In the work, a PID design method with desired settling time is proposed for plants with
transfer functions of the following types: first-order system, second-order system and integral
first-order system. The method provides better performances than both the magnitude optimum
method and the symmetric optimum method in the sense of overshoot and settling time. In
addition, the settling time can be achieved as expected for a larger frequency band. Some
examples for DC motor control are used to illustrate the method.
Future works will focus on robustness, effects of model error and other types of transfer
functions, may be higher-order system with time delay
11 trang |
Chia sẻ: honghp95 | Lượt xem: 527 | Lượt tải: 0
Bạn đang xem nội dung tài liệu Settling time assignment with pid - Nguyen Doan Phuoc, để tải tài liệu về máy bạn click vào nút DOWNLOAD ở trên
Vietnam Journal of Science and Technology 55 (3) (2017) 357-367
DOI: 10.15625/2525-2518/55/3/8808
SETTLING TIME ASSIGNMENT WITH PID
Nguyen Doan Phuoc, Nguyen Hoai Nam
Hanoi University of Science and Technology, 1 Dai Co Viet Street, Hanoi
*Email: nam.nguyenhoai@hust.edu.vn
Received: 24 October 2016; Accepted for publication: 21 February 2017
ABSTRACT
In this paper, the settling time of closed-loop systems controlled with PID, which is
established by using magnitude and symmetric optimum methods, is analyzed. These methods
have some limitations such as limited frequency bands, existence of overshoot and larger settling
time. Thus, our aim is to improve these performances of the closedloop system for a certain class
of transfer functions. The proposed method will be tested on a DC motor by simulation. It can be
concluded that the method provides a larger range of frequency, smaller settling time and no
overshoot.
Keywords: PID controller, magnitude optimum criterion, symmetric optimum criterion, first-
order system, first-order plus integral system.
1. INTRODUCTION
Many PID tuning methods have been developed since 1940s. These methods can be
classified into groups: practical approach [1 - 4]; optimal approach [5 - 8]; gain and phase
margin specification based approach [9 - 13] and others [14 - 18]. However, there have not been
yet any method that concurrently dealing with both the settling time and the overshoot of the
closed-loop system.
It is known that when plants, having transfer functions of first-order, second-order and
third-order systems the magnitude optimum method based parameter determination , p Ik T and
DT [5 - 8] for the PID controller seems to be one of the most effective methods, validated in
reality. However, there has been no fundament until now to conclude that the magnitude
optimum method is the only one to provide PID controller parameters to make the closed-loop
system stable and the settling time short in the sense that the closed-loop system with transfer
function
( ) ( )( )
1 ( ) ( )= +c
R s G s
G s
R s G s
has
Nguyen Doan Phuoc, Nguyen Hoai Nam
358
{ } ( ) 1R+Ω = ∈ ≈cG jω ω (1)
in small range of frequency, which is large enough.
A method, proposed in this work, has an approach that the closed-loop system, obtained as
follows,
/
1( )
1
=
+
cG s
T s
(2)
has as small /T as possible (when /T is smaller, Ω is larger), which will demonstrate the
conclusion.
2. PID CONTROLLER DESIGN
2.1. Control of First-Order Systems
Firstly, let’s assume that the plant has a form of first-order system and a PI controller is
used. Then, the transfer function of the plant is represented as follows
( )
1
=
+
k
G s
Ts
and the PI controller is
( )11( ) 1 + = + =
p I
p
I I
k T s
R s k
T s T s
Thus, the open-loop transfer function is
( )
( )
1( ) ( ) ( )
1
+
= =
+
p I
o
I
k k T s
G s R s G s
T s Ts
(3)
and the closed-loop transfer function is
2
( ) 1( )
1 ( ) 11 1
+
= =
+
+ + +
o I
c
o I
I
p p
G s T s
G s
G s TT
s T s
k k k k
(4)
Since the first-order system (2) with unit gain is always stable, the steady state error is zero and
the settling time is smaller if the time constant is shorter, in this case the two parameters pk and
IT are determined such that the closed-loop transfer function (4) has the expected form as (2).
In order to do that, pk and IT are chosen to cancel out a zero by a pole. The zero is 1= − Is T .
