Proceedings of the 2010 International Conference on

Modelling, Identification and Control, Okayama, Japan, July 17-19,2010

Stabilizing Parameters of PID Controllers for Second-Order Systems

with Time- Delay
De-Jin Wang, Jiang-Hui Zhang

Abstract-This article focuses on the stabilization, using PID analytical approaches, including the Hermite-Biehler
controllers, of second-order systems with a pair of complex theorem [12], [13], linear programming [13], and
poles plus time-delay. The parameters constraint of the D-decomposition approach [14].
controlled plant for the existence of a stabilizing PID controller In the present article, we focus our attention on the
is first provided. And the stabilizing range of proportional-gain
second-order time-delay systems with a pair of complex poles,
is derived in terms of a version of Hermite-Biehler Theorem
which is different from our previous consideration in [10],
applicable to quasi- polynomial. Then, based on a graphical
where only real poles of the model are presented. The
stability criterion for time-delay systems, the stabilizing region
in integral-derivative space is drawn and identified, and an
importance of our work lies in (1) many processes in control
algorithm for finding the stabilizing parameter set of PID engineering can be approximated by the second-order model
controller is also developed. Finally, examples are given to plus time-delay, especially, the higher-order systems
illustrate the design steps and the shapes of stabilizing regions. possessing a pair of