0% found this document useful (0 votes)
16 views14 pages

A Novel Fuzzy Logic Scheme For PID Controller Auto-Tuning

This paper introduces a novel fuzzy logic scheme for the auto-tuning of PID controllers, utilizing expert knowledge within a fuzzy inference system to iteratively enhance closed-loop system performance. The method focuses on key step response characteristics such as rise time, overshoot, and settling time, and operates in three modes to balance system response. The effectiveness of the proposed scheme is validated through computer simulations across various processes.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
16 views14 pages

A Novel Fuzzy Logic Scheme For PID Controller Auto-Tuning

This paper introduces a novel fuzzy logic scheme for the auto-tuning of PID controllers, utilizing expert knowledge within a fuzzy inference system to iteratively enhance closed-loop system performance. The method focuses on key step response characteristics such as rise time, overshoot, and settling time, and operates in three modes to balance system response. The effectiveness of the proposed scheme is validated through computer simulations across various processes.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Automatika

Journal for Control, Measurement, Electronics, Computing and


Communications

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/taut20

A novel fuzzy logic scheme for PID controller auto-


tuning

Ilija Kamenko, Velimir Čongradac & Filip Kulić

To cite this article: Ilija Kamenko, Velimir Čongradac & Filip Kulić (2022) A novel
fuzzy logic scheme for PID controller auto-tuning, Automatika, 63:2, 365-377, DOI:
10.1080/00051144.2022.2043988

To link to this article: https://doi.org/10.1080/00051144.2022.2043988

© 2022 The Author(s). Published by Informa


UK Limited, trading as Taylor & Francis
Group

Published online: 23 Feb 2022.

Submit your article to this journal

Article views: 2518

View related articles

View Crossmark data

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=taut20
AUTOMATIKA
2022, VOL. 63, NO. 2, 365–377
https://doi.org/10.1080/00051144.2022.2043988

A novel fuzzy logic scheme for PID controller auto-tuning


Ilija Kamenko , Velimir Čongradac and Filip Kulić
Faculty of Technical Sciences, Department of Computing and Control Engineering, University of Novi Sad, Novi Sad, Serbia

ABSTRACT ARTICLE HISTORY


This paper presents a novel method for PID (proportional–integral–derivative) controller auto- Received 3 August 2020
tuning based on expert knowledge incorporated into a fuzzy logic inference system. The pro- Accepted 10 September 2021
posed scheme iteratively tries to improve the performance of the closed-loop system. As perfor- KEYWORDS
mance measures, the proposed scheme uses the characteristics of the step response (rise time, PID; fuzzy logic; expert
overshoot, and settling time). PID parameters in the first iteration can be calculated based on the knowledge; auto-tuning
basic open-loop step response experiment or it is possible to use current parameters. In each suc-
cessive iteration, step response characteristics are measured and the relative changes expressed
in the percentage of value in the first iteration are calculated and converted into linguistic values.
The fuzzy expert system computes fuzzy values that are used after defuzzification as multiply-
ing factors for current PID parameters. To achieve a balance between the aggressive and robust
closed-loop response, as well as between the slower and the faster one, the fuzzy expert system
works in three operating modes: the one for speeding up the system, the one for reducing the
overshoot, and the one for a balanced reduction of rise time and overshoot. The performance and
robustness are verified by computer simulation using an extensive range of different processes.

1. Introduction and Hägglund proposed a revitalization of the origi-


All automatic control systems are characterized by a nal method in [4,5] while maintaining simplicity but
specific control law or control laws. For basic control improving performance for a broader set of differ-
laws, their effect may linearly depend on the error (P ent processes. Not long after, Šekara and Mataušek [6]
law), integral of error (I law), or the first derivative of also suggest a revitalization of the original Ziegler and
the error in time (D law). Combining basic control laws Nichols work. A comprehensive overview of develop-
renders the PID (proportional–integral–derivative) law ments in PID controllers theory and practice can be
that is the most commonly used control law today due found in [7], while a summary of tuning methods in
to its simple implementation in modern programmable a systematic way is given in [8].
logic controllers used in industry, as well as home To reduce the time and knowledge necessary for tun-
and building automation. The PID controller has three ing PID parameters, Åström and Hägglund proposed
adjustable parameters, one for each control law. The the first automatic tuning method [9]. Such automatic
values of the parameters are closely related to closed- tuning was accomplished by combining the relay exper-
loop system performance and robustness. However, in iment as a method for determining process dynam-
addition to its simple implementation, setting up three ics with methods for computing the parameters of a
adjustable PID parameters is not an easy task and has PID controller. In recent years, the authors proposed
posed a challenge for many authors for years. different approaches to the automatic tuning of PID
The first systematic way to tune PID parameters was controller parameters [10–15]. The most recent com-
proposed by Ziegler and Nichols in 1942 [1]. Inspired parison of PID auto-tuners is presented in [16,17].
by this work, many authors suggested improvements After a parameter of the PID controller is tuned, it
to meet different performance or robustness require- is often necessary to manually fine-tune parameters.
ments [2,3]. These methods are given in the form of The tuning is achieved by applying little corrections
simple expressions that formalize previously experi- to the parameters based on practical experience. Fuzzy
mentally determined characteristics of process dynam- logic has been recognized as an appropriate method
ics. Despite its imperfection, the original Ziegler and for implementing the operator’s experience into the
Nichols method is still widely used today because of PID controller design. Zadeh proposed a fuzzy set the-
its simplicity and ease of use. In the early 2000s, more ory in 1965 as an extension of the classical notion of
than 60 years later than the original work, Åström a set [18]. Inspired by this work, in 1974 Mamdani

CONTACT Ilija Kamenko [email protected] University of Novi Sad, Faculty of Technical Sciences, Department of Computing and Control
Engineering, Trg Dositeja Obradovića 6, Novi Sad, Serbia
© 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted
use, distribution, and reproduction in any medium, provided the original work is properly cited.
366 I. KAMENKO ET AL.

