Fuzzyfication

Fuzzification
Fuzzification refers to the representation of a crisp value by a membership
function. As seen before, this is needed prior to applying the composition (CRI),
when the data (measurements) are crisp values, as common in control
applications. In the example, measured value of temperature is fuzzified simply
by picking the grade value of the membership function of the fuzzy variable
temperature (T) corresponding to the particular measured value. This is the
method of fuzzy singleton, and is perhaps the most common method of
fuzzification.
It may be argued that the process of fuzzification amounts to giving up the
accuracy of crisp data. This is not so in general. The reason is, a measured
piece of data may not be known to be 100% accurate. If the measured variable
has some inherent fuzziness (similar to inherent randomness), it is particularly
appropriate to assign a grade of membership (similar to assigning a probability)
based on some prior knowledge regarding the measured quantity. This prior
knowledge is available in the form of a membership function in the case of a
fuzzy variable (similar to a probability distribution function in the case of a
random variable).
 Singleton method

Consider a crisp measurement y0 of a fuzzy variable Y. If it is known that the

measurement is perfectly accurate, it may be represented by a fuzzy quantity
F with the singleton membership function
Since the measured data are not perfectly accurate, a more appropriate
method of using fuzzy singleton to fuzzify a crisp value is given now. As before,
suppose that a crisp measurement y0 is made of a fuzzy variable Y. Let it be
known that Y can assume n fuzzy states Y1, Y2, . . . , Yn. Since Y = Y1 OR Y2 OR . .
. OR Yn , the membership function of Y is given as the union of the membership
functions of the individual fuzzy states.
Then, the membership function of the fuzzified quantity F is given according to
the extended singleton method by a set of fuzzy singletons where, for state j,
the fuzzified quantity is:
 Triangular function method

Instead of a singleton, an appropriately scaled (shaped) triangular

membership function may be used to represent the fuzzified quantity
for each fuzzy state, similar to what is given in equation (3.18).
Specifically, note that a triangular membership function (continuous
case) may be expressed as
Note that y0 is the peak point, where the membership grade is 1, and s is
the base length (support set). For example, a sharper peak is realized by
making s smaller.
Again, suppose that a crisp measurement y0 is made of a fuzzy variable Y,
which can assume n fuzzy states Y1, Y2, . . . , Yn . Then, the membership
function of the fuzzified quantity F is such that, for state j, the membership
function of the fuzzified quantity is given by:
 Gaussian function method

In the previous approach of fuzzification, the shaping function of the

fuzzified membership function need not be triangular and can be quite
general. The triangular shape is quite common, however. Another
appropriate shaping function is given by the Gaussian membership
function:
As before, the smaller the s the sharper (or less fuzzy) the membership function
represented by equation (3.21). In the present method of fuzzification, the
membership function of the fuzzified quantity F is such that, for state j, the
membership function of the fuzzified quantity is given by: