Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

Paper received: 14.03.2010

DOI:10.5545/sv-jme.2010.081 Paper accepted: 23.03.2011

An Intelligent Electro-Hydraulic Servo Drive Positioning

Detiček, E. – Župerl, U.
Edvard Detiček* - Uroš Župerl
University of Maribor, Faculty of Mechanical Engineering, Slovenia

The goal of the research is to develop a closed loop control system for position control of electro-
hydraulic servo drive useful for practical application. The research was performed based on the theory
of adaptive feedback systems. The proposed new control structure consists of three parts. The first part is
actually a feed-forward digital filter in the form of an inverse model of the system, which enables a perfect
reference tracking behaviour. The second part is a position feedback fuzzy logic controller designed as a
self-learning system, which possesses an ability of automatic tuning and is able to cope with parameter
and load changes of the control loop. The third part is the so called switching integrator involved to
improve steady state errors and ensure precise final positioning. The effectiveness is proven by laboratory
experimental investigations.
©2011 Journal of Mechanical Engineering. All rights reserved.
Keywords: servo-hydraulic, linear actuators, position control, nonlinear and fuzzy control

0 INTRODUCTION proportional and servo valves, as well as with

position transducers and force sensors, creating
High-speed/high accuracy positioning is together closed control loops. To fulfil another
a key element in modern mechatronics systems demand of the above mentioned machines,
such as computer controlled machine tools, namely the accurate and precise position and force
metal-forming machines, manipulators, injection control, appropriate control strategies are needed.
moulding plastic machines, assembly and From the control engineering point of view, a
transport devices, material testing devices, etc. good reference tracking for the step, ramp, and
To obtain this rapid movement, hydraulic drives sinusoidal reference signals, as well as rejection
controlled by resistance in impressed pressure on external disturbances, have to be achieved.
network are used very often. In addition to the However, the disadvantages of hydraulic
excellent dynamics, the main advantage of such systems such as the low damping ratio,
systems is the parallel use of multiple drives nonlinearities of valve cylinder combination,
fed by one pressure net combined with energy friction forces, and fluid compressibility need to
recycling into the hydraulic accumulators. In be overcome. This is successfully obtainable only
particular, the hydraulic cylinders should be by implementation of modern control theory and
stressed at this point due to their ability of direct sophisticated digital control algorithms [2]. Backe
transformation of hydraulic energy into linear [2] summarized the development of hydraulic
movements and forces. Fluid power actuators applications from 1975 to 1995 and forecast the
are characterized by their high power density need for advanced control techniques.
and excellent dynamic response. The hydraulic In other words, to meet the requirements
actuators in particular are capable of very high of modern mechatronics systems the electro-
output power levels combined with very compact hydraulic drive systems have to possess
drive unit dimensions, with very good positioning increasingly more intelligence. The objective
speed being achieved. They are ideally suited to of many researchers is therefore to develop an
many high dynamic drive applications in modern algorithm which would be able to control the drive
machines and mechatronics systems. without any a-priori knowledge of geometrical,
The servo-hydraulic cylinder system is operating or any other parameters of the system.
also a mechatronic system [1], which can be A contribution to achieve the above objective is
divided into the electronic part and the mechanical presented in this paper. A very promising tool
part. It is quite normal nowadays for cylinders to achieve such desired goals is the fuzzy logic
to be equipped with electronically controlled control. Fuzzy control technology is based on
*Corr. Author’s Address: University of Maribor, Faculty of Mechanical Engineering,
394 Smetanova 17, Maribor, Slovenia, [email protected]
 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