proposed the first application of the fuzzy set theory in where e(t) = r(t)-y(t) is a system error signal that rep-
the fuzzy logic control algorithm to synthesize the lin- resents the difference between setpoint value r(t) and
guistic control mechanism of an experienced operator current system output value y(t); u(t) is the control vari-
[19]. In the following years, fuzzy logic has been suc- able, Kp is a proportional gain, Ti is an integral time
cessfully used in many applications and has resulted in constant, and Td is a derivative time constant.
significant achievements [20–27]. A concise overview Classical tuning methods like Ziegler and Nichols
of fuzzy control research breakthrough results is given and related methods are model-based methods. These
in [28], showing the evolution of this field. Tzafestas and methods are based on the approximation of process
Papanikopoulos were among the first ones to propose dynamics using a simple model. A simple process
an approach in which the performance of the closed- model can be obtained from an open-loop response.
loop system is improved by slightly changing the values The first-order process with time delay (FOPTD) pro-
of PID parameters based on a fuzzy matrix that contains cess model that is commonly used in process control
the experience of an operator in the condensed rule has the following transfer function:
form [29]. The authors in [30] proposed an auto-tuning
K
algorithm based on a fuzzy inference mechanism for G(s) = e−sL (2)
tuning PID parameters online using the response error 1 + sT
signal of the closed-loop system as input. Many authors where K is the static gain, T is time constant (lag),
have applied a similar principle and introduced slightly and L is time delay. There are some processes with the
different schemes to implement expert knowledge in dynamics that contain integration or very long-time
the PID tuning process by using the fuzzy logic con- constants and that will not reach a steady state under
cept [31–39]. A comparison of fuzzy logic auto-tuning open-loop conditions. A combination of an integrator
methods is presented in [40]. and a time delay makes the first-order integral process
The purpose of this study is to develop a novel with a time delay (FOIPTD) process model with the
scheme for automatic tuning of a PID controller based transfer function:
on expert knowledge incorporated into a fuzzy infer-
ence system. The primary aim of the scheme is to Kv −sL
G(s) = e (3)
enhance the performance of the arbitrary closed-loop s
system while maintaining robustness within accept-
where Kv is velocity gain and L is the time delay. The
able limits with minimal investment of time, effort, and
parameters of the PID controller are then expressed in
required knowledge.
terms of the obtained process parameters using simple
The novelty of the approach is that the scheme in an
formulas [8].
iterative procedure modifies the parameters of the PID
The present research adopted the revitalization
controller based on response characteristics. Unlike
method proposed by Åström and Hägglund [5] as an
other previously mentioned methods [30,31,36] that
initial tuning for a new proposed fuzzy expert system
use the error signal and the first derivate of the error
method for PID parameter auto-tuning. This method
signal, this method uses the rise time, overshoot, and
was first used in [4] for PI controller and the same
the settling time as input in the fuzzy expert system.
approach was later used for PID control with additional
In addition, this method works offline as the classical
constraints in [5]. For simplicity, the Åström and Häg-
auto-tuning method and is not intended for continu-
glund method will be denoted as the A-H method in the
ous adaptation, unlike most methods that utilize the
rest of this paper. For stable processes, according to this
fuzzy concept. Besides, this approach makes it possi-
method, PID parameters can be expressed as:
ble to select an operating mode and, in that way, make
an influence on setting performance and robustness  
1 T
tuning goals. Kp = 0.2 + 0.45 (4)
K L
The motivation for this research comes from the
0.4L + 0.8T
real problems that the authors encountered during the Ti = L (5)
implementation of the automatic control system over L + 0.1T
the years in industry and building automation. 0.5LT
Td = (6)
0.3L + T

2. Classical tuning methods For integrating processes, Equations (4)-(6) can be


written as:
The well-known expression of a PID controller is the
0.45
following: Kp = (7)
Kv
  t  Ti = 8L (8)
1 de(t)
u(t) = Kp e(t) + e(τ )dτ + Td (1) Td = 0.5L (9)
Ti 0 dt
AUTOMATIKA 367

Figure 1. Block diagram of the proposed FES scheme.

3. Proposed fuzzy expert system 3.1. Test batch


A block diagram of the proposed fuzzy expert sys- The test batch consists of 133 processes whose trans-
tem (FES) for auto-tuning PID controller parameters is fer functions are given in Equations (16)–(24). The
shown in Figure 1. test batch is discussed in [5] and it is here used prac-
It consists of the performance evaluation part, com- tically unchanged. The test batch is used to derive
parator, fuzzy interference system, and parameter cal- expert knowledge and test final solutions. The test batch
culation part. The main purpose of the performance includes delay dominated, lag dominated, and integrat-
evaluation part is to evaluate the closed-loop step ing processes.
response based on current value y(t) and setpoint value
r(t). The closed-loop response is evaluated using ris- e−s
ing time tr (time from 10% to 90% steady-state value), G1 (s) = ,
1 + sT
overshoot o, and settling time ts (time of entering in 2%
region of steady-state value) as performance measures. T = 0.02, 0.05, 0.1, 0.2, 0.3, 0.5,
The comparator compares the current values of perfor- 0.7, 1, 1.3, 1.5, 2, 4, 6, 8, 10,
mance measures with reference values that are saved 20, 50, 100, 200, 500, 1000 (16)
in the first iteration. The comparator renders the per-
centage ratios of rise time tr_%−rat , overshoot o_%−rat , e−s
G2 (s) = ,
and settling time ts _%−rat , which can be expressed as (1 + sT)2
follows: T = 0.01, 0.02, 0.05, 0.1, 0.2,
tr 0.3, 0.5, 0.7, 1, 1.3, 1.5,
tr_%−rat = 100% (10)
tr_ref
2, 6, 8, 10, 20, 50, 100,
o
o_%−rat = 100% (11) 200, 500 (17)
o_ref
ts 1
ts_%−rat = 100% (12) G3 (s) = ,
ts_ref (1 + s)(1 + sT)2
T = 0.005, 0.01, 0.02,
The fuzzy interference system based on percentage
ratios calculates PID parameter multipliers Kp_multp , 0.05, 0.1, 0.2, 0.5,
Ti_multp , and Td_multp using expert knowledge within 2, 5, 10 (18)
the rule base and fuzzy inference mechanism. The
1
parameter calculation part calculates new PID param- G4 (s) = , n = 3, 4, 5, 6, 7, 8 (19)
eters Kp , Ti , and Td using multipliers calculated in the (1 + s)n
previous part and the current value of PID parameters, 1
G5 (s) = ,
according to the following equations: (1 + s)(1 + αs)(1 + α 2 s)(1 + α 3 s)
α = 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9
Kp = Kp + Kp Kp_multp (13)
(20)

