FKBC DESIGN

PARAMETER
 INTRODUCTION
• Fuzzy Logic is a particular area of concentration in the study of Artificial Intelligence and is
based on the value of that information which is neither definitely true nor false.

• The information which humans use in their everyday lives to base intuitive decisions and
apply general rules of thumb can and should be applied to those control situations which
demand them.

• Acquired knowledge can be a powerful weapon to combat the undesired effects of the
system response.
• Fuzzy logic controllers use a very flexible set of if-then rules. The solution is then
applied to appropriate membership functions. Referring to figure above, values which
lie within the shaded area are called true beyond a shadow of a doubt.

• Those values which lie within the cross hatched area are called false beyond a shadow
of a doubt. If all data falls to one side or the other of the overlap area, then Fuzzy Logic
probably would be of little benefit.
COMPONENTS
The principal structure of a FKBC, as illustrated in previous consists of the
following components:

• Fuzzification module

• Knowledge Base

• Inference Engine

• Defuzzification Module
 Fuzzification Module
The fuzzification module (FM) performs the following functions:

• FM-F1: Performs a scale transformation (i.e., an input normalization)

which maps the physical values of the current process state variables into
a normalized universe of discourse (normalized domain).

It also maps the normalized value of the control output variable onto its physical domain
(output denormalization). When a non-normalized domain is used then
there is no need for FM-Fi.

• FM-F2: Performs the so-called fuzzification which converts a point-wise

(crisp), current value of a process state variable into a fuzzy set, in order
to make it compatible with the fuzzy set representation of the process
state variable in the rule-antecedent.
 Knowledge base
• The knowledge base of a FKBC consists of a data base and a rule base.

• The basic function of the data base is to provide the necessary information for the
proper functioning of the fuzzification module, the rule base, and the defuzzification
module.

• This information includes:

• Fuzzy sets (membership functions) representing the meaning of the linguistic

values of the process state and control output variables.

• Physical domains and their normalized counterparts together with the

normalization/ denormalization (scaling) factors.
The design parameters involved in the construction of
the rule base include:

• Choice of process state and control output variables,

• Choice of the contents of the rule-antecedent and the rule-consequent

• Choice of term-sets (ranges of linguistic values) for the process state

and control output variables

• Derivation of the set of rules.

 Inference engine
• The basic function of the inference engine of the second type is to compute the overall value
of the control output variable based on the individual contributions of each rule in the rule
base.

• Each such individual contribution represents the values of the control output variables as
computed by a single rule.

• The output of the fuzzification module, representing the current, crisp values of the process
state variables, is matched to each rule-antecedent, and a degree of match for each rule is
established.

• This degree of match represents the degree of satisfaction of a fuzzy proposition

The inference engine or rule firing can be of two basic types:

• Composition based inference:

• In this case, the fuzzy relations representing the meaning of each individual rule are
aggregated into one fuzzy relation describing the meaning of the overall set of
rules.

• Then inference or firing with this fuzzy relation is performed via the operation
composition between the fuzzified crisp input and the fuzzy relation representing
the meaning of the overall set of rules.

• As a result of the composition one obtains the fuzzy set describing the fuzzy value
of the overall control output .
• Individual-rule based inference:

In this case, first each single rule is fired. This firing can be simply described by:

(1) computing the degree of match between the crisp input and the fuzzy sets
describing the meaning of the rule-antecedent
and
(2) "clipping" the fuzzy set describing the meaning of the rule-consequent to the
degree to which the rule-antecedent has been matched by the crisp input. Finally,
the "clipped" values for the control output of each rule are aggregated, thus forming
the value of the overall control output.

Usually, the second type of inference is preferred since it is computationally very

efficient and saves a lot of memory. The FKBC described and analyzed in the
subsequent chapters all use the individual-rule type of inference unless stated
otherwise.
• The first choice determines uniquely the second one, i.e., once the representation
of the meaning of a single rule is decided upon, then there is a unique way of
aggregating the individual representations into a global one.

• The two basic approaches to the representation of the meaning of a set of rules, are
based on the Mamdani and GCidel type of representation of the meaning of a single
rule.
In this context the design parameters for the inference engine
are as follows:

• Choice of representing the meaning of a single production rule

• Choice of representing the meaning of the set of rules

• Choice of inference engine

• Testing the set of rules for consistency and completeness.

 Defuzzification module
The functions of the defuzzification module (DM) are as follows:

• DM-F1: Performs the so-called defuzzificaiion which converts the set of modified control
output values into a single point-wise value.

• DM-F2: Performs an output denormalizaiion which maps the point-wise value of the control
output onto its physical domain. DM-F2 is not needed if non-normalized domains are used.

The design parameter of DM is

• Choice of defuzzification operators.
 Rule base
• Its basic function is to represent the control policy of an experienced process
operator &/or control engineer in a structured way.

• The info is produced in the form of set of rules as

IF <process state> THEN <control output>

The design parameters of the rule base include:

• Choice of process state and control output variables

• Choice of the content of the rule-antecedent and the rule-consequent

• Choice of term-sets for the process state and control output variables

• Derivation of the set of rules.

• Choice of Variables and Content of Rules
• If one has made the choice of designing a P-, PD-, PI- or PID-like FKBC this already
implies the choice of process state and control output variables, as well as the
content of the rule-antecedent and the rule-consequent for each of the rules.

• The process state variables representing the contents of the rule antecedent
(if-part of a rule) are selected amongst

• error, denoted by ∆,
• change-of-error, denoted by ∆e or ė,
• sum-of-errors, denoted by δe.
The control output (process input) variables representing the contents of the rule-
consequent (then-part of the rule) are selected amongst

• change-of-control output, denoted by ∆u or ů,

• control output, denoted by u.

Furthermore, by analogy with a conventional controller, we have that

• e(k) = ysp - y(k),

• ∆e(k) = e(k) - e(k -1),
• ∆u(k) = u(k) - u(k - 1).

In the above expressions, ysp stays for desired process output or set-point: y is
the process output variable (control variable); k is the sampling time.
 Data base
• Its basic function is to provide the necessary information for the proper functioning of the
fuzzification module, the rule base, & the defuzzification module.

• This info includes:

• Fuzzy sets (MFs) representing the meaning of the linguistic values of the process state
and control o/p variables.

• Physical domains & their normalized counterparts together with the scaling factors
Choice of above info. constitutes the design parameters for DB
The two design parameters involved in the construction of the data base are

• Choice of membership functions,

• Choice of scaling factors.

Thanks