Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

2015 4th International Conference on Computer Science and Network Technology (ICCSNT 2015)

Comparison of Time Delay Processing Methods in


Control System

Lin Xueyan Ye Zheng


Automation School Automation School
Beijing University of Posts and Telecommunications Beijing University of Posts and Telecommunications
Beijing, postcode: 100876, China Beijing, postcode: 100876, China
xylin@bupt.edu.cn yexx12@outlook.com

is a small delay system and when  >0.5, the system is a large


Abstract—Time delay exists in the actual control system and
presents some difficulties in processing of transfer function for
T
process control. Approximation for time delay is commonly used delay system. When a system inevitably contains time delay,
in control system. This paper analyses the effects of different
some measures are needed with a reasonable mathematical
approximation including Taylor, Pade and all-pole
approximation for different orders closed loop system. Second
method to approximate time delay to overcome the effects of
order Pade approximation can fit the time delay in first-order time delay in order to make control quality of the system meet
closed-loop system. Taylor approximation cannot apply to the the requirements. Time delay approximation methods currently
system with large time delay. First order all-pole approximation used are Taylor approximation, Pade approximation and
gets the smallest over shoot, adjusting time both in first and all-pole approximation. In this paper, these three kinds of
second-order large delay system. approximate methods will be compared in different  and
orders system.
Keywords—time delay; Taylor approximation; Pade
approximation; All-pole approximation; closed-loop system; Fig.1 is the typical closed-loop system with unit negative
performance indexes feedback. G0(s) and G(s) are the transfer function of the time
delay and control objects respectively. r(t) and y(t) are the input
I. INTRODUCTION and output of the system respectively.
The variation of output and input in time delay is the same r(t) y(t)
except that the initial response of output falls behind input with G(s) - G0(s)
1
time  . Time delay transfer function is defined as e s  es ( is -
the delay time constant). In real control systems, time delay is
common [1-4].
Time delay does not have any effects on output amplitude Fig. 1. Closed-loop system flowchart
of control system. But it will make the stability of the system
worse, response speed of the system slower, accuracy of W ( s) 
Y ( s)

GG0

steady-state response lower[5]. To be more specific, Closed-loop transfer function is R( s) 1  GG0 . G(s) is
Amplitude-frequency characteristic of time delay is the origin K
G( s) 
frequently accepted as the first order objects Ts  1 , or
as the Center, with 1 as the radius of the circle. As the delay
K
constant increases, there will be two undesirable consequences: G(s) 
2
T 2s  s 1
a) Lower crossover frequency. b) Less critical gain. the second order objects .
It is necessary to deal
T

with G0(s)  es with some appropriate approximation


Lower crossover frequency will make the system more methods, on the basis of the performance indexes.
sensitive to low-frequency periodic disturbances in closed-loop
response. And less critical gain will affect the accuracy of Control system performance is inferred by its dynamic
steady-state response. performance indexes. This paper chooses rise time, overshoot,
and overshoot time adjusting time to describe performance of
The effect of time delay on the quality of control systems different systems.
does not depend on the absolute value of , but the value of
ratio  (T is inertia time constant). The ratio is used to
T II. USUAL TIME DELAY APPROXIMATIONS
describe whether the controlled object has a large time delay [6].
s
It is generally believed that when the control system  <0.3, it The Taylor expansion of Time delay e is as follows:
T

978-1-4673-8173-4/15/$31.00 ©2015 IEEE 1502 Harbin, China


s  2s 2  3s 3  ns n (1) Approxi Closed-loop transfer function
e s  1 -    ...(1)n mation First order Second order
1! 2! 3! n!
Taylor 1  s  2s 2
W a(s )  1  s 
The variant Taylor expansion of Time delay e s is as (5   )s  2 W b(s )  2
 2s 2
follows:  (5   )s  2
2
1 1
e s   All-pole 1
e s s  2s 2  3s 3  ns n Wc(s )  1
1    ... 5s2  (5   )s  2 W d (s )  5 2 2 2
1! 2! 3! n! (2) 2
s3  (5 
2
)s  (5   )s  2

Equation (2) is also known as the all-pole approximation.