Ti = Ti + Ti Ti_multp (14) e−sL


G6 (s) = ,
s(1 + Ts)
Td = Td + Td Td_multp (15) T = 0.01, 0.02, 0.05, 0.1, 0.3, 0.5, 0.7,
368 I. KAMENKO ET AL.

Figure 2. Tuning map for PID control of a process G4 (19) with parameter n = 4 (blue solid Td = 0.2; red dotted Td = 0.7; green
dashed Td = 1.1).

0.9, 1.0 T + L = 1 (21) are modified. The modifications are based on simple
rules that can be expressed as tuning maps. The tuning
T
G7 (s) = e−sL , maps are a matrix of transient responses correspond-
(1 + sT)(1 + sT1 )
ing to systematic variations in PID parameters whose
L = 0.01, 0.02, 0.05, 0.1, 0.3, 0.5, 0.7, 0.9, 1.0 essential function is to provide insight into how changes
T = 1, 2, 5, 10 T1 + L = 1 (22) in controller parameters influence the behaviour of the
closed-loop system.
1 + αs
G8 (s) = , To speed up the synthesis of expert knowledge with-
(1 + s)3 out losing the generality, only a few processes from the
α = 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, test batch given in Equations (16)-(24) are observed.
Each process represents one of the four ways to clas-
0.9, 1.0, 1.1 (23)
sify the processes of the test batch. The systems with
1 the integrator are classified into the first group rep-
G9 (s) = ,
(1 + s)((sT) + 1.4sT + 1))
2
resented by process G6 given in Equation (21) with
T = 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0 parameter L = 0.1. Process G5 given in Equation (20)
(24) with parameter α = 0.5 represents the second group of
lag-dominated processes. The third group includes the
balanced lag and delay processes represented by process
3.2. Expert knowledge G4 given in Equation (19) with parameter n = 4. Pro-
Since many methods give only rouge values, it is often cess G2 given in Equation (17) with parameter T = 0.02
necessary to apply manual tuning to obtain better represents the last group of delay-dominated processes.
closed-loop properties. This is typically done through Tuning maps are created for all four processes. As an
experiments with the closed-loop process. The closed- example, a tuning map for process G4 given by Equation
loop response is observed and controller parameters (18) with parameter n = 4 is shown in Figure 2. The
AUTOMATIKA 369

Figure 3. Block diagram of a fuzzy interference system.

Table 1. Basic relations between increasing PID controller


parameters and step response characteristics.
Parameter Rise time Overshoot Settling time
Kp decrease increase minor decrease
Ti increase decrease increase/decrease
Td minor increase decrease decrease

figure illustrates how step response is influenced by


changes in the proportional gain (Kp ), integral time
(Ti ), and derivative time (Td ). Parameter Kp increases
along abscissa and parameter Ti along the ordinate,
while the response for each value of parameter Td is
represented with different line types and colours.
Based on the created tuning maps it is possible to
derive the relations between the increasing PID con- Figure 4. Membership functions for three inputs: rise time,
troller parameter and the values of step response char- overshoot, and settling time ratios.
acteristics (rise time, overshoot, and settling time). The
derived fundamental relations are given in Table 1.

3.3. Fuzzy interference system


This study adopted the Mamdani fuzzy interference
system with three inputs and three outputs. A block
diagram of the proposed system is shown in Figure 3.
The rise time ratio tr_%−rat , overshoot ratio o_%−rat ,
and settling time ratio ts_%−rat were chosen as input
parameters. The value of input parameters was set to
range from 0% to 200%. For the ratios greater than
200%, the values of current measures have already been
certainly larger than those of referent values, so the
upper limit was saturated at level 200. The input vari-
ables have been decomposed into three different fuzzy
linguistic levels: low (L), medium (M), and high (H). Figure 5. Membership functions for three outputs: Kp , Ti , and
The shape of membership functions for all three inputs Td multipliers.
is shown in Figure 4.
Multipliers Kp_multp , Ti_multp, and Td_multp were cho- negative medium (NM), negative low (NL), zero (Z),
sen as output parameters. The value of the output positive low (PL), positive medium (PM), and positive
parameters was set to be in the range from −0.5 to high (PH). The shape of membership functions for all
0.5. Therefore, PID parameters could be adjusted by three multipliers can be seen in Figure 5.
up to 50% of their value up or down in each iteration. A comparative analysis of relations obtained from
The output variables have been decomposed into seven tuning maps of four representative processes was used
different fuzzy linguistic levels: negative high (NH), to create a rule base. Since each input is described
370 I. KAMENKO ET AL.

Table 2. The rule base of a fuzzy interference system.