the linguistic variables ˗ the human thought optimization algorithm to determine the effective
process; the input signal is fuzzified at first, then control parameters for an electro-hydraulic
goes through the fuzzy reasoning process under position control system. In his algorithm the
operational experience and expert knowledge, variations of the system dynamics can be properly
and finally the control signal is defuzzified and compensated. Xiang performed a design study
sent out [3] and [4]. In the literature [5] to [7], of adaptive Fuzzy PD controller for pneumatic
servo-hydraulic system control has been executed servo-system [16].
by fuzzy controller successfully; however, This paper presents a new hybrid-fuzzy
which control parameters to use to satisfy the control strategy for position control of the electro-
performance requirement is determined through hydraulic linear drive. An adaptability is obtained
trial and error method. Berger [8] presented the by fuzzy logic controller designed as a self-
idea of designing fuzzy controllers according to learning system, while the reference tracking and
the classification of the controlled systems. Thus, position accuracy are improved by conventional
fuzzy controllers can be classified as Fuzzy-PID, control measures such as an inverse feed-forward
Fuzzy-PI and Fuzzy-PD controllers. Yamazaki controller and switching integrator. The algorithm
[9] and Mudi [10] tested the self-tuning of PI is experimentally investigated and implemented
controllers using fuzzy logic and demonstrated on the hydraulic device for testing mechanical
that fuzzy control was able to reduce the initial constructions-load simulator.
cost of building an automatic tuning system. An
optimal-tuning PID control has also been devoted 1 ADAPTIVE CONTROL STRUCTURE
to a proportional control system including a
4/3-way proportional valve and a differential There are clear design goals for electro-
cylinder with computer data acquisition system hydraulic drive control system to achieve fast
[11] and [12]. Branco [13] has demonstrated response to step command signals with minimal
the feasibility of using fuzzy control to reduce overshoot (dead-beat response) and approximate
the influence of unmodeled nonlinearities and zero steady state errors for step, ramp and
parameter uncertainties in hydraulic systems. sinusoidal reference inputs. Those goals have to
Jelali [14] summarized recent developments be obtainable in a rough industrial environment,
in nonlinear identification, nonlinear control where significant system behaviour changes
and the application of both to hydraulic servo- may occur due to nonlinear characteristics of
systems. Kim [15] proposed an experimental system elements, temperature changes, noise

Fig. 1. Shematic presentation of electro-hydraulic servo drive

An Intelligent Electro-Hydraulic Servo Drive Positioning 395

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

and other environmental influences. Therefore, dpB (t ) E ( pB ) dy (t )

the controller has to be able to adapt itself to the dt
=
V0 B − AB y (t )
[ AB dt − QB ( pB , xV , t ) (4)
changing properties of controlled process and its + K Li ( p A (t ) − pB (t )) ] ,
signals.
where AA, AB are piston areas, pA - left chamber
1.1 Description of Electro-Hydraulic Servo pressure, pB - right chamber pressure, V0A, V0B -
Drive Dynamics dead volumes, KLi - leakage coefficient, and y is
the piston rod position e.g. load position.
The schematic of electro-hydraulic servo The dynamics of the moving parts of the
system is depicted in Fig. 1. It is composed of system is expressed by the force equilibrium:
a single rod cylinder, a 4/3 way servo valve,
a sliding load mass, a position transducer, an d2y
m = p A (t ) AA − pB (t ) AB − FR (t ) − FL (t ), (5)
electronic interface and a control computer. dt 2
Since the high response servo valve is used it is where m is the moving mass, FR(t) is friction force
assumed that the valve spool displacement xV is and FL(t) is external load force.
directly proportional to the electric current i from Finally, by rearrangement of the above
the electronic amplifier driven by control input equations and introduction of some new terms,
voltage u. such as velocity gain Vs(pA, pB, y), eigen frequency
The relation is described by the following ω(pA, pB, y) and damping D(pA, pB, y), leads to a
equation where KV is valve spool gain: common mathematical expression:
xV(t) = KV u(t) . (1)
The servo valve second stage output flows d 3 y (t ) d 2 y (t )
3
+ 2 D( p A , pB , y )ω ( p A , pB , y ) +
QA, QB, by symetric zero overlapping spool and dt dt 2
pump flow Q0 (rated), supply pressure p0, tank +ω 2 ( p A , pB , y ) dy (t ) = ω 2 ( p A , pB , y ) ⋅ (6)
pressure pT ≈ 0, according to Bernoullie’s equation dt
are [14]: dF
⋅Vs ( p A , pB , y ) KV u (t ) − L .
for xV ≥ 0 : dt m
Q0 p0 − p A
QA = xV sign( p0 − p A ) , After linearization of Eq. (6) in particular
xV max p0 operating point, using Laplace transformation,
the transfer function of the entire system can be
Q0 pB
QB = − xV sign( pB ) , derived. It generally has the form of a third order
xV max p0 system (second order system and integrator),
(2)
for xV < 0 : which numerical parameter varies among different
operating points. A major simplification is done
Q0 pA by neglecting the second order term (neglected
QA = xV sign( p A ) ,
xV max p0 oil compressibility and inertial forces) and
representing the system by transfer function of an
Q0 p0 − pB integrator:
QB = − xV sign( p0 − pB ) .
xV max p0
y ( s ) VS KV K I
Following the continuity condition, with G=
(s) = = . (7)
u (s) s s
regard to oil compressibility E(p), the pressure
changes in both cylinder chambers, can be It should be pointed out that integrator
expressed as: gain KI significantly depends on piston movement
direction.
dp A (t ) E ( pA )
= QA ( p A , xV , t ) −
dt V0 A + AA y (t )  1.2 Adaptive Control Structure
(3)
dy (t )
− AA − K Li ( p A (t ) − pB (t )) ] , A successfully applied control strategy
dt
based on conventional control theory is depicted