Pade 2  s 2
Pade approximation of e s is [8]: We ( s) 
5s 2  10s  4 Wf ( s) 
2  s 
6
s2
5 2 3 2 2
s  (5  ) s  10s  4
s  2s 2 n!  n s n 6 3
1-   ...(1)n
 s 2 12 (2n )!
e 
s  2s 2  ns n The unit-step response curve of closed-loop system with
1   ...
2 12 n!2n (3) different approximations are shown in Fig. 2 to 5.
Closed-loop step response of first-order time delay system with approximation
Different approximations processing time delay are shown
in tableⅠ.
0.8

TABLE I. EXPRESSION AFTER DIFFERENT ORDER APPROXIMATION


0.6
Approximation expression after different approximate
First order Second order 0.4

value
Taylor e s  1 - s e s
 1 - s 
 2s2
0.2 first order Taylor
2
second order Taylor
first order Pade
All-pole 1 1 1 1
e s 
0
e s    second order Pade
es 1  s es  2s2 first order All-pole
1  s  second order All-pole
2 -0.2
not treated

Pade 1 - s / 2 2 - s 
 s 2 2 0 5 10 15 20 25 30 35 40
e s  time s (seconds)
1  s / 2 e s  6
 2s2
2  s 
6 Fig. 2. =8, Closed-loop step response of first-order time delay
system with approximations
Both Taylor approximation and Pade approximation
increase the zero points of open-loop system in the right TABLE III. WHEN=8, THE PERFORMANCE OF FIRST-ORDER TIME DELAY
half-plane of s domain equating to add a non-minimum-phase SYSTEM WITH APPROXIMATION PROCESSING
system. Then the closed-loop step response of the system will
Approximation Rise time Overshoot Overshoot adjustin
be slower than the original. All-pole approximation and Pade (tr) time (tp) g time
(%)
approximation add system open-loop poles of the left (ts)
half-plane of s domain. That means the system adds a negative Not treated 2.45 63.31% 17.14 80.20
angle in root-locus equation of phase, which is not conducive First order Taylor —* — — —
to the stability of the system and its dynamic performance Second order Taylor — — — —
improvement. The system reduces the overshoot, but the First order Pade 3.32 37.76% 13.26 38.32
adjusting time will be increased. For a second order system, Second order Pade 2.55 60.54% 16.58 78.10
when the real part of pole-zero is close from the original First order all-pole 9.64 3.58% 20.40 26.07
damping coefficient value, the additional pole-zero has great Second order 8.40 26.73% 20.27 58.49
impact on the system. all-pole
* - Indicates that the value does not exist, and time unit is second.
III. COMPARISON OF FIRST-ORDER DELAY SYSTEM WITH
As shown Figure 2, step response of time delay closed-loop
APPROXIMATION
system with no approximation has 8 seconds decaying
1 oscillation. Taylor approximation system is unstable, and
G(s) 
As shown in Figure 1, let 5s  1 , is selected 8, 4, 2.5,
closed-loop unit-step response is divergent. Overshoot and
and 1 respectively. When the >2.5,   0.5 , the system is a adjusting time of first order Pade approximation system are
T less than second order Pade approximation system, but rise
large delay system. When <2.5, the system is a small delay time of these two systems is relatively similar. Second order
system. Processing time delay with different approximations to Pade approximation system has the longest adjusting time.
get the closed-loop transfer function of the system is shown in Overshoot and adjusting time of first order all-pole
tableⅡ. approximation system are the minimum. Adjusting time of first
order all-pole approximation system is similar to second order
TABLE II. THE CLOSED-LOOP TRANSFER FUNCTION OF THE TIME all-pole approximation. The response curve of the system with
DELAY SYSTEM

1503
second order Pade approximation is closed to the original Step responses of three approximations are similar in small
system. delay systems while the step responses are quite different in
large delay systems. The difference between different
Closed-loop step response of first-order time delay system with approximation
approximations in large delay system is shown in table IV:
0.8
TABLE IV. RESPONSE CURVE CHARACTERISTICS OF FIRST ORDER LARGE
DELAY SYSTEMS
0.6