Rule tr_%−rat o_%−rat ts_%−rat Kp_multp Ti_multp Td_multp
1 L L L Z Z Z
2 L L M Z Z Z
3 L L H NL NM Z
4 L M L NL PM PL
5 L M M NL PM PL
6 L M H NL PM NL
7 L H L NL PM PL
8 L H M NL PM PL
9 L H H NL NM PL
10 M L L PM PL NL
11 M L M PM PL NL
12 M L H PM PL NL
13 M M L PM PM Z
14 M M M PM PM Z
15 M M H PM PM Z
16 M H L NM PH PM
17 M H M NM PH PM
18 M H H NM PH PM
19 H L L PH NL NL Figure 6. Positions of membership functions for two different
20 H L M PH NL NL values of a parameter P – solid lines P = 50, dashed lines
21 H L H PH NL NL P = 75.
22 H M L PH NM NM
23 H M M PH NM NM
24 H M H PH NM NM
25 H H L NH NH NH
26 H H M NH NH NH algorithm can be obtained. This study proposes three
27 H H H NH PH NH operating modes of the FES tuning method.
The first operating mode is Speed up tuning, whose
primary objective is to decrease the rise time of the
with three fuzzy sets, a well-defined fuzzy inference closed-loop system in the first place, and then decrease
system should have 3 × 3×3 = 27 fuzzy rules. Each the settling time and overshoot while keeping sensi-
rule defines how PID parameters should be changed tivity within reasonable limits. Therefore, the value of
depending on the values of ratios between current and parameter P for the first input tr_%−rat was set to 40, for
referent performance measures. The overview of fuzzy the second input o_%−rat it was set to 100, and for the
rules is shown in Table 2. third input ts_%−rat it was set to 100.
The second operating mode is Reduce overshoot
3.4. Operating modes tuning, whose primary objective is to decrease the over-
shoot of the closed-loop system in the first place, and
The trade-off between performance and robustness then decrease rise time and settling time while keeping
varies among different control problems, which is why sensitivity within reasonable limits. Therefore, the value
it is desirable to have a design parameter to change the of parameter P for the first input tr_%−rat was set to 100,
properties of the closed-loop system. The positions of for the second input o_%−rat it was set to 40, and for the
membership functions of fuzzy sets L, M, and H for third input ts_%−rat it was set to 100.
all three inputs have a considerable influence on FES The third operating mode is Balanced tuning, whose
operation, which can be used to select an objective for primary objective is to balance the decrease overshoot
PID tuning. Triangular and trapezoidal membership and rise time of the closed-loop system in the first place,
functions can be expressed by a set of their vertexes. and then reduce the settling time while keeping sensi-
The position of the membership function depending tivity within reasonable limits. Therefore, the value of
on parameter P can be described by the following parameter P for each of three inputs (tr_%−rat , o_%−rat ,
equations: and ts_%−rat ) was set to 75.
L = [0, 0, P, max((50 + P), 100)] (25)
M = [P, max((50 + P), 100), max((100 + P), 150)] 3.5. Tuning procedure
(26)
The flow chart of the proposed FES auto-tuning scheme
H = [max((50 + P), 100), 200, 200] (27) is shown in Figure 7. If there are no initial PID param-
eters, or the closed-loop system has sluggish perfor-
Figure 6 shows the position of membership func- mance, new initial parameters can be calculated in
tions for two different values of parameter P. From the pre-tuning phase. Alternatively, pre-tuning can be
Equations (25)-(27) it can be seen that when P = 0, skipped if closed-loop performances are satisfactory
membership function positions come down to the basic (Step 2).
version shown in Figure 4. With the proper selection of In the pre-tuning phase, the PID controller is discon-
parameter P, different operating modes of the proposed nected. Therefore, initial tuning is performed offline.
AUTOMATIKA 371

Based on the open-loop response, the system is approx-


imated with the model given in Equations (2) or (3),
depending on whether the system possesses an inte-
grator (Steps 3-4). After successfully obtaining initial
parameters (Step 6), the PID controller is connected
again, and the second phase can begin.
The second phase is performed online, with initial
parameters adopted as current parameters (Step 7).The
first iteration of the fine-tuning phase includes setpoint
perturbation (Step 8) and calculation of performance
measures of the step response (Step 9). The computed
performance measures are saved as referent values for
this iteration and further ones (Step 10). The relative
values of performance measures expressed as the per-
centage of saved referent values are calculated in Step
11. The fuzzy expert system generates multiplier fac-
tors for PID parameters using calculated relative values
of performance measures as input in Step 12. Based on
computed multiplier factors and current values of PID
parameters, new parameters are calculated in Step 13.
These steps are repeated until multipliers are greater
than tolerance, and the number of iterations does not
exceed the maximum number (Step 14). When mul-
tipliers become smaller than tolerance, the last PID
parameters are adopted as final (Step 15) and the proce-
dure is finished successfully (Step 16). If the number of
iterations is exceeded, parameters are set back to initial
values (Step 15) and the procedure ends unsuccessfully
(Step 16). This stopping criterion based on the num-
ber of iterations is beneficial if the process does not
converge.

4. Results and discussion


In addition to the performance measures used in the
tuning process, additional performance and robustness
measures are used to enable a better comparison of the
obtained results. The integral of absolute error IAE is a
beneficial measure given in:
 t
IAE = |e(t)|dt (28)
0

where e(t) = r(t)-y(t) is a system error that represents


the difference between setpoint value r(t) and current
system output value y(t). The integral of the absolute
error for load disturbance response IAEd is the same as
the previously defined measure but it is used in load dis-
turbance response. Robustness is specified through the
maximum value of the sensitivity function:
 
 1 
Ms = max    (29)
ω 1 + Gc (jω)G(jω) 

where Gc is the transfer function of the controller, while


G is the transfer function of the process. Value Ms can
be interpreted as the inverse of the shortest distance
between the loop transfer function and the critical point
Figure 7. Flow chart of the proposed FES tuning method. in the Nyquist plot [5].
372 I. KAMENKO ET AL.

Figure 8. Procedure of tuning PID controller parameters for process G6 (21) with parameter T = 0.5 using the proposed FES tuning
method in the speed up tuning mode.

