## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

**: #44-116-201-5531; fax: #44-116-
**

2015464.

E-mail address: guoping.liu@energy.alstom.com (G.P. Liu)

Control Engineering Practice 7 (1999) 821}830

Optimal-tuning PID controller design in the frequency domain with

application to a rotary hydraulic system

G.P. Liu*, S. Daley

Energy Technology Centre, ALSTOM, Cambridge Road, Leicester LE8 6LH, UK

Received 13 November 1998; accepted 25 March 1999

Abstract

In this paper a new PID controller design scheme that uses optimisation in the frequency domain is proposed for industrial process

control. An optimal-tuning PID controller is designed to satisfy a set of frequency-domain performance requirements: gain margin,

phase margin, crossover frequency and steady-state error. Using an estimated process frequency response, the method can provide

optimal PID parameters even in cases where the process dynamics are time variant. This scheme is demonstrated through its

application to a rotary hydraulic system and its performance is compared with six alternative PID tuning rules. 1999 Elsevier

Science Ltd. All rights reserved.

Keywords: Optimal-tuning control; PID controller; Frequency-domain; Hydraulic system

1. Introduction

The proportional, integral and derivative (PID) con-

trol algorithm remains the most popular approach for

industrial process control despite continual advances in

control theory. This is not only due to the simple struc-

ture, which is conceptually easy to understand and,

which makes manual tuning possible, but also to the

fact that the algorithm provides adequate performance

in the vast majority of applications. However, for a var-

iety of reasons optimal setting of the gains is di$cult

and as a result many PID design techniques have

been developed in the literature (see, for example,

Astrom & Hagglund, 1984; Zhuang & Atherton, 1993;

Daley & Liu, 1998; Liu, Dixon & Daley, 1998;

McCormack & Godfrey, 1998).

Most of the PID tuning rules developed in the last 50

years use frequency-response methods. Examples in-

clude, Ziegler}Nichols rule (Ziegler & Nichols, 1942),

symmetric optimum rule (Kessler, 1958; Voda & Landou,

1995), Ziegler}Nichols' complementary rule (Mantz

& Tacconi, 1989), some-overshoot rule (Seborg, Edgar

& Mellichamp, 1989), no-overshoot rule (Seborg et al.,

1989), re"ned Ziegler}Nicholes rule (Hang, Astrom

& Ho, 1991), integral of squared time weighted error rule

(Zhuang & Atherton, 1993), and integral of absolute

error rule (Pessen, 1994). These methods are straightfor-

ward to apply since they provide simple tuning formulae

to determine the PID controller parameters. However,

since only a small amount of information on the dynamic

behaviour of the process is used, in many situations they

do not provide good enough tuning or produce a satis-

factory closed-loop response. For example, in practice,

the Ziegler}Nichols rule often leads to a rather oscilla-

tory response to setpoint changes.

In an e!ort to improve the performance of PID tuning

for processes with changing dynamic properties, several

automatic tuning and adaptive strategies have been pro-

posed (Kraus & Mayron, 1984; Astrom & Hagglund,

1984; Radke & Issermann, 1987). These controllers have

self-initialisation and recalibration features to cope with

little a priori knowledge and signi"cant changes in the

process dynamics, based on the automatic measurement

of the ultimate gain and period. Various techniques, such

as relay excitation feedback (Astrom & Hagglund, 1984)

and rule-based autotuning (McCormack & Godfrey,

1998), have been developed. However, the PID controller

parameters are still computed using the classic tuning

formulae and, as noted above, these do not provide good

control performance in all situations.

0967-0661/99/$- see front matter 1999 Elsevier Science Ltd. All rights reserved.

PII: S 0 9 6 7 - 0 6 6 1 ( 9 9 ) 0 0 0 4 7 - 7

Table 1

Six PID tuning rules

PID rule PID parameters

Ziegler}Nichols K

A

"0.60K

S

, ¹

G

"0.5¹

S

, ¹

B

"0.125¹

S

Integral of absolute

error

K

A

"0.70K

S

, ¹

G

"0.4¹

S

, ¹

B

"0.150¹

S

Some-overshoot rule K

A

"0.33K

S

, ¹

G

"0.5¹

S

, ¹

B

"0.330¹

S

No-overshoot rule K

A

"0.20K

S

, ¹

G

"0.5¹

S

, ¹

B

"0.330¹

S

Integral of squared

time weighted error

K

A

"0.509K

S

, ¹

G

"0.051(3.302K

N

K

S

#1)¹

S

n ¹

B

"0.125¹

S

Symmetric optimum

K

A

"

4#B

8(2

B

K

¹``

, ¹

G

"

4#B

B

¹``

, ¹

B

"

B

(4#B)

¹``

Fig. 1. Standard PID control system.

In order to address this problem a PID controller

design based on direct optimisation in the frequency-

domain is developed in this paper. A set of frequency-

domain performance functions are considered, which are

the gain-margin, phase-margin, crossover-frequency and

steady-state error. The PID controller is then designed to

directly satisfy the requirements set for these functions.

Using a non-parametric model represented by the system

frequency response, a method is derived that can pro-

vide optimal PID parameters for di!erent working con-

ditions. This controller is demonstrated through its

application to a rotary hydraulic system and its perfor-

mance evaluated through comparison with six other PID

tuning rules.

2. Rule-based PID controller design

A large number of industrial processes can be charac-

terised by a "rst-order plant with dead time (FOPDT).

The transfer function of a FOPDT model is described by

G(s)"

K

N

eQO

1#s¹

, (1)

where K

N

is the gain, ¹ the time constant and the dead

time. It is assumed that the ideal transfer function of

a PID controller is given by

K(s)"K

A

1#

1

¹

G

s

#¹

B

s

, (2)

where K

A

, ¹

G

and ¹

B

are the PID parameters.

Based on the FOPDT model, there are a number of

PID tuning formula available. Six PID tuning rules are

introduced and assessed in later sections. These tuning

rules are Ziegler}Nichols (NZ) (Ziegler & Nichols, 1942),

integral of absolute error (IAE) (Pessen, 1994), some-

overshoot rule (SOR) (Seborg et al., 1989), no-overshoot

rule (NOR) (Seborg et al., 1989), integral of squared time

weighted error (ISTWE) (Zhuang & Atherton, 1993),

symmetric optimum rule (SO) (Kessler, 1958; Voda

& Landou, 1995). A summary of the six PID tuning rules

are given in Table 1.

