1. Dipl.Ing (FH) Mario Sader
FachHochschule Lausitz
University of Applied Sciences
THE SEMINAR OF MASTER PROJECT
“The Implementation of Feedforward-Feedback
Fuzzy Logic Algorithm for Level Control System at Process Mini-Plant
Measurement Laboratory FH-Lausitz “ by
R.Danu Setyo Nugroho
(Matrikel.Nr )
Supervisor
Prof.Dr.Ing. E.Stein
Co-Supervisor
Dipl.Ing (FH) Mario Sader
Senftenberg, 7th July 2004

2. 3. Fuzzy Logic Algorithm Theory
FachHochschule Lausitz
University of Applied Sciences
Topic Discussion :
1. Introduction
2. Basic Control Theory
3. Fuzzy Logic Algorithm Theory
4. Fuzzy Logic Control Design
5. Validation
6. Summary

3. INTRODUCTION Background FachHochschule Lausitz ON/OFF Control Mode
University of Applied Sciences
INTRODUCTION
Background
ON/OFF Control Mode
Set Point
Process Mini-Plant at Measurement Laboratory

4. To keep the set point : qin = qout
FachHochschule Lausitz
University of Applied Sciences
INTRODUCTION
The Goal
CONTINUOUS CONTROL
qin
Set Point
qout
To keep the set point : qin = qout

5. ? ? INTRODUCTION Problems Lack of Parameter Systems Information Gc Gm
FachHochschule Lausitz
University of Applied Sciences
INTRODUCTION
Problems
Lack of Parameter Systems Information ? ∑ Gc Gm Gt
Process SP + - e
controller
control valve
level transmitter
mini plant
Block Diagram of Close Loop Systems ?

6. qin should be increased qin should be reduced qin > qout
FachHochschule Lausitz
University of Applied Sciences
INTRODUCTION
Problems
Set point Changing
Load Change
Error
New Set Point
Set Point
Load Change
Normal Load
qin should be increased
qin should be reduced
qin > qout

7. Feedforward – Feedback
FachHochschule Lausitz
University of Applied Sciences
INTRODUCTION
SOLUTION
Feedforward – Feedback
Fuzzy Logic Control

8. process characteristics
FachHochschule Lausitz
University of Applied Sciences
BASIC CONTROL SYSTEMS
process characteristics
dead time τd
time constant τc
response time τr
Step Response
95% τr
63% τc
actual level τd
time

9. criteria of good control
FachHochschule Lausitz
University of Applied Sciences
BASIC CONTROL SYSTEMS
criteria of good control
minimum absolute error
quarter amplitude decay
critical damping
minimum absolute error
SP1
CV1
time
SP2 t0 a
a/4
CV2
∫|E|dt= minimum

10. criteria of good control
FachHochschule Lausitz
University of Applied Sciences
BASIC CONTROL SYSTEMS
criteria of good control
critical damping
SP2
CV2
SP1
CV1
time t0
under damping
over damping
critical damping

11. feedforward control systems
FachHochschule Lausitz
University of Applied Sciences
BASIC CONTROL SYSTEMS
control systems
feedforward control systems
feedback control systems
disturbance
controller
process sp
valve
sprayed
90o
180o
270o 5m
10m
15m
calibration set

12. free of mathematic modeling systems
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
advantages
history
free of mathematic modeling systems
The concept of Fuzzy Logic (FL) was conceived by Lotfi Zadeh, a professor at the University of California at Berkley, and presented not as a control methodology, but as a way of processing data by allowing fuzzy set membership rather than crisp set membership or non-membership.
( e.g Laplace transform, transfer function systems are not required)
empirically-based on operator’s experience rather than technical
understanding of control systems
( The advance knowledge of control theory is not required)
flexible and easy in design
(e.g MIMO,MISO,SISO, rule base determination, simple aritmethic)
fun

13. Upss , it is warm !, it isn’t hot at all !
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Fuzzy Set Vs Crisps Set
700C
warm
hot
Fs : X  [0,1]
[1]
[0]
Crisps Set (FS )
Fuzzy Set (μf)
How about T = 69oC ?
warm
250C
750C
650C
850C
hot
Upss , it is warm !, it isn’t hot at all !
Are you happy with that ?
μf : X  |0,1|