For all processes from the test batch given in Equa- Table 3. Overview of PID parameters through iterations of the
tions (16)-(24), excluding the four ones used for expert proposed FES tuning method in the speed up tuning mode for
knowledge synthesis, PID parameters are initially tuned process G6 (21) with parameter T = 0.5.
by the A-H method. The processes are then fine-tuned Iteration Kp Ti Td
by the FES tuning method with all three modes. 1 0.45 8.00 0.50
The results obtained using both methods are com- 2 0.56 8.80 0.45
3 0.68 9.49 0.41
pared below. 4 0.79 10.08 0.39
5 0.90 10.52 0.37
6 0.97 10.77 0.36
7 0.99 10.83 0.36
4.1. Fes tuning method in the speed up mode
Consider process G6 (21) with parameter T = 0.5 as
the first example. Fitting the model given in Equation Table 3 shows how parameters are changed through
(3) to the process, it was found that the apparent time iterations, from the first one to the last one. For all seven
delay and velocity gain were T = 1 and Kv = 1. By iterations, the responses of the system to step changes
applying A-H tuning, the following PID parameters are in setpoint and load are shown in Figure 9. From the
obtained: Kp = 0.45, Ti = 8, and Td = 0.5. The FES figure, it can be seen that the rise time is shorter and
tuning method in the speed up mode is applied and the shorter with each iteration. At the same time, the over-
procedure through iterations can be seen in Figure 8. shoot hardly changes while the rise time is slightly
The procedure starts at moment t1 and ends at moment improved. It can also be seen that the response to the
t8 . After seven iterations, the following PID parameters unit step changes in load improves in each subsequent
are obtained: Kp = 0.99, Ti = 10.83, and Td = 0.36. iteration.

Figure 9. Response to a unit step change in setpoint (a) and load (b) for process G6 (21) with parameter T = 0.5. The PID controller
is tuned by the A-H method and then fine-tuned by the proposed FES tuning method in the speed up tuning mode.
AUTOMATIKA 373

Table 4. Performance and robustness of the PID controller for process G6 (21) with parameter T = 0.5 tuned by the A-H method and
fine-tuned by the FES tuning method in the speed up mode.
tr [s] o [%] ts [s] IAE IAEd Ms
A-H method 2.52 18.89 20.77 3.67 17.79 1.22
FES tuning – speed up mode 1.01 18.69 17.58 2.24 10.93 1.62
Improvement [%] −60.02 −1.10 −15.34 −38.98 −38.57 32.64

Table 5. Average improvement of performance and robustness measures for all processes achieved by the FES tuning method in
the speed up mode after A-H initial tuning.
Process type Count tr [%] o[%] ts [%] IAE [%] IAEd [%] Ms [%] Iterations
G1 (16) 21 −60.92 −27.97 −18.99 −25.05 −24.24 20.48 3.57
G2 (17) 20 −61.08 −17.80 −8.28 −32.94 −31.50 22.56 11.60
G3 (18) 10 −59.64 −3.28 −11.65 −44.59 −31.46 20.98 23.50
G4 (19) 5 −57.67 −2.19 72.44 −9.64 −8.93 46.05 49.00
G5 (20) 8 −57.68 −21.02 37.60 −25.88 −16.66 25.85 37.25
G6 (21) 8 −60.15 −14.21 −21.99 −42.47 −40.64 21.79 8.63
G7 (22) 36 −60.35 −14.73 −39.89 −44.56 −42.00 18.94 10.75
G8 (23) 11 −60.68 −48.21 21.78 −22.44 −18.58 36.72 19.09
G9 (24) 10 −52.50 10.41 169.67 58.61 161.35 39.95 46.60
All 129 −59.64 −17.25 2.37 −27.06 −15.97 24.71 17.19

Performance and robustness measures are shown in Table 6. Overview of PID parameters through iterations of the
Table 4 to enable a better comparison of results. It can proposed FES tuning method in the reduce overshoot mode for
be seen from the table that the proposed method meets process G5 (20) with parameter α = 0.1.
the previously set goals but with the degradation of Iteration Kp Ti Td
sensitivity, which is expected as the system speeds up. 1 4.29 0.44 0.05
The procedure is repeated for all other processes, and 2 3.86 0.55 0.06
3 4.20 0.64 0.06
improvements are calculated. Average improvements 4 4.76 0.71 0.06
grouped by process type are displayed in Table 5. The 5 5.38 0.74 0.05
FES tuning method in the speed up mode achieved a
significant average decrease of nearly 60% in rise time,
which is the primary goal of this operating mode. At (2) to the process, it was found that the apparent time
the same time, a visible average improvement of over- delay and time constants were L = 0.11 and T = 0.99.
shoot and settling time can also be observed, excluding Applying the A-H method, the following PID parame-
a couple of groups of processes with the degradation ters are obtained: Kp = 4.29, Ti = 0.44, and Td = 0.05.
of settling time. The values of the integral of the abso- The FES tuning method in the reduce overshoot mode
lute error of the reference step response and load step is applied and the procedure through iterations can be
response are also noticeably reduced on average. Speed- seen in Figure 10. The procedure starts at moment t1
ing up the system degradation of sensitivity is expected and ends at moment t6 . After five iterations, the follow-
and is almost 25% on average, which is satisfactory. ing PID parameters are obtained: Kp = 5.38, Ti = 0.74,
Considering all the processes, the average duration of and Td = 0.05. Table 6 shows how the parameters are
the procedure is about 17 iterations. changed through iterations, from the first one to the
last one. For all five iterations, the responses of the
system to step changes in the setpoint and load are
shown in Figure 11. The figure shows that overshoot
4.2. Fes tuning method in the reduce overshoot
decreases throughout iterations. At the same time, the
mode
rise time changes little while the settling time shows vis-
Consider process G5 (20) with parameter α = 0.1 as the ible improvement. It can also be seen that the response
second example. Fitting the model given in Equation to unit step changes in load is improved.

Figure 10. Procedure of tuning PID controller parameters for process G5 (20) with parameter α = 0.1 using the proposed FES tuning
method in the reduce overshoot mode.
374 I. KAMENKO ET AL.