In Table 1, K

S

and ¹

S

are the inverse of the system gain

and frequency at which the phase is !1803. K

¹``

and

¹``

correspond to the gain of the system and frequency

at which the system phase is !1353. B3[1, 2] is an

acceleration factor. It can be seen from the table that the

six PID design methods do not require a parametric

transfer function model of the process and only need

either one or two frequency response measurements of

the process.

3. Optimal-tuning PID control

The design of feedback control systems in industry

using frequency-response methods is more popular than

any other. This is primarily because the frequency re-

sponse method provides good designs in the face of

uncertainty in the plant model and can easily use experi-

mental information for design purposes.

3.1. Frequency response estimation

To estimate the system frequency response, the design

of the excitation signal is very important. The excitation

signal utilised here is the commonly used multi-sine

which has the following form:

x(t)"

,

I¯¹

A

I

sin(2 f

G

t#

I

) , (3)

where A

I

,

I

and f

I

are the amplitude, phase and fre-

quency of the signal components. The phase chosen after

the desired power distribution is determined has an im-

portant in#uence on the time-domain signal shape. For

example, the maximal peak can be signi"cantly com-

pressed, enabling greater energy to be injected for the

given input range of the measurement device, and the

system is kept in the linear working region.

During the identi"cation, the closed-loop system with

the PID controller is assumed to be of the structure

shown in Fig. 1. The multi-sine signal is added to the

reference r(t). For this system, the unbiased estimate of

the frequency-response of the process using spectral anal-

ysis is given by (Wellstand, 1981; McCormack & God-

frey, 1998)

GK( j)"

PK

WP

( j)

PK

SP

( j)

, (4)

822 G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830

where

PK

WP

( j)"

1

m

K

G¯¹

RH

G

( j)>

G

( j), (5)

PK

SP

( j)"

1

m

K

G¯¹

RH

G

( j);

G

( j) (6)

with R

G

( j), ;

G

( j) and >

G

( j) representing the

Fourier coe$cients of the reference input, the control

input and the output, respectively. If the averaging of

every quantity is done in a recursive way, then, for

k"1, 2,

2

GK( j)"

PK'I'

WP

( j)

PK'I'

SP

( j)

, (7)

where

PK'I'

WP

( j)"

k!1

k

PK'I¹'

WP

( j)#

1

k

RH

I

( j)>

I

( j), (8)

PK'I'

SP

( j)"

k!1

k

PK'I¹'

SP

( j)#

1

k

RH

I

( j);

I

( j) (9)

with PK'"'

WP

( j)"0 and PK'"'

SP

( j)"0. For the open-loop

identi"cation case, the multi-sine signal is directly added

to the control input of the process. The estimation of the

frequency response of the process is still the same as the

above with R

I

( j)"1, ∀k. Generally speaking, the accu-

racy of the open-loop estimation is better than the

closed-loop case. But, in practice, the latter is more con-

venient for implementation than the former.

3.2. Optimal PID controller design

Although the six PID design methods give simple

tuning rules for the controller parameters using either

one or two measurement points of the system frequency

response, their control performance may not satisfy the

desired requirements. To overcome this disadvantage, an

optimal PID controller design is proposed in the fre-

quency domain.

In the frequency domain, there are two quantities used

to measure the stability margin of the system. One is the

gain margin, which is the factor by which the gain is less

than the neutral stability value. The other is the phase

margin, which is the amount by which the phase of the

system exceeds !1803 when the system gain is unity.

The gain and phase margins are also related to the

damping of a system. In addition to the stability of

a design, the system is also expected to meet a speed-of-

response speci"cation like bandwidth. The crossover fre-

quency, which is the frequency at which the gain is unity,

would be a good measurement in the frequency domain

for the system's speed of time response. Also, the larger

the value of the magnitude on the low-frequency asym-

ptote, the lower the steady-state errors will be for the

closed-loop system. This relationship is very useful in the

design of suitable compensation.

After the above considerations, it is assumed that

the major requirements in the frequency domain are

those on the gain-margin, the phase-margin, the crossover-

frequency and the steady-state error. Thus, the follow-

ing performance functions need to be considered during

the design of a PID controller:

¹

(K

A

, ¹

G

, ¹

B

)"

"K( j)G( j)"

G

+

,

LK( j)G( j)"!1803

, (10)

`

(K

A

, ¹

G

, ¹

B

)"

2!

180#LK( j)G( j)

P

+

,

"K( j)G( j)""1

, (11)

`

(K

A

, ¹

G

, ¹

B

)"

2 f

B

, "K( j) G( j)""1

, (12)

"

(K

A

, ¹

G

, ¹

B

)"

1

e

QQ

"K( j)G( j)"

, "1

, (13)

where

G

(K

A

, ¹

G

, ¹

B

), for i"1, 2, 3, 4, are the normalised

gain-margin, phase-margin, crossover-frequency and the

steady-state error functions with the desired values

G

+

, P

+

, f

B

and e

QQ

, respectively. Thus, the following

performance criteria should be satis"ed:

G

(K

A

, ¹

G

, ¹

B

))1, i"1, 2, 3, 4. (14)

If the above inequalities are met, then the problem is

solved. Clearly, the design problem is to "nd a PID

controller to make Eq. (14) hold. There are a number of

methods to solve the performance criteria problem (14).

Here, two methods are brie#y introduced: the minimax

optimisation method and the method of inequalities.

Using the minimax optimisation method (Gill, Murray

& Wright, 1981), the performance criteria (14) can be

satis"ed if

min

)A 2G 2B

max

G¯¹ ` ` "

+

G

(K

A

, ¹

G

, ¹

B

),)1. (15)

Clearly, the above minimises the worst case values of the

performance functions.

The method of inequalities (Zakian & Al-Naib, 1973;

Liu, 1992; Whidborne & Liu, 1993) uses optimisation

algorithms (e.g., moving boundaries algorithm) to "nd

the admissible or feasible set of parameter vectors, for

which all the performance inequalities hold. The admiss-

ible set is de"ned as

"

¹

5

`

5

`

5

"

, (16)

G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830 823

Fig. 3. Schematic of hydraulic rig.

