Você está na página 1de 50

1

Fuzzy Logic and Control


A. Homaifar

Autonomous Control & Information Technology
(ACIT) Center
Department of Electrical & Computer Engineering
North Carolina A&T State University
2
OUTLINE
Fundamentals
Fuzzy Sets, Operators,
GMP and GMT
Fuzzy Engines
Hybrid Fuzzy-PID Controllers
Generalized Sugeno Controllers
Applications
3

1.
A B
(u) = max {
A
(u),
B
(u) }
2.
A B
(u) = min {
A
(u),
B
(u) }
3.
A
(u) = 1.0 -
A
(u)
Set Theoretic Operations
4
Union Of Fuzzy Sets
0
1
A
X

A B
(X)
B
Union of fuzzy sets A & B
5
Intersection Of Fuzzy Sets
0
1
A
X

A B
(X)
B
Intersection of fuzzy sets A & B
6
Complement Of Fuzzy Set
0
1
A
X
A

A
(X)

A
(X) = 1 -
A
(X)
7
Law of Contradiction
0
1
A
X
A
Fuzzy A A =
8
Law Of Excluded Middle
0
1
A
X
A
Fuzzy A A = X
9
Definition of t-norms
t-norms are two-valued functions from [0,1]x[0,1]
that satisfy the following conditions:
vity) (associati (x)) (x)), (x), t(t( (x))) , (x) t( (x), t( 4.
vity) (commutati (x)) (x), t( (x)) (x), t( 3.
ity) (monotonic (x) (x) and (x) (x) if
(x)) (x), t( (x)) (x), t( 2.
X x (x), (x)) t(1, (x),1) t( 0; t(0,0) 1.
C
~
B
~
A
~
C
~
B
~
A
~
A
~
B
~
B
~
A
~
D
~
B
~
C
~
A
~
D
~
C
~
B
~
A
~
A
~
A
~
A
~
=
=
s s
s
e = = =
10
Definition of t-conorms
t-conorms or s-norms are associative,
commutative, and monotonic two-placed functions
s that map from [0,1]x[0,1] into [0,1] that satisfy the
following conditions:
vity) (associati (x)) (x)), (x), s(s( (x))) , (x) s( (x), s( 4.
vity) (commutati (x)) (x), s( (x)) (x), s( 3.
ity) (monotonic (x) (x) and (x) (x) if
(x)) (x), s( (x)) (x), s( 2.
X x (x), (x)) s(0, (x),0) s( 1; s(1,1) 1.
C
~
B
~
A
~
C
~
B
~
A
~
A
~
B
~
B
~
A
~
D
~
B
~
C
~
A
~
D
~
C
~
B
~
A
~
A
~
A
~
A
~
=
=
s s
s
e = = =
11
Relationship between t-norms and t-conorms
t-norms and t-conorms are related in a sense of
logical duality.
t-conorm as a two-placed function s mapping from
[0,1] x [0,1] in [0,1] such that the function t defined
as
(x)) - 1 (x), - s(1 - 1 (x)) (x), t(
B
~
A
~
B
~
A
~ =
Is a t-norm.
12

R
(x,y) defines a relation between x and y:
Y
1
Y
2
Y
3

X
1
0.2 1 0.4
X
2
0 0.6 0.3
X
3
0 1 0.8
Composition of R and S
R o S = { [(u,w), sup
V
(
R
(u,v) *
S
(v,w) ], u e U,
v e U, w e W }
Fuzzy Relation
13
Example:
R =
R
(x,y) = 0.7 0.5
0.8 0.4
S =
S
(y,z) = 0.9 0.6 0.2
0.1 0.7 0.5
T =
T
(x,z) = V
y e Y
(
R
(x,y)
S
(y,z) )
= 0.7 0.6 0.5
0.8 0.6 0.4
For Example:

T
(x,z) = max [ min (0.7,0.9), min (0.5, 0.1) ]
= max [ 0.7, 0.1 ] = 0.7
14
Max Product
T =
T
(x,z) = V
y e Y
(
R
(x,y) .
S
(y,z) )
= 0.63 0.42 0.25
0.72 0.48 0.24
For Example:

T
(x,z) = max [ (0.7)*(0.9) , (0.5)*(0.1) ]
= max [ 0.63, 0.05 ] = 0.63

15
Max Average
T =
T
(x,z) = 0.5 V
y e Y
(
R
(x,y) +
S
(y,z))
= 0.8 0.65 0.5
0.85 0.7 0.5
For Example:

T
(x,z) = 0.5 max [ (0.7)+(0.9) , (0.5)+(0.1) ]
= 0.5 max [ 1.6, 0.6 ] = 0.8

Typical dual pairs of
nonparameterized t-
norms and t-conorms
)) x ( ), x ( (
B
~
A
~

=
otherwise .......... .......... .......... 0
1 )} x ( ), x ( max{ if )}... x ( ), x ( min{
B
~
A
~
B
~
A
~
drastic product
drastic sum
s
w
)) x ( ), x ( (
B
~
A
~
=

