2018 3rd International Conference on Mechanical, Control and Computer Engineering

The Research On The Parameter Tuning Method Of

An Analytic Type Fuzzy Integral Hybrid Controller

LIANG Shutian 1st Affiliation(Author) MENG Dedong 3rd Affiliation(Author)

Wuhan Institute of Marine Electric Propulsion Wuhan Institute of Marine Electric Propulsion
Wuhan, China Wuhan, China
[email protected] [email protected]

LIU Wenda 2nd Affiliation(Author)

Wuhan Institute of Marine Electric Propulsion
Wuhan, China
[email protected]

Abstract: This paper proposed the method of Analytic Type comparing to the traditional PID control system. In
Fuzzy Integral Hybrid Control (ATFIHC)regarding the reference[3], a designing method of a parameter adaptive fuzzy
difficulties of realization of fuzzy control algorithm and the controller was advanced based on the conventional fuzzy
steady state error of the fuzzy control system. The parameter controller ,the performance of the basic fuzzy controller was
tuning equations of ATFIHC are obtained on the basis of the boosted. In reference[4], fuzzy control algorithm based on
corresponding relations of parameters between the analytic fuzzy parameter auto-tuning with gain scheduling was proposed and
integral controller and the classic PID controller, then the grid applied to the control of a single linear inverted pendulum ,the
search method was applied to obtain the coefficients of the
experiment result showed the algorithm was relatively robust
parameter tuning equations. The result of simulation shows that
with quick response and good anti-interference ability.
the control effect of the ATFIHC is superior to the optimized PID
controller in integral norms such as IAE 、 ISE 、 ITAE and Two problems require special attention in the application of
control norms such as Overshoot and Response Time, and it is fuzzy control system, one is that the amount of fuzzy rules
more adaptive to the changes of parameters and structure of the grows exponentially as the amount of controlled object and the
controlled objects. domain classification increase, which causes great difficulty to
realize the fuzzy control algorithm with the excessive quantity
Key words: Fuzzy Control; Analytic type Fuzzy Integral Hybrid of fuzzy rules and to meet the need of real-time control.
Controller(ATFIHC); Parameter tuning; Intelligent control Another problem is that the steady-state errors exist in the
I. INTRODUCTION fuzzy control system because it has no integration loop. To
solve the first problem, the analytic type fuzzy controller could
The auto control stems from the need of the automation of be adopted .This type of controller approach the fuzzy control
the producing process, which reduces the tense of labor and rules by analytic expression with faster running speed and
elevates the production efficiency and quality of products. With stronger adaptability. To solve the second problem, the fuzzy
the rapid development of science and technology, it became integral hybrid controller can be adopted, i.e. adding
increasingly complicated and difficult to build a precise integration loop in the fuzzy controller to eliminate the steady-
mathematical model of the controlled object because of the state errors. By integrating the aforementioned two controllers
nonlinearity, time-variance and uncertainty. Even if the we obtain the Analytic Type Fuzzy Integral Hybrid Controller
mathematical models are built, it is difficult to apply them to (ATFIHC). Although ATFIHC has many advantages, the
the design of real control systems. Therefore, the classic parameter tuning of it is a critical problem to solve. This paper
control theories and modern control theories, which are based studied the parameter tuning method of ATFIHC.
on the precise mathematic model of the control objects, face
great challenge[1]. II. ANALYTIC TYPE FUZZY INTEGRAL HYBRID
CONTROLLER
The intelligent control theory is a new generation of control
theory arose in the context of lack of precise model; it is a The basic structure of an ATFIHC is illustrated in figure 1.
computational digital control of the combination of artificial Similar to classic controller, the input of the Fuzzy Integral
intelligence and feedback control theory, as an important and Hybrid Controller is error e, and the control signal of the
efficient form of intelligent control, the fuzzy control controlled object is u.
developed rapidly. In reference[2], an adaptive fuzzy-PID
control DC brushless motor speed regulating system was
designed, the simulation and test show that the adaptive fuzzy
PID control has less overshoot and better control effect

978-1-5386-8481-8/18/$31.00 ©2018 IEEE 50

DOI 10.1109/ICMCCE.2018.00018
  K F = K p (1 + Td )

