Anda di halaman 1dari 7

Tuning of PID Controllers for First Order Plus

Time Delay Unstable Systems


Saxena Nikita and M. Chidambaram*
Department of Chemical Engineering, Indian Institute of Technology Madras, Chennai 600036
e-mail:chidam@iitm.ac.in

Abstract- The focus of the current study is to propose an improved Zeigler-


Nichols method for tuning the PID controllers for unstable First Order plus Time
Delay (FOPTD) systems. In the proposed method, the controller settings are
obtained by solving the magnitude and phase angle criteria. The addition of the
controller modifies the overall gain of the system, subsequently changing the
values of the controller gain. A second iteration for calculating the updated
controller gain is conducted by solving the system's magnitude and the phase
margin criteria, incorporating the Proportional-Derivative-Integral controller
transfer function, with unity proportional gain, and with the values of reset time
and derivative time. A number of FOPTD systems with varying time delay to
time constant ratios are simulated by using the proposed method. A non-linear
model of a bioreactor is also simulated to show the enhanced performance over
the other methods present in literature. The proposed method significantly
improves the performances for the servo and regulatory problems.

Keywords: continuous cycling, Ziegler-Nichols method, unstable systems, PID

1.1. Introduction
Over past 50 years, the Proportional-Derivative-Integral (PID) controller
remains a central element in the process industries. The Proportional (P)
mode adjusts controller output according to the error size, the Integral (I)
mode eliminate offset at the steady state whereas the Derivative (D) mode
increases the speed of the response. Due to these useful functions
alongwith its lack of complexity, efficient online tuning and superior
performance approximately 97% of process industries use PID controllers.
Efficient rules and automatic methods for tuning the PID controllers are
desired. Many methods, such as gain margin/phase margin method, pole
placement technique, optimization technique (minimizing
IAE/ISE/ITAE), direct synthesis method, internal model control method,
equating coefficient method and robust loop shaping have been reviewed
by Vilanova and Visioli (2012); Yu (1999). One of the methods of tuning
the PID controllers based on stability analysis is the Ziegler-Nichols
method (1942). In the method, the system is brought to the condition of
marginal instability. At the point, the system exhibits a sustained
oscillation commonly known as limit cycles in the response. The ultimate
gain and the ultimate frequency (referred as critical points) determines the
PID controller settings. It is a analytical method of designing the PID
controller. The method is also automated by using relay auto-tuning
method which is a single step method. The PID tuning rules proposed
were based on the simulation of a large number of processes. As the key
criterion is a quarter decay ratio, a large overshoot and an oscillatory
response is obtained. Many researchers proposed methods based on the
critical points and proposed different setting to improve the method.
Astrom and Hagglund (1988, 1995, 2006), Tyreus-Luyben (1992), Yu
(1999) and Smith (2003) proposed tuning formulae to improve the ZN
method. Many such controller tuning rules are reviewed by O'Dwyer
(2009).

But the scope of tuning PID controllers for unstable system based on
critical points is restricted. The following study includes description on
the general controller design procedure alongwith the performance and
therobustness measures in section 1.2. Section 1.3 discusses the
fundamentals of improved method for determining the PID settings for
unstable FOPTD systems for varying time delay to time constant ratios,
followed by section 1.4, where the proposed method is applied on first
order non-linear bioreactor model and is compared with the different
reported methods. Section 1.5 summarises the study.

