control

parabola fitting of draining through valve V4

draining through valve V 4
h1=0.0086*t 2-4.95*t+622
700
700
h1
600 600 parabola

500 500

400
400
h1 [mm]

h1 [mm]
300
300
200

200
100

100 0

0 -100
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
time [s] time [s]

Fig. 4. Draining of tank T1 through valve V4 Fig. 6. Parabola fitting to the draining course

It is obvious that parabola depicted in Fig. 4 Values of k and h0 can be easily obtained from
would continue below zero contrary to (8). A term polynomial coefficient according to (9). Valve can
corresponding to the vertical length of outflow pipe be closed to different positions at the beginning of
h0 was added to the model. The vertical length h0 is draining experiment and relation between valve
depicted in Fig. 5. position and value of k can be achieved. This
relation for one set of experiment on valve V4 is
presented in Fig. 7 where circles represent
tank individual experiments. The characteristic is not
strictly linear. It contains saturation of fully closed
and fully opened valve. Transitions to saturation
h valve states are smooth.
valve opening
0.3

reservoir 0.25
h0
0.2

0.15
k4

Fig. 5. Vertical length of outflow pipe (h0)

0.1
Due to mechanical configuration of the plant, the
value of h0 for outflow valves V3, V4, V5, and V6
0.05
cannot be measured directly. But it can be identified
from draining course (Fig. 4). To encapsulate h0 into
0
model, equations (7) and (8) were superseded: -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
position of valve V4 [MU]
dh ( t )
= − k h ( t ) + h0 Fig. 7. Relation between valve position and k4
dt (9)
k2 4.3 Valve flow parameter for interconnection
h ( t ) = ⋅ t 2 − k h ( 0 ) + h0 ⋅ t + h ( 0 ) valves
4
Similar approach to obtaining values of k as
A second order polynomial (parabola) was fitted presented in previous subsection can be used also
to an appropriate interval of draining data in least for valves V1 and V2 which interconnects tanks T1
mean squares sense. The MATLAB function polyfit and Ts, and Ts and T2, respectively. Flow from the
was used for this task. Parabola fitting is presented full tank T1 to the empty tank Ts was used to
in Fig. 6. measure valve constant k1. The other valves were

ISBN: 978-1-61804-004-6 198

 Recent Researches in Automatic Control

closed during the experiment. According to (3), the 4.4 Valve hysteresis
flow can be described by two differential equations: The experiments presented in previous subsections
were preformed for opening of a valve only. At the
dh1 ( t )
= − k1 h1 ( t ) − hs ( t ) ⋅ sign ⎡⎣ h1 ( t ) − hs ( t ) ⎤⎦ beginning, the valve was fully closed and
dt
(10) subsequently was partially opened to a given
dhs ( t ) position. In this section a problem of closing of a
= k1 h1 ( t ) − hs ( t ) ⋅ sign ⎡⎣ h1 ( t ) − hs ( t ) ⎤⎦ valve is studied. Performed experiments are similar
dt
except initial part. The experiment starts with full
Since the value of h1 is always higher or equal to tank and closed valve too, but then the valve was
hs, the term inside absolute values is always fully opened and then partially closed to the desired
nonnegative. As the water flow just from T1 to Ts position. Therefore the same valve position (value
and the geometry of both tanks is the same, of analogue signal from a valve) was reached but
according to mass conservation law the sum of h1 from opposite direction.
and hs remain the same during the experiment. Then valve 2
0.25
the course of draining T1 and filling Ts can be closing 1
described by two independent differential equations. closing 2
closing 3

dh1 ( t )
0.2 closing 4
= − k1 2h1 ( t ) − hΣ opening 1
opening 2
hΣ = h1 ( t ) + h2 ( t ) (11)
dt opening 3
dhs ( t )
0.15

= − k1 hΣ − 2h2 ( t )
k2

dt
0.1
Solving these equations lead to time course
described by second order polynomial.
0.05
k2
h1 ( t ) = ⋅ t 2 − k 2h1 ( 0 ) − hΣ ⋅ t + h1 ( 0 )
2 (12) 0
k2 2 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
h2 ( t ) = − ⋅ t − k hΣ − 2h2 ( 0 ) ⋅ t + h2 ( 0 ) valve position [MU]
2 Fig. 9. Hysteresis of valve V2
An example of courses and corresponding
parabolas are depicted in Fig. 8. Experiments unveiled a hysteresis present in all
parabola fitting of flow through valve V1
valves. The characteristics for opening and for
h1=0.0209*t2 -4.95*t +570 closing of valve V2 is presented in Fig. 9. There are
hs =-0.0212*t 2 +5.01*t -20
600
four sets of experiments for closing of the valve
h1 performed in various time in the figure. Individual
500 hs experiments correspond to stars in the figure. Three
parabola h1 set of experiment are presented for closing of the
400 parabola hs
valve where individual experiments are represented
300
by circles. The figure shows that hysteresis plays a
big role in the experiments. The value of position
h [mm]

