IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO.

2, MARCH 1998 221

On Fuzzy Logic Applications for Automatic

Control, Supervision, and Fault Diagnosis
Rolf Isermann, Member, IEEE

Abstract— The degree of vagueness of variables, process de-

scription, and automation functions is considered and is shown
where quantitative and qualitative knowledge is available for
design and information processing within automation systems.
Fuzzy-rule-based systems with several levels of rules form the
basis for different automation functions. Fuzzy control can be
used in many ways, for normal and for special operating condi-
tions. Experience with the design of fuzzy controllers in the basic
level is summarized, as well as criteria for efficient applications.
Different fuzzy control schemes are considered, including cascade,
feedforward, variable structure, selftuning, adaptive and quality
control leading to hybrid classical/fuzzy control systems. It is then
shown how fuzzy logic approaches can be applied to process
supervision and to fault diagnosis with approximate reasoning
on observed symptoms. Based on the properties of fuzzy logic
approaches the contribution gives a review and classification of
the potentials of fuzzy logic in process automation.
Fig. 1. Multilevel process automation.
I. INTRODUCTION
been going on concerning the merits of fuzzy control versus
B ECAUSE of the progress of automatic control theory
and technology during the last 60 years, many technical
processes operate automatically under certain operating condi-
conventional control. This article tries to contribute to this
discussion by considering not only basic control loops but
tions. There exists a large variety of design and implementation also other functions in process control systems and process
approaches of continuous feedforward and feedback control automation.
or discontinuous logic and sequence control for any kind of
processes with all kind of behaviors. Most of the control II. VAGUENESS IN AUTOMATIC CONTROL
algorithms are based either on differential resp. difference
As fuzzy logic provides a systematic framework to process
equations or Boolean logic and are suitable to solve most
vague variables and knowledge the various automation tasks
control problems quite sufficiently.
will first be discussed with regard to the sources and the
The development of fuzzy logic theory [1]–[3] now stimu-
degree of vagueness. The goal is to determine some automation
lated alternative ways to solve automatic control problems.
functions for which fuzzy logic offers attractive possibilities
Based on these basic ideas of fuzzy logic Mamdani and
and advantages.
Assilian [4] proposed fuzzy controllers which describe human
Fig. 1 shows a general scheme for the organization of
control in linguistic form. Consequently the first applications
process automation tasks in several levels. Usually mea-
of fuzzy control replaced a human operator.
surement and manipulation takes place in the lower level.
For about 20 years, contributions on fuzzy control were
The next level comprises feedforward and feedback control.
presented at conferences and in the control literature, but
Here certain variables are adjusted according to conditions or
the field of fuzzy control did not obtain high attention for
reference variables. The second level comprises supervision
a longer period. It is only recently—since about 1990—that
which serves to indicate undesired or unpermitted process
the interest in fuzzy control has increased strongly because
states (monitoring) and to take appropriate actions such as fail-
of successes and advertisement of applications in Japanese
safe, shut-down, or triggering of redundance or reconfiguration
consumer products such as washing machines and camcorders.
schemes. The higher levels may perform tasks such as opti-
Moreover, successful fuzzy control of industrial processes,
mization, coordination or general management in order to meet
such as, e.g., a cement kiln in Denmark [5], in train operation
economic demands or scheduling. The lower levels usually
[6], or simulations for ship steering [64], was shown (see
need a fast reaction and act locally whereas the higher levels
also [7] and [8]). Since then, a controversial discussion has
are dedicated to tasks which do not require a fast response
Manuscript received February 27, 1995; revised February 5, 1997. and act globally.
The author is with the Darmstadt University of Technology, Institute of
Automatic Control, 64283 Darmstadt, Germany. Table I shows estimates of degrees of sharp and vague
Publisher Item Identifier S 1083-4427(98)00137-4. information for the main elements in process information: the
1083–4427/98$10.00  1998 IEEE
222 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998
TABLE I
DEGREES OF SHARPNESS AND VAGUENESS FOR MAIN ELEMENTS IN PROCESS AUTOMATION. INTENSITY OF A PROPERTY: +++ LARGE; ++ MEDIUM; + SMALL
input variables, the process classes, the automation functions
and the output variables. Sharp information is labeled to be
“crisp” and vague information to be “fuzzy.”
A. Input Variables
With regard to the type of input signals precise and impre-
cise values are distinguished. Usual measurement equipment
is designed to deliver signals with precise mean values and
low standard deviations. Low cost sensors may have imprecise
outputs with biased values and large standard deviations.
Nondirectly measurable variables like slip, material stress,
or comfort or quality measures can only be calculated or
estimated based on other measurement variables and analytical
models and are therefore imprecise to some extent (This is also
known as sensor fusion). Also linguistic values of a human
operator may be processed, like the color of fire in a cement
kiln, a brush fire of a dc motor or the type of noise of an
engine. Hence, the degree of precision and imprecision of the
input variables may vary within fairly broad ranges.
B. Process Classes
The description of the static and dynamic behavior in form
of mathematical models plays a crucial role as well for the
design of the technical processes as for a systematic design of
high performance control systems. As an example continuous
processes with continuous input and output signals are con-
sidered. If the internal process behavior can be described by
basic physical principles, theoretical modeling can be applied
by stating balance equations, physical-chemical state-equations
(constitutive relations) and phenomenological laws. Then the
structure of the mathematical model is obtained either as a
lumped parameter or distributed parameter system (ordinary
or partial differential equations). By this way quantitative
models based on physical laws result and one can distinguish
the following cases of analytical process models (quantitative
models):
A.1) equations with known parameters;
A.2) equations with partially unknown parameters;
A.3) equations with unknown parameters.
In practice, case A.2 appears more frequently than A.1 and
A.3. If the structure is known, the unknown parameters can
be determined experimentally by parameter estimation and
identification methods, e.g., [9]. Hence, quantitative process
models result from proper combination of theoretical and
experimental modeling. The field of process modeling has
reached a fairly mature status and is packaged in attractive
software tools.
However, if the internal behavior cannot be expressed by
physical laws in the form of equations, as for “not well
defined” processes like drying, vulcanizing, special chemical
reactions or biological processes, some qualitative information
on the causalities may be expressed as rules:
(1)
The condition part then contains as inputs the memberships
of facts to the premise and possibly their logical connection
by AND, OR and NEGATION, and the conclusion part de-
scribes the logical consequence on the output. Then one can
distinguish heuristic process models (qualitative models):
B.1) rules with known conditions, known conclusions and
known relations between conditions and conclusions;
B.2) rules with unknown logical connections within the
condition, but known conclusion;
B.3) rules with known condition and unknown conclusion.
Hence, a gradual decrease of knowledge on the process
can be observed from analytical process models, case A.1,
to heuristic process models, case B.3. This is also shown in
ISERMANN: FUZZY LOGIC APPLICATIONS
Fig. 2. Reachable model accuracy in dependence on the degree of quantita-
tive or qualitative knowledge of processes.
Fig. 3. Reachable model accuracy for different processes in dependence on
the dimension of its elements.
Fig. 2 where the degree of knowledge ranges from precise
to vague. Precise quantitative models can be obtained for
mechanical and electrical processes, fairly precise, and less
precise models for thermal and chemical processes. Less vague
qualitative models correspond, e.g., to biological processes and
fairly vague and vague models to socioeconomic processes,
compare Table I.
However, the model accuracy also depends on the size of
elements or components, i.e., the dimension of a process. With
increasing dimensions, in general, the model accuracy de-
creases (see Fig. 3). If processes with only qualitative knowl-
edge are of high dimension, complex processes result which
naturally have a low reachable model accuracy.
C. Automation Functions
The functions of process automation are usually organized
according to the levels shown in Fig. 1. Feedforward control
systems are characterized by an open signal flow and can be
divided in continuous and discontinuous signal flow systems.
In the last case one distinguishes between logic control for
which input and output signals, according to Boolean algebra,
storages and time functions are related to each other and
sequence control for which the switching from one step to the
following depends on the switching condition. The outputs are
usually binary variables (on-off-signals). Hence, crisp inputs
and outputs are processed.
223
Feedback control systems have a closed-loop structure to
adjust control variables to given reference variables. Usually
both the control variables and the manipulated variables are
crisp values. The known controller equations such as PID-
controllers or state-controllers with observers lead to precise
algorithms. Because the knowledge on the processes is vague
in most cases the tuning of PID-controller parameters is
frequently performed by tuning rules. Only for designing
sophisticated and safety critical control systems precise math-
ematical process models are required usually.
For supervision a basic task is to check if variables ex-
ceed tolerances. These tolerances are generally a compromise
between small values to early detect faulty behavior and
large values to avoid false alarms subject to normal process
fluctuations. Hence, in reality these tolerances are not crisp
but fuzzy variables. Moreover, in many cases fault diagnosis
means to treat vague process knowledge and therefore to apply
rules.
The process management includes optimization with re-
gard to economical performance or coordination of several
processes. Here both the knowledge on the process and the
criteria are frequently not precise and therefore a good part
of qualitative knowledge has to be treated. The man-machine
interface is designed for human interaction. Basic tasks of the
human operator are higher level tasks as, e.g., supervision,
management and redundancy for all control functions. The
actions of human operators are based on qualitative knowledge
and may be better described by rules than by precise algo-
rithms. Hence, the kind of information in process automation
generally changes from using more precise algorithms at the
lower levels to using more rules at the higher levels, as
summarized in Table I.
D. Output Variables
Actuators with electrical, pneumatic, or hydraulic drives
need crisp inputs. Also on-off actuators like switches have
crisply defined switching points. Outputs to the human oper-
ator through the man-machine interface can be either crisp, if
numbers are indicated or fuzzy, if they are linguistic.
In summary this discussion and Table I show that vagueness
appears in different form and degree dependent on the type
of variables, process classes and automation tasks. For the
different automation functions Table II considers the degree of
vagueness if a fuzzy logic approach is applied. The vagueness
shows up at the input and output variables and in the rules.
Because the degree of vagueness increases from the left to
the right, it can be stated that the vagueness (qualitative
knowledge) in general increases with increasing automation
level.
III. INTELLIGENT CONTROL SYSTEMS
After the recent development of model-based and adaptive
control systems the present development in control systems
can be summarized with “intelligent” control. This new devel-
opment now amplifies the treatment of qualitative knowledge.
Various definitions do exist for intelligent control (see, e.g.,
[10]–[16]). In comparison to human beings first developments
approach a low degree of intelligence. A “low-degree intel-
224 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998

TABLE II
DEGREES OF VAGUENESS FOR PROCESS AUTOMATION FUNCTIONS WITH A FUZZY LOGIC APPROACH. 11 VAGUE; 1 LESS VAGUE;  CRISP; 0 NOT DEFINED

ligent” controller has, e.g., the “ability to model, reason and

learn about the process and its control within a given frame,”
[17]. The intelligent controller then adapts the control to the
mostly nonlinear process behavior (adaptation) and stores its
controller parameters in dependence on situations (learning),
supervises all relevant elements and performs a fault diagnosis
(supervision) to request for maintenance, to shut off faulty
components, or to fail safe (decision on actions).
Fig. 4 shows the organization of an “advanced intelligent”
automatic control system. It consists of multicontrol levels,
a knowledge base, inference mechanisms, and communica-
tion interfaces. Both the knowledge base and the inference
mechanism treat quantitative and qualitative knowledge. The
processing of qualitative knowledge may be performed by
methods of soft-computing [3], including fuzzy logic, neuronal
networks and genetic optimization algorithms. This figure
indicates that intelligent control systems can be considered
as on-line, real-time expert systems. It can also be stated that,
in general, the hierarchical structure of an intelligent system
shows a decreasing precision with increasing intelligence [11].
This also means that with decreasing quantitative knowledge
and increasing qualitative knowledge in general more intelli-
gence is required. Hence, as well for conventional automation
systems (section II) as for intelligent control systems it holds
that the potentials of fuzzy logic approaches in general increase
with higher automation levels.
IV. FUZZY-RULE BASED SYSTEM
Fig. 5 shows a general scheme of a fuzzy IF-THEN rule- Fig. 4. Advanced intelligent automatic system with multicontrol levels,
knowledge base, inference mechanism, and interfaces.
based system [4], [8], [18], [19] consisting of the following
modules:
• fuzzfication: (converting of crisp facts into fuzzy sets • inference: The fuzzy IF-THEN rule expresses a fuzzy
described by linguistic expressions); implication relation between the fuzzy sets of the premise
ISERMANN: FUZZY LOGIC APPLICATIONS
Fig. 5. General scheme of a fuzzy IF-THEN rule-based system.
Fig. 6. Signal flow in a fuzzy rule-based system with two levels of rules and max/min operations and singletons as inputs.
and the fuzzy sets of the conclusion. If the rule is inter-
preted by the Mamdani-implication and the compositional
rule of inference by Zadeh [3] is applied, the following
steps can be distinguished:
1) matching of the facts with the rule premises (determina-
tion of the degrees of fulfillment of the facts to the rule
premises fit values);
2) if the rule contains logic connections between several
premises by fuzzy AND or fuzzy OR the evaluation
is performed by t-norms or t-conorms (The result gives
then the degree of fulfillment of the facts to the complete
premise);
3) the evaluation of the resulting conclusion follows the
max-min composition [3]. (The degree of fulfillment of
the premise restricts the fuzzy set of the conclusion.)
• accumulation of all conclusion fuzzy sets if several rules
contribute to the output (a union operation yields a global
fuzzy set of the output).
• defuzzification to obtain crisp outputs (various defuzzifi-
cation methods can be used, as, e.g., center of gravity,
maximum-height, and mean of maximum, to obtain a
crisp numerical output value).
The computational effort of a fuzzy-rule-based system in-
creases strongly if the number of fuzzy sets and the number of
225
rules grow. By special assumptions the computations can be
simplified considerably [20], [21] for example by singletons
at the input and output or universal sets at the output.
Fig. 6 shows the signal flow for a fuzzy-rule-based system
with two levels of rules and singletons as inputs. In this case
fuzzification and matching reduces to directly determining the
degree of fulfillment of the premises. The evaluation of the
rule conclusion then consists of applying the respective -
norms and -conorms (e.g., min-max or prod-sum operation)
for the logic operations in the premises and determination of
the fuzzy sets of the conclusion.
The chaining of rules can be performed with or without
defuzzification. In Fig. 6 no defuzzification is considered
between the two levels of rules.
According to the knowledge base the rules can be repre-
sented by a network, building up several levels. For fuzzy
control usually only one level of rules is required. However,
fault diagnosis leads to tree structures with several levels. The
contributions of all rules to the last level have then to be
accumulated, e.g., by a union operation. This usually results in
combined fuzzy sets. A crisp output as a final result is obtained
by defuzzification, if required. In the next two sections different
kinds of using fuzzy logic approaches in control systems are
considered for both normal and special operating conditions.
226 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998
Fig. 7. Signal flow of a fuzzy controller with two inputs (2 from 6 rules
are shown).
V. FUZZY-FEEDBACK CONTROL FOR
NORMAL OPERATING CONDITIONS
With regard to automatic feedback control normal operat-
ing conditions are understood where automatic operation is
possible without manual control.
A. Fuzzy Control in the Basic Level (Direct Fuzzy Control)
Before discussing more general aspects of fuzzy control
applications some practical aspects of fuzzy control are first
considered. For the design of a fuzzy controller the signal
flow of a general fuzzy-rule-based system as shown in Fig. 5
reduces to a rule activation block with one rule level, as shown
in Fig. 7. The two inputs and are, e.g., the control
deviation and its derivative
(2)
and are assumed to be crisp values.
The membership functions of the inputs (attributes) are
usually symmetric triangles with equal distribution over the
universe of discourse or ramps, such that the sum is always
one. The formulation of the attributes of the output variable
shows frequently more variety [22]. Now a special controller
configuration is considered. The effect of choosing different
output attributes like rectangles, triangles, ramps, etc., has been
investigated in [21]. For two ramps as inputs and and
center of gravity defuzzification, the analysis of the controller’s
transfer behavior yielded the following results.
1) In selecting the output attributes the position of their
center of gravity is mainly important.
2) The shape, the width of the output attributes and the
inference method (max-min or max-prod.) are of minor
importance. They influence the interpolation between the
centers.
3) The width of rectangles and triangles has no influence, as
long as the membership functions do not overlap each
other.
4) Overlapping triangles or ramps (mountains) generate S-
shaped nonlinear characteristics which are, however, not
much different from a linear one.
5) Formulation of the linguistic output attributes as sin-
gletons leads to a drastic reduction of the computational
effort. The activation of an output attribute which
is a singleton at by a crisp input becomes
(3)
Hence, a singleton of height at position
results. Then it makes no difference any more, if as -
norm the minimum- or product-operator is applied. For
the accumulation the maximum of all singletons is taken
and defuzzification follows via the center of singletons
(4)
A similar procedure follows for multiple inputs, leading
to simple vector multiplications.
6) Drastic nonlinear characteristics can mainly be reached
by distributing the output attributes unequally over the
universe of discourse.
B. Criteria for the Application of Fuzzy
Control in the Basic Level
To discuss the potentials of fuzzy controllers in the basic
level (direct control) one has to consider the following basic
control engineering viewpoints.
1) The proper design of the static and dynamic behavior
of the process is the most important prerequisite for
reaching a good control performance.
2) The engineering design includes the right specification
and design of the actuator and the measurement equip-
ment, such that the process can at least be controlled
manually.
3) The solution of these problems presupposes the exis-
tence of models of the process, at least rough static
characteristics. Hence, simple models frequently exist
for important processes.
4) In most cases linear PI- and PID-controllers are ap-
plied and the parameters are tuned by tuning rules.
Mathematical models are not required for the selection
of the controller type and the parameter tuning. A
rough understanding of the process’ control behavior is
sufficient in most cases.
5) With modern microprocessor-based controllers it is no
problem to generate a static nonlinear behavior like non-
linear characteristics, dead zones, direction dependent
control actions, etc.
Fuzzy-rule-based controllers as in Fig. 7 and Fig. 8(a) gen-
erate a three-dimensional static mapping of the output in
dependence on the two inputs (look-up-table). Depending on
the selection of the membership functions this mapping may
become nonlinear and depending on the number of inputs
multidimensional. By adding dynamic elements at the input
or the output (e.g., derivative or integrating terms) the fuzzy
controller gets dynamic properties.
ISERMANN: FUZZY LOGIC APPLICATIONS
(a)
(c)
(e)
(g)
(i)
Fig. 8. Control system structures with different fuzzy control components (hybrid classical/fuzzy control). F: fuzzification; I: inference; and D: defuzzification.
Nonlinear characteristic surfaces can of course also be
generated directly. The advantage of the fuzzy-rule-based
approach lies in the formulation of the input-output behavior
in form of linguistically expressed rules. By this way the
practical experience of human operators can be implemented
in another form than by (simple) equations, which may be
easier to understand for some operators and nonexperts.
227
(b)
(d)
(f)
(h)
(j)
The control engineering aspects for the application of fuzzy
controllers are still under investigation. Some advantages and
disadvantages can be summarized as follows.
1) Nonlinear processes. For processes which can be lin-
earized and processes with analytically known nonlinear
behavior, such as nonlinear characteristics or hysteresis,
many solutions do exist with classical control sys-
228 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998

tems like PID-, state space controllers with observers, TABLE III
cancellation-, deadbeat-, minimum-variance and predic- CONTROLLER STRUCTURE IN DEPENDENCE
ON THE PROCESS BEHAVIOR AND SIGNALS
tive controllers [23]. In most cases only rough process
model knowledge is required, especially, if self-tuning
or adaptive control methods are applied [9].
For processes where no sufficient control perfor-
mance can be reached with PI- and PID-controllers
and parameter adjustment by tuning rules, fuzzy control
may be an alternative. This is especially valid for
processes with difficultly understandable behavior, if
only qualitative knowledge is available and strongly
nonlinear properties do exist. Some examples are milling
auxiliary controller and a main controller [Fig. 8(c)] [23]. The
plants, glass melting, cement kilns, food processing.
auxiliary controller can, e.g., be a classical P- or PI-controller
2) Knowledge of human operators. If manual control of
and the main controller can be a fuzzy controller which is
a difficult process is possible the operator knowledge
especially designed for the nonlinear behavior of process part
may be formulated in fuzzy rules with linguistic inputs
and outputs (This is in many cases not an easy task).
Another interesting possibility is a cascade control system
The parametrization can be improved especially by
with a fuzzy main controller for two different control variables
trial and error and also additional inputs can be used,
[Fig. 8(d)]. Here fuzzy rules decide which one of both control
e.g., in the sense of feedforward actions. However,
variables is in action with a smooth transition from one to the
the fine parametrization may take a long time, since
other. An example is the velocity and distance control in car
much more parameters have to be adjusted than for
driving [31].
PID-controllers. Another application is the imitation
of human controllers in the loop, as “fuzzy control D. Fuzzy Feedforward Control
involves a representation of what is basically a human If an external disturbance of a process can be measured
solution in the form of fuzzy if-then rules,” [3]. This before it acts on the output variable the control perfor-
kind of application is, e.g., interesting for car driving, mance with respect to this disturbance can be improved by
commanding aircraft, operating a crane or workshop feedforward control. Then the fuzzified measured disturbance
activities. Here the actions of a human can be formulated is just added to the premise of the rules of the fuzzy feed-
in rules and can be adapted by identification methods back controller [Fig. 8(e)]. Thus, proportional acting nonlinear
[21], [24]. Hence, fuzzy controllers can be used to feedforward control can be realized.
investigate human’s behavior.
3) Fast prototype design. A further advantage of the fuzzy E. Variable Control Structures with Fuzzy Weighting
control concept is a fast prototyping by some few rules,
if a classical standard controller is not suitable. However, The selection of classical controllers depends among others
the fine tuning for reaching a good control performance on process properties such as the transfer behavior of the plant,
usually lasts too long. the size of control deviations and the kind of disturbances [23].
Therefore the type of controller (and its parameters) should
A still holding disadvantage of direct fuzzy control is the
be altered, if such properties change. This is the case if for
difficulty of stability analysis. In fact, the proof of stability
example the size of control errors change drastically or if wide
for arbitrary fuzzy controllers is not a simple task due to
process operating ranges have to be governed.
their various structures and nonlinear characteristics. Several
Table III shows some examples for changing (classical)
approaches for stability analysis have been presented in recent
controller structures. Usually only one controller type with
years. [25]–[27] give an overview of existing methods. Most
robust properties is implemented or two controller types are
of them are adopted from nonlinear control theory, e.g.,
switched on and off alternatively. A fuzzy logic approach now
Lyapunov method, harmonic balance, Popov criterion etc. The
offers a gradual change from one controller type to another by
examples show that both time domain and frequency domain
weighting the control deviation by for controller 1 and
approaches are applied. It can be distinguished between meth-
for controller 2 based on the (e.g., identified) process
ods considering the fuzzy controller as nonlinear input-output
behavior [Fig. 8(f)]. Dependent on the process behavior rules
mapping, e.g., [8], and those regarding the internal structure of
can be established to weight the controllers action (and to
the fuzzy controller such as [28]–[30]. The latter approaches
change its parameters). One example is IF disturbances have
can also be used in a constructive manner in order to design
higher frequencies AND control deviation is medium THEN
robust fuzzy controllers.
decrease weight of PI-controller AND increase weight of
The next sections show other forms of applying fuzzy
PD-controller (see also [32]).
concepts for control systems.
C. Fuzzy Cascade Control Systems F. Fuzzy-Tuning of Classical Controller Parameters
The control performance of a single loop system can be In practice the parameters of P-, PI- and PID- con-
improved by expanding it to a cascade control system with an trollers are mostly adjusted by tuning rules based on
ISERMANN: FUZZY LOGIC APPLICATIONS
closed loop oscillation experiments with a P-controller
(Ziegler–Nichols rule) or open loop transfer function mea-
surement (Chien–Hrones–Reswick rule). In [24] and [33], a
parameter tuning procedure with if-then fuzzy rules [Fig. 8(h)]
is shown for PI- and PID-controllers. The controller starts with
small values of the parameters. Measured features of the closed
loop transient function such as the overshoot and the ratio rise
time/settling time are fuzzified and rules for the changes of the
controller parameters allow to adjust the controller parameters
after some few (2–8) step response experiments. One example
of the 24 tuning rules for a PI-controller is:
IF overshoot is too small AND settling ratio is alright
THEN increase gain by 30% The tuning rules hold for
a large variety of process behaviors and also for stochastic
disturbances.
G. Fuzzy Adaptive Control
Adaptive control systems adapt their control behavior to
the changing properties of the controlled processes and their
signals. Most adaptive control systems can be divided into
feedforward and feedback adaptive controllers [9]. Fig. 8(g).
shows a scheme for a feedforward adaptive control which
is based on the fact that the changing properties of the
plant can be grasped by the measurement of a signal
on the process (e.g., mass flow, speed, load). If rules are
known how the controller parameters of a classical (or fuzzy
controller) must be changed in dependence on the measured
signal, a fuzzy rule adaptation can be realized. If the process
behavior cannot be determined directly by the measurement
of external process signals, feedback adaptive control has to
be used [Fig. 8(h)]. The changing properties of the process
have to be determined by the measurement of different internal
control loop signals. Classical feedback adaptive control is
then based on the identification of a mathematical process
model (model identification adaptive control) or a comparison
with a reference model (model reference adaptive control)
with the goal to design a controller automatically. A fuzzy
feedback adaptive controller does not use a mathematical
process model but rules for the adaptation of the controller
based on measured closed loop signals. For example the
controlled and manipulated signals are used to determine
measures for the control performance. After the fuzzification
of these measures tuning rules of the controller determine
the adjustment of the parameters of a classical or a fuzzy
controller.
A further possibility results from using a fuzzy model of
the process. The expert knowledge is here used to model the
process by rules and measured signals to refine the model
with identification procedures [34]–[36]. The controller design
then is performed automatically. The controller types may be
classified as predictive controller [35], PID-controller [36],
feedback linearization, etc. or fuzzy controllers. This results
then in adaptive control with fuzzy models.
H. Fuzzy Control of Nonmeasurable Quantities
Sometimes a control variable is not measurable or a sensor
should be saved because of economic reasons. Then one can
229
try to use already available other sensor signals as inputs