=
otherwise . .......... .......... .......... 1
0 )} x ( ), x ( min{ if )}... x ( ), x ( max{
B
~
A
~
B
~
A
~
} 1 ) ( ) ( , 0 { max )) ( ), ( ( t ~ ~ ~ ~ 1 + = x x x x
B A B A

bounded different
t
w
=
)} ( ) ( , 1 min{ )) ( ), ( ( ~ ~ ~ ~
1
x x x x s
B A B A
+ =
bounded sum
18
)) x ( ), x ( (
B
~
A
~ t
1.5 =
)] x ( ). x ( ) x ( ) x ( [ 2
) x ( ). x (
B
~
A
~
B
~
A
~
B
~
A
~
+
Einstein
product
s
1.5
)) x ( ), x ( (
B
~
A
~
=
Einstein
sum
) x ( ). x ( 1
) x ( ) x (
B
~
A
~
B
~
A
~
+
+
t
2
)) x ( ), x ( (
B
~
A
~
=
s
2
)) x ( ), x ( (
B
~
A
~
=
) x ( ). x (
B
~
A
~
) x ( ). x ( ) x ( ) x (
B
~
A
~
B
~
A
~ +
Algebraic product
Algebraic sum
19
t
2.5
)) x ( ), x ( (
B
~
A
~
=
) x ( ). x ( ) x ( ) x (
) x ( ). x (
B
~
A
~
B
~
A
~
B
~
A
~
+

Hamacher
product
s
2.5
)) x ( ), x ( (
B
~
A
~
=
) x ( ). x ( 1
) x ( ). x ( 2 ) x ( ) x (
B
~
A
~
B
~
A
~
B
~
A
~

+
Hamacher
sum
t
3
)) x ( ), x ( (
B
~
A
~
=
s
3
)) x ( ), x ( (
B
~
A
~
=
min
max
)} x ( ), x ( {
B
~
A
~
)} x ( ), x ( {
B
~
A
~
minimum
maximum
20
Theorem 1:
All t-norm operators, are bounded below by drastic product t
w
,
and bounded above by t
3

t
w

s
t
1

s
t
1.5

s
t
2

s
t
2.5

s
t
3

Theorem 2:
All s-norm operators, are bounded below by s
3
, and bounded
above by drastic sum s
w
s
w
>s
1
> s
1.5
> s
2
> s
2.5
> s
3

21
Fuzzy Logic & Approximate Reasoning
1- Generalized Modus Ponens (GMP):
Premise 1: x is A (meaning not exactly A),
Premise 2: if x is A then y is B,
consequence: y is B (i.e. not exactly B)
22
Fuzzy Logic & Approximate Reasoning
1- Generalized Modus Tollens (GMT):
Premise 1: y is B (meaning not exactly B),
Premise 2: if x is A then y is B,
consequence: x is A (i.e. not exactly A)
23
Fuzzy Logic Controller
Fuzzification Interface
Measures the values of input variables,
Performs a scale mapping (if necessary) that
transfers the range of values of input variables into
corresponding universes of discourse
Performs the function of fuzzification that
converts the crisp real world input data into suitable
linguistic values, which may be viewed as labels of
fuzzy sets.
24
Knowledge Base
Fuzzy rules consist of a premise with one or
more antecedents, and a conclusion with one
or more consequences. The individual rules in
the rule base are connected through the
operator also.
25
Decision-making Logic
The two most common types are Min of
Mamdani and product of Larson [lee,1990].
The Min operator takes the minimum of all
fuzzy membership values in the "if-side" for the
rule being evaluated, and clips the corresponding
output membership at this level.
The product operator scales the output
membership as opposed to clipping it.
26
Defuzzification Interface
The defuzzification interface performs the
following functions:
a scale mapping, which converts the range of
values of output variables into corresponding
universes of discourse
defuzzification which yields a nonfuzzy control
action from an inferred fuzzy control action.
27
Defuzzification Interface
Two methods of defuzzification are most often
used:
maximum membership: chooses the output value
corresponding to the maximum degree of
membership in the output fuzzy set.
centroid or center of gravity: is the most
commonly used defuzzification method. In the
case of a discrete universe, the centroid method
yields a weighted sum of the output values [Lee,
1990].
28
KB
Fuzzification
Interface
Defuzzificatio
Interface
Decision
Making
Logic
Fuzzy Fuzzy
Process Output & State
Actual Control
Non-Fuzzy
Controled
System
(Process)
Block Diagram
29
Hybrid Fuzzy PID
30
HYBRID FUZZY-PID CONTROLLER
A Hybrid Fuzzy-PID controller is a type of fuzzy
controller in which the fuzzy engine is placed above a
conventional PID controller in the control hierarchy. This
is shown by the picture below.
The control inference of the HFPID is of the following:
IF e is big and de/dt is small and edt is small THEN P is big and D is
small and I is medium
31
Sugeno Engines
32
GENERALIZED SUGENO CONTROLLER
A Generalized Sugeno Controller (GSC) is a fuzzy
controller which maps the input space to the output
space by the following fuzzy control
inference:
R
i
: IF x
1
is A
1
i
and ... x
n
is A
n
i
THEN y = P
i
(x); 1<i<k
P
i
(x) is a polynomial which is given by:
33
PROBLEM STATEMENT
The Generalized Sugeno Controller (GSC) requires the
determination of several unknown parameters for its design
and implementation. The search for these parameters can
be quite exhaustive.
Therefore, the purpose of this research is to develop a
simple approach to determine the unknown parameters of a
Generalized Sugeno Controller.
Our approach is applied to the following systems:
A Hybrid Fuzzy-PID controlled robot manipulator arm
An approximation of an optimal feedback control law
for a ship tracking problem
34
GSC vs. HFPID
The GSC has more desirable properties than the HFPID
The GSC is easier to design since the consequence
of each rule is caluclated automatically.
Rule evaluation and defuzzification is easier for the
GSC than the HFPID.
It is easier to analyze the GSC in a qualitative
manner for stablility, controllability, observability,
and other issues of control systems.
35
APPLICATIONS
36
Applicability of Using A Fuzzy Controller
for DC-DC Converters