Figure 11. Response to a unit step change in setpoint (a) and load (b) for process G5 (20) with parameter α = 0.1. The PID controller
is tuned by the A-H method and then fine-tuned by the FES tuning method in the reduce overshoot mode.

Table 7. Performance and robustness of the PID controller for process G5 (20) with parameter α = 0.1 tuned by the A-H method
and fine-tuned by the FES tuning method in the reduce overshoot mode.
tr [s] o [%] ts [s] IAE IAEd Ms
A-H method 0.32 15.74 1.50 0.31 0.11 1.14
FES tuning – red. overshoot mode 0.32 5.41 1.42 0.23 0.14 1.09
Improvement [%] −1.35 −65.63 −5.31 −25.16 25.57 −3.90

Performance and robustness measures are shown in of an absolute error on the reference step response is
Table 7. It can be seen that the proposed method meets reduced by nearly 17% on average. At the same time,
the previously set goals with a little improvement of there is a small average degradation of the integral of an
sensitivity. absolute error on the load step response and sensitivity.
The procedure is repeated for all processes, and The average duration of the procedure considering all
improvements are calculated. The results are grouped the processes is nearly ten iterations.
by process type and average improvements for all
groups are displayed in Table 8. The FES tuning method
4.3. Fes tuning method in the balanced mode
in the reduce overshoot operating mode achieved a sig-
nificant average decrease of nearly 64% in overshoot, Consider process G7 (22) with parameter T = 1 and
which is the primary goal of this operating mode. At the L = 0.1 as the third example. Fitting the model given
same time, there is a minimal average improvement of in Equation (2) to the process, it was found that the
rising time and settling time. The value of the integral apparent time delay and time constants were L = 0.6

Table 8. Average improvement of performance and robustness measures for all processes achieved by the FES tuning method in
the reduce overshoot mode after A-H initial tuning.
Process type Count tr [%] o[%] ts [%] IAE [%] IAEd [%] Ms [%] Iterations
G1 (16) 21 −25.49 −25.96 14.60 −14.28 21.84 33.82 16.33
G2 (17) 20 −19.54 −26.00 −4.50 −15.68 −7.02 26.37 13.50
G3 (18) 10 −2.10 −72.66 −11.78 −21.52 21.53 −2.79 5.30
G4 (19) 5 −3.33 −99.15 −18.30 −9.85 −5.52 1.38 4.00
G5 (20) 8 −1.52 −90.85 −31.24 −17.49 −1.41 −2.35 4.13
G6 (21) 8 −6.10 −57.55 53.03 −11.62 84.49 6.02 35.38
G7 (22) 36 −4.97 −75.99 −22.13 −23.43 5.63 1.93 4.78
G8 (23) 11 −3.84 −99.97 −15.33 −12.51 −7.24 5.35 4.27
G9 (24) 10 −1.78 −94.33 −10.70 −14.39 1.03 −1.47 4.40
All 129 −9.80 −63.98 −6.91 −17.33 10.11 10.54 9.81
AUTOMATIKA 375

Figure 12. Procedure of tuning PID controller parameters for process G7 (22) with parameters T = 1 and L = 0.1 using the proposed
FES tuning method in the balanced mode.

Table 9. Overview of PID parameters through iterations of the in Figure 13. The figure shows that overshoot, rise time,
proposed FES tuning method in the balanced mode for process and settling time decrease throughout iterations. It can
G7 (22) with parameters T = 1 and L = 0.1. also be seen that the response to the unit step changes
Iteration Kp Ti Td in load improves.
1 1.24 1.10 0.27 Based on the performance and robustness measures
2 1.36 1.28 0.27 shown in Table 10, it can be concluded that the pro-
3 1.63 1.38 0.25
4 1.88 1.45 0.23 posed method meets the previously set goals.
5 2.05 1.49 0.23 The procedure is repeated for all processes, and
improvements are calculated. The average improve-
ments for all groups of processes are displayed in Table
and T = 1.39. Applying A-H tuning, the following 11. The FES tuning method in the balanced mode
PID parameters are obtained: Kp = 1.24, Ti = 1.1, and achieved a significant average improvement of over 30%
Td = 0.27. The FES tuning method is applied in the in settling time and overshoot, which is the main goal of
balanced mode, and the procedure through iterations this mode. At the same time, the average improvement
can be seen in Figure 12. The procedure starts at in the settling time and integral of the absolute error
moment t1 and ends at moment t6 . After five iterations, of the reference and load step response is significant
the following PID parameters are obtained: Kp = 2.05, with approximately equal improvements across groups.
Ti = 1.49, and Td = 0.23. Table 9 shows how parame- A degradation of sensitivity is visible, with a highly sat-
ters are changed through iterations, from the first one to isfactory value of 12% on average. Considering all the
the last one. For all five iterations, the responses of the processes, the average duration of the tuning procedure
system to step changes in setpoint and load are shown is about seven iterations.

Figure 13. Response to a unit step change in setpoint (a) and load (b) for process G7 (22) with parameters T = 1 and L = 0.1. The
PID controller is tuned by the A-H method and then fine-tuned by the proposed FES tuning method in the balanced mode.
376 I. KAMENKO ET AL.