[see Fig. 8(i)]. If analytical relations are known the control
variable can be calculated, e.g., a heat stress from measured
temperature variations. However, if mathematical models are
unknown or too complicated, one can formulate rules. An
example is the room temperature heating control without
indoor and outdoor sensors. The basic water temperature is
determined by the average temperatures over the year and
during a day (yearly and daily temperature profile). This rough
forecast is corrected by AND connections in the premise
with measured inputs as: average energy consumption of the
previous day and present energy consumption [37].
I. Fuzzy Quality and Comfort Control
The quality of products is mostly based on several proper-
ties. Examples are as follows.
1) Metal products: geometrical precision, roughness of sur-
face, texture, stiffness, hardness.
2) Ceramic products: geometry, color, surface, acoustic
sound.
3) Textile products: elasticity, strength, texture, color, pat-
tern, homogenity, bulking, skin reception, washing, and
iron properties.
4) Cement: particle size and its distribution, specific sur-
face (Blaine number), sieve residues, early strength and
strength after 28 days of binding.
5) Tyre: stiffness, traction, durability, noise, rain and aqua-
planing sensitivity, temperature sensitivity.
The various product properties which accumulate to an
overall quality measure can be influenced during processing by
changing of reference values of control loops, reaction times,
load variables, ingredients, etc. As frequently no analytical
models are known, but only rules, a multivariable quality fuzzy
controller with quality attributes as inputs and manipulated
process variables as outputs can be formulated [Fig. 8(b)].
A similar situation can be found for the control of comfort
where the human perception plays a role. Examples are as
follows:
1) The comfort of the climate in air conditioning where
the temperature, humidity and kind of labor of human
beings follow certain rules.
2) The comfort of the passengers during car driving where
the received frequency spectra of accelerations can be
formulated as rules with the adjustable shock absorbers
as outputs and road conditions and accelerations as
inputs. In [38] the simulation of a fuzzy controller is
shown where rules are given for the adjustment of the
shock absorber dependent on the relative speed of the
damper and the speed of the chassis.
3) The comfort of the start-up of automobiles with manip-
ulation of the clutch and gear shifting [39].
J. Neuro-Fuzzy Control
An increasing activity can be observed in the field of neuro-
fuzzy control systems [40]–[42]. The combination of fuzzy
logic and neural networks integrates the advantages of both
approaches in one controller. On the one hand, the controllers
230 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998
can be designed in the fuzzy sense using heuristic knowledge,
on the other hand, they can be trained from learning data by
means of algorithms developed for neural networks. Various
training algorithms have been presented for both on-line and
off-line training. Two general strategies in neuro-fuzzy control
can be distinguished: The most simple proposals use two
separate components (one fuzzy system, one neural network)
in a parallel control structure. However, approaches utilizing
the structural equivalence between certain types of fuzzy
systems and artificial neural networks [43]–[45] seem to be
more promising. One single controller can be trained from data
(interpretation as neural network) and still remain transparent
due to the internal fuzzy-rule structure (interpretation as fuzzy
systems).
VI. FUZZY-CONTROL FOR SPECIAL OPERATING
CONDITIONS AND LARGE-SCALE DYNAMIC EFFECTS
Many processes are automatically controlled for normal
operating conditions with sufficient control performance. With
regard to feedback control systems normal operating condi-
tions mean that stable or not too sluggish control is possible
with a fixed controller or an adaptive controller (mostly
feedforward adaptation of controller parameters). This is for
example the case for the control of flow or pressure, heat
exchangers or steam boilers within a range of about 20%
to 100% of the load. Especially for small loads closed loop
operation with the normal controller is frequently not possible
as the gain of the process and its time constants increase too
much or the valves operate in nearby closed position with
undefined characteristics.
Other examples of abnormal process behavior are
1) the start-up and shutdown phase of engines, machines,
industrial processes;
2) mode changes in chemical processes;
3) spinning of cars outside of normal tire/road slip (unstable
steering behavior);
4) large slip of tires and blocking of wheels after heavy
braking;
5) stalling of an aircraft after falling short of speed;
6) control of processes after failures (emergency control).
In these cases the process behavior is frequently not well
known because some physical laws do not hold any more or
are not known (e.g., change from turbulent to laminar flow in
heat exchangers, rotating of combustion engines with irregular
speed, chattering in metal cutting processes). Hence, there is
need to control processes with large uncertainties and large-
scale dynamic effects [46]. The usual way for governing the
abnormal process is manual control with experienced operators
by bringing the process back into normal or safe conditions.
Now a special fuzzy controller can be designed by making
use of the rules gained by the operator [Fig. 8(j)]. It is typical
for these abnormal operating conditions that the manipulation
has to be performed differently in comparison to the normal
operation, e.g., by using fast and large actions as in the case
of spinning cars or the recovery of an aircraft from the post-
stalling phase. The special fuzzy controller may then use an
additional manipulated variable and controlled variable
[Fig. 8(j)].
The discussion in Sections III and IV has shown that the
potentials for fuzzy logic approaches in the feedback control
domain are mainly
1) fuzzy control in the basic level for nonlinear processes
with only qualitative knowledge or if imitation of human
controllers is required;
2) addition of fuzzy control to classical control (cascade-
control, feedforward control, variable control structures);
3) fuzzy control in the adaptation level;
4) fuzzy control for nonmeasurable quantities, quality or
comfort measures;
5) fuzzy control for abnormal operating conditions and
large-scale dynamic effects.
Hence, fuzzy control approaches may be especially advan-
tageous as an addition to classical control systems, forming
hybrid classical fuzzy control systems. Then different fuzzy
control components can be specified, as shown in Fig. 8.
VII. SUPERVISION, FAULT DETECTION AND DIAGNOSIS
A. Knowledge-Based Fault Detection
Within automatic control of technical systems supervisory
functions serve to indicate undesired or unpermitted process
states and to take appropriate actions in order to maintain the
operation and to avoid damages or accidents. The following
functions can be distinguished:
a) monitoring: measurable variables are checked with re-
gard to tolerances and alarms are generated for the operator.
b) automatic protection: in the case of a dangerous process
state, the monitoring function initiates automatically an appro-
priate counteraction.
c) supervision with fault diagnosis: based on measured
variables features are calculated, symptoms are generated via
change detection, a fault diagnosis is performed and decisions
are made for counteractions.
Fig. 9 shows an overall scheme of knowledge based fault
detection and diagnosis. The main tasks can be subdivided in
fault detection by analytic and heuristic symptom generation
and fault diagnosis [47], [48].
1) Analytic Symptom Generation: The analytical knowl-
edge on the process is used to produce quantitative, analytical
information. Based on measured process variables a data
processing has to be performed to generate first characteristic
values by
1) limit value checking of directly measurable signals.
Characteristic values are exceeded signal tolerances.
2) signal analysis of directly measurable signals by use of
signal models like correlation functions, frequency spec-
tra, autoregressive moving average models (ARMA).
Characteristic values are, e.g., variances, amplitudes,
frequencies or model parameters.
3) process analysis by using mathematical process mod-
els together with parameter estimation, state estimation
and parity equation methods. Characteristic values are
parameters, state-variables or residuals.
ISERMANN: FUZZY LOGIC APPLICATIONS 231

