Professional Documents
Culture Documents
net/publication/343812958
CITATIONS READS
0 288
1 author:
Sanjeev Pandey
KPIT
27 PUBLICATIONS 112 CITATIONS
SEE PROFILE
All content following this page was uploaded by Sanjeev Pandey on 19 September 2020.
Abstract—This paper examines the implementation issue of parameters and a fast change in proportional gain may skip the
PID auto-tuning as applied for temperature control. An exhaus- sustained oscillation and plant may go to an unstable operating
tive review of various tuning methods, PID control, anti windup region. Moreover, this method drives plant to a marginal stable
and derivative filters are first reported. A practical PID auto-
tuning algorithm is developed for first order plus dead time region, which is not good from a safety point of view. These
system. First, PID parameters are calculated based on ON/OFF issues are valid for Ziegler-Nichols, Modified Ziegler-Nichols
control. Once these parameters are obtained, the control mode and Tyreus Luyben tuning methods. In a Damped oscillation
changes from ON/OFF to PID control. The developed algorithm method, proportional gain is increased until the decay ratio
only requires plant output to auto-tune PID parameters. The of 1/4th is achieved. The gain at which the decay ratio is
auto tuning is developed in MATLAB/Simulink and verified for
various first order plus dead time systems. obtained, is known as ultimate gain Kd and the time period
Index Terms—Temperature control, FOPDT system, tuning is known as oscillation time period Pd . The advantage of this
methods, auto-tuning, anti windup, derivative filter, PID control, method over previously discussed methods is that it does not
set point tracking, MATLAB/Simulink. drive plant to be in a marginal stable region.
It can be observed from Table II that all the open loop
I. I NTRODUCTION based tuning methods must know the plant parameters to tune
PID control is most common control algorithm in industries. PID parameters. The recently developed internal model control
It is used in process industries [1], power converters control (IMC) [10] is robust but it requires plant parameters and design
[2], [3], motor control [4] and electro-hydraulic system control of filter λ. The guideline to design filter is given in [8], [10].
[5]. Based on a survey, for regulatory control, 97% of process Similarly, minimum error criteria based tuning method also
industries use PID algorithm [1]. It is widely used in industries requires knowledge of plant parameters. Therefore, to develop
due to its simplicity and robustness. The proportional, integral an auto-tuning algorithm, plant parameters must be identified
and derivative are three components of PID control and it is first. As it is well known that the plant gain can be identified
not easy to find the ideal value of these parameters without by applying a step input and observing the steady state output.
knowing about the plant. In the last six decades, various open Similarly, the time constant can be identified by measuring the
and closed loop based tuning methods are proposed. All the time taken by the plant to reach the steady state output. The
tuning methods suggest how to find the optimal value of dynamics of temperature system is slow and it becomes slower
PID parameters. A review of tuning methods is presented in when it reaches close to the steady state value. Therefore,
[6]–[8]. The objective of all the tuning methods is to find auto-tuning based on the open loop method is a slow process.
out PID parameters such that the plant has less overshoot, However, open loop based tuning methods have one advantage,
settling time and robust against disturbances. A summary of it does not drive the system to a marginal stable region.
closed and open loop based tuning methods is given in Table The auto-tuning PID is a desirable feature in process control
I and Table II, respectively. There are a few more tuning applications. The first paper on auto-tuning algorithm is pro-
methods which are not included in the tables. Out of various posed in [11] in which the plant parameters are first estimated
tuning methods, the closed loop Ziegler-Nichols [9], Damped and then PID parameters are auto-tuned. In [12], Newton-
oscillation method and open loop Ziegler-Nichols and Cohen Raphson based auto-tuning is proposed to control flow rate.
Coon are very famous in industries. However, development of In [13], adaptive based auto-tuning is proposed to control first
the auto-tuning algorithm using these methods is difficult. order plus dead time (FOPDT) system. In [14], a robust auto-
To implement closed loop Ziegler-Nichols method, the PID tuning PID is proposed to control the level of a fluid tank
control is set in proportional mode by making integral and system.
derivative control equal to zero. Now, the proportional gain is In this paper, a simple PID auto-tuning algorithm is de-
gradually increased until the plant reaches to sustain oscilla- veloped for temperature control. The control is designed for
tion. The gain at which plant reaches to sustain oscillation, is following transient and steady state requirements:
known as ultimate gain Ku and the time period of sustained 1) In auto-tuning mode, the plant should not be driven
oscillation is known as ultimate time period Pu . The PID towards damped or sustained oscillation.
parameters are tuned by these two values as shown in Table 2) The undershoot should be less than 2% when control
I. The issues with this method is that there is no guideline changes from auto-tuning mode to PID mode.
on how to increase the proportional gain. A slow increase 3) The steady state error should be less than 1% of the
in proportional gain will take more time to auto-tune PID desired temperature.