1.2. Controller design method


In this paper, the single loop feedback controller structure is used, where
Gp is the process transfer function, Gc is the PID controller transfer
function of the form given by
𝐾𝐼
𝐺𝑐 (𝑠) = 𝐾𝑐 + + 𝐾𝐷 𝑠 (1)
𝑠
𝑘𝑐
where, 𝐾I = 𝜏 𝐼,
; 𝐾𝐷 = 𝑘𝑐 ∗ 𝜏𝐷
The proposed method is implemented on unstable processes. It should be
recalled in the context of unstable system that they have a peculiar
propoerty of exhibiting the maximum value of controller gain (𝐾𝑐,𝑚𝑎𝑥 ) as
well as the minimum value of controller gain (𝐾𝑐,𝑚𝑖𝑛 ).
The tuning rule proposed by Ziegler and Nichols (1942) for the PID
controllers are based on the systems critical values. In the present work,
simple equations for calculating PID settings for unstable FOPTD are
proposed based on the reported work of Nikita and Chidambaram (2016).
The amplitude ratio (Ar) and phase angle (𝜑) for the FOPTD system are
realized by solving the magnitude criterion and the phase angle criterion.
The controller gain is set equal to 1/Ar so as to make the gain unity of the
system at π radian phase angle . At crossover frequency, the closed loop
system will attain the condition of marginal stability and will give limit
cycle with frequency equal to 𝜔𝑚𝑎𝑥 . To eliminate offset and to expedite
the response, the integral mode and the derivative mode is added to the
controller.
It is observed that once the integral and derivative modes become active,
the controller gain is affected. For stable systems based on the bode plots,
Douglas (1972) proposed a trial and error procedure to ascertain the
updated values of controller constants. The integral and derivative times
are considered respectively as 5/ωc and 1/ωc. Enhanced results are
obtained when the controller is tuned around these values.
To compare the performance of the system based on proposed method
over the other methods present in literature, Integral of the Time weighted
Absolute Error (ITAE), maximum sensitivity as defined by Vilanova and
Visioli (2012), gain margin and phase margin in controller output values
are considered for unit step change in the input and in the disturbance.
1.3. Proposed method
The main objective of the control system design is the fast set point
tracking while also ensuring the system is insensitive to the uncertainties
in the model.The FOPTD system is given as:
K p exp −𝜀𝑠 (2)
𝐺𝑝 𝑠 =
𝑠−1
where, ε is the time delay to time constant ratio and Kp =1. The
corresponding amplitude ratio and the phase angle are written as:
1 (3)
𝐴𝑟 = ; 𝜑 = −𝜀𝑠 − 𝜋 + tan−1 (𝜔)
𝜔2 + 1
At the crossover frequency, phase angle is equal to -180. Hence, eq. (3)
can be solved to obtain the minimum frequency and the maximum
frequency 𝜔c,min , 𝜔c,max , for different values of ε. On substituting the
𝜔c,min , 𝜔c,max values in eq. (3) , the minimum and the maximum values
of amplitude ratio is calculated, and corresponding controller gain can be
1
determined as by using the relation 𝐾𝑐 = 𝐴 .
𝑟
The values of reset time (𝜏𝐼 ) and derivative time 𝜏𝐷 constant are
2𝜋 𝑃 𝑃
calculated using 𝜔c,max as: 𝑃𝑢,𝑚𝑎𝑥 = ; 𝜏𝐼 = 𝑢 ,𝑚𝑎𝑥 ; 𝜏𝐷 = 𝑢 ,𝑚𝑎𝑥
𝜔 𝑐,𝑚𝑎𝑥 2 8
The controller is designed based on the Zeigler-Nichols method. For the
proposed method, 𝜏𝐼 and τD constant are calculated using the equation:
5 0.8
𝜏𝐼 ′ = 𝜔 ; 𝜏𝐷 ′ = 𝜔 (4)
𝑐,𝑚𝑎𝑥 𝑐,𝑚𝑎𝑥
The 𝜏𝐼 ′ and τD' values obtained in eq. (4) are corresponding to the series
PID controller. The corresponding values for the parallel PID form (Yu,
1999) are computed by:
𝜏 ′∗ 𝜏 ′
𝜏𝐼𝑝 = 𝜏𝐼 ′ + 𝜏𝐷 ′; 𝜏𝐷 𝑝 = 𝐼 𝐷 (5)
𝜏 𝐼 ′ +𝜏 𝐷 ′
Once controller is designed and integral plus derivative action becomes
effective, the value of Kc changes. It needs to be updated again. For that
purpose, the transfer function of PID controller is added to the FOPTD
system as:
 1  exp( s) 
𝐺𝑝 𝑠 = 1    D p s   (6)
  I ps  s 1 
 