Fig. 9. Scheme of a knowledge-based fault detection and diagnosis.

In some cases special features can be extracted from these
characteristic values, e.g., physical defined process coefficients
or special filtered or transformed residuals. These features
are then compared with the normal features of the nonfaulty
process and methods of change detection and classification are
applied. The resulting changes (discrepancies) of the described
direct measured signals, signal models or process models are
then analytic symptoms of faults.
2) Heuristic Symptom Generation: In addition to the symp-
tom generation with quantifiable information heuristic symp-
toms can be produced by using qualitative information from
human operators. Through human observation and inspection
heuristic characteristic values in form of special noise, color,
smell, vibration, wear and tear, etc. are obtained. The process
history in form of performed maintenance, repair, former
faults, and lifetime, load measures constitute a further source
of heuristic information. Statistical data (e.g., MTBF, fault
probabilities) achieved from experience with the same or sim-
ilar processes can be added. By this way heuristic symptoms
are generated which can be represented as linguistic variables
(e.g., small, medium, large) or as vague numbers (e.g., around
a certain value).
B. Model-Based Fault Detection Methods
Different approaches for fault detection by using mathemat-
ical models were developed in the last 20 years. The task
consists in the detection of faults in the technical process
including actuators and sensors by measuring the available
input and output variables and . The process is
considered to operate in open loop. A distinction can be
made between static and dynamic, linear and nonlinear process
models.
Three important model-based fault detection methods and
their residuals are:
1) parameter estimation: changes of parameter estimates
;
2) state estimation: changes of state estimates or
output errors ;
3) parity equations: output error or polynomial error
.
For more details, see [47]. The measured or estimated
quantities like signals, parameters, state-variables or residuals
are usually stochastic variables with mean value and
variance
(5)
as normal values for the nonfaulty process. Analytic symptoms
are then obtained as changes
(6)
with reference to the normal values. Fig. 10(a) shows an ex-
ample for a stochastic symptom If a fixed threshold
is used, a compromise can be made between the detection of
small faults and false alarms because of short term exceeding
[see Fig 10(b)]. Methods of change detection which estimate
the change of the mean value and relate it to the standard
deviation may improve the (binary) decision. Examples are
likelihood-ratio test, Bayes-decision, run-sum test, two probe
-test. Now a triangle-membership function is specified
with the center as the mean and the lower width or
because for a Gaussian distribution 68% or 95% of all
values lie within those intervals [Fig. 10(c)], thus obtaining
a combined symptom representation of the mean and the
232 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998

(a) (b) (c)