 K I = K p Ti （6）
 α = 1 1+ T
 d
Thus, with the corresponding relationship of an ATFIHC
and a classic PID controller, the ATFIHC parameter can be
determined by search approach based on the class PID
Figure 1. The basic structure of a Fuzzy Integral control system controller parameters.
A Fuzzy controller can be categorized into one-dimension III. THE PARAMETER TUNING OF AN ANALYTIC FUZZY
controller, two-dimension controller, three-dimension INTEGRAL HYBRID CONTROLLER (ATFIHC)
controller and multi-dimension controller. Mostly, a Fuzzy
controller adopt two- dimension form, i.e. the inputs are the A. The empirical equation of an optimized PID controller
error and the error variation(first-order derivative), if we also The parameter tuning algorithm of a controller is usually
use the variation of the error variation(second-order derivative) advanced based on a controlled object model with First-order
as an input, we get three dimension controller. The dimension Plus Dead-time(FOPDT),the expression of which is as
of the input variables of a fuzzy controller is determined by the following:
feature of controlled object’s dynamic behavior. One
K −τ s
dimension or two dimension is applied to a first-order process Go ( s ) = e （7）
with slow changes, and the two dimension or three dimension Ts + 1
is applied to a controlled object with fast changes in dynamic where K is the amplification coefficient, T is the integral
features. time constant, τ is the time delay and e is a natural constant.
This paper study on two dimension fuzzy controller, the The optimized control normally takes error integral as the
expression of which is as following[5]: performance index, and the error integral is typically as follow:
u = K F [α e + (1 − α )e ] （1）
IAE =  e ( t ) dt （8）
where K F is the proportion coefficient of the fuzzy
ISE =  e 2 ( t )dt （9）
controller and α ∈ (0,1) is the adjustment coefficient that
represent the weight of error . ITAE =  t e ( t ) dt （10）
The equation(1) is expressed in transfer function as
following: where e ( t ) is the error function.

u( s) By meeting the criterion of the minimizing the

GF ( s ) = = K F (α + (1 − α ) s ) （2） forementioned error integral value ,we yield the parameter
e( s ) tuning method of an optimize PID controller as following[6-8]:
The integral loop transfer function is:
 b
a1  T  1
K  p K =  
GI ( s ) = I （3） K τ 
s 
where K I is the amplification coefficient of integral loop.  T T  2
b

 iT =   （11）
Apparently, the transfer function of an analytic fuzzy  a2  τ 
integral Hybrid controller is:  b
Td = a3T  
T 3
KI
GFC ( s ) = GF ( s ) + GI ( s ) = K F (α + (1 − α ) s ) +  τ 
s the coefficients in the equation(11) take the values in the
（4） following table[9]:
the transfer function of a classic PID controller is:
1
Gc ( s ) = K p (1 + + Td s ) （5）
Ti s
comparing equation(4)and equation(5), we yield the
corresponding relationship between an ATFIHC and a classic
PID controller as following:

51
 TABLE 1. THE PARAMETER TUNING INDEX OF A PID CONTROLLER
 a1  T  1 
b
T  
b3

IAE ISE ITAE  K F =   1 + a3T   

 K  τ    τ  
a1 1.435 1.495 1.357
 b −b
b1 0.921 0.945 0.947  a1a2  T  1 2
 KI =   （12）
a2 0.878 1.101 0.842
 TK  τ 
b2 0.749 0.771 0.738  T 
b3

a3  α = 1 1 + a3T  
0.482 0.560 0.381
 τ 
b3 -1.137 -1.006 -0.995 
based on equation(12), we modify the coefficients based on
B. The empirical parameter tuning equation of an ATFIHC table 1 by following steps:
From equation (6) and equation (11), we yield ATFIHC step(1):Take typical objects
parameter tuning equation based on optimized PID controller
parameter tuning method, which is as following: From the equation(12) it is clear that the influence that the
object model has on parameter is mainly embodied on
parameters K 、 T and T τ . Take three set of parameter
where in each set T and T τ remain the same and K take
values 0.5,1 and 2; in different sets either T or T τ changes
while other parameter remain the same(see table 2). This
method can cover all the change trends of parameters.