Table 10. Performance and robustness of the PID controller for process G7 (22) with parameters T = 1 and L = 0.1 tuned by the
A-H method and fine-tuned by the FES tuning method in the balanced mode.
tr [s] o [%] ts [s] IAE IAEd Ms
A-H method 1.65 12.21 5.60 1.43 1.05 1.27
FES tuning – balanced mode 1.23 8.71 3.94 0.99 0.73 1.25
Improvement [%] −25.46 −28.65 −29.67 −30.36 −30.35 −1.08

Table 11. Average improvement of performance and robustness measures for all processes achieved by the FES tuning method in
the balanced mode after A-H initial tuning.
Process type Count tr [%] o[%] ts [%] IAE [%] IAEd [%] Ms [%] Iterations
G1 (16) 21 −37.81 −12.67 3.02 −17.04 −8.57 28.23 12.19
G2 (17) 20 −38.51 −13.07 −16.99 −23.78 −23.12 17.53 13.70
G3 (18) 10 −26.03 −35.00 −20.57 −28.66 −20.26 3.02 5.40
G4 (19) 5 −28.23 −82.89 −15.56 −20.06 −17.75 9.75 4.60
G5 (20) 8 −26.18 −30.72 −34.57 −26.98 −25.13 5.06 5.00
G6 (21) 8 −27.13 −32.29 9.00 −20.67 −2.49 5.71 4.25
G7 (22) 36 −30.17 −43.41 −25.21 −28.94 −22.29 6.05 4.69
G8 (23) 11 −29.02 −80.88 −28.73 −22.46 −19.76 12.07 4.45
G9 (24) 10 −25.91 −37.03 −16.44 −25.75 −19.37 8.44 6.40
All 129 −31.45 −35.80 −16.68 −24.40 −18.36 11.97 7.47

5. Conclusions Acknowledgment
This paper studied a novel fuzzy expert scheme for PID The authors would also like to express their sincere gratitude
to prof. Milan Rapaić from University of Novi Sad, Faculty
auto-tuning. The proposed scheme utilizes the power of
of Technical Sciences and prof. Veljko Milutinović from the
the fuzzy logic concept to enhance the tuning of the PID University of Belgrade, School of Electrical Engineering for
controller. This approach incorporates the control engi- the constructive advice that has contributed to the quality of
neer’s experience into a fuzzy expert system that would this paper.
take over the tuning and deliver improved performance
over the PID controller tuned using the classical tuning Disclosure statement
method. The proposed scheme updates the PID con-
No potential conflict of interest was reported by the author(s).
troller parameters through iterations based on response
characteristics. The tuning process is guided by the
scheme that analyzes process data much faster than a Funding
human, which results in a significant acceleration of The authors are grateful for the support of the Ministry of
manual tuning. Education, Science and Technological Development of the
The scheme has been tested through simulations on Republic of Serbia, grant number TR32018.
a broad set of different processes. The extensive set of
processes was used to emphasize the power of general- ORCID
ization of the proposed solution. The A-H method as a Ilija Kamenko http://orcid.org/0000-0003-3352-7637
representative of classical tuning methods was used as a Velimir Čongradac http://orcid.org/0000-0002-2246-8945
reference point for comparison of obtained results. The Filip Kulić http://orcid.org/0000-0002-0982-3017
A-H method already gives a good result for all observed
processes, which leaves less room for improvement.
References
However, the results showed that the proposed scheme
could produce improvements in performance and at [1] Ziegler JG, Nichols NB. Optimum settings for automatic
controllers. Trans Am Soc Mech Eng. 1942;64:759–768.
the same time keep the robustness within reasonable
[2] Chien KL, Hrones JA, Reswick JB. On the automatic
ranges. The operating modes raise the usability of the control of generalised passive systems. Trans Am Soc
proposed scheme to a higher level by allowing it to pri- Mech Eng. 1952;74:175–185.
oritize the goals of the tuning procedure. The results [3] Cohen GH, Coon GA. Theoretical considerations of
demonstrated that all three operating modes achieved retarded control. Trans Am Soc Mech Eng. 1953;75(1):
performance improvement following the initial set pri- 827–834.
[4] Hägglund T, Åström KJ. Revisiting the Ziegler-Nichols
orities (speeding up, reduce overshot, or balanced) tuning rules for PI control. Asian J Control. 2002;4(4):
in a relatively short time with little degradation of 364–380. doi:10.1111/j.1934-6093.2002.tb00076.x.
robustness. [5] Åström KJ, Hägglund T. Revisiting the Ziegler–Nichols
Future work will focus on applying the proposed step response method for PID control. J Process Con-
scheme to real controllers used in the industrial and trol. 2004;14(6):635–650. doi:10.1016/j.jprocont.2004.
01.002.
building automation application, and verification of the
[6] Šekara TB, Mataušek MR. Revisiting the Ziegler–
scheme in a real environment. Nichols process dynamics characterization. J Process
AUTOMATIKA 377