Approximation Large delay systems


0.4 =8 =4
value

not treated decaying oscillation decaying oscillation,


0.2 first order Taylor
smaller overshoot and
second order Taylor
first order Pade
adjusting time
0 second order Pade
first order All-pole First order Taylor Divergence, unstable, large initial reverse
-0.2 second order All-pole large initial reverse response, no overshoot
not treated
response
0 5 10 15 20 25 30 35 40
time s (seconds)
Second order Taylor Divergence, unstable, initial reverse
initial reverse response response, large
Fig. 3. =4, Closed-loop step response of first-order time delay system with overshoot and
approximations adjusting time
Closed-loop step response of first-order time delay system with approximation
First order Pade Small initial reverse Small initial reverse
first order Taylor
second order Taylor
response, fitting with response, fitting with
0.8 first order Pade not treated well not treated well
second order Pade
first order All-pole Second order Pade Fitting with not treated Fitting with not treated
0.6 second order All-pole best best
not treated
0.4 First order all-pole No overshoot, small No overshoot, small
value

adjusting time, slow adjusting time, slow


0.2 response response

0 Second order all-pole Large adjusting time fitting with not treated
well
-0.2

0 5 10 15 20 25 30 35 40 IV. COMPARISON OF SECOND-ORDER DELAY SYSTEM WITH


time s (seconds)
Fig. 4. =2.5, Closed-loop step response of first-order time delay system with
APPROXIMATION
approximations 1
As shown in Figure 1, let G  , T=3 and
9s 2  5s  1
The rise time, overshoot and adjusting time for different damping ratio =0.83.  is respectively selected 8, 2.5 and 1.
approximation methods are reduced with decreasing delay time 
. But performance indexes of first order all-pole When the >1.5,  0.5 , the system is a large delay system.
T
approximation systems have little change with decreasing . When <1.5, the system is a small delay system. Processing
Closed-loop step response of first-order time delay system with approximation
time delay with different approximation methods to get the
first order Taylor
second order Taylor
closed-loop transfer function of the system is shown in tableⅤ.
0.8 first order Pade
second order Pade
TABLE V. CLOSED-LOOP TRANSFER FUNCTION OF THE TIME DELAY SYSTEM
first order All-pole
0.6
second order All-pole
not treated Approximation Closed-loop transfer function
0.4 not treated e s
W 0(s ) 
value

9s  5s  1  e s
2

0.2 First order Taylor 1  s


W a(s ) 
9s 2  (5   )s  2
0
Second order  2s 2
1  s 
Taylor 2
-0.2 W b(s ) 
2
0 5 10 15 20 25 30 35 40
(9  )s 2  (5   )s  2
2
time s (seconds)
First order 1
Wc(s ) 
Fig. 5. =1, Closed-loop step response of first-order time delay all-pole 9s 3  (9  5 )s2  (5   )s  2
system with approximations Second order Wd(s ) 
1
9 2 4 5 2 3 2
all-pole  s  (9  )s  (9  5  )s 2  (5   )s  2
2 2 2
In = 1 small delay system, different approximation First order Pade 2  s
W e(s ) 
processing methods have similar performance indexes 9s 3  (18  5 )s 2  10s  4
including adjusting time, rise time and overshoot. Except the Second order 2  s 
2
s2
initial moment, the response curve is coincident basically. Pade W f(s )  6
3 2 4 5 2
2 2
 s  (9  )s3  (18  5  )s  10s  4
2 6 3

1504
The unit-step response curve of closed-loop system with For small delay second-order systems( = 1), closed-loop
different approximations are shown in Fig. 6 to 8. step response curves for different approximation methods are
almost the same. However, the step response curve of the
Closed-loop step response of second-order time delay system with approximation
system with not approximation has no overshoot and has fast
0.8
rise time. Three approximate methods can not fit well.