396 Detiček, E. – Župerl, U.

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

in Fig. 2. It consists of a digital self-tuning, The final form of an adaptive P-controller

parameter adaptive controller to cope with with initial model value bm (Koek) is:
parameter changes, a cancellation feed-forward   
bm (k ) b1 (k ), b1 (k ) ≥ b1 min
controller to minimize steady-state errors of ramp u AC (k ) = , b1 (k ) =  . (12)
and sinusoidal reference inputs and a switching b1 (k ) b1 (k − 1), else
integrator to improve precise final positioning of
hydraulic drive. The computer control algorithm also
contains a decision logic to separate computing
of (bm,left , b 1,left  ,bm,right , b 1, right  ), depending on
the movement direction. The effectiveness of the
above adaptive P-controller is experimentally
proven by step change of supply pressure (see Fig.
9). Supply pressure change has been avoided from
Fig. 2. Adaptive control structure for position any further test, while this is not a part of normal
control of electro-hydraulic drive operating conditions.
The feed-forward controller (UNbeha) is
It is important to obtain the digital realized on the basis of the inverse model Gm-1(z).
parameter adaptive controller, to be realizable on It has non-causal form but is realizable in control
a control computer, inside a sampling interval, computer for a partial compensation of steady-
which needs to be small enough to represent state errors by ramp and sinusoidal reference
electro-hydraulic servo drive dynamics well (in signals:
our case T = 2 ms). Therefore, the self-tuning
1
approach is used and a simplified mathematical w′(k ) =
bm
[ w(k + 1) + w(k )] . (13)
model Eq. (7) is taken to be the identification
model of the system. However, it shows that all It also depends on movement direction
changes are reflected on load pressure changes (bm,left , bm,right ). Switching integrator in form:
and therefore on KI as well. In that way it is the
function, not only of systems state variables, servo uSI (k − 1) + ∆uSI (k ), if emin ≤ e(k ) ≤ emax
uSI (k ) =  , (14)
valve rated flow, piston area and supply pressure  0 , else
but also of inertial, friction and external forces. is switched on to improve final position accuracy.
Putting Eq. (7) in to discrete form it gives: It is switched off just before the desired position
KI TK I b1 arrives to prevent limit cycle oscillations.
Gm ( z ) = (1 − z −1 ) z ( 2
)= −1
= . (8) Among the several advantages of the above
s 1− z 1 − z −1
control structure the main disadvantage is the
Therefore, a realizable, digital parameter inability to obtain automatic first start because the
adaptive (self tunning) controller, with only one initial value of the estimated parameter must be
unknown parameter b1 is obtained using recursive known in advance.
least square estimation (Iserm). The equation for To overcome this problem our further
unknown parameter vector estimation has the investigations lead to the replacement of the
general form: conventional adaptive controller by appropriate
 ( k + 1) = Θ
Θ  (k ) + γ (k )e(k + 1), fuzzy logic self learning (self organizing)
(9) algorithm.
with correction vector γ(k) :
P(k )Ψ (k + 1) 2 FUZZY POSITION CONTROL OF
γ (k ) = T , (10) ELECTRO-HYDRAULIC LINEAR DRIVE
Ψ (k + 1) P(k )Ψ (k + 1) + λ (k + 1)
where Ψ is the signal vector and P is the 2.1 Fuzzy Control
covariance matrix, with λ - forgetting factor:
Fuzzy logic, developed by [3] in 1960s,
P(k + 1) = 1 . (11)
λ (k + 1)  I − γ (k )Ψ T (k + 1)  P(k ) is much closer to human reasoning and natural