Fig. 2. Optimal-tuning PID control structure.

where

G

is the set of parameter vectors for which the ith

functional inequality is satis"ed, that is

G

"+(K

A

, ¹

G

, ¹

B

),

G

(K

A

, ¹

G

, ¹

B

))1,. (17)

Similar algorithms for solving the problem de"ned in (14)

now exist in standard libraries of optimisation software,

for example, the optimisation toolbox for use with MAT-

LAB (Grace, 1994).

3.3. Optimal-tuning PID control scheme

When a system has di!erent operating points with

widely di!ering dynamic properties, it is not always pos-

sible to exercise control with a "xed parameter control-

ler, even if this is a highly robust controller. For this case,

the optimal-tuning PID control scheme shown in Fig. 2

is proposed. It mainly consists of four parts: frequency

response estimation, desired system speci"cations,

optimal-tuning mechanism and PID controller. The fre-

quency response estimated using frequency-domain iden-

ti"cation methods provides a non-parametric model for

the process. The desired system speci"cations includes

a set of requirements in the frequency domain: gain

margin, phase margin, crossover frequency and steady-

state error. The optimal-tuning mechanism uses the pro-

cess frequency response to "nd optimal parameters for

the PID controller so that the desired system speci"ca-

tions are satis"ed.

The operating procedure of the optimal-tuning PID

control is as follows. When the system's operating-point

or dynamics change, the new process frequency response

is re-estimated by switching on the excitation signal.

Then, using this updated frequency-response, the tuning

mechanism searches for the optimal parameters for the

PID controller to satisfy the desired system speci"ca-

tions. Finally, the PID controller is set to the obtained

optimal parameters. In this way, the PID controller may

cope with all operating-points of the system and the

closed-loop system will have similar optimal-control

performance. But, compared with a "xed parameter con-

trol, the disadvantage of this strategy is that it needs

slightly more computation to search for the optimal

parameters.

4. Application to a rotary hydraulic system

The optimal-tuning PID control scheme is applied to

a rotary hydraulic test rig (Daley, 1987), which is repre-

sentative of many industrial systems that utilise #uid

power. This is a particularly opposite application of the

method since hydraulic systems are often very conserva-

tively tuned, due to the fact that the cost of getting the

tuning wrong can be highly destructive and costly. To

e!ectively assess the performance of the proposed tuning

method the other six tuning rules which were introduced

in Section 2 are also applied to the rig.

4.1. Rotary hydraulic system

The rotary hydraulic test rig, as shown in Fig. 3,

comprises an electro-hydraulic servo control valve driv-

ing a "xed displacement hydraulic motor up to 8000 rpm

with a maximum operating pressure of 21 MPa. The

motor is coupled by a rigid shaft to a hydraulic pump of

the same displacement as the motor and a solenoid

controlled relief valve is used to simulate variations in

load.

824 G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830

Fig. 4. Implementation of the optimal-tuning PID control of the hydraulic rig.

Fig. 5. Multisine excitation signal.

This type of hydraulic system is typically applied to

mixer drives, centrifuge drives and machine tool

drives where accurate speed control with fast response

times is required, and large changes in load can be

expected.

4.2. Implementation structure

The optimal-tuning PID control scheme is imple-

mented using the MathWorks Real-Time Workshop

connected to a dSPACE DSP board based around the

TMS320, MATLAB and SIMULINK. With this system

a powerful implementation strategy is possible, as shown

in Fig. 4. It mainly comprises an external control loop

and an internal control loop. The external control loop

includes the hydraulic system and a PID controller. The

internal control loop includes the adaptive model up-

dated by an on-line system identi"cation algorithm and

a PID controller.

The parameters of the two PID controllers are ad-

justed by the optimal-tuning PID algorithm. Roughly

speaking, the implementation of PID control strategies

consists of two stages. The "rst stage is to run a

PID controller on the identi"ed on-line model to predict

the performance of the PID controller before it is used

on the rig. The second stage is to apply the PID control-

ler to the rotary hydraulic system using the PID

parameters that have been veri"ed as being safe on

the adaptive model. In this way, any unnecessary

damage resulting from the wrong PID parameters can

be prevented through a slight increase in computational

load.

4.3. Experimental results

During the experiment, the rotary hydraulic system

was operated in two working conditions: with load and

without load. So, the experimental results for the two

cases are presented.

4.3.1. Case I: with load

For the sake of simplicity, a periodic multi-sine excita-

tion signal, as shown in Fig. 5, was directly applied to the

hydraulic system in an open-loop way. Based on the

input}output data, the frequency response of the system

G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830 825

Fig. 6. Frequency-response of the hydraulic motor with load.

Table 2

PID parameters for case I

PID rule K

A

¹

G

¹

B

Ziegler}Nichols 0.1662 0.0606 0.0152

Integral of squared time

weighted error

0.1410 0.0423 0.0152

Integral of absolute error 0.1939 0.0485 0.0182

Symmetric optimum 0.1268 0.1141 0.0065

Some-overshoot rule 0.0914 0.0606 0.0400

No-overshoot rule 0.0554 0.0606 0.0400

Optimal design rule 0.0632 0.0372 0.0351

was estimated using the frequency-domain identi"cation

methods discussed in Section 3. The estimated magnitude

and phase of the system with respect to frequency are

shown in Fig. 6.

Following the six PID tuning rules in Section 2 and the

optimal PID design rule proposed in this paper provides

the parameters which are given in Table 2. The speed

responses of the hydraulic motor using seven PID

controllers are shown in Figs. 7 and 8.

4.3.2. Case II: without load

In this case, there was no load on the hydraulic motor.

This means the process dynamics are di!erent. When the

PID parameters for Case I were still used, three PID

controllers failed to stabilise the system, namely the

Ziegler}Nichols, integral of squared time weighted error

and integral of absolute error rules. The speed responses

using the other four PID control rules are given in Fig. 9.

Note that the symmetric optimum rule and the some-

overshoot rule give a signi"cant oscillatory response.