TABLE 2. TYPICAL CONTROLLED OBJECT MODEL AND THE CORRESPONDING OPTIMIZED PID PARAMETER
Model of Controlled Object IAE ISE ITAE

set K T T τ KF KI α KF KI α KF KI α
0.5 8 4 18.5 0.40 0.56 23.4 0.52 0.47 17.8 0.38 0.57

1 1 8 4 9.3 0.20 0.56 11.7 0.26 0.47 8.9 0.19 0.57

2 8 4 4.6 0.10 0.56 5.9 0.13 0.47 4.5 0.10 0.57

0.5 8 2 15.0 0.35 0.36 18.6 0.46 0.31 13.2 0.33 0.40

2 1 8 2 7.5 0.18 0.36 9.3 0.23 0.31 6.6 0.17 0.40

2 8 2 3.7 0.09 0.36 4.7 0.12 0.31 3.3 0.08 0.40

0.5 4 2 10.2 0.71 0.53 12.2 0.93 0.47 9.2 0.66 0.57

3 1 4 2 5.1 0.35 0.53 6.1 0.46 0.47 4.6 0.33 0.57

2 4 2 2.6 0.18 0.53 3.0 0.23 0.47 2.3 0.17 0.57

step(2): Calculate the values of parameters where the step range of K F was 0- K F +5， with step length
According to equation(12)and table 1, we calculate the of 0.1, and the step range of K I was 0- K I +1 with step length
parameter value of ATFIHC listed in the following table2.
of 0.01 and the step range of α was 0-1 with step length of
step(3):Search for optimized parameter 0.01 . The search result value of optimized parameter is listed
We used the data on table 2 as search center and searched in table 3.
for the optimized parameter by means of Grid Search(GS) [10],

TABLE 3. TYPICAL CONTROLLED OBJECT MODEL AND THE CORRESPONDING OPTIMIZED PARAMETER

IAE ISE ITAE

Model of Controlled Object

set K T T τ KF KI α KF KI α KF KI α
0.5 8 4 14.7 0.85 0.52 15.6 1.02 0.45 13.1 0.79 0.54
1
1 8 4 7.6 0.43 0.51 7.8 0.51 0.45 6.0 0.39 0.57

52
 2 8 4 3.6 0.21 0.52 3.9 0.26 0.45 3.0 0.19 0.56

0.5 8 2 10.5 0.41 0.38 12.6 0.50 0.31 9.2 0.39 0.41

2 1 8 2 5.3 0.20 0.37 6.3 0.25 0.31 4.5 0.19 0.41

2 8 2 2.6 0.10 0.38 3.1 0.12 0.31 2.2 0.10 0.41

0.5 4 2 7.9 0.89 0.54 8.3 1.02 0.46 6.4 0.78 0.59

3 1 4 2 3.9 0.44 0.54 4.2 0.52 0.46 3.2 0.39 0.59

2 4 2 2.0 0.22 0.54 2.1 0.26 0.46 1.6 0.19 0.58

step(4):Parameter determination
With the data in table3 based on the basic form of
equation(12),we determine the coefficients of the empirical
equation of ATFIHC parameter tuning by data fitting. The
result is on table 4 as shown below.
Thus, equation(12) is the empirical equation of parameter
tuning of ATFIHC ,the coefficients of the equation are listed on
table 4.

TABLE4. THE PARAMETER TUNING INDEX OF AN ATFIHC

IAE ISE ITAE Figure 2. The comparison of the optimized IAE index

a1 1.0264 1.0867 0.9953

b1 0.9725 0.8458 0.8904

a2 0.7251 0.9022 0.7440

b2 -0.1318 -0.1828 -0.1471

a3 0.3605 0.5067 0.3241

b3 -0.8217 -0.8648 -0.8496

IV. SIMULATION VALIDATION

Assume the controlled object model with delay is as follow: Figure 3. The comparison of the optimized ISE index

1
G1 ( s ) = e −2 s (13)
3s + 1
We calculated the optimized PID controller parameter
according to equation(11)and table 1 and calculate the
ATFIHC parameter according to equation(12)and table 4.The
results are in the following table 5 .