An Intelligent Electro-Hydraulic Servo Drive Positioning 397

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

language. In contrast to the traditional set the controller, translating the designer’s desired
theory, where elements are only, either complete control rules from linguistic representation to a
members of a set, or complete non-members of numeric computation. The inference engine can
a set. It provides an effective means of capturing be divided into three elements. The first step is
the approximate and inexact nature of the real known as “aggregation”. During this step the
world. Fuzzy sets have no precise criterion for premise (IF statement) of each rule in the rule
membership. base is calculated using the fuzzified controller
To distinguish between members that are inputs. With fuzzification, each condition in the
more probable than those which are less probable premise is assigned a degree of membership in
in belonging to the set, the grade of membership the corresponding input fuzzy set. It should be
denoted by μ which lies in the range of 0 to 1 is noted that the traditional formulation of logical
used. The fuzzy sets can have different shapes statements such as AND/OR has been modified
such as triangular, trapezoidal, bell-shape, etc. to accommodate the use of membership functions.
The elements of the fuzzy set are taken from the In particular, the result of an AND operation is
universe of all elements. often defined as either the minimum or a product
A principal characteristic of fuzzy control of two fuzzy values compared. Similarly, an OR
is that it works according to linguistic rules such operation is often defined as either the maximum
as “if the room temperature is low then increase or probabilistic sum. Any premise with a value
heating” rather than obeying mathematical models greater than zero means that its corresponding rule
and functional relationship. The fuzzy controller is active. The second step of the inference process
consists of four main elements: a fuzzifier, an is known as “composition” or “implication”,
inference engine, a rule base and a defuzzifier. where the consequent (THEN statement) of each
The first step of the fuzzy controller design rule is created using the premises calculated
is to select the number of controller inputs and in the first step. The output of the composition
corresponding number of input membership step is not a single value for each rule in the
functions required for the fuzzification process. rule base, but rather one modified output fuzzy
The fuzzification is the process of translating set for each rule. These modified output sets are
crisp input values into fuzzy linguistic values known as “implied” fuzzy sets. The modification
through the use of membership functions. In other is controlled by the premise calculated in the
words, determining how much each real value aggregation step. There are two fundamental
belongs to each fuzzy set using the corresponding methods of creating the fuzzy sets that are the
membership value. results of composition min and prod. The min
The fuzzy rule base is simply a data base operation truncates the output fuzzy set at the
of the desired control rules for the system. It is value of the premise while the prod operation
most equivalent to the controller of a traditional scales the output fuzzy set according to the
control system and formalizes the designer’s premise. The third and final step of the inference
“expert knowledge” of what control output should process is known as “accumulation” or “results
result from a given combination of system states, aggregation”. In this step the implied fuzzy sets
expressed in a linguistic manner. It often takes the that are the output of the composition process are
form of a truth table consisting of rules constructed combined into an accumulated fuzzy set, which is
in the following form: IF “condition 1” AND/ the input to the defuzzification. Defuzzification
OR “condition 2” THEN “consequence 1”. In is the process of converting the fuzzy output set
fuzzy logic terminology, the statement following which is the result of inference of the process into
the IF condition is known as the “premise”, a discrete number. Two fundamental methods are
“antecedent” or “condition”. The corresponding known as the Mean of Maxima and the Center of
statement following THEN is known as the gravity. The latter, also known as the Center of
“conclusion” or “consequent”. The inference Area method involves two steps:
engine is the heart of a fuzzy logic controller. • multiply the membership degree μ(xi) of
It acts as the bridge between the fuzzification each element i by the singleton value xi of the
input stage and defuzzification output stage of membership function.