Though the no-overshoot rule and the optimal design

rule do not give a speed response as good as the optimal

design rule for Case I, their performance is reasonably

good and for some practical cases may be considered

acceptable.

Since the process dynamics have changed signi"cantly

a more appropriate response would be to re-tune. The

periodic multi-sine excitation signal, as shown in Fig. 5,

was directly applied to the hydraulic system again. The

estimated frequency response of the load-free case is

shown in Fig. 10. Comparing Fig. 10 with Fig. 6, it is

clear that the system frequency responses for these two

cases are di!erent, with the loading change being seen

mainly as an apparent gain change. In terms of the seven

PID tuning rules, the parameters of the PID controllers

for this case are given in Table 3.

The performance of the seven PID controllers for this

case are shown in Figs. 11 and 12. It is clear that the

relative performance is similar to that displayed for

Case I.

5. Conclusions

In this paper an optimal PID controller design scheme

based in the frequency domain has been presented. The

scheme mainly consists of four parts: frequency-response

826 G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830

Fig. 7. The speed of the hydraulic motor using NZ, ISTWE, IAE and SO controllers (Case I).

Fig. 8. The speed of the hydraulic motor using SOR, NOR and ODR controllers (Case I).

estimation, a de"nition of desired system speci"cations,

an optimal-tuning mechanism and a PID controller.

The frequency response estimated using frequency-

domain identi"cation methods provides a non-paramet-

ric model for the process. The desired system speci"ca-

tions includes a set of requirements in the frequency

G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830 827

Fig. 9. The speed of the hydraulic motor without load using SOR, NOR, SO and ODR controllers for Case I.

Fig. 10. Frequency response of the hydraulic motor without load.

Table 3

PID parameters for case II

PID rule K

A

¹

G

¹

B

Ziegler}Nichols 0.1012 0.0549 0.0137

Integral of squared time

weighted error

0.0858 0.0330 0.0137

Integral of absolute error 0.1180 0.0439 0.0165

Symmetric optimum 0.0844 0.0994 0.0056

Some-overshoot rule 0.0556 0.0549 0.0362

No-overshoot rule 0.0337 0.0549 0.0362

Optimal design rule 0.0317 0.0262 0.0385

domain: gain-margin, phase-margin, crossover-frequency

and steady-state error. The optimal-tuning mechanism

uses the process frequency response to "nd optimal para-

meters for the PID controller so that the desired system

speci"cations are satis"ed. This scheme has been success-

fully applied to a rotary hydraulic system and has also

been compared with another six auto-tuning PID con-

trol rules. The experimental results have shown that this

optimal-tuning PID controller can signi"cantly improve

system performance, and copes well with changes in the

process dynamics.

828 G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830

Fig. 11. The speed of the hydraulic motor using NZ, ISTWE, IAE and SO controllers (Case II).

Fig. 12. The speed of the hydraulic motor using SOR, NOR and ODR controllers (Case II).

Acknowledgements

The authors are grateful to the management of the

ALSTOM Energy Technology Centre for giving per-

mission to publish this work.

References

Astrom, K. J., & Hagglund, T. (1984). Automatic tuning simple regu-

lators with speci"cations on phase and amplitude margins. Auto-

matica, 20(5), 645}651.

G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830 829

Daley, S. (1987). Application of a fast self-tuning control algorithm to

a hydraulic test rig. Proceedings of the Institution of Mechanical

Engineers, 201, 285}295.

Daley, S., & Liu, G. P. (1998). Optimal PID tuning using direct search

algorithm. Tuning-in to increase proxt-developments in PID tuning,

IMechE Seminar, London.

Hang, C. C., Astrom, K. J., & Ho, W. K. (1991). Re"nements of the

Ziegler}Nichols tuning formula. IEE Proceedings-D, 138(2), 111}118.

Gill, P. E., Murray, W., & Wright, M. H. (1981). Practical optimisation.

New York: Academic Press.

Grace, A. (1994). Optimisation toolbox for use with MATLAB. The Math-

Works Inc.

Kessler, C. (1958). Das symmetrische optimum. Regelungstetechnik,

6(11), 395}400.

Kraus, T. W., & Mayron, T. J. (1984). Self-tuning PID controllers based

on a pattern recognition approach. Control Engineering, 106}111.

Liu, G. P. (1992). Theory and design of critical control systems. Ph.D.

thesis, Control Systems Centre, University of Manchester Institute

of Science and Technology, U.K.

Liu, G. P., Dixon, R., & Daley, S. (1998). Multiobjective optimal-tuning

PI controller design for a gasi"er. The MEC benchmark challenge on

gasixer control, IMechE Seminar, Coventry.

Mantz, R. J., & Tacconi, E. J. (1989). Complementary rules to Ziegler

and Nichols' rules for a regulating and tracking controller. Interna-

tional Journal of Control, 49(5), 1465}1471.

McCormack, A. S., & Godfrey, K. (1998). Rule-based autotuning based

on frequency domain identi"cation. IEEE Transactions on Control

Systems Technology, 6(1), 43}61.

Pessen, D. W. (1994). A newlook at PID-controller tuning. Transactions

of the American Society of Mechanical Engineers, Journal of Dynamic

Systems, Measurement and Control, 116, 553}557.

Radke, F., & Isermann, R. (1987). A parameter-adaptive PID controller

with stepwise parameter optimisation. Automatica, 23, 449}457.

Seborg, D. E., Edgar, T. F., & Mellichamp, D. A. (1989). Process

dynamics and control. New York: Wiley.

Voda, A., & Landau, I. D. (1995). A method for the auto-calibration of

PID controllers. Automatica, 31(1), 41}53.

Wellstead, W. E. (1981). Nonparametric methods of systems identi"ca-

tion. Automatica, 17(1), 55}69.

Whidborne, J. F., & Liu, G. P. (1993). Critical control systems: Theory,

design and applications. New York: Research Studies Press Ltd and

Wiley.

Zhuang, M., & Atherton, D. P. (1993). Automatic tuning of optimum

PID controllers. IEE Proceedings-D, 140(3), 216}224.

Ziegler, J. G., & Nichols, N. B. (1942). Optimum settings for automatic

controllers. Transactions of ASME, 64, 759}768.