0.6
In large delay systems, first order Pade approximation can
be used to reproduce the response curve of the system that is
0.4 not treated, fitting a higher degree. From the control point of
value

0.2
view, the first order all-pole approximation recommended can
first order Taylor
second order Taylor take both overshoot and adjusting time well, with more gentle
0
first order Pade
second order Pade
curves and no jitter phenomenon. First or second order Pade
first order All-pole
second order All-pole
approximation system have fluctuation in the initial delay time;
-0.2
not treated Taylor approximation has strong inverse response at the initial
0 5 10 15 20
time s (seconds)
25 30 35 40
moment. In the control system, fluctuation and inverse
Fig. 6. =8, Closed-loop step response of second-order time delay response will bring bad effects on the certain actuator, so for
system with approximations second-order delay system, it is not recommended to use
low-order Pade and Taylor approximations.
When  = 8, Taylor approximation system remains unstable,
and the dynamic performance indexes do not exist. Combined V. CONCLUSIONS
with figure 2, Taylor approximation cannot be applied to the Only in a small delay first-order system, all three
system with large time delay. When =2.5, initial fluctuation approximations can fit the original system well. Systems with
and overshoot of first-order Taylor approximation system that Taylor approximation have initial inverse response, and Pade
has reverse fluctuation in the initial moment decreases with approximation has response vibration in the initial delay time.
decreasing , and Pade approximation system has response In large delay systems, the step response of second-order
vibration in the initial delay time. Pade approximation and Taylor approximation is damped oscillation, response speed is
second order all-pole approximation have the similar rise time, accelerated with the decrease of , the overshoot decreases with
adjusting time, overshoot. Except a slightly larger rise time, the
decreasing . Precision in the initial delay time of Pade
overshoot and adjusting time of the first order all-pole
approximation are minimized. approximation varies while the order varies. The higher the
Closed-loop step response of second-order time delay system with approximation
order is, the more accurate the approximation is. However, the
calculation would increase according to the high order, and the
0.8
system becomes more complex, more difficult to control.
Approximate effect of all-pole approximation is good. The
0.6 overshoot and adjusting time of first order all-pole
0.4
approximation are the smallest. Obviously, the all-pole
value

approximation in a first or second-order system is more stable


0.2 first order Taylor than the other two approximations. It can be an ideal control
second order Taylor
first order Pade effect with stabler step response curve.
0
second order Pade
first order All-pole
-0.2 second order All-pole
not treated REFERENCES
0 5 10 15 20 25 30 35 40
time s (seconds) [1] J.E.Marshall. “Control of time delay system,” J. The Institution of
Fig. 7. =2.5, Closed-loop step response of second-order time Electrical Engineers,1979.
delay system with approximations [2] Jiang Quan Yuan, “Consider delays affect the stability of power system
Closed-loop step response of second-order time delay system with approximation analysis and research progress in Wide-area control,” J. Electric Power
Systems, March 2005,pp. 1-7.
[3] Zhang Guangxin. “Fully automated control and its application of beer
0.8
production process,” D. Zhejiang University, 2002.
0.6 [4] Huang Can, “Several studies on alumina continuous carbonation
decomposition process in delay system,” D. Central South University,
0.4 2012
value

[5] Shi jie, Wang Chen Shan, “Impact of time delay on the performance of
0.2 first order Taylor power system stabilizer,” J. Relay, 2006,13: 21-24.
second order Taylor
first order Pade [6] John Gerry, “How to control proeess with large dead time,” J. Control
0
second order Pade Engineering, March 1998,pp.145-146.
first order All-pole
-0.2 second order All-pole [7] Jin Qibing, “Base on all-pole approximation a new Internal Model PID
not treated Control method for the system with time delays,” CA: IEEE
0 5 10 15 20 25 30 35 40 International Conference on Mechatronics and Automation. 2009,pp.
time s (seconds)
268-273
Fig. 8. =1, Closed-loop step response of second-order time delay
system with approximations [8] Li Chengkai, “Research of. Internal Model Control method to unstable
delay systems,” D. Beijing University of Chemical Technology, 2011.

1505

You might also like