This is equivalent to
2 11 1 0
+ + + =
I
I
p p
TT
s T s
k k k k
as
1
= −
I
s
T
or
( )0 1= − + +I p I pT T k k T k k
Settling time assignment with PID
359
Thus .=IT T
Next, the closed-loop system becomes
/
2
1 1( )
111 1
+
= =
+
+ + +
I
c
I
I
p p
T s
G s
T sTT
s T s
k k k k
with the selected parameters as above, if the time constant /T satisfies the following equation:
( )( ) ( )( )
( )
/ / 2
2
2
11 1 1 1 1 1
1
1
+ + = + + = + + +
+
= + +
I
I I
p p
p
p p
TT
T s T s T s Ts s T s
k k k k
T k kT
s s
k k k k
⇔ ( ) ( )2/ / 2 2 11 1++ + + = + +p
p p
T k kT
T T s T Ts s s
k k k k
⇔
( )/ 1+
= − =
p
p p
T k k T
T T
k k k k
This implies
( )2
1 1( )
1 1 1
+
= =
+ + + +
I
c
p I I p
p
T s
G s
Tk k TTs T k k s s
k k
Hence, in the case of =IT T the bigger pk is chosen, the smaller settling time of closed-
loop system will be. Furthermore, the closed-loop system is stable and has no overshoot.
Obviously, in comparison to the performance of the magnitude optimum criterion method,
the proposed PI controller provides a more flexible band of frequency (Ω). When the parameter
pk is bigger, the frequency range Ω becomes wider:
lim ( ) 1,
→∞
= ∀
p
c
k
G s s
Therefore, the desired settling time 5%T of closed-loop system can be assigned by choosing
5%
ln 20
=p
T
k
kT
which is clearly obtained from its unit step response
/( ) 1 −= − t Th t e or /5%0.95 1 −= − T Te
2.2. Control of Second-Order Systems
When the plant has the form of a second-order system, we use a PID controller. The plant model
Nguyen Doan Phuoc, Nguyen Hoai Nam
360
and the controller are represented, respectively, as follows
− Plant: ( )( )1 2( ) 1 1= + +
k
G s
Ts T s
− PID controller:
( )( )1 11( ) 1 + + = + + =
p A B
p D
I I
k T s T s
R s k T s
T s T s
where + =A B IT T T and =A B I DT T TT , it means that whenever AT and BT are known we can
obtain IT and DT . The open-loop system is
( )( )
( )( )
( ) ( )
( ) ( )
/
1 2 1 2
1 1 1 1( ) ( ) ( )
1 1 1 1
+ + + +
= = =
+ + + +
p A B p A B
o
I A
k k T s T s k k T s T s
G s R s G s
T s Ts T s T s Ts T s
(5)
where
/
=
p A
p
I
k T
k
T
Thus, if it is chosen that 2=BT T then the transfer function (5) has the same form as (3) shown
in the first case
( )
( )
/
1
1( )
1
+
=
+
p A
o
A
k k T s
G s
T s Ts
where the parameters IT and pk are replaced by the parameters AT and
/
pk , respectively. So,
with the set of parameters in the case 1: 1=AT T and
/
pk arbitrary, the closed-loop transfer
function becomes a first-order system with unity gain and the time constant
/ 1 1
/= = =
I I
p A pp
T TT T
T
kk T kkk k
and therefore, the bigger pk is chosen, the smaller its time constant
/T will be. Thus, we have
1 2
1 2
1 2
and = + = + = =
+
A B
I A B D
I
T T TT
T T T T T T
T T T
In summary, the larger pk is optionally selected the smaller the settling time associated with
/T
will be. The assignment of desired settling time 5%T for the closed-loop system can be easily
realized with
( )1 2
5%
ln 20+
=p
T T
k
kT
In addition, there is no overshoot in the closed-loop system.
2.3. Extended Method to Integral First-Order Systems
Settling time assignment with PID
361
For the basic control theory of linear systems, it is well known that the symmetric optimum
method is applied to find the parameters , p Ik T and DT of the PID controller
1( ) 1 = + +
p D
I
R s k T s
T s
(6)
for integral first-order systems and integral second-order systems, in order to make the closed-
loop system stable and keep the phase margin larger. However, the disadvantage of the method is
big overshoot. Although the overshoot can be decreased with the usage of an input filter, this
causes an additional cost.