TABLE I control is implemented along with a low pass filter (LPF). In
C OMPARISON OF CLOSED LOOP PID TUNING FORMULAS literature, two types of anti windup scheme is used, namely,
Tuning methods Kp Ti Td (i) conditional integration and (ii) tracking anti windup. In
Ziegler-Nichols method 0.6 Ku Pu Pu conditional integration when u is not in the range of usat then
2 8
Modified Ziegler-Nichols method 0.33 Ku Pu
2
Pu
8
the integration process is stopped. In tracking anti windup,
Ku Pu integral control is given by
Tyreus – Luyben method 3.2
2.2Pu 6.3
Pd Pd
Damped oscillation method 1.1Kd 3.6 9 Zt
Kp
ui = + Kaw (usat − u) e(τ )dτ (2)
Ti
4) The overshoot for set point change should be less than 0
10%.
where
5) The PID parameters should not be required to be re-
tuned if set point changes. u for umin < u < umax
usat = umax for u ≥ umax
The paper is divided into six parts. In Section II, the system
umin for u ≤ umin
under study and control objectives are stated. In Section III, a
review of PID control is discussed. The proposed auto-tuning umin and umax are constants, explained in detail in Section
algorithm is introduced in Section IV. Simulation results are IV. In [15], [16], various anti windup schemes are discussed.
provided for three different FOPDT systems in Section V. The In this paper, tracking anti wind up is used.
paper is concluded in Section VI. The conventional PID control or Type A is defined as
II. M ATHEMATICAL M ODEL OF FOPDT S YSTEM Zt
The transfer function of a heating or cooling system is a Kp
u = Kp e(t) + + Kaw (usat − u) e(τ )dτ
FOPDT. In a most common form, transfer function of FOPDT Ti
0
system can be defined as
de(t)
K + Kp Td (LP F ) (3)
Gp (s) = e−θs (1) dt
τs + 1
where Kaw is anti windup gain. If generated control u
where K is system gain, τ is time constant and θ is time saturates, then rate of convergence of u to usat is decided by
delay. For a heating system K > 0 and for a cooling system Kaw . It may be noted from (3) that when umin ≤ u ≤ umax ,
K < 0. It may be noted here that for the developed algorithm, anti wind up term Kaw (usat − u) = 0 i.e. (3) becomes
the value of these parameters is unknown. Broadly, the control a conventional
objective can be divided into two parts (i) auto-tuning of PID √ PID control. In [1], recommended value of
Kaw = 1/ Ti Td for PID control and Kaw = 1/Ti for PI
parameters and (ii) regulation and tracking of the desired set control.
point. Two types of LPF is used for derivative control (i) Type 1:
III. PID C ONTROL N1 /(N1 + Td s) and (ii) Type 2: N2 /(N2 + s). where N1 > 0
and N2 > 0 are LPF time constants. Generally, they are in
In this section, various forms and types of PID are reviewed
the range of 2 ≤ N1 ≤ 20 and 0 < N2 ≤ 1. It may be
and commonly used PID are industries is presented. There are
noted here that one of the LPF has process dependent time
three forms of PID control, namely, (i) standard or ideal form
constant Td . In this paper, Type 2 LPF is implemented. The
(ii) parallel form and (iii) classical form. The control formulas
major disadvantage of Type A controller is that when the set
are: point changes in a step fashion, proportional and derivative
1) Standard or ideal form u = Kp 1 + T1i s + Td s e(s) control makes (3) behave violently [17] therefore, these two
controllers are modified as
2) Parallel form u = Kp + Ksi + Kd s e(s)
Zt
Kp
3) Classical form u = Kp 1 + T1i s 1 + Td s e(s) u = Kp e(t) + + Kaw (usat − u) e(τ )dτ
Ti
0
where Kp is proportional gain, Ti is integral time, Td is dy(t)
derivative time, Ki (= Kp /Ti ) is integral gain, Kd (= Kp Td ) − Kp Td (LP F ) (4)
dt
is derivative gain, u is generated PID control and error
e(s) is difference between set point and output of the plant. and
Advantages and disadvantages of above forms are discussed
Zt
extensively in literature therefore, they are not discussed here.