Control. 2010;20(3):360–363. doi:10.1016/j.jprocont. [25] Devi Vidhya S, Balaji M. Hybrid fuzzy PI controlled
2009.08.004. multi-input DC/DC converter for electric vehicle
[7] Åström KJ, Hägglund T. Advanced PID control. application. Automatika. 2020;61(1):79–91. doi:10.1080/
Research Triangle Park (NC): ISA - Instrumentation, 00051144.2019.1684038.
Systems, and Automation Society; 2006. [26] Du P, Pan Y, Li H, et al. Nonsingular finite-time
[8] O’Dwyer A. Handbook of PI and PID controller tuning event-triggered fuzzy control for large-scale non-
rules. 3rd ed London: Imperial College Press; 2009. linear systems. IEEE Trans. Fuzzy Syst. 2020: 1.
[9] Åström KJ, Hägglund T. Automatic tuning of sim- doi:10.1109/TFUZZ.2020.2992632.
ple regulators with specifications on phase and ampli- [27] Farsizadeh H, Gheisarnejad M, Mosayebi M, et al.
tude margins. Automatica (Oxf). 1984;20(5):645–651. An intelligent and fast controller for DC/DC con-
doi:10.1016/0005-1098(84)90014-1. verter feeding CPL in a DC microgrid. IEEE Trans
[10] Huang H-P, Lin C-H. A stable on-line self-tuning opti- Circuits Syst II Express Briefs. 2020;67(6):1104–1108.
mal PID controller for a class of unknown systems. doi:10.1109/TCSII.2019.2928814.
Asian J Control. 2007;9(2):151–162. doi:10.1111/j. [28] Nguyen AT, Taniguchi T, Eciolaza L, et al. Fuzzy control
1934-6093.2007.tb00318.x. systems: past, present and future. IEEE Comput Intell
[11] Dey C, Mudi RK. An improved auto-tuning scheme Mag. 2019;14(1):56–68. doi:10.1109/MCI.2018.2881
for PID controllers. ISA Trans. 2009;48(4):396–409. 644.
doi:10.1016/j.isatra.2009.07.002. [29] Tzafestas S, Papanikolopoulos NP. Incremental fuzzy
[12] Romero JA, Sanchis R, Balaguer P. Pi and PID auto- expert PID control. IEEE Trans Ind Electron. 1990;37(5):
tuning procedure based on simplified single parameter 365–371. doi:10.1109/41.103431.
optimization. J Process Control. 2011;21(6):840–851. [30] Jigang H, Jie W, Hui F. An anti-windup self-tuning fuzzy
doi:10.1016/j.jprocont.2011.04.003. pid controller for speed control of brushless dc motor.
[13] Pavković D, Polak S, Zorc D. Pid controller auto- Automatika. 2017;58(3):321–335. doi:10.1080/000511
tuning based on process step response and damp- 44.2018.1423724.
ing optimum criterion. ISA Trans. 2014;53(1):85–96. [31] He SZ, Tan S, Xu FL, et al. Fuzzy self-tuning of
doi:10.1016/j.isatra.2013.08.011. PID controllers. Fuzzy Sets Syst. 1993;56(1):37–46.
[14] Matić D, Kamenko I, Nikolić P, et al. Optimal control of doi:10.1016/0165-0114(93)90183-I.
fruit and vegetables drying process. J. Process. Energy [32] Zhao ZY, Tomizuka M, Isaka S. Fuzzy gain schedul-
Agric. 2014;18(3). ing of PID controllers. IEEE Trans Syst Man Cybern.
[15] Cetin M, Iplikci S. A novel auto-tuning PID con- 1993;23(5):1392–1398. doi:10.1109/21.260670.
trol mechanism for nonlinear systems. ISA Trans. [33] Petrov M, Proychev T, Topalov A. Expert PID controller
2015;58:292–308. doi:10.1016/j.isatra.2015.05.017. with fuzzy self-tuning. IFAC Proc. 1995;28(5):367–372.
[16] Kumar R, Singla SK, Chopra V. Comparison among doi:10.1016/S1474-6670(17)47254-5.
some well known control schemes with different tun- [34] Almeida OdM, Coelho AAR. A fuzzy logic method for
ing methods. J. Appl. Res. Technol. 2015;13(3):409–415. autotuning a PID controller: SISO and MIMO systems.
doi:10.1016/j.jart.2015.07.007. 15th triennial World Congress; 2002, pp. 1–6.
[17] Berner J, Soltesz K, Hägglund T, et al. An experimen- [35] Niewiara Ł, Zawirski K. Auto-tuning of PID con-
tal comparison of PID autotuners. Control Eng Pract. troller based on fuzzy logic. Comput Appl Electr Eng.
2018;73:124–133. doi:10.1016/j.conengprac.2018.01. 2013;11:230–240.
006. [36] Wang Y, Jin Q, Zhang R. Improved fuzzy PID controller
[18] Zadeh LA. Fuzzy sets. Inf Control. 1965;8(3):338–353.. design using predictive functional control structure.
[19] Mamdani EH. Application of fuzzy algorithms for con- ISA Trans. 2017;71:354–363. doi:10.1016/j.isatra.2017.
trol of simple dynamic plant. Proc Inst Electr Eng. 09.005.
1974;121(12):1585–1588. doi:10.1049/piee.1974.0328. [37] Sivalingam R, Chinnamuthu S, Dash SS. A modi-
[20] Čongradac V, Prica M, Paspalj M, et al. Algorithm for fied whale optimization algorithm-based adaptive fuzzy
blinds control based on the optimization of blind tilt logic PID controller for load frequency control of
angle using a genetic algorithm and fuzzy logic. Sol autonomous power generation systems. Automatika.
Energy. 2012;86(9):2762–2770. doi:10.1016/j.solener. 2017;58(4):410–421. doi:10.1080/00051144.2018.146
2012.06.016. 5688.
[21] Zdešar A, Cerman O, Dovžan D, et al. Fuzzy control of [38] Khooban MH. Hardware-in-the-loop simulation for
a helio-crane: Comparison of two control approaches. the analyzing of smart speed control in highly non-
J Intell Robot Syst Theory Appl. 2013;72(3–4):497–515. linear hybrid electric vehicle. Trans Inst Meas Control.
doi:10.1007/s10846-012-9796-0. 2019;41(2):458–467. doi:10.1177/0142331218764784.
[22] Bačkalić T, Bugarski V, Kulić F, et al. Adaptable fuzzy [39] Gheisarnejad M, Khooban MH. Design an optimal
expert system for ship lock control support. J. Navig. fuzzy fractional proportional integral derivative con-
2016;69(6):1341–1356. doi:10.1017/S03734633160 troller with derivative filter for load frequency con-
00242. trol in power systems. Trans Inst Meas Control.
[23] Blažič S, Zdešar A. An implementation of an evolving 2019;41(9):2563–2581. doi:10.1177/0142331218804
fuzzy controller. 2017 evolving and Adaptive Intelligent 309.
Systems (EAIS); 2017, pp. 1–7, doi:10.1109/EAIS.2017. [40] Visioli A. Comparison of fuzzy logic based tuning
7954830. methods for PID controllers. European Control Con-
[24] Gao Y, Xiao F, Liu J, et al. Distributed soft fault detec- ference, ECC 1999 - Conference Proceedings; 1999, pp.
tion for interval type-2 fuzzy-model-based stochastic 497–502, doi:10.23919/ecc.1999.7099353.
systems with wireless sensor networks. IEEE Trans Ind
Informatics. 2019;15(1):334–347. doi:10.1109/TII.20
18.2812771.

You might also like