The proposed method for determination of PID parameters later will guarantee that the
closed-loop system has no overshoot and arbitrary small settling time. It is similar to first-order
or second-order systems, the idea of this approach is also to find parameters , p Ik T and DT of
the PID controller for the integral first-order plant
( )( ) 1= +
k
G s
s Ts
(7)
such that the closed-loop system has the following form:
( )( )1 2
1( )
1 1
+
=
+ +
t
c
m m
Ts
G s
T s T s
(8)
Consider a closed-loop system, which includes an integral first-order system (7) and a PID
controller (6)
( )( )1 11( ) 1 + + = + + =
p A B
p D
I I
k T s T s
R s k T s
T s T s
with + =A B IT T T and =A B I DT T TT . The open-loop system is as follows
( )( )
( )2
1 1( ) ( ) ( )
1
+ +
= =
+
p A B
o
I
kk T s T s
G s R s G s
T s Ts
Thus, if =BT T then = +I AT T T and the closed-loop system is
( )
( ) ( )2
1( )( )
1 ( ) 1
+
= =
+ + + +
p Ah
c
h A p B
kk T sG s
G s
G s T T s kk T s
So, the closed-loop system will be the same as the expected form (8) if there the following
equation must be satisfied
( )
( ) ( ) ( )( )2 1 2
1 1
,
1 11
+ +
= ∀
+ ++ + +
p A t
m mA p A
kk T s Ts
s
T s T sT T s kk T s
⇔ ( ) ( )( ) ( ) ( ) ( )21 21 1 1 1 1 , + + + = + + + + ∀ p A m m A p A tkk T s T s T s T T s kk T s Ts s
Nguyen Doan Phuoc, Nguyen Hoai Nam
362
Controlled
Object PID
Input
filter
Figure 1. Control with PID and input filter.
⇔ ( )
( )
1 2 1 2
1 2
1 2
+
+ + = +
+ = + +
+
=
A
A m m m m A t
p
A t A m m
A t
m m A
p
T T
T T T T T T T
kk
T T T T T
T T T
T T T
kk
⇔
( )
1 2
1 2
1 2
+
=
= +
+
=
A
m m
p
t m m
A t
m m A
p
T T
T T
kk
T T T
T T T
T T T
kk
⇔
1 2
1 2
+
=
= = +
A
m m
p
A t m m
T T
T T
kk
T T T T
This implies that
( )
1 2
1 2 1 2
1 2
1 2
1 2
+ + +
= =
= + = + +
+
= =
+ +
A m m
p
m m m m
I A m m
m mA B
D
I m m
T T T T T
k
kT T kT T
T T T T T T
T T TT T
T
T T T T
(9)
It can be seen that with the above equations, 1 mT and 2mT are freely chosen. The expected
transfer function (8) becomes
( )
( )( )
1 2
1 2
1( )
1 1
+ +
=
+ +
m m
c
m m
T T s
G s
T s T s
Thus, with an addition of an input filter as illustrated in Fig.1:
( )1 2
1( )
1
=
+ +m m
M s
T T s
to obtain the closed-loop system:
( )( )1 2
1( ) ( )
1 1
=
+ +c m m
G s M s
T s T s
we always can choose 1mT and 2mT such that this closed-loop system has a settling time 5%T as
desired. Obviously, there are two possible combinations of time constants 1mT and 2.mT The
first case is 1 2= =m m mT T T and the second
one is 1 2≠m mT T .
Now some steps, about how to choose these
time constants such that the desired settling
time can be achieved, will be shown. Firstly,
Settling time assignment with PID
363
let’s consider the case one. With the usage of input filter as shown in Fig. 1, the closed-loop
transfer function is as follows
( )2
1( )
1
=
+
c
m
G s
T s
So, the unit step response is
/ /( ) 1 − −= − −m mt T t T
m
t
h t e e
T
Thus, the desired settling time 5%T will be the solution to the following equation
/ / 0.05 0− −+ − =m mt T t T
m
t
e e
T
Let 5%= mx T T . Then
( ) (1 ) 0.05 0−= + − =xf x e x (10)
As 0, (0) 0.95= =x f and when = ∞x it is obtained ( ) 0.05.∞ = −f In addition,
/ ( ) 0, 0.−= − xf x e x x That means there must be a unique solution to the equation (10).
The solution is approximately equal to 4.7439. Hence,
5%
1 2= 4.7439
= =m m m
T
T T T (11)
Then, the PID parameters are determined from Eq. (9).
For the second case, the closed-loop system with input filter is
( )( )1 2
1( )
1 1
=
+ +c m m
G s
T s T s
Without loss of generality, let 1 2m s mT k T= where 0 1.sk< < Doing the same procedure as the
first case, the following equation is obtained
1 2/ /
1 2
1 2
0.05 0
− −
−
− =
−
m mt T t T
m m
m m
T e T e
T T
(12)
Obviously, the expected settling time will be the solution to the equation (12). Let
5% 2= mx T T , then substitute it and 1mT into the equation (12) to have
/
( , ) 0.05 0
1
− −
−
= − =
−
sx k x
s
s
s
k e e
f x k
k
(13)
It can be easily verified that
(0, ) 0.95, ( , ) 0.05= ∞ = −s sf k f k and / ( , ) 0<sf x k .