()
Fig. 10. Crisp and fuzzy threshold for a stochastic variable S t with standard deviation  . (a) Time history with a change of mean by 1 S for t > Tf .
(b) Membership function of a crisp threshold. (c) Membership function of a fuzzy threshold and matching with current value.

standard deviation. By matching the current value with

the symptom’s membership function “increased”

one obtains a gradual measure for exceeding a threshold, a

fuzzy threshold. (a) (b)

C. Fault Diagnosis Methods

The task of fault diagnosis consists of determining the type
of fault with as much details as possible such as the fault
size, location and time of detection. The diagnosis procedure
is based on the observed analytical and heuristic symptoms and
the heuristic knowledge of the process, as shown in Fig. 9.
In this section the heuristic part of the knowledge and
(c)
inference mechanisms for diagnosis are described in order to
Fig. 11. Examples for membership functions of features S resp. symptoms
build up on-line expert systems for fault diagnosis.
1) Unified Symptom Representation: For the processing of
1 S represented as fuzzy sets (a) symptom increases, (b) symptom decreases,
and (c) symptom increases or decreases with linguistic terms (labels).
all symptoms in the inference mechanism it is advantageous
to use a unified representation. One possibility is to present
The condition part (premise) contains facts in the form
the analytic and heuristic symptoms with confidence numbers
of observed symptoms as inputs and the conclusion
and treatment in the sense of probabilis-
part includes events and faults as a logical cause
tic approaches known from reliability theory [49]. Another
of the facts. This procedure results in fault-symptom trees
possibility is the representation as membership function
(“directed graphs”), relating symptoms to events and faults.
of fuzzy sets [50] (Fig. 11). By these fuzzy sets and
Then symptoms or events are associated by AND and OR
corresponding membership functions all analytic and heuristic
connectives and the rules become of form
symptoms can be represented in a unified way within the range
[0, 1]. These integrated symptom representations are the inputs
for the inference mechanism (Fig. 9).
2) Heuristic Knowledge Representation: For establishing
(7)
heuristic knowledge bases for diagnosis several approaches
do exist. In general specific rules are applied in order In the classical fault-tree analysis the symptoms and events
to set up logical interactions between observed symptoms are considered as binary variables with for happened
(effects) and unknown faults (causes). The propagation from and for not happened [51].
appearing faults to observable symptoms in general follows 3) Diagnostic Reasoning: Based on the available heuristic
physical cause-effect relationships where physical properties knowledge in form of heuristic process models and weighting
and variables are connected to each other quantitatively and of effects different diagnostic forward and backward reason-
also as functions of time. However, the underlying physical ing strategies can be applied. Finally the diagnostic goal is
laws are frequently not known in analytical form or too achieved by a fault decision which specifies the type, size and
complicated for calculations. Therefore heuristic knowledge location of the fault as well as its time of detection. A review
in form of qualitative process models can be expressed in of recent developments in reasoning oriented approaches for
form of IF-THEN rules. diagnosis was given in [52]. By using the strategy of forward
ISERMANN: FUZZY LOGIC APPLICATIONS 233