200 itself does not give sufficient information about

current value of parameter k2. For example, if the
100
position is 0 MU the value of k2 can be anywhere in
0
range 0.03 to 0.13. Especially in case of using the
valve as an actuator the hysteresis should be taken
-100 into account. Otherwise control process can easily
0 20 40 60 80 100 120 140 160
time [s] become unstable.
Since the shape of curves corresponding to
Fig. 8. Parabola fitting to the flow course through V1
opening and closing of the valve is similar, an
average difference between them in direction of
A similar approach as presented for valve V1 was
position axis can be computed. This value can be
used to measure characteristics of valve V2.
used as measure of hysteresis. Values of hysteresis,
h0 as well as maximal value of k for each valve is
presented in Table 2.

ISBN: 978-1-61804-004-6 199

 Recent Researches in Automatic Control

5 Conclusion
Table 2 The paper presented a development of the model of
Valve positions for important states valves of a hydraulic system. The Amira DTS200
Valve hysteresis three tank system was considered but used
maximal k h0 [mm]
no. [MU] techniques can be easily generalized to wide set of
1 0.0219 0.2180 - hydraulic systems. The real system contains several
2 0.1783 0.2237 - nonlinearities which incorporate complexity to the
3 0.0310 0.2601 126.6 system. Total number of experiments concerning
4 0.0426 0.2976 127.9 valves reached 433 taking altogether more than 113
5 0.1307 0.2735 121.2 hours. Resulting model includes all major
6 0.0800 0.2688 97.1 nonlinearities and can be integrated into a Simulink
model of whole three tank system.

4.5 Modelling of valve characteristics

The course of relation between valve position in Acknowledgement
MU and k is similar to step responses of dynamical This work was supported by the Ministry of
system and therefore it was modeled in similar way. Education of the Czech Republic under grant
Other types of approximation functions, like 1M0567.
sigmoids, were also tested, but did not achieve
better results. A model based on transfer of 4th order
aperiodic system produced satisfactory results. Thus References:
relation between position and k was as follows: [1] V. Bobál, J. Böhm, J. Fessl & J. Macháček:.
⎡ 1 −ab b3 + 3b2a + 6ba2 + 6a3 ⎤ Digital Self-tuning Controllers: Algorithms,
pos < pos0 : k = kmax ⎢1− e ⎥ Implementation and Applications (Springer -
⎣ 6 a3 ⎦ Verlag London Ltd., 2005).
b = pos0 − pos (13) [2] G. P. Liu, Nonlinear identification and control
– A neural network Approach (Springer -
pos ≥ pos0 : k =0
Verlag London Ltd., London, 2001). 
where pos is valve position in MU and parameters a [3] L. Ljung, System identification : theory for the
and pos0 were obtained by nonlinear regression. The user (Upper Saddle River, NJ : Prentice Hall
regression for valve V2 is presented in Fig. 10. PTR, 1999)
Behavior of system inside hysteresis area was [4] D. M. Himmelblau, J. B. Riggs, Basic
studied as well. This task was time consuming principles and calculations in chemical
because a performed set of experiments took more engineering (Upper Saddle River, N.J. :
than 45 hours. The parameter k did not change its Prentice Hall, 2004)
value till it reaches a border of hysteresis area, i.e. [5] Amira, DTS200 Laboratory Setup Three - Tank
the curve of either opening or closing of the valve. - System (Amira GmbH, Duisburg, 2002).
[6] L.Li & D. Zhou, Fast and robust fault
0.25
data close diagnosis for a class of nonlinear systems:
model close
data open
detectability analysis, Computers &Chemical
0.2 model open Engineering, 28, 2004, 2635-2646.
[7] D. Henry & A. Zolghadri, Norm-based design
0.15 of robust FDI schemes for uncertain system
under feedback control: Comparison of two
2
k

0.1
approaches, Control engineering practice, 14,
2006, 1081-1097.
[8] S. Blazic, I. Skrjanc & D. Matko, Globally
0.05
stable direct fuzzy model reference adaptive
control, Fuzzy Sets and Systems, 2003, 3-33.
0
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
valve pos [MU]

Fig. 10. Model of parameter k for valve V2

ISBN: 978-1-61804-004-6 200