Zakian, V., & Al-Naib, U. (1973). Design of dynamical and control

systems by the method of inequalities. IEE Proceedings, 120(11),

1421}1427.

830 G.P. Liu, S. Daley / Control Engineering Practice 7 (1999) 821}830

¹ "0.5¹ .20K .1. crossover-frequency and steady-state error. someovershoot rule (SOR) (Seborg et al. S. no-overshoot rule (NOR) (Seborg et al. phase-margin. ¹ the time constant and the dead N time. Standard PID control system. ¹" . The phase chosen after the desired power distribution is determined has an important in#uence on the time-domain signal shape. 1981. and f are the amplitude. (3) I G I I where A . there are a number of PID tuning formula available. ¹ "0. 3. ¹ "0. 1993).60K . 1. symmetric optimum rule (SO) (Kessler. It can be seen from the table that the six PID design methods do not require a parametric transfer function model of the process and only need either one or two frequency response measurements of the process. 1994). Daley / Control Engineering Practice 7 (1999) 821}830 In order to address this problem a PID controller design based on direct optimisation in the frequencydomain is developed in this paper. B3[1. the maximal peak can be signi"cantly compressed. 1995). the closed-loop system with the PID controller is assumed to be of the structure shown in Fig. 2] is an acceleration factor. This controller is demonstrated through its application to a rotary hydraulic system and its performance evaluated through comparison with six other PID tuning rules. ¹ "0. Liu. which are the gain-margin. Voda Table 1 Six PID tuning rules PID rule Ziegler}Nichols Integral of absolute error Some-overshoot rule No-overshoot rule Integral of squared n time weighted error Symmetric optimum PID parameters K "0.051(3..4¹ . A summary of the six PID tuning rules are given in Table 1. 1998) P (j ) K G( j )" WP K . Optimal-tuning PID control The design of feedback control systems in industry using frequency-response methods is more popular than any other. 1989). A G B Based on the FOPDT model.509K . Six PID tuning rules are introduced and assessed in later sections. The PID controller is then designed to directly satisfy the requirements set for these functions. K P (j ) SP (4) 2..P. K and correspond to the gain of the system and frequency at which the system phase is !1353. 1#s¹ (1) where K is the gain. 1942).5¹ . This is primarily because the frequency response method provides good designs in the face of uncertainty in the plant model and can easily use experimental information for design purposes. (2) A B ¹ s G where K . ¹" G B B (4#B) A 8(2 K Fig. A set of frequencydomain performance functions are considered. McCormack & Godfrey. ¹ and ¹ are the PID parameters. the unbiased estimate of the frequency-response of the process using spectral analysis is given by (Wellstand. ¹ "0. integral of squared time weighted error (ISTWE) (Zhuang & Atherton.330¹ A S G S B S K "0. x(t)" A sin(2 f t# ) .70K . ¹ "0. integral of absolute error (IAE) (Pessen. Using a non-parametric model represented by the system frequency response.5¹ .822 G. In Table 1. . the design of the excitation signal is very important. These tuning rules are Ziegler}Nichols (NZ) (Ziegler & Nichols.150¹ A S G S B S K "0. It is assumed that the ideal transfer function of a PID controller is given by 1 K(s)"K 1# #¹ s . The transfer function of a FOPDT model is described by K e\QO G(s)" N . 1. The multi-sine signal is added to the reference r(t).125¹ B S 4#B B 4#B B K" . ¹ "0. Frequency response estimation To estimate the system frequency response. 3. During the identi"cation. 1958. enabling greater energy to be injected for the given input range of the measurement device.33K . and the system is kept in the linear working region. ¹ "0. For this system. a method is derived that can provide optimal PID parameters for di!erent working conditions. K and ¹ are the inverse of the system gain S S and frequency at which the phase is !1803. ¹ "0. The excitation signal utilised here is the commonly used multi-sine which has the following form: . & Landou. For example.125¹ A S G S B S K "0. Rule-based PID controller design A large number of industrial processes can be characterised by a "rst-order plant with dead time (FOPDT). phase and freI I I quency of the signal components. 1989).330¹ A S G S B S K "0.302K K #1)¹ A S G N S S ¹ "0.

Optimal PID controller design Although the six PID design methods give simple tuning rules for the controller parameters using either one or two measurement points of the system frequency response. If the averaging of every quantity is done in a recursive way. 1992. Whidborne & Liu. ¹ . the multi-sine signal is directly added to the control input of the process. 2. the system is also expected to meet a speed-ofresponse speci"cation like bandwidth. there are two quantities used to measure the stability margin of the system. ¹ )" . the performance criteria (14) can be satis"ed if max + (K . ∀k.2 PI(j ) K . 3. G( j )" WP K PI(j ) K SP where k!1 1 P I ( j )" K P I\ ( j )# RH( j )> ( j ). the following performance functions need to be considered during the design of a PID controller: (K . which is the amount by which the phase of the system exceeds !1803 when the system gain is unity. 3. would be a good measurement in the frequency domain for the system's speed of time response. 2. K WP WP I k k I k!1 1 P I ( j )" K P I\ ( j )# RH( j ). ( j ) K SP SP I k k I (7) closed-loop system. ¹ ))1. Liu. To overcome this disadvantage.2. are the normalised G A G B gain-margin. Thus.. the crossoverfrequency and the steady-state error.G. ¹ ). Clearly. 180#LK( j )G( j ) (K . (14) hold. A G B P + (10) (8) (9) "K( j )G( j )""1 .)1. . G A G B (14) If the above inequalities are met. P . S. Murray & Wright. for k"1. for i"1. There are a number of methods to solve the performance criteria problem (14). 4. (5) WP G G m G 1 K P ( j )" K RH( j ). Daley / Control Engineering Practice 7 (1999) 821}830 823 where 1 K P ( j )" K RH( j )> ( j ). Thus. Generally speaking. The gain and phase margins are also related to the damping of a system. ( j ) (6) SP G G m G with R ( j ).P. respectively. ¹ . 4. the larger the value of the magnitude on the low-frequency asymptote. their control performance may not satisfy the desired requirements. ¹ )" 2! . then. then the problem is solved. (11) (12) with P ( j )"0 and P ( j )"0. ¹ . This relationship is very useful in the design of suitable compensation. ¹ . which is the factor by which the gain is less than the neutral stability value.g. G + LK( j )G( j )"!1803 . which is the frequency at which the gain is unity. The estimation of the frequency response of the process is still the same as the above with R ( j )"1. A G B e "K( j )G( j )" QQ (13) where (K . Also. ¹ ). In addition to the stability of a design. ¹ . 2. moving boundaries algorithm) to "nd the admissible or feasible set of parameter vectors. But. After the above considerations. A G B 2 f B "K( j ) G( j )""1 . in practice. f and e . The method of inequalities (Zakian & Al-Naib. the control input and the output. For the open-loop K K WP SP identi"cation case. min G A G B )A 2G 2B G (15) Clearly. 3. In the frequency domain. the above minimises the worst case values of the performance functions. the phase-margin. the design problem is to "nd a PID controller to make Eq. Liu. the accuI racy of the open-loop estimation is better than the closed-loop case. (K . 1993) uses optimisation algorithms (e. crossover-frequency and the steady-state error functions with the desired values G . an optimal PID controller design is proposed in the frequency domain. 1981). the latter is more convenient for implementation than the former. two methods are brie#y introduced: the minimax optimisation method and the method of inequalities. the following + + B QQ performance criteria should be satis"ed: (K . Using the minimax optimisation method (Gill. The admissible set is de"ned as " 5 5 5 . phase-margin. respectively. One is the gain margin. it is assumed that the major requirements in the frequency domain are those on the gain-margin. (16) . ¹ )" A G B "K( j )G( j )" . 1973. the lower the steady-state errors will be for the 1 (K . i"1. ( j ) and > ( j ) representing the G G G Fourier coe$cients of the reference input. ¹ )" . The crossover frequency. "1 . The other is the phase margin. ¹ . ¹ . for which all the performance inequalities hold. Here.