This implies the equation (13) must have only one solution with respect to .sk The Tab. 1 shows
solutions for some values of sk .
Table 1. Solution with respect to sk .
sk 0.01 0.3 0.5 0.7 0.99
*x 3.005 3.352 3.6761 4.067 4.7202
Nguyen Doan Phuoc, Nguyen Hoai Nam
364
Thus,
5%
2 *=m
T
T
x
and 5%1 * .m s
T
T k
x
= (14)
Hence, given a desired settling time, first we have to choose sk , then time constants 1mT and
2mT are determined by using Eq. 14, and finally the PID parameters are calculated from Eq. 9. It
can be concluded that as 1→sk then
* 4.7439→x . Thus, the second case becomes the first case.
3. SIMULATION RESULTS
3.1. DC Motor Speed Control
Given a speed transfer function of a DC motor as follows [19]
( )
1
=
+
k
G s
Ts
where 100=k , 1.9568=T , the input is the current applied to the DC motor and the output is
the speed of the rotor. The I controller parameter is determined by using the magnitude optimum
method is 2 391.4= =IT kT , whereas the proposed method provides values for the PI
controller as 1.9568= =IT T and 0.02=pk ( 0.1=pk for the second case) which can be
arbitrarily chosen such that the settling time is smaller than any expected value. Simulation
results for two cases are shown in Fig. 2 and Fig. 3. For the first case, the reference is a step
function 0.5=r rad/s, but for the second case, the reference is a sine function
0.5sin(0.3 )=r t . It can be seen that the proposed method gives much better performances in
terms of settling time and overshoot. Especially, when the reference signal is a sine function, the
magnitude optimum method failed to keep the stability and other performances of the close-loop
system.
To compare performances obtained by using different methods, the IAEs (integral absolute
error) at a set point change and IAEd at process disturbance change are used [4]. IAEs is
calculated as follow
0 5 10 15 20 250
0.1
0.2
0.3
0.4
0.5
seconds
ra
d/
s
The speed of rotor
Magnitude optimum method
Proposed method
Setpoint signal
Figure 2. Responses to a step function.
0 5 10 15 20 25 30-0.5
0
0.5
seconds
ra
d/
s
The speed of rotor
Magnitude optimum
Proposed method
Reference signal
Figure 3. Responses to a sine function.
Settling time assignment with PID
365
IAE ( )
f
i
t
s
t
e t dt∫= (15)
where e is the error between the plant output and the set point, it is the initial time, which is the
time of the step change, ft is an appropriate final time. For the IAEd, the formula is the same as
Eq. (15) but the initial time will be the time of a disturbance change.
The IAEs and IAEd for the two cases are shown in Tab. 2. A step disturbance with amplitude
of 0.01 is applied to the input of the plant at the time of 30 seconds to measure the IAEd. In
conclusion, the proposed method provides better performances in term of disturbance
compensation and set point change for both cases.
Table 2. IAE and IAEd for the two cases.
Methods Case 1 Case 2 IAE IAEd IAE IAEd
Magnitude Optimum 2.23 4.265 9.254 15.046
Proposed Method 0.4892 0.9788 0.5405 0.7585
3.2. DC Rotor Position Control
In this section, the PID controllers are designed to control the angular position of the rotor
[19] which has transfer function as below
( )( ) 1= +
k
G s
s Ts
where the values of parameters are the same as in the previous section. The PI controller
parameters are chosen by using the symmetric optimum method as follows
9, 9 17.61= = =Ia T a and
1 0.0017= =pk
kT a
but our method gives
( )
1 2
1 2
1 2
1 2
1 2
0.1575, 2.8,
0.5893
+ +
= = = + + =
+
= =
+ +
m m
p I m m
m m
m m
D
m m
T T T
k T T T T
kT T
T T T
T
T T T
where the desired settling time 5% 2=T seconds and 1 2 0.4216.= =m mT T For the second case,
if 0.5=sk and 5% 2=T , then 2 0.5441=mT and 1 0.2720.=mT Thus, 37.4724=pk ,
=2.7729IT and 0.5759.=DT
For zero overshoot, an input filter ( ) M s is used to cancel the zero of the close-loop
system out. The two proposed PID controllers provide similar performances as shown in Fig. 4.