chaining a rule, the facts are matched with the premise and 5) fault decision by maximal possibility;
the conclusion is drawn based on the logical consequence 6) defuzzification to obtain a crisp size of the most possible
(modus ponens). Therefore with the symptoms as inputs fault.
the possible faults are determined using the heuristic Multiple rule activation from all symptoms to all faults
causalities. In general the symptoms have to be considered (vertical procedure) [53]:
as uncertain facts. Therefore a representation of all observed 1) chaining by
symptoms as membership functions of fuzzy sets in the
interval [0, 1] is feasible, especially in unified form. multiple composition to obtain overall relations
4) Approximate Reasoning with Fuzzy Logic: With the in form of a matrix ;
structure of a fault symptom tree a fuzzy-rule-based system 2) matching of with all current symptoms
with multiple levels of rules can be established, as shown ;
in Fig. 6. The symptoms of the features are now 3) accumulation to obtain the global possibility for each
represented by fuzzy sets with linguistic fault;
4) fault decision by maximal possibility;
meanings like “normal,” “less increased,” “much increased,”
5) defuzzification to obtain a crisp size of the most possible
etc. Then the following analogies to Fig. 6 hold:
fault.
— symptoms In both approaches, a defuzzification can be performed after
— events each level or not.
To reduce the computational effort, simplifying assumptions
— faults may help. The following simplifications are possible:
In contrast to fuzzy control in the basic level fuzzy fault 1) singletons for the symptoms: matching is reduced to
diagnosis differs by direct fuzzification of crisp values;
2) singletons for the events and faults: the conclusions are
1) inputs are mostly no crisp measurements, but detected
singletons with a height of the matching degree between
symptoms represented as fuzzy sets;
facts and rule premise;
2) not only one level of rules does exist, but mostly several
3) universal fuzzy sets for the events and faults: the con-
levels;
clusion is a universal set with height of the matching
3) frequently it is difficult to specify the membership
degree between facts and rule premise. But then the size
functions of the intermediate events because of very
of the event or fault cannot be determined.
vague knowledge.
In the cases of singletons or universal sets for events
The approximate reasoning follows the steps described in
and faults the conclusions are identical with the membership
Section IV: fuzzification or matching, rule activation with eval-
degree of the facts with the logical operation in the premise.
uation by the compositional rule of inference, accumulation
This means the whole procedure reduces to an aggregation of
and defuzzification. The dimensions of the fuzzy-rule-based
the facts with appropriate t-norms and t-conorms, and chaining
system is given by
of the results [47]. Fuzzy fault diagnosis can be expanded to
— number of symptoms time dependent faults and symptoms, including incipient and
— number of levels intermittent faults and multiple faults [53].
— number of rules per level 5) Neuro-Fuzzy Systems for Diagnosis: The knowledge
— number of faults acquisition and tuning for fault diagnosis usually takes
The overall dimension may therefore blow up strongly even much effort and time. Therefore adaptation mechanisms of
for small components or processes. Therefore the software the establishment of rules and membership functions are
implementation is important. Two procedures to perform the required. Considering now the similarity between fuzzy logic
reasoning are known. and neuronal networks adaptive neuro-fuzzy systems can be
Sequential rule activation from level to level developed [54]–[59]. In [54] it is shown that a combination
(horizontal procedure): of the bounded operator for fuzzy-AND and the bounded sum
for fuzzy-OR leads to an adjustable ANDOR-operator ( -
1) matching of symptoms with the premises of rules
operator), which can be approximated by a sigmoidal activity
of level (including their logical operations) to function. Furthermore, by applying Gaussian membership
determine the events ; functions adjustable symptom membership functions can be
2) matching of events with the obtained. By measurement of faults and symptoms the neuro-
fuzzy diagnosis system is adapted by a backpropagation
premises of rules of level to determine the events
method, see also [55].
;
3) matching of events with the faults fuzzy subsets VIII. CONCLUSION
; Fuzzy logic provides a systematic framework to process
4) accumulation of fault fuzzy subsets to obtain the global vague variables and vague knowledge. As the vagueness
possibility for each fault; in automation generally increases with increasing functional
234 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 28, NO. 2, MARCH 1998
hierarchical level the potentials of fuzzy logic approaches
naturally grow with higher levels. Fuzzy control in the basic
level may be attractive for processes where only qualitative
knowledge is available and strongly nonlinear properties do
exist, or if human control has to be imitated. Interesting
possibilities for fuzzy control exist in the higher control
levels, as e.g., cascade control, variable control structures
with fuzzy weighting, fuzzy controller parameter tuning, fuzzy
adaptive control, fuzzy control of nonmeasurable quantities
and fuzzy quality and comfort control. Another application
area is the fuzzy control for special (abnormal) conditions
where today mostly only manual control is applicable. Thus
hybrid classical/fuzzy control system structures result.
The supervision of processes requires the treatment of
quantitative and qualitative knowledge. Here fuzzy approaches
are especially attractive for symptom generation with fuzzy
thresholds, linguistically described observed symptoms and
the approximate reasoning with multilevel fuzzy-rule-based
systems for fault-symptom tree structures.
Further applications of fuzzy logic methods may arise
within the qualitative knowledge processing of intelligent
or multiobjective control systems including higher level con-
trol functions and learning. Especially the combination with
nonlinear approximation (identification) and classification by
artificial neural networks may open many new possibilities.
REFERENCES
[1] L. A. Zadeh, “Outline of a new approach to the analysis of complex
systems and decision processes,” IEEE Trans. Syst., Man, Cybern., vol.
SMC-3, pp. 28–44, 1973.
[2] , “A rationale for a fuzzy control,” J. Dyn. Syst., Meas., Contr.,
vol. 94, Series G, pp. 3–4, 1972.
[3] , “The role of fuzzy logic in modeling and control,” Model.,
Identif. Contr., vol. 15, no. 3, pp. 191–203, 1994.
[4] E. H. Mamdani and S. Assilian, “An experiment in linguistic synthesis
with a fuzzy logic controller,” Int. J. Man-Mach. Stud., vol. 7, pp. 1–13,
1975.
[5] L. Holmblad and I. I. Ostergaard, “Control of a cement kiln by fuzzy
logic,” in Fuzzy Information and Decision-Processes, M. M. Gupta and
E. Sanchez, Eds. Amsterdam, The Netherlands: North-Holland, 1982,
pp. 389–399.
[6] S. Yasunobu and S. Miyamoto, “Automatic train operation by predictive
fuzzy control,” in Industry Applications of Fuzzy Control, M. Sugeno,
Ed. Amsterdam, The Netherlands: North-Holland, 1985, pp. 1–18.
[7] H. P. Preuß, “Fuzzy-control-heuristische Regelung mit unscharfer
Logik,” Automatisierungstechnische Praxis, vol. 34, pp. 176–184, 1992.
[8] D. Driankov, H. Hellendoorn, and M. Reinfrank, An Introduction to
Fuzzy Control. Berlin, Germany: Springer-Verlag, 1993.
[9] R. Isermann, K.-H. Lachmann, and D. Matko, Adaptive Control Systems.
Hertfordshire, U.K.: Prentice-Hall, 1992.
[10] G. N. Saridis, Self Organizing Control Of Stochastic Systems. New
York: Marcel Dekker, 1977.
[11] G. N. Saridis and K. P. Valavanis, “Analytical design of intelligent
machines,” Automatica, vol. 24, pp. 123–133, 1988.
[12] K. J. Åström, “Intelligent control,” in Proc. European Control Conf.,
Grenoble, France, 1991
[13] Handbook of Intelligent Control, D. A. White and D. A. Sofge, Eds.
New York: Van Nostrand Reinhold, 1992.
[14] P. Antaklis, Defining intelligent control. IEEE Trans. Contr. Syst. Tech-
nol., pp. 4–66, June 1994.
[15] M. M. Gupta and N. K. Sinha, Intelligent Control Systems. New York:
IEEE, 1996.
[16] Advances in Intelligent Control, C. J. Harris, Ed., London, U.K.: Taylor
and Francis, 1994.
[17] R. Isermann, “Toward intelligent control of mechanical processes,”