The amplitude ratio and phase angle of the new system are written:
 1 
  tan 1  D        tan 1 () (7)
  I  
2
 1 
 D   1
  I   (8)
𝐴𝑟 =
 2 1
At the crossover frequency, eq. (7) is solved for the updated ωc,min and
ωc,max corresponding to which the updated Kc,min and Kc,max for different
values of ε are obtained (Table 1.1). The Kc,des is taken to be the average
of Kc,min and Kc,max. The following correlations are proposed for the
unstable FOPTD system:
𝐾𝑐 𝐾𝑝 = 1.045𝜀 −0.96 when 0.01 < 𝜀 < 0.1 (9a)
𝐾𝑐 𝐾𝑝 = 1.041𝜀 −0.94 when 0.1 < 𝜀 < 0.8 (9b)
𝜏𝐼
𝜏
= 8.059𝜀 3 − 4.511𝜀 2 + 3.620𝜀 − 0.036 (9c)
𝜏𝐷
= 1.414𝜀 3 − 0.791𝜀 2 + 0.635𝜀 − 0.006 (9d)
𝜏
The controller performance is evaluated for three different value of ε
(0.05, 0.5 and 0.8) given in Table 1.1 and the performance is evaluated in
the closed loop system for a unit step change in the set point and unit
change in load as shown in Fig 1.1. The proposed method gives enhanced
results (Table 1.2). The maximum sensitivity (MS), gain margin (Gm) and
phase margin (φm) for different values of ε are given in Table 1.3. It can
be clearly seen that the proposed method is more robust than the ZN
method indicating the closed loop system is less sensitive to the variations
in the process dynamics.

Figure 1.1: Process response for step change(i) ε = 0.05; (ii) ε = 0.5; (iii) ε = 0.8
(Dotted –ZN method; Solid – proposed method)

Table 1.1: Controller parameters for FOPTD system for different values of ε
ZN Method Proposed Method
ε Kc,des τI τD Kc,des τI τD
0.05 15.8911 0.1021 0.0255 22.6045 0.1850 0.0224
0.5 1.6183 1.3477 0.3369 2.0562 2.4882 0.2959
0.8 1.8935 3.3100 0.8275 1.2167 6.1111 0.7266
Table 1.2: ITAE value for different cases
SERVO RESPONSE REGULATORY
RESPONSE
METHOD ZN Proposed Imp* ZN Proposed Imp*
ε = 0.05 0.0493 0.0187 61% 0.0031 0.0015 52%
ε = 0.5 158.2 8.035 95% 85.92 4.487 94%
ε = 0.8 2202 440.3 80% 1656 358 78%
*Imp denontes percentage reduction in the values
Table 1.3: Maximum sensitivity, gain margin & phase margin for different values of ε
ε = 0.05 ε = 0.5 ε = 0.8
METHOD ZN Proposed ZN Proposed ZN Proposed
MS 1.984 3.017 10.267 2.173 13.052 8.671
Gm 2.106 1.495 1.1079 1.851 1.083 1.130
Φm 60.532 38.713 11.178 54.775 8.788 13.244
1.4. CASE STUDY: UNSTABLE NON-LINEAR BIOREACTOR
The proposed method is applied to a Non-Linear continuous bioreactor
that exhibits output multiplicity. The model given by Agarwal and Lim
(1984) is considered. The system is assumed to be at unstable steady state
condition. At t = 0 , X = 0.9951; S = 1.5122 and D = 0.3 l/hr. The method
is applied to the local linearised model with the parameters: Kp = -0.5898,
τ= 5.6125 and θ = 1.0. The controller parameters are calculated for the
Non-Linear system using the same method as discussed in section 3. The
method is compared with the methods present in literature. The controller
parameters obtained for Majhi and Atherton method (2000); Thyagarajan
and Yu method (2003) and Tyreus-Luyben method (1992) along with the
proposed method are shown in Table 1.4. Appreciable reduction in ITAE
values in servo response is obtained (step change in X from 0.9951 to
1.294). Also when disturbance is added to the substrate feed (± 10%), the
proposed method gives enhanced performance as seen in Table 1.4.
Table 1.4: Controller parameter for different methods & corresponding ITAE values
Parameters Proposed Majhi and Thyagarajan Tyreus and
method Atherton and Yu Luyben
method method method
Kc -0.92915 -1.1855 -0.9439 -0.8157
τI 3.9874 4.7255 5.5363 9.5031
τD 0.4741 0.4377 0.5235 0.6857
ITAE (servo) 3.271 13.16 4.522 9.462
ITAE 0.7871 0.8173 1.284 3.315
(Sf decreased)
ITAE 0.7289 0.7515 1.185 3.002
(Sf Increased)

