In general, tuning a fuzzy expert system takes much more time and effort
than determining fuzzy sets and constructing fuzzy rules. Usually a reasonable
solution to the problem can be achieved from the first series of fuzzy sets and
fuzzy rules. This is an acknowledged advantage of fuzzy logic; however, improving
the system becomes rather an art than engineering.
Tuning fuzzy systems may involve executing a number of actions in the
following order:
1 Review model input and output variables, and if required redefine their
ranges. Pay particular attention to the variable units. Variables used in the
same domain must be measured in the same units on the universe of
discourse.
2 Review the fuzzy sets, and if required define additional sets on the universe
of discourse. The use of wide fuzzy sets may cause the fuzzy system to
perform roughly.
3 Provide sufficient overlap between neighbouring sets. Although there is no
precise method to determine the optimum amount of overlap, it is
suggested that triangle-to-triangle and trapezoid-to-triangle fuzzy sets
should overlap between 25 and 50 per cent of their bases (Cox, 1999).
4 Review the existing rules, and if required add new rules to the rule base.
5 Examine the rule base for opportunities to write hedge rules to capture the
pathological behaviour of the system.
6 Adjust the rule execution weights. Most fuzzy logic tools allow control of the
importance of rules by changing a weight multiplier.
In the Fuzzy Logic Toolbox, all rules have a default weight of (1.0), but
the user can reduce the force of any rule by adjusting its weight. For
example, if we specify
If (utilisation_factor is H) then (number_of_spares is L) (0.6)
then the rule’s force will be reduced by 40 per cent.
7 Revise shapes of the fuzzy sets. In most cases, fuzzy systems are highly
tolerant of a shape approximation, and thus a system can still behave well
even when the shapes of the fuzzy sets are not precisely defined.
But how about defuzzification methods? Should we try different
techniques to tune our system?
The centroid technique appears to provide consistent results. This is a wellbalanced
method sensitive to the height and width of the total fuzzy region as
well as to sparse singletons. Therefore, unless you have a strong reason to believe
that your fuzzy system will behave better under other defuzzification methods,
the centroid technique is recommended.
124 FUZZY EXPERT SYSTEMS
4.8 Summary
In this chapter, we introduced fuzzy logic and discussed the philosophical ideas
behind it. We presented the concept of fuzzy sets, considered how to represent a
fuzzy set in a computer, and examined operations of fuzzy sets. We also defined
linguistic variables and hedges. Then we presented fuzzy rules and explained the
main differences between classical and fuzzy rules. We explored two fuzzy
inference techniques – Mamdani and Sugeno – and suggested appropriate areas
for their application. Finally, we introduced the main steps in developing a fuzzy
expert system, and illustrated the theory through the actual process of building
and tuning a fuzzy system.
The most important lessons learned in this chapter are:
. Fuzzy logic is a logic that describes fuzziness. As fuzzy logic attempts to model
humans’ sense of words, decision making and common sense, it is leading to
more human intelligent machines.
. Fuzzy logic was introduced by Jan Lukasiewicz in the 1920s, scrutinised by
Max Black in the 1930s, and rediscovered, extended into a formal system of
mathematical logic and promoted by Lotfi Zadeh in the 1960s.
. Fuzzy logic is a set of mathematical principles for knowledge representation
based on degrees of membership rather than on the crisp membership of
classical binary logic. Unlike two-valued Boolean logic, fuzzy logic is multivalued.
. A fuzzy set is a set with fuzzy boundaries, such as short, average or tall for men’s
height. To represent a fuzzy set in a computer, we express it as a function
and then map the elements of the set to their degree of membership. Typical
membership functions used in fuzzy expert systems are triangles and
trapezoids.
. A linguistic variable is used to describe a term or concept with vague or fuzzy
values. These values are represented in fuzzy sets.
. Hedges are fuzzy set qualifiers used to modify the shape of fuzzy sets. They
include adverbs such as very, somewhat, quite, more or less and slightly. Hedges
perform mathematical operations of concentration by reducing the degree of
membership of fuzzy elements (e.g. very tall men), dilation by increasing the
degree of membership (e.g. more or less tall men) and intensification by
increasing the degree of membership above 0.5 and decreasing those below
0.5 (e.g. indeed tall men).
. Fuzzy sets can interact. These relations are called operations. The main
operations of fuzzy sets are: complement, containment, intersection and
union.
. Fuzzy rules are us