TABLE 5. THE CONTROL PARAMETER ( G1 AS CONTROLLED OBJECT)

Optimized PID controller ATIFHC

KP KI
Index
Ti Td KF α
Figure 4. The comparison of the optimized ITAE index

IAE 2.1 4.63 0.91 2.7 0.39 0.56 From the simulation result it is clear that by using the
parameter tuning method proposed in this paper ,the control
ISE 2.2 3.72 1.11 3.2 0.50 0.48
effect of ATFIHC is superior to optimized PID controller in
ITAE 2.0 4.81 0.76 2.4 0.38 0.59 integral norms such as IAE、ISE、ITAE and control norms
The comparison of simulation results is inllustrated as such as overshoot and response time.
figure2-4 Next, we validated the adaptability of the controller by
changing the parameter and the structure of the transfer
function of the controlled object. We changed the parameter of
the controlled object as in equation(14) and obtain simulation

53
result as in figure5-7,then we changed the structure of the
controlled object as in equation(15) with the simulation result
shown in figure8-10.
1
G2 ( s ) = e −2 s (14)
2.5s + 1

1
G3 ( s ) = e −2 s (15)
(3s + 1)( s + 1)

Figure 8. The comparison of simulation result when the structure changes(IAE

optimized parameter)

Figure 5. The comparison of simulation result when the parameters

change(IAE optimized parameter)

Figure 9. The comparison of simulation result when the structure changes(ISE

optimized parameter)

Figure 6. The comparison of simulation result when the parameters

change(ISE optimized parameter)

Figure 10. The comparison of simulation result when the structure

changes(ITAE optimized parameter)

From the simulation result it is apparent that the control

effect of the PID controller based on precise mathematic model
of the controlled object became weaker, unstable and less
adaptable after the parameter and structure change, while the
ATFIHC was more adaptable with better control effect after the
changes.
V. CONCLUSION
The advantage of Fuzzy control and intelligent control is
Figure 7. The comparison of simulation result when the parameters that it lowers the dependency level that of a controller on the
change(ITAE optimized parameter)
model of the controlled object, which enables the design of a
controller with good control effect when the model of the
controlled object is imprecise. It can also mimic the intelligent
control decision behavior of an experienced operator of the
system to achieve a better control effect. By the parameter
tuning method proposed in this paper, the simulation of
ATFIHC proves the advantages mentioned above. The
ATFIHC has better control effect when the model of the

54
controlled object is given(see figure2-4),whereas when the [4] ZHENG Haiping, ZHONG Chenxing, and WU Liping, “Fuzzy control
model of the controlled object is imprecise, i.e. the parameter based on parameter self-tuning with gain scheduling for an inverted
pendulum,” J. Huazhong Univ. of Sci. & Tech.(Natural Science Edition),
and structure of the model of the controlled object changes ,the 2013, pp. 5-8.
ATFIHC is more adaptable with better control effect(see [5] LI Shiyong, and LI Yan, “The intelligent control,” Beijing: Tsinghua
Figure5-10). university press, 2016, pp. 3–5.
In real applications, the initial parameter of a controller [6] XUE Dingyu, “Computer aided control systems design-using MATLAB
language(Third edition),” Beijing: Tsinghua university press, 2012, pp.
could be calculated with the parameter tuning equation and 330-332.
index proposed in this paper and then the controller parameters [7] O Dwyer A, “Handbook of PI and PID controller tuning rules,”
could be further modified by searching offline or tuning online. London:Imperial College Press, 2003, pp. 86–91.
[8] Visioli A, “Optimal tuning of PID controller for integral and unstable
REFERENCES processes,” Proceedings of IEE, Part D, 2001, pp.180-184
[1] LEI Binbin, BAO Hong, and XU Qian, “Design and structural analysis [9] HUANG Zhonglin, “Calculation and simulation of the conctol system
of interval Type-2 fuzzy PI /PD controller,” Electric Machines and besed on MATLAB,” Beijing: National defense industry press, 2001, pp.
Control, 2016, pp. 50-62. 180–184.
[2] CAI Hongping, WU Lebin, and XU Fang, “Speed control system of [10] ZHANG Gongrang, and WAN Fei, “Application of support vector
brushless DC motor based on fuzzy adaptive-PID control,” Journal of machine in network intrusion detection based on grid search,” Computer
Mechanical & Electrical Engineering, 2016, pp. 991-996. Technology and Development, 2016, pp. 97-100.
[3] LIANG Teicheng, and JIANG Changhong, “Design and simulation of
parameter self-regulation fuzzy control system,” Journal of System
Sinulation, 2006, pp. 628-633.

55