398 Detiček, E. – Župerl, U.

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

• sum the values obtained in the first step of the decision table are for error e and its change
and divide total by the sum of the output e respectively. Each axis has 13 entries from -6
membership degrees. through 0 to +6. The points 1 to 6 can be taken to
mean: very small, small, medium, quite large and
2.2 Fuzzy PD Controller very large. The fuzzy sets have, for computational
simplicity, a triangular shape with μ - values of 0,
For stability reasons the general form of 0.1, 0.2, ..., 0.9, 1. The consequent of each rule in
PD-controller was used at the beginning: the matrix also has one of 13 values and represents
the contribution of each rule to the output action
u = f (e, e) . (15) set. The collective effect of all rules can be given
By fuzzification of two input variables as: (rule 1) ELSE (rule 2) ELSE (rule 3) ELSE...
(e, e ) and one output variable (u) of the controller, “ELSE” can be regarded as an inclusive OR
a two- dimensional rule base has been constructed. function or MAX function. Finally, the centre of
Linguistic variables e, e , and u are shown in Fig. gravity of the output action set was used, as the
3. The universe of discourse of each variable is method of defuzzification, to obtain unique real
selected in range 1:α, where α=AA/AB is piston output.
area ratio. From Fig. 4 it can be seen that the rule base
is diagonally symmetrical and the rules along the
diagonal change equally.

Fig. 3. Linguistic variables represented as fuzzy Fig. 4. Rule base or decision table
variables; a) e, b) e , c) u
The consequence is a linear dynamic
The rule base or decision table, shown behaviour of the controller. In other words,
in Fig. 4, contains linguistic rules of the form: the fuzzy controller emulates an exactly
IF (A and B ) THEN C. In the present context conventional PD-controller. This was practically
the antecedents A and B refer to linguistic proved by experiments. Due to fuzzyffication
statements about the error between set point and of measurement signals and robustness of fuzzy
actual position of the moving mass and the rate rules, the fuzzy controller is less sensitive to noise
of change of this error. The logic conclusion of and small parameter variations. It is therefore, to a
each rule is about the servo valve opening. Each certain extent, more robust than the conventional
non zero entry in the decision table corresponds one.
to one rule. Two axes or universes of discourse

An Intelligent Electro-Hydraulic Servo Drive Positioning 399

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

Fig. 5. a) Self-organizing mechanism of fuzzy logic controller; b) Self-learning mechanism of fuzzy logic
controller

On the other hand, it is possible to obtain hierarchical structure, which consists of two rule
nonlinear characteristics of the fuzzy PD- bases.
controller by modification of the symmetrical rule The first one is constructed as a base
base. of instructions and a base of performances
The following paragraph describes an simultaneously. It creates and modifies the
attempt of automatic rule base creation. main rule base according to the desired overall
performance of the system. At the first start of
2.3 Fuzzy Self-Learning Controller the controller, the main rule base contains zero
values and it begins to fill according to base of
The theory of fuzzy control seems to be instructions.
a suitable mathematical tool for both modelling The quality of control action is checked
and control of complex systems [3]. However, the by a comparison of present values of control
error e and e with expected values from the
primitive form of fuzzy control sometimes fails
base of performances. The linguistic terms in the
in dealing with these complex systems mainly
base of performances show how the rules must
because it lacks enough adaptability. Several
be changed. The question is which rule must be
researchers have devoted themselves to developing
changed, if we know that the present e and e are
fuzzy controller with more “intelligence” [1], [17] caused by a past control action. The solution is
and [18]. Unfortunately, most of them are difficult determining the time constant τ, which express that
for real time implementation. the control result at nth discrete moment of time
Therefore, we decided to implement the so is a consequence of control action activated at the
called reinforcement learning, proposed in [9] and (n ‒ τ)th discrete moment in the past. Determination
[19]. of τ is done on the base of heuristic knowledge
Self-learning fuzzy logic controller tries of dynamic behaviour of the control process.
to emulate human decision making behaviour As the investigations show, the selection of the
and learning, namely, an ability to make rules τ is not critical and can be made inside a certain
and to modify them based on experience. It has a range of values. The self learning fuzzy controller