the PID controller may cope with all operating-points of the system and the closed-loop system will have similar optimal-control performance. ¹ .. When the system's operating-point or dynamics change. Fig. which is representative of many industrial systems that utilise #uid power. (17) G A G B G A G B Similar algorithms for solving the problem de"ned in (14) now exist in standard libraries of optimisation software. ¹ . 3. Rotary hydraulic system The rotary hydraulic test rig. that is "+(K . To e!ectively assess the performance of the proposed tuning method the other six tuning rules which were introduced in Section 2 are also applied to the rig. ¹ ))1. 2. The optimal-tuning mechanism uses the process frequency response to "nd optimal parameters for the PID controller so that the desired system speci"cations are satis"ed. the optimisation toolbox for use with MATLAB (Grace. crossover frequency and steadystate error. for example. (K . 3. even if this is a highly robust controller. It mainly consists of four parts: frequency response estimation. the disadvantage of this strategy is that it needs slightly more computation to search for the optimal parameters. the new process frequency response is re-estimated by switching on the excitation signal. as shown in Fig. using this updated frequency-response. desired system speci"cations. In this way. Optimal-tuning PID control structure. This is a particularly opposite application of the method since hydraulic systems are often very conservatively tuned. the optimal-tuning PID control scheme shown in Fig. phase margin. Schematic of hydraulic rig. 4. But. 3. . The operating procedure of the optimal-tuning PID control is as follows. 1994). Daley / Control Engineering Practice 7 (1999) 821}830 where is the set of parameter vectors for which the ith G functional inequality is satis"ed. optimal-tuning mechanism and PID controller.824 G. Finally. The frequency response estimated using frequency-domain identi"cation methods provides a non-parametric model for the process. the tuning mechanism searches for the optimal parameters for the PID controller to satisfy the desired system speci"cations. Fig. it is not always possible to exercise control with a "xed parameter controller. 2 is proposed. due to the fact that the cost of getting the tuning wrong can be highly destructive and costly.1. ¹ ).3. the PID controller is set to the obtained optimal parameters. S. compared with a "xed parameter control. The desired system speci"cations includes a set of requirements in the frequency domain: gain margin. The motor is coupled by a rigid shaft to a hydraulic pump of the same displacement as the motor and a solenoid controlled relief valve is used to simulate variations in load. 4. 1987).P. Then. Optimal-tuning PID control scheme When a system has di!erent operating points with widely di!ering dynamic properties. For this case. Application to a rotary hydraulic system The optimal-tuning PID control scheme is applied to a rotary hydraulic test rig (Daley. comprises an electro-hydraulic servo control valve driving a "xed displacement hydraulic motor up to 8000 rpm with a maximum operating pressure of 21 MPa. Liu.

any unnecessary damage resulting from the wrong PID parameters can Fig. The external control loop includes the hydraulic system and a PID controller.P.2. a periodic multi-sine excitation signal. 4.3. With this system a powerful implementation strategy is possible. and large changes in load can be expected. Implementation of the optimal-tuning PID control of the hydraulic rig. This type of hydraulic system is typically applied to mixer drives. was directly applied to the hydraulic system in an open-loop way. Daley / Control Engineering Practice 7 (1999) 821}830 825 Fig.G. the rotary hydraulic system was operated in two working conditions: with load and without load. It mainly comprises an external control loop and an internal control loop. as shown in Fig. 4. The parameters of the two PID controllers are adjusted by the optimal-tuning PID algorithm. Multisine excitation signal. The internal control loop includes the adaptive model updated by an on-line system identi"cation algorithm and a PID controller. S. Case I: with load For the sake of simplicity. Implementation structure The optimal-tuning PID control scheme is implemented using the MathWorks Real-Time Workshop connected to a dSPACE DSP board based around the TMS320. be prevented through a slight increase in computational load. MATLAB and SIMULINK. 4. Based on the input}output data.3. 4. In this way. The second stage is to apply the PID controller to the rotary hydraulic system using the PID parameters that have been veri"ed as being safe on the adaptive model. 5. the implementation of PID control strategies consists of two stages. 5. Experimental results During the experiment. Liu. 4.1. Roughly speaking. as shown in Fig. the frequency response of the system . the experimental results for the two cases are presented. The "rst stage is to run a PID controller on the identi"ed on-line model to predict the performance of the PID controller before it is used on the rig. centrifuge drives and machine tool drives where accurate speed control with fast response times is required. So.

