Fuzzy Systems

Example of a Mamdani/Larsen
fuzzy controller

Defuzzification How can fuzzy systems be used in a world where

measurements and actions are expressed as crisp
values?

* Fuzzy Systems Toolbox, M. Beale and H Demuth

Fuzzy Systems (Cont.) Fuzzy Systems (Cont.)

Input Fuzzifier Fuzzy System Defuzzifier output

* Fuzzify crisp inputs to get the fuzzy inputs

It is too hot! Set the fan at
* Defuzzify the fuzzy outputs to get crisp outputs
90% speed
90 Degree F. Turn the fan on high

* Fuzzy Systems Toolbox, M. Beale and H Demuth

Fuzzification Fuzzification
Process of making a crisp Fuzzification example Fuzzy singleton
quantity fuzzy

µF(x)
If it is assumed that input data Membership
do not contain noise of grade
vagueness, a fuzzy singleton of crisp input x
can be used x0

in the fuzzy set

µF(x)

If the data are vague or

perturbed by noise, they
should be converted into a Crisp input x
fuzzy number x0

base

1
 Fuzzy Approximation Theorem Build a Fuzzy Controller
3 Steps
1. Pick the linguistic variables
• Example: Let temperature (X) be input and
motor speed (Y) be output
2. Pick the fuzzy sets
• Define fuzzy subsets of the X and Y
3. Pick the fuzzy rules
• Associate output to the input

*Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko *Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko

Example: Build a Fuzzy Controller Example: Build a Fuzzy Controller

Goal: Design a motor speed controller for air Step 2: Pick fuzzy sets
conditioner Define linguistic terms of the linguistic variables
temperature (X) and motor speed (Y) and
Step 1: assign input and output variables associate them with fuzzy sets
For example, 5 linguistic terms / fuzzy
Let X be the temperature in Fahrenheit sets on X
Let Y be the motor speed of the air conditioner • Cold, Cool, Just Right, Warm, and Hot
Say 5 linguistic terms / fuzzy sets on Y
• Stop, Slow, Medium, Fast, and Blast
*Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko *Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko

Example: Build a Fuzzy Controller Example: Build a Fuzzy Controller

Input Fuzzy sets Output Fuzzy sets

*Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko *Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko

2
 Example: Build a Fuzzy Controller Example: Build a Fuzzy Controller

Step 3: Assign a motor speed set to each

temperature set
A Fuzzy Relation
• If temperature is cold then motor speed is stop expressed by a
• If temperature is cool then motor speed is slow rule
• If temperature is just right then motor speed is medium
• If temperature is warm then motor speed is fast
• If temperature is hot then motor speed is blast

*Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko *Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko

Example: Build a Fuzzy Controller Example: Build a Fuzzy Controller

A Fuzzy controller with 5 patches ! "
# $ %

&

*Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko *Fuzzy Thinking:The new Science of Fuzzy Logic, Bart Kosko

Example: temp. = 65 degree F. Example: temp. = 63 degree F.

If temperature is just right then motor speed is medium

*Fuzzy Thinking, Bart Kosko *Fuzzy Thinking, Bart Kosko

3
 Example: t = 63 degree F. Example: t = 63 degree F.
Summed (MAXed) of the partially fired then-part
fuzzy sets

OR OUTPUT

*Fuzzy Thinkring, Bart Kosko *Fuzzy Thinkring, Bart Kosko

Example: t = 63 degree F. Defuzzification

Defuzzify to find the output motor speed

Converts a fuzzy set into a crisp output.

Defuzzification is a process to get a non-fuzzy value
that best represents the possibility distribution of an
inferred fuzzy control action.
There is no systematic procedure for choosing a good
defuzzification strategy.
Selection of defuzzification procedure depends on the
Question: how to convert a fuzzy set into a crisp value? properties of the application.
*Fuzzy Thinkring, Bart Kosko

Defuzzification Defuzzification
& 0
k zj
12 3 z0 =
j =1 k

'( " "

)*+,-! ")**./ 4

& 0

zj: control action whose membership functions reach the maximum.

k: number of such control actions.

*Fuzzy Logic with Engineering Applications, Timothy J. Ross 5

56 6 65 5 78.5 04 5

4
 Defuzzification Defuzzification
Max-membership Weighted average
principal, also known as method
height method
• Valid for symmetrical
output membership
functions
• Produces results very close
to the COA method
• Less computationally
intensive Formed by weighting each
functions in the output by
its respective maximum
membership value

*Fuzzy Logic with Engineering Applications, Timothy J. Ross *Fuzzy Logic with Engineering Applications, Timothy J. Ross

Defuzzification Defuzzification
Bisector of the Area First (or last) of maxima

• The BOA generates the • Determine the smallest

action (z0) which partitions value of the domain with
the area into two regions maximized membership
with the same area degree

z0 β
µC ( z ) dz = µC ( z ) dz
α z0

α = min{z | z ∈ W }
β = max{z | z ∈ W }

*Fuzzy Logic with Engineering Applications, Timothy J. Ross *Fuzzy Logic with Engineering Applications, Timothy J. Ross

Example: Defuzzification Example: Defuzzification

Find an estimate crisp output from the following 3 CENTROID
membership functions

*Fuzzy Logic with Engineering Applications, Timothy J. Ross *Fuzzy Logic with Engineering Applications, Timothy J. Ross

5
 Example: Defuzzification Example: Defuzzification
Weighted Average Mean-Max

Z* = (6+7)/2 = 6.5

*Fuzzy Logic with Engineering Applications, Timothy J. Ross *Fuzzy Logic with Engineering Applications, Timothy J. Ross

Example: Defuzzification Defuzzification

First and Last of maxima Which defuzzification method is the best?
The answer is context or problem-dependent.

4 criteria against which to measure the methods:

• #1 Continuity. small change in the input should

not produce the large change in the output

*Fuzzy Logic with Engineering Applications, Timothy J. Ross *Fuzzy Logic with Engineering Applications, Timothy J. Ross

Defuzzification Summary
#2 Disambiguity. Defuzzification method
should always result in a unique value, I.e. 1. Fuzzification of inputs.
no ambiguity.
2. Linguistic variables, linguistic terms and
#3 Plausibility. Z* should lie approximatly in the associated fuzzy sets
middle of the support region and have high
degree of membership.
3. Rules

#4 Computational simplicity. 4. Defuzzification

*Fuzzy Logic with Engineering Applications, Timothy J. Ross *Fuzzy Logic with Engineering Applications, Timothy J. Ross