400 Detiček, E. – Župerl, U.

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

therefore, uses control computer memory buffer to Fig. 6 shows the original forms of rule
save past information and contains the mechanism base and base of instructions at the beginning
for changing rules (Fig. 5a) or creating new rules of the control process. The rule base and the
(Fig. 5b), if they do not yet exist. It can be said base of instructions has integer entries ranging
that the controller also possesses the ability to self from -6 to +6. These entries are positive or
organize. negative reinforcements, which are given to the
Let us suppose the above mentioned corresponding (e, e ) pair in the rule base.
learning process is convergent, then the general Reinforcements also implicate rules
creation, in case they do not yet exist. Fig. 6b
rule base become constant after certain number
shows the matrix developed by [9], although
of experiments. As long as the changes in system
the other versions could be used. Learning
parameters are not significant, the rule base stays
proceeds as follows. Let us assume a delay of τ
constant; otherwise a new adaptation is needed.
samples (in our case τ = 4) on controlled system,
The self-learning and self-organizing then the present e and e values are mainly
mechanism is explained comprehensively in Fig. the consequences of controller output from τ
5, with only three linguistic terms for the sake of samples ago. Each past state consisting of a triplet
simplicity. (eτ, e τ, uτ) is stored in a control computer “history”
buffer. As the present values of e and e point to
the value p in the performance matrix, then the
rule entry for eτ, e τ in the rule base become uτ,+p,
whatever the rule entry may have been up to the
present time.
The self-organizing controller comprises
both the blank rule base fuzzy controller and the
above learning process.

3 EXPERIMENTAL RESULTS

The laboratory experiments were

performed on laboratory testing equipment shown
in Fig. 7. First the self tuning parametric adaptive
controller, according to Eq. (12), was tested. It
was compared with an ordinary P-controller (Fig.
8a), by ramp reference signals.
Adaptive ability was proven by a sudden
supply pressure drop of 20%, at the time of 0.3
s (Fig. 8b). The next experiments were step
responses of fuzzy logic controller. First the self-
organizing fuzzy logic controller was started with
zero rule base (Fig. 9a), while the learning process
was currently involved in the following tests (Fig.
9b).
The rule base becomes mostly constant
after the second and third step. Finally such an
”experienced” fuzzy controller, together with
feed-forward controller Eq. (13) and switching
Fig. 6. a) Rule base or decision table, b) base of integrator Eq. (14) were joined in, so called,
instructions “hybrid” control structure to obtain a ramp
response of the hydraulic drive (Fig. 10).

An Intelligent Electro-Hydraulic Servo Drive Positioning 401

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

a) b)

Fig. 7. a) Structure of the testing system (cylinder40/28 mm, slider mass 70 to 300 kg, lin. pot. 300 mm
±1%), b) Servo valve (MOOGD-769-233, Q = 19 l/min)

a) b)

Fig. 8. a) P-controller ramp response by 20% supply pressure drop, b) Self-tuning controller ramp
response by 20% supply pressure drop

Fig. 9. a) First step of learning process, b) Third step of learning process

The proposed hybrid fuzzy control first starting conditions [20] and [21]. It is also
structure is shown in Fig. 11. able to cope with non-linearity and parameter
The experiments show that the self- changes and therefore, possess a certain level
organizing fuzzy controller enables a successful of adaptability as well. However, the reference
control of electro-hydraulic drive without any tracking behaviour [22] and final positioning
a-prior knowledge of the system dynamics in the accuracy are still poor.

402 Detiček, E. – Župerl, U.

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

Fig. 10. Ramp response of electro-hydraulic servo drive with hybrid fuzzy position controller

Fig. 11. Hybrid fuzzy control structure of electro-hydraulic servo drive position control

4 CONCLUSIONS adaptation to parameter changes and to therefore,

deal with nonlinear dynamic behaviour associated
Systems having structural uncertainties or with the drive motion.
a known complicated structure such as hydraulic Unfortunately, there is no systematic way
systems are difficult to control. to prove the convergence of a learning mechanism
The dynamic characteristics of such
and overall stability of the control system. Trial
systems are usually very complex and highly
and error method was used with these experiments.
nonlinear. For a practical control system, it is
usually desired to have a fast accurate response Therefore, additional research will be required in
with a small over-shoot. the future.
In this paper, a possibility to apply a self-
organising and self learning fuzzy algorithm 5 REFERENCES
for position control of hydraulic servo drive is
represented. In addition to self-organisation it also [1] Shih, M.C., Tsai, C.P. (1995). Servohydraulic
possesses the ability of self learning. It is able of cylinder position control using a neuro-fuzzy

An Intelligent Electro-Hydraulic Servo Drive Positioning 403

 Strojniški vestnik - Journal of Mechanical Engineering 57(2011)5, 394-404

controller. Mechatronics, vol. 5, no. 5, p. 497- mechanisms in an electro-hydraulic system

512. ‒ Part I: Actuator’s fuzzy modeling.
[2] Backe, W. (1993). The present and future of IEEE Transactions on Systems, Man, and
fluid power. Proceedings of the Institution Cybernetics ‒ Part C: Applications and
of Mechanical Engineers, Part I: Journal of Reviews, vol. 30, no. 3, p. 305-316.
Systems and Control Engineering, vol. 207, [14] Jelali, M., Kroll, A. (2003). Hydraulic servo-
p. 193-212. system modeling, identification and control.
[3] Zadeh, L.A. (1965). Fuzzy Sets, Informat Springer, London, New York.
Control, vol. 4, no. 8, p. 126-174. [15] Kim, M.Y., Lee, C.O. (2006). An experimental
[4] Bouslama, F., Ichikawa, A. (1992). Fuzzy study on the optimization of controller gains
control rules and their natural control laws. for an electro-hydraulic servo system using
Fuzzy Sets Syst., vol. 48, p. 65-86. evolution strategies. Control Engineering
[5] Shih, M.C., Chen, P.C. (1991). An Practice, vol. 14, no. 2, p. 127-147.
experimental study on the position control [16] Xiang, G., Zheng-Jin, F. (2005). Design
of a hydraulic cylinder usinga fuzzy logic study of an adaptive Fuzzy-PD controller for
controller. JSME, Series III, vol. 34, no. 4, p. pneumatic servo system. Control Engineering
481-489. Practice, vol. 13, no. 1, p. 55-65.
[6] Chen, C.Y., Liu, L.Q., Cheng, C.C., [17] Čuš, F., Župerl, U. (2007). Adaptive self-
George, T. (2008). Fuzzy controller learning controller design for feedrate
design for synchronous motion in a dual- maximization of machining process.
cylinder electro-hydraulic system. Control
Advances in Production Engineering &
Engineering Practice, vol. 16, no. 6, p. 658-
Management, vol. 2, no. 1, p. 18-27.
673.
[18] Milfelner, M., Župerl, U., Čuš, F. (2004).
[7] Kastrevc, M., Pusenjak, R. (2005). Fuzzy
Optimisation of cutting parameters in high
pressure control of hydraulic system with
speed milling process by GA. International
gear pump driven by variable speed induction
Journal of Simulation Modelling, vol. 3, no.
electro motor. Exp. Tech. (Westport Comm.),
4, p. 121-131.
vol. 29, no. 3, p. 57-62.
[19] Lee, C.C. (1990). Fuzzy logic in control
[8] Berger, M. (1996). Self-tuning of a PI
controller using fuzzy logic for a construction systems. IEEE Transaction on Systems, Man
unit testing apparatus. Control Engineering and Cybernetics, vol. 20, no. 2, p. 24-32.
Practice, vol. 4, no. 6, p. 785-790. [20] Herakovič, N. (1995). Piezoaktorbetätigung
[9] Yamazaki, T. (1982). An improved algorithm für ein einstufiges hochdynamisches
for self-organising controller and its Servoventil = [Piezoactuator for a single-
experimental analisys. John Hopkins, stage servovalve with high dynamic
University Press, Baltimore. response]. O+P, Ölhydraul. Pneum., vol. 39,
[10] Mudi, R.K., Pal, N.K. (2000). A self-tuning no. 8, p. 601-605.
fuzzy PI controller. Fuzzy Sets Systems, vol. [21] Župerl, U., Čuš, F., Kiker, E., Milfelner,
115, no. 2, p. 327-338. M. (2005). A combined system for off-line
[11] Liu, G.P., Daley, S. (2000). Optimal-tuning optimization and adaptive adjustment of the
nonlinear PID control of hydraulic systems. cutting parameters during a ball-end milling
Control Engineering Practice, vol. 8, no. 9, p. process. Strojniški vestnik – Journal of
1045-1053. Mechanical Engineering, vol. 51, no. 9, p.
[12] Kim, S.M., Han, W.Y. (2006). Induction 542-559.
motor servo drive using robust PID-like [22] Roy, S.S. (2010). Modelling of tool life,
neuro-fuzzy controller. Control Engineering torque and trust force in drilling: a neuro-
Practice, vol. 14, no. 5, p. 481-487. fuzzy approach. International Journal of
[13] Branco, P.J.C., Dente, J.A. (2000). On Simulation and Modelling, vol. 9, no. 2, p.
using fuzzy logic to integrate learning 74-85.

404 Detiček, E. – Župerl, U.