14. The Properties of Fuzzy Set
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
The Properties of Fuzzy Set
Label 1
0.5
μf(x) 50 30 70
100
cold
warm
hot oC
membership function
degre of fuzzy
universal of discorse
scope domain
crisps input

15. The most common used operation logic
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Operation Logic
Boolean
Logical
Fuzzy
The most common used operation logic

16. Fuzzy Inference Engine
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Fuzzy Inference Engine
fuzzyfication
defuzzyfication
rule evaluation
rule base
crisps
inputs
outputs
controller

17. Simple Fuzzy Logic Application Home sprinkler system
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Simple Fuzzy Logic Application
Home sprinkler system
How long the watering duration should take?
It depends on the air temperature and soil moisture
Air temperature
Soil moisture
..FUZZY
duration

18. Fuzzification for air temperature
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Fuzzification for air temperature
cool
warm
hot 1 μ 30 50 80
Temp/C
fuzzification C W H
T_in 60
0.66
0.66 30
0.33
0.33
-30 60
μw = (temp_in – 80) / gradient
μh = (temp_in – 50) / gradient
μh = (60 – 50) / 30
μw = (60 – 80) / -30
μh = 0.33
μw = 0.66

19. Fuzzification for soil moisture
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Fuzzification for soil moisture
0.43
0.56
dry
moist
wet 1 μ 15 25
Moist%
fuzzification D M W
T_in
8 % 8 30
-30

20. Which knowledge base should be used ? operator’s experiences
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Rule Evaluation
Which knowledge base should be used ?
operator’s experiences
If temperature is hot AND moister is wet THEN watering duration is short
If temperature is hot AND moister is moist THEN watering duration is medium
If temperature is hot AND moister is dry THEN watering duration is long
If temperature is warm AND moister is wet THEN watering duration is short
If temperature is warm AND moister is moist THEN watering duration is medium
If temperature is warm AND moister is dry THEN watering duration is long
If temperature is cool AND moister is wet THEN watering duration is short
If temperature is cool AND moister is moist THEN watering duration is medium
If temperature is cool AND moister is dry THEN watering duration is long
If “antecedence 1” AND “antecedence 2 “ THEN “consequent”

21. Mamdani Min-Max Operation
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Rule Evaluation
Mamdani Min-Max Operation
temp
cool
warm
hot
rule
strength
0.66
0.33
moisture
wet
0.56
0.43 S M L = S M L
0.43
0.33
moist
0.56
0.56
0.43
dry
0.33
If temperature is warm (0.66) AND moisture is moist (0.56) THEN watering duration is medium (0.56)
Y = Max (a,b)
If temperature is hot (0.33) AND moisture is dry (0.43) THEN watering duration is long (0.33)
Y = Min (a,b)

22. Center Of Gravity = 0 x 20 + 0.56 x 40 + 0.43 x 60
FachHochschule Lausitz
University of Applied Sciences
FUZZY LOGIC ALGORITHM
Singleton Defuzzification
watering duration
(min) 1 μ
short
medium
long S M L
time
defuzzification
0.56
0.43
0.56
0.43
COG 20 40 60
Center Of Gravity = 0 x x x 60
(COG)
= 48.2 minute

23. Strategy of Control Design
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design ∑
Process SP + - e
Feedback
Fuzzy Controller
Feedforward
diagram block system

24. Strategy of Control Design
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design
fuzzification of feedforward systems
Level Membership Function
μ(f)
19,30
29,35
37,90
50,35
very low
low
medium
high
very high
h(cm) 1
12,60
The degree of membership function |1,0|
very low
low
medium
high
very high 18
0,21
0,782 20
0,949
0,05 25
0,543
0,45 28
0,137
0,86 36
0,22
0,77
Level
(cm)

25. Strategy of Control Design
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design
rule evaluation of feedforward systems
1. IF the level is “very high” THEN the opening of valve is “very big”
2. IF the level is “high” THEN the opening of valve is “big”
3. IF the level is “medium” THEN the opening of valve is “medium”
4. IF the level is “low” THEN the opening of valve is “small”
5. IF the level is “very low” THEN the opening of valve is “very small”
THEN
level 1
1/2
“high” IF 1
1/2
opening of valve
“big”
1/2
“big” 1
opening of valve
fuzzification.vi
defuzzification.vi
5 rule