37
Dynamic Control of Paralleled
DC-DC Converters
(Current Sharing)
Desirable Characteristic:
Steady State Load Sharing
Transient Load Sharing
38
BASIC CELL STRUCTURE OF DC-DC POWER MODULE
V
ref
+
-
V
in
I
L
d
V
C
+
-
H
i
H
v
V
o
Basic Cell
I V
+
-
39
DIAGRAM OF TWO PARALLELED MODULES WITH MSC
WITH A DEDICATED MASTER
Module I
Module n
Cell #1
Cell #2
LO
A
D
V
o
I
o1
I
o2
V
ref1
+
-
H
v
+
+
+
-
H
v
+
-
H
i
V
ref2
Module #1
Module #2
CS_Bus
40
BLOCK DIAGRAM OF #N PARALLELED
CONVERTERS WITH MASTER-SLAVE CONTROL
Cell #1
Cell #n
Module #1
Module #n
-
-
-
-
-
-
L
O
A
D
V
o
V
in

I
o1
I
on
V
ref1
+
+
+
-
H
v
+
-
H
i
+
+
+
-
H
v
+
-
H
i
V
refn
CS Bus
41
Classical Control
Complexity of analysis:
Huge number of loops and transfer
functions
Stability analysis is almost impossible
42
Fuzzy Control
Linguistic Rules:
Ease of analysis
Ease of implementation
43
Fuzzy Implementation

Input and output selection for the FLC
Membership function definition
Fuzzy rules definition
(If x is A and y is B, then z is C)
44

INPUT


NB


NS


Z


PS


PB


OUTPUT


NB


NS


Z


PS


PB

ATTRIBUTES OF THE MEMBERSHIP FUNCTION
IN DESIGNING FLC OF LOAD SHARING
NB: Negative Big NS: Negative Small
Z: Zero
PS: Positive Small PB: Positive Big
45
BASIC MEMBERSHIP FUNCTION OF INPUT AND
OUTPUT FOR THE CSC FUZZY LOGIC CONTROLLER
(err) NB NS Z PS PB







-2 -1 0 1 2 err
46
The centroid method is determined from the
fuzzy set as follow:

'
'
-
=
i
i i c
i
i i c i
z
z z
z
) (
) (
0

Defuzzification
47
If the error (difference of output currents)is PB,
then the duty cycle change of slave should be PB
If the error is NS, then the duty cycle change of
slave should be NS
If the error is PS and its derivative is NB, then the
duty cycle change of slave should be Z
Example of Fuzzy Control Rules
master-slave problem
48
Block Diagram
49
DEFFERENCE BETWEEN TWO OUTPUT CURRENTS
OUTPUT VOLTAGE V
O

SYSTEM USING A CLASSICAL CSC SYSTEM USING A FUZZY LOGIC CSC
two paralleled buck converters output waveforms
Results
50
DEFFERENCE BETWEEN TWO OUTPUT CURRENTS
OUTPUT VOLTAGE V
O

SYSTEM USING A CLASSICAL CSC SYSTEM USING A FUZZY LOGIC CSC
STEP LOAD RESPONSES FOR 50% OF THE LOAD
Results
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
x 10
-3
-1.5
-1
-0.5
0
0.5
1
T
O
u
t
p
u
t

C
u
r
r
e
n
t

1
-
O
u
t
p
u
t
C
u
r
r
e
n
t

2

1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
x 10
-3
13
14
15
16
17
18
19
T
O
u
t
p
u
t
V
o
l
t
a
g
e

Você também pode gostar