Contr. Eng. Practice, vol. 1, no. 2, pp. 233–252, 1993.
[18] W. Pedrycz, Fuzzy Control And Fuzzy Systems, 2nd ed. New York:
Wiley, 1993.
[19] R. Yager, Essentials of Fuzzy Modeling And Control. New York: Wiley,
1994.
[20] B. Kosko, Neural Networks and Fuzzy Systems. London, U.K.:
Prentice-Hall, 1992.
[21] B.-M. Pfeiffer, Einsatz von Fuzzy-Logik in lernfähigen digitalen
Regelsystemen. Fortschr.-Ber. VDI-Reihe 8 Nr. 500. Düsseldorf:
VDI-Verlag, 1995.
[22] , Selbsteinstellende klassische Regler mit Fuzzy-Logik,” Automa-
tisierungstechnik, vol. 42, pp. 69–73, 1994.
[23] R. Isermann, Digital Control Systems. New York: Springer-Verlag,
1989, vol. 1–2.
[24] B.-M. Pfeiffer and R. Isermann, “Fuzzy logic applications in learning
digital control systems,” in Proc. 3rd European Congr. Fuzzy Intelligent
Techniques Soft Computing, Aachen, Germany, 1995, pp. 847–852.
[25] G. Bretthauer and H.-P. Opitz, “Stability of fuzzy systems—a sur-
vey,” in Proc. European Congr. Fuzzy Intelligent Technologies, Aachen,
Germany, 1994, pp. 283–290.
[26] J. P. Marin and A. Titli, “Comparative analysis of stability methods
for fuzzy controllers,” in Proc. European Congr. Fuzzy Intelligent
Technologies, Aachen, Germany, 1994, pp. 1183–1187.
[27] W. J. Jansen et al., “Stability properties of fuzzy controllers,” J. A, vol.
36, no. 3, pp. 27–37, 1995.
[28] H. Kiendl and J. J. Rüger, “Stability of fuzzy control systems using
facet functions,” Fuzzy Sets Syst., vol. 70, no. 2–3, pp. 275–285, 1995.
[29] H. O. Wang, K. Tanaka, and M. Griffin, “An analytical framework
of fuzzy modeling and control of nonlinear systems: Stability and
design issues,” in Proc. American Control Conf., Evanston, IL, 1995,
pp. 2272–2276.
[30] T. A. Johansen, “Fuzzy model based control: Stability, robustness, and
performance issues,” IEEE Trans. Fuzzy Syst., vol. 2, pp. 221–234, 1994.
[31] S. Germann and R. Isermann, “Nonlinear distance and cruise control for
passenger cars,” in Proc. IFAC-Workshop Advances Automotive Control,
Ascona, Switzerland, Mar. 13–17, 1995, pp. 203–208.
[32] R. Stenz and U. Kuhn, “Automation of a batch destillation column using
fuzzy and conventional control,” IEEE Trans. Contr. Syst. Technol., vol.
3, no. 2, pp. 171–176, 1995.
[33] B.-M. Pfeiffer and R. Isermann, “Selftuning of classical controllers with
fuzzy logic,” Math. Comput. Simul., vol. 37, pp. 101–110, 1994.
[34] L. X. Wang, Adaptive Fuzzy Systems and Control. Englewood Cliffs,
NJ: Prentice-Hall, 1994.
[35] R. Babuska, J. Sousa, and H. B. Verbruggen, “Model-based design of
fuzzy control systems,” in Proc. 3rd European Congr. Fuzzy Intelligent
Technologies, Aachen, Germany, 1995.
[36] M. Fischer, O. Nelles, and D. Füssel, “Tuning of PID-controllers for
nonlinear processes based on local linear fuzzy models,” in Proc. 4th
European Congr. Fuzzy Intelligent Technologies, Aachen, Germany,
1996.
[37] D. Pfannstiel, “Einsatz adaptiver und fuzzybasierter Regelungsstrategien
in der Heizungstechnik,” Automatisierungstechnische Praxis, vol. 37,
pp. 42–48, 1995.
[38] A. Heinrich, J. Kraft, F. Dörrscheidt, M. Boll, “Fuzzy-Regelung eines
semiaktiven Pkw-Fahrwerks,” Automatisierungstechnische Praxis, vol.
36, pp. 23–30, 1995.
[39] K. Pfeiffer and R. Isermann, “Automatic control of vehicle starting for
emission tests on engine dynamometers,” in Proc. American Control
Conf., Seattle, WA, 1995.
[40] H. Takagi, “Fusion technology of fuzzy theory and neural net-
works—survey and future directions,” in Proc. Int. Conf. Fuzzy Logic
Neural Networks, Iizuka, Japan, 1990, pp. 13–26.
[41] M. Brown and C. Harris, Neurofuzzy Adaptive Modeling and Control.
Hertfordshire, U.K., Prentice-Hall, 1994.
[42] H.-P. Preuß and V. Tresp, “Neuro-fuzzy,” Automatisierungstechnische
Praxis, vol. 36, no. 5, pp. 10–24, 1994.
[43] J. S. R. Jang and C. T. Sun, “Functional equivalance between radial basis
function networks and fuzzy inference systems,” IEEE Trans. Neural
Networks, vol. 4, no. 1, pp. 156–158, 1993.
[44] K. Hunt, R. Haas, and R. Murray-Smith, “Extending the functional
equivalence of radial basis function networks and fuzzy inference
systems,” IEEE Trans. Neural Networks, vol. 7, no. 3, pp. 776–781,
1996.
[45] S. Halgamuge, Advanced Methods for Fusion of Fuzzy Systems and
Neural Networks in Intelligent Data Processing, Fortschr.-Ber. VDI
Reihe 10 Nr. 401. Düsseldorf, VDI-Verlag, 1996.
[46] J. Perkins, “Trends in process systems engineering,” Model., Identif.
Control, vol. 15, no. 3, pp. 171–177, 1994.
[47] R. Isermann, “Integration of fault detection and diagnosis methods,”
in Proc. IFAC Symp. Fault Detection, Supervision, Safety for Processes,
Espoo, Finland, June 13–16, 1994, pp. 597–612.
ISERMANN: FUZZY LOGIC APPLICATIONS
[48] , “Model based fault detection and diagnosis methods,” in Proc.
American Control Conf., Seattle, WA, June 21–23, 1995.
[49] R. Isermann and B. Freyermuth, “Process fault diagnosis based on
process model knowledge. Part I: Principles for fault diagnosis with
parameter estimation. Part II: Case study experiments,” J. Dyn. Syst.,
Meas., Contr., vol. 113, p. 627, 1992.
[50] R. Isermann, “On the integration of fault detection methods via unified
fuzzy symptom representation,” 1st European Congr. Fuzzy Intelligent
Technologies, Aachen, Germany, Sept. 7–10, 1993, pp. 400–407.
[51] W. S. Lee, D. L. Grosh, F. A. Tillmann, and C.H. Lie, “Fault tree
analysis, methods and applications—a review,” IEEE Trans. Rel., vol.
R-34, no. 3, pp. 194–202, 1985.
[52] R. Isermann and M. Ulieru, “Integrated fault detection and diagnosis,”
in IEEE/SMC Conf. Systems Engineering Service Humans, Le Touquet,
France, 1993.
[53] M. Ulieru, “Fuzzy logic in diagnostic decision: Possibilistic networks,”
Ph.D. dissertation, Tech. Univ. Darmstadt, Darmstadt, Germany, 1996.
[54] M. Ayoubi, “Fuzzy systems design based on a hybrid neural structure
and application to the fault diagnosis of technical processes,” Contr.
Eng. Practice, vol. 4, no. 1, pp. 35–42, 1996.
[55] R. Isermann and M. Ayoubi, “Fault detection and diagnosis with
neuro-fuzzy-systems,” in Proc. 4th European Congr. Fuzzy Intelligent
Technologies, Aachen, Germany, 1996.
[56] H. Berenji, “Learning and tuning fuzzy logic controllers through rein-
forcements,” IEEE Trans. Neural Networks, vol. 3, no. 5, 1992.
[57] R. Kruse, J. Gebhardt, and F. Klawonn, Foundations of Fuzzy Systems.
New York: Wiley, 1994.
[58] T. Pfeufer and M. Ayoubi, “Fault diagnosis of electro-mechanical
actuators using a neuro-fuzzy network,” in Proc. 3rd Workshop Fuzzy-
Neuro-Systeme ’95, Gesellschaft für Informatik, Darmstadt, Germany,
1995.
[59] H. Ushida, T. Yamaguchi, K. Goto and T. Takagi, “Fuzzy-neuro control
using associative memories, and its applications,” Contr. Eng. Practice,
vol. 2, no. 1, pp, 129–145, 1994.
[60] J. H. Kim, K. C. Kim, and E. Chong, “Fuzzy precompensated PID-
controllers,” IEEE Trans. Contr. Syst. Technol., vol. 2, pp. 406–411,
1994.
235
[61] H. Kiendl, “Fuzzy-control,” Automatisierungstechnik, vol. 41, pp.
A1–A8, 1993.
[62] G. V. S. Raju and J. Zhow, “Adaptive hierarchical fuzzy controller,”
IEEE Trans. Syst., Man, Cybern., vol. 24, p. 973, July/Aug. 1993.
[63] H. J. Zimmermann, Fuzzy Set Theory and Its Applications, 2nd ed.
Boston: Kluwer, 1991.
[64] J. van Amerongen, H. R. van Nauta Lemke, and J. C. T. van der
Veen, “An autopilot for ships designed with fuzzy sets,” in IFAC/IFIP
Conf. Digital Comp. Applications. Amsterdam, The Netherlands:
North-Holland, 1977.
Rolf Isermann (M’92) received Dipl.-Ing degree
in mechanical engineering and the Dr.-Ing. degree,
both from the University of Stuttgart, Stuttgart,
Germany, in 1962 and 1965, respectively.
He became Professor at the University of Stuttgart
in 1972. Since 1977, he has been Head of the
Laboratory of Control Engineering and Process Au-
tomation at the Institute of Automatic Control, Tech-
nical University of Darmstadt, Darmstadt, Germany.
His main area of research interest covers process
modeling and identification, digital control systems,
adaptive control systems, fault diagnosis, and mechatronic systems. He has
also written several books and papers on these topics.
Dr. Isermann was the Chairman of the international program committees of
several IFAC Symposia and of the 10th IFAC-World Congress on Automatic
Control in Munich, Germany, in 1987. He was awarded by the Dr. h.c. degree
from L’Université Libre de Bruxelles, Bruxelles, Belgium, in 1989 and from
the University of Bucharest, Bucharest, Romania, in 1996, and received the
highest scientific award from VDE in 1996. Since 1996, he has been IFAC
Vice President (Executive Board).