26. The degree of membership function [1,0]
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design
defuzzification of feedforward system
Set Point
The degree of membership function [1,0]
Control
Signal
very low
low
medium
high
very high 18
0,938
0,06
5,78 V 20
0,54
0,36
6,05 V 25
0,4
0,6
6,55 V 28
6,86 V 30
0,1
0,9
7,77 V 8 9
very small
small
medium
big
very big 6 7
volt 1 5
μ(f)
Opening Valve Membership Function
Level (cm)

27. The degree of membership function [1,0]
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design
fuzzification of feedback system NB NS ZE PS PB e 1
Error membership function
μ(f)
LNB
PVB
XLNB
-0,5
0.5
1,5
-1,5 -1 -2
The degree of membership function [1,0]
XLNB
LNB NB NS ZE PS PB
LPB -2 1
-0,9
0,8
0,2
-0,2
0.4
0.6
0.5
1,2
0,4
0,6
1,5
error
(cm)

28. Strategy of Control Design
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design
rule evaluation of feedback system
Based on P Controller
IF error is “XL Negative Big” THEN corrective valve is “XL Negative Big”
IF error is “Large Neg. Big” THEN correction valve is “Large Neg. Big “
IF error is “Negative Big” THEN correction valve is “Negative Big”
IF error is “Negative Small” THEN correction valve is “ Negative Small”
IF error is “Zero Error” THEN correction valve is “Zero”
IF error is “Positive Small” THEN correction valve is “Positive Small”
IF error is “Positive Big” THEN correction valve is “Positive Big”
If error is “Large Pos Big” THEN correction valve is “Large Pos Big”

29. The degree of membership function [1,0]
FachHochschule Lausitz
University of Applied Sciences
Strategy of Control Design
defuzzification of feedback system NB NS ZE PS PB
volt 1
Defuzzification of Corrective Valve
μ(f)
LNB
LPB
XLNB
-1,5
0.1 3
-2,2
1.0 -2 -3
The degree of membership function [1,0]
XLNB
LNB NB NS ZE PS PB
LPB -2 1 -3
-0,9
0,8
0,2
-1,34
-0,2
0.4
0.6
-0.28
0.5
0,5
1,2
0,4
0,6
1,8
1,5 3
Error
(cm)
correcting
signal
(v)

30. FachHochschule Lausitz
University of Applied Sciences

31. Note : Parameter PI Controller are P = 1,33 and I = 120/s.
FachHochschule Lausitz
University of Applied Sciences
VALIDATION
THE COMPARISON OF PERFOMANCE CONTROLLER τs
95% of 25
FUZZY LOGIC Vs PI CONTROLLER
STEP RESPONS TESTING
SET POINT CHANGING
LOAD CHANGE
Note : Parameter PI Controller are P = 1,33 and I = 120/s.

32. 25 cm level is the normal level without load change
FachHochschule Lausitz
University of Applied Sciences
VALIDATION
LOAD CHANGE TESTING
error
Note :
25 cm level is the normal level without load change
Error open loop in 450% load change is 7 cm

33. SET POINT CHANGE TESTING
FachHochschule Lausitz
University of Applied Sciences
VALIDATION
SET POINT CHANGE TESTING

34. in steady state response
FachHochschule Lausitz
University of Applied Sciences
SUMMARY
The performance of Fuzzy Logic Control here is better then PI Controller
in transient response.
The performance of PI Controller here is better then Fuzzy Logic Control
in steady state response
The number of fuzzy membership’s label that is used influence the smoothness
of the controller’s reaction.
Fuzzy Logic Control is able to avoid both of overshoot and undershoot condition
Even plant has two tank, it is catagorized as first order systems.
Because the second tank doesn’t act as capacitive element during normal process.

35. THANK YOU FOR YOUR ATTENTION
FachHochschule Lausitz
University of Applied Sciences
SUMMARY
RECOMMENDED FUTURE RESEARCH TOPICS
Fuzzy Logic Control based on the PI controller
THANK YOU FOR YOUR ATTENTION
Adaptive Neuro-Fuzzy Logic Control
Self Tuning or Gain Scheduling PI Controller using Fuzzy Logic Algorithm