Although the proposed method presented here is based on the transfer


function model, it can be automated using relay auto-tuning method where
identification of transfer function model is not neccessary for controller
design. The work can be further extended for application in industrial
auto-tuners.
1.5. CONCLUSIONS
An improved ZN method is proposed for tuning the PID controllers. The
simulation results of unstable FOPTD model with different ε values are
given. The correlation for the unstable FOPTD system based on the
simulaiton studies are proposed. The method is implemented to a Non-
Linear bio reactor model and compared with three other methods present
in literature. Significant improvement in performances along with the
robust behaviour is seen for the proposed method.

References
Agarwal, P. And H.C. Lim: Analysis of various control schemesfor continuous
bioreactors,Advances in Biochemical Eng./Biotechnology, 30,61-90, (1984)
Astrom,K.J. and Hagglund, T.: Automatic Tuning of PID controllers. Instument
Society of America, North Carolina, (1988)..
Astrom,K.J. and Hagglund, T.: PID controllers: Theory, Designand tuning, 2 nd
Edition, Instument Society ofAmerica, North Carolina, (1995).
Astrom,K.J. and Hagglund, T.: Advanced PID control, Instument Society of
America, North Carolina, (2006)
Douglas, J.M.: Process dynamics and control, 2 Control system synthesis. NJ:
Prentice Hall, Englewood Cliffs, (1972)
Majhi,S. and Atherton, D.P.: Online tuning of controllers for an unstable FOPTD
process, IEEE Proc. CTA, 146,415-25, (2000).
O'Dwyer: A. Handbook of PI/PID controller Tuning Rules, 3rd Ed.: Imperial
College Press, (2009)
Smith, C.L.: Intelligently tune PID controllers, Chemical Engineering, January,
pp. 56-62, (2003)
Nikita, S. And M. Chidambaram: Improved continuous cycling method of tuning
PID controllers for unstable systems, Indian Chemical Engineer, (2016) in
press
Thyagarajan, T. and Yu, C.C.: Improved auto-tuning using shape factorfrom relay
feedback, Ind, Eng. Chem.Res., 42,4425-440, (2003).
Tyreus, B.D. and Luyben,W.L.: Tuning PI Controllers for Integrator / Dead-Time
Processes. Ind.Eng.Chem. Res., ,31,2625, (1992).
Vilanova, R. and Visioli: A. PID Control in Third Millennium,. London: Springer
verlag Ltd., London, (2012)
Yu, C.C.: Auto Tuning of PID Controllers. Berlin: Springer-Verlag, (1999).
Ziegler, J.G. and Nichols, N.B.: Optimum Settings for automatic Controllers.
Trans. ASME, ,64, 759-765, (1942)