there was no load on the hydraulic motor. 7 and 8. The scheme mainly consists of four parts: frequency-response . 6.0914 0.0554 0.1939 0.0372 ¹ B 0.1141 0. In terms of the seven PID tuning rules. Table 2 PID parameters for case I PID rule Ziegler}Nichols Integral of squared time weighted error Integral of absolute error Symmetric optimum Some-overshoot rule No-overshoot rule Optimal design rule K A 0.0065 0. with the loading change being seen mainly as an apparent gain change. The estimated magnitude and phase of the system with respect to frequency are shown in Fig. Frequency-response of the hydraulic motor with load. their performance is reasonably good and for some practical cases may be considered acceptable. the parameters of the PID controllers for this case are given in Table 3.0485 0.0400 0. Liu. 9. 4. 5. Note that the symmetric optimum rule and the someovershoot rule give a signi"cant oscillatory response.0423 0. 6.0152 0. 10 with Fig.0351 was estimated using the frequency-domain identi"cation methods discussed in Section 3.0182 0.0606 0. When the PID parameters for Case I were still used. This means the process dynamics are di!erent. it is clear that the system frequency responses for these two cases are di!erent. The periodic multi-sine excitation signal. The performance of the seven PID controllers for this case are shown in Figs.3. The estimated frequency response of the load-free case is shown in Fig. as shown in Fig. Since the process dynamics have changed signi"cantly a more appropriate response would be to re-tune. S. 5. Following the six PID tuning rules in Section 2 and the optimal PID design rule proposed in this paper provides the parameters which are given in Table 2. The speed responses using the other four PID control rules are given in Fig. three PID controllers failed to stabilise the system. Daley / Control Engineering Practice 7 (1999) 821}830 Fig.2.P.826 G.0606 0. Though the no-overshoot rule and the optimal design rule do not give a speed response as good as the optimal design rule for Case I.0632 ¹ G 0. Comparing Fig.1268 0. 6.0400 0. integral of squared time weighted error and integral of absolute error rules. Case II: without load In this case. 10.0606 0. Conclusions In this paper an optimal PID controller design scheme based in the frequency domain has been presented.0152 0. 11 and 12. It is clear that the relative performance is similar to that displayed for Case I. The speed responses of the hydraulic motor using seven PID controllers are shown in Figs.1410 0.1662 0. was directly applied to the hydraulic system again. namely the Ziegler}Nichols.

P. Liu. a de"nition of desired system speci"cations. IAE and SO controllers (Case I). The frequency response estimated using frequency- domain identi"cation methods provides a non-parametric model for the process. an optimal-tuning mechanism and a PID controller. Fig.G. S. The speed of the hydraulic motor using NZ. The desired system speci"cations includes a set of requirements in the frequency . estimation. 8. The speed of the hydraulic motor using SOR. ISTWE. 7. NOR and ODR controllers (Case I). Daley / Control Engineering Practice 7 (1999) 821}830 827 Fig.

.0056 0.828 G. Fig. SO and ODR controllers for Case I.0385 domain: gain-margin.0556 0. S.0362 0.0337 0.0330 0. 9.0549 0.0858 0. The optimal-tuning mechanism uses the process frequency response to "nd optimal parameters for the PID controller so that the desired system speci"cations are satis"ed. and copes well with changes in the process dynamics. crossover-frequency and steady-state error.0439 0. The speed of the hydraulic motor without load using SOR.1180 0. phase-margin.0844 0.0549 0.0362 0. 10.0262 ¹ B 0.0549 0.0994 0. Daley / Control Engineering Practice 7 (1999) 821}830 Fig. NOR.0137 0.1012 0.P. Frequency response of the hydraulic motor without load. The experimental results have shown that this optimal-tuning PID controller can signi"cantly improve system performance. Liu.0137 0. Table 3 PID parameters for case II PID rule Ziegler}Nichols Integral of squared time weighted error Integral of absolute error Symmetric optimum Some-overshoot rule No-overshoot rule Optimal design rule K A 0. This scheme has been successfully applied to a rotary hydraulic system and has also been compared with another six auto-tuning PID control rules.0317 ¹ G 0.0165 0.

12. The speed of the hydraulic motor using SOR. IAE and SO controllers (Case II). Automatica. . 20(5). J. NOR and ODR controllers (Case II).P. Fig. & Hagglund. Automatic tuning simple regulators with speci"cations on phase and amplitude margins. ISTWE. Daley / Control Engineering Practice 7 (1999) 821}830 829 Fig. K. 11. Acknowledgements The authors are grateful to the management of the ALSTOM Energy Technology Centre for giving permission to publish this work. The speed of the hydraulic motor using NZ. Liu. 645}651. T.G.. S. (1984). References Astrom.