The results show that the proposed method gives better performances, such as smaller settling
time as desired. When parameters 1mT and 2mT increase the settling time is larger but the
Nguyen Doan Phuoc, Nguyen Hoai Nam
366
overshoot is smaller and vice versa. The IAEs and IAEd are shown in Tab. 3, where a step
disturbance with amplitude of 0.01 is applied to the input of the plant at the time of 70 seconds
to measure the IAEd. Again, the IAEs and IAEd proved that the proposed method gives much
better performance than the symmetric optimum method.
0 10 20 30 40 50 60 700
1
2
3
4
5
6
seconds
ra
d
The angular position of rotor
Symmetric optimum
Proposed method
Setpoint signal
Figure 4. Responses to a step function.
Table 3. IAEs and IAEd.
Methods Performance
IAEs IAEd
Symmetric Optimum 24.66 104.94
Proposed Method 0.7858 0.2080
4. CONCLUSIONS AND FUTURE WORKS
In the work, a PID design method with desired settling time is proposed for plants with
transfer functions of the following types: first-order system, second-order system and integral
first-order system. The method provides better performances than both the magnitude optimum
method and the symmetric optimum method in the sense of overshoot and settling time. In
addition, the settling time can be achieved as expected for a larger frequency band. Some
examples for DC motor control are used to illustrate the method.
Future works will focus on robustness, effects of model error and other types of transfer
functions, may be higher-order system with time delay.
REFERENCES
1. Ziegler J. G. and Nichols N. B. - Optimum settings for automatic controllers, Trans.
ASME 64 (1942) 759-765.
2. Oldenbourg R. C. and Hans S. - A uniform approach to the optimum adjustment of control
loops, Trans. ASME 18 (1954) 1-9.
3. Hang C. C., Astrom K. I. and Ho W. K. - Refinements of the Ziegler-Nichols tuning
formula, IEE Proc. D. 138 (1991) 111-118.
Settling time assignment with PID
367
4. Finn H. and Bernt L. - Relaxed Ziegler-Nichols Closed-Loop Tuning of PI Controllers,
Modeling, Identification and Control 34 (2013) 83-97.
5. Kessler C. - Das symmetrische optimum, Regelungstechnik 11 (1958) 395-400.
6. Jeffrey W. U. and Mohammed S. - Magnitude and symmetric optimum criterion for the
design of linear control systems: What is it and how does it compare with the others?
IEEE Trans. Indus. Appl. 26 (1990) 489-497.
7. Konstantinos G. P. and Nikolaos I. M. - Extending the symmetrical optimum criterion to
the design of PID type-p control loops, Journal of Process Control 22 (2012) 11-25.
8. Konstantinos G. P., Nikolaos D. T. and Nikolaos I. M. - Revisiting the magnitude
optimum criterion for robust tuning of PID type-I control loops, Journal of Process
Control 22 (2012) 1063-1078.
9. Astrom K. J. and Hagglund T. - Automatic tuning of simple regulators with specifications
on phase and amplitude margins, Automatica 20 (1984) 645-651.
10. Weng K. H., Chang C. H. and Lisheng S. C. - Tuning of PID controllers based on gain
and phase margin specifications, Automatica 31 (1995) 497-502.
11. Sheng Y. C. and Ching C. T. - Tuning of PID controllers based on gain and phase margin
specification using fuzzy neural network, Fuzzy Sets and Systems 101 (1999) 21-30.
12. Cominos P. and Munro N. - PID controllers: recent tuning methods and design to
specification, IEE Proc. Control Theory Appl. 149 (2002) 46-53.
13. Keyu L. - PID tuning for optimal closed-loop performance with specified gain and phase
margins, IEEE Trans. Control Syst. Tech. 21 (2013) 1024-1030.
14. Tyreus B. D. and Luyben W. L. - Tuning PI controllers for integrator/dead time processes,
Ind. Eng. Chem. 31 (1992) 2626-2628.
15. Astrom K. J. and Hagglund T. - PID controllers: Theory, design and tuning, ISA 1995.
16. Astrom, K. J. and Hagglund T. - Advanced PID control, The Instrumentalion, Systems,
and Automation Society, 2005.
17. Phuoc, N. D. - Linear Control Theory, Science and Technology Publisher, 2007.
18. Ntogramatzidis L. and Ferrante A. - Exact tuning of PID controllers in control feedback
design, IET Cont. Theo. Appl. 5 (2011) 565-578.
19. Siri W. and El S. - Identification and control of a DC motor using back-propagation neural
networks. IEEE Transactions on Energy Conversion 6 (1991) 663-669.
Các file đính kèm theo tài liệu này:
- 8808_37438_1_pb_8116_2061747.pdf