Kp
In this paper, standard form of PID is used. It may be noted u = −Kp y(t) + + Kaw (usat − u) e(τ )dτ
Ti
from above PID forms that anti windup and derivative filter 0
are not used in the control. In industrial application, integral dy(t)
− Kp Td (LP F ) (5)
control is implemented with anti windup scheme and derivative dt
TABLE II
C OMPARISON OF OPEN LOOP PID TUNING FORMULAS
Tuning methods Kp Ti Td
1.2τ
Open loop Ziegler-Nichols method Kθ
2θ 0.5θ
1.2τ
C−H−R method Kθ
2θ 0.42θ
32+ 6θ
τ
θ
τ 4 θ 4θ
Cohen and Coon method + 4τ
Kθ 3 13+ 8θ
τ
11+ 2θ
τ
2τ +θ
IMC 2K(λ+θ)
τ + θ2 λθ
2τ +θ
1
(b) Derivative control with LPF (inside PID control block of Fig. 2d)
0.8
ON/OFF control
0.6
0.4
0.2
0
0 50 100
time (sec)
(c) Inside ON/OFF control of Fig. 2d
(b) Output of ON/OFF control
Fig. 3. Plant response during auto-tuning.
1
1
0.8
ON/OFF control
0.8
ON/OFF control
0.6
0.6
0.4
0.4
0.2
0.2
0
0 100 200 300 400
0
time (sec) 0 500 1000 1500
(b) Output of ON/OFF control time (sec)
(b) Output of ON/OFF control
Fig. 5. Response under developed auto-tuning PID control for system (7).
2
PID is auto-tuned at 0.3 ◦ C and then the set point changed to 0.5 ◦ C.
1 TABLE III
control
some different value of Kaw . new set point without any steady state error. A summary of
6) The set point is changed to 60 ◦ C at 220 sec and auto- auto-tuned PID parameters for all the three systems is given
tuned PID control can drive the system to the new set in Table III.
point. It can be concluded that PID auto-tuned at a set
point, need not to be re tune when the set point changes. VI. C ONCLUSION
In Fig. 5 and Fig. 6, the performance of the developed In this paper, an auto-tuning algorithm is developed for PID
algorithm for (7) and (8) are shown respectively. In both the control parameters. The proposed algorithm is very simple
cases, the system is first auto-tuned for a set point and then to implement and applicable to heating and cooling process.
the set point is changed to a new value. It can be observed The PID control is implemented with tracking anti windup
that the developed algorithm is able to drive the system to the and derivative filter such that when control changes from auto
[4] V. Kumar, P. Gaur, and A. Mittal, “ANN based self tuned PID like
adaptive controller design for high performance PMSM position con-
trol,” Expert Systems with Applications, vol. 41, no. 17, pp. 7995–8002,
2014.
[5] K. Sinthipsomboon, I. Hunsacharoonroj, J. Khedari, W. Pongaen, and
P. Pratumsuwan, “A hybrid of fuzzy and fuzzy self-tuning pid controller
for servo electro-hydraulic system,” in 2011 6th IEEE Conference on
Industrial Electronics and Applications, pp. 220–225, IEEE, 2011.
[6] S. Skogestad, “Probably the best simple PID tuning rules in the world,”
in AIChE Annual Meeting, Reno, Nevada, vol. 77, 2001.
[7] M. W. Foley, R. H. Julien, and B. R. Copeland, “A comparison of
PID controller tuning methods,” The Canadian Journal of Chemical
Engineering, vol. 83, no. 4, pp. 712–722, 2005.
[8] W. Tan, J. Liu, T. Chen, and H. J. Marquez, “Comparison of some
well-known PID tuning formulas,” Computers & chemical engineering,
vol. 30, no. 9, pp. 1416–1423, 2006.
[9] J. G. Ziegler and N. B. Nichols, “Optimum settings for automatic
(a) Plant output (black) and set point (red) controllers,” trans. ASME, vol. 64, no. 11, 1942.
[10] C. E. Garcia, D. M. Prett, and M. Morari, “Model predictive control:
theory and practice-a survey,” Automatica, vol. 25, no. 3, pp. 335–348,
1 1989.
[11] K. J. Åström and T. Hägglund, “Automatic tuning of simple regula-
tors with specifications on phase and amplitude margins,” Automatica,
0.8
ON/OFF control