G. Automatic tuning of optimum PID controllers. Process dynamics and control. Whidborne.. Optimal PID tuning using direct search algorithm. C. J. 64. Self-tuning PID controllers based on a pattern recognition approach.K. D. (1989). Gill. (1973). A. & Daley. (1991). Pessen. Edgar. Liu.. P. Murray.. Hang. D.. & Atherton. & Landau. The MEC benchmark challenge on gasixer control. D. G. 116. G. I. 49(5). (1984). U..830 G. J. Grace. Regelungstetechnik. J. Ziegler. S. Nonparametric methods of systems identi"cation. T. & Al-Naib. & Liu. Critical control systems: Theory. Proceedings of the Institution of Mechanical Engineers. J. thesis. (1995). G. & Wright. New York: Research Studies Press Ltd and Wiley. R. Kessler. Design of dynamical and control systems by the method of inequalities. 41}53. . M.. Transactions of ASME. 285}295. S.P. A parameter-adaptive PID controller with stepwise parameter optimisation. A method for the auto-calibration of PID controllers. P. 31(1). Astrom. Zakian.. J. 120(11). R. IEE Proceedings-D. G. (1993). (1998). T. Automatica. E. & Mayron. Automatica. 140(3). IMechE Seminar. (1958). IEEE Transactions on Control Systems Technology. 55}69. 23.. Dixon. W. 106}111. 111}118. Multiobjective optimal-tuning PI controller design for a gasi"er. 43}61.. design and applications. (1989). B. H. Daley. P.. IEE Proceedings. E. K. Control Engineering. U. 553}557. A. IEE Proceedings-D. K. Automatica. & Liu. & Tacconi. 1465}1471. Rule-based autotuning based on frequency domain identi"cation. (1994). New York: Academic Press.. (1987). (1993).. Voda. (1998). (1981). London. 449}457. S. The MathWorks Inc. 201. S... K. Mantz. (1942). Complementary rules to Ziegler and Nichols' rules for a regulating and tracking controller. 6(1). Practical optimisation. Liu. Wellstead. & Nichols. 6(11). T. Seborg. M. Re"nements of the Ziegler}Nichols tuning formula. (1987). N. & Mellichamp. (1992).D. Daley. & Godfrey. A. E. Ph. & Ho. F. P. 138(2). Optimum settings for automatic controllers. A. R. (1998). IMechE Seminar. D. F. University of Manchester Institute of Science and Technology. (1981). Journal of Dynamic Systems. F. W. 17(1). 216}224. Coventry. Theory and design of critical control systems. New York: Wiley. Optimisation toolbox for use with MATLAB. D. Transactions of the American Society of Mechanical Engineers. Liu. 395}400. Control Systems Centre. W. S.. Zhuang. 1421}1427. E. Das symmetrische optimum. & Isermann. W... C. (1994). V. W. Tuning-in to increase proxt-developments in PID tuning. J. Application of a fast self-tuning control algorithm to a hydraulic test rig. Radke. International Journal of Control. A new look at PID-controller tuning. 759}768. C. P.. Measurement and Control. Daley / Control Engineering Practice 7 (1999) 821}830 McCormack. Kraus. P.

- 13_Modelling and Simulation -good.pdf
- Lecture 18
- roni-121121211857-phpapp02
- Furnace Temperature Control
- advanced process control10.5923.j.control.20130301.03
- v45-55
- Chap08
- k 1055628376
- Fixed Order Controllers
- wp_apc_statusReport
- ion for Pid Controller & Closed Loop Control
- Labo 1 Control Fuente
- ho1999.pdf
- Chapter 12
- C13-1-An-Li-ROBIO
- Injection Molding Machine
- Process Dynamics and Control
- MIMO Fuel Cell
- S5 Load Sharing Rev0
- DC Motor Control
- Advanced Control Basics
- 412
- Control Engineering 2
- HW5 2003 Solution
- Control System
- Chap 08 Marlin 2002
- Automatic Load Frequency Control of Two Area Power System With Conventional and Fuzzy Logic Control
- 06190539
- FANAEI
- adaslide4

- UT Dallas Syllabus for ee4310.001.11f taught by Nicholas Gans (nrg092000)
- tmp4D3F.tmp
- Tuning of PI Controller using Integral Performance Criteria for FOPTD System
- Position Control of Satellite In Geo-Stationary Orbit Using Sliding Mode Control Algorithm
- A Review on Model Based Methods to Degrade Nonlinearity Effects for Control Prosthetic Arm
- tmpEF5E.tmp
- UT Dallas Syllabus for ee4310.501.11f taught by Gerald Burnham (burnham)
- UT Dallas Syllabus for mech4310.001.11f taught by James Hilkert (jmh011500)
- UT Dallas Syllabus for ee4310.501.09f taught by Gerald Burnham (burnham)
- UT Dallas Syllabus for ee4310.001.11f taught by Nicholas Gans (nrg092000)
- UT Dallas Syllabus for opre7346.001 06s taught by Alain Bensoussan (axb046100)
- UT Dallas Syllabus for ee4310.501.11s taught by Charles Bernardin (cpb021000)
- UT Dallas Syllabus for ee4310.001.08s taught by Gerald Burnham (burnham)
- UT Dallas Syllabus for ee4310.002.08s taught by Charles Bernardin (cpb021000)
- UT Dallas Syllabus for mech4310.002.11s taught by James Hilkert (jmh011500)
- UT Dallas Syllabus for ee4310.001.09f taught by Charles Bernardin (cpb021000)
- tmp5C4F.tmp
- UT Dallas Syllabus for ee4310.021.07u taught by (jmh011500)
- UT Dallas Syllabus for ee4310.5u1.09u taught by (jmh011500)
- UT Dallas Syllabus for ee4310.0u1.10u taught by James Hilkert (jmh011500)
- UT Dallas Syllabus for ee4310.5u1.08u taught by (jmh011500)
- UT Dallas Syllabus for ee4310.501.11f taught by Gerald Burnham (burnham)
- Design And Fabrication Of A.C.V. -A Review
- UT Dallas Syllabus for ee4310.001.07f taught by Louis Hunt (hunt)
- UT Dallas Syllabus for ee4310.501.08f taught by (jmh011500)
- tmp67C0.tmp
- PSO based optimization of a PI controller for a Real time Pressure process
- UT Dallas Syllabus for ee4310.0u1.11u taught by Nicholas Gans (nrg092000)
- UT Dallas Syllabus for ee4310.001 06f taught by Gerald Burnham (burnham)
- UT Dallas Syllabus for ee4310.001.07s taught by Gerald Burnham (burnham)

- null
- Centurion Industries, Inc. And Eric F. Burtis v. Warren Steurer and Associates and Instructional Materials and Equipment Distributors, Cybernetic Systems, Inc., a New Mexico Corporation, Deponent-Appellant, 665 F.2d 323, 10th Cir. (1981)
- null
- Disabled People With Neuro Sky Mind Wave In Brain Computer Interface
- null
- null
- Employee Attendance Monitoring using Video Streaming Face Recognition
- tmp9AC7
- null
- Image Processing Techniques For Quality Checking In Food Industry
- AUTONOMOUS SURFACE CLEANING ROBOT EAMBEDED WITH AT89S52
- Comparison of Fuzzy C-Means and Hierarchical Agglomerative Clustering Algorithms for Data Mining

Sign up to vote on this title

UsefulNot usefulClose Dialog## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

Close Dialog## This title now requires a credit

Use one of your book credits to continue reading from where you left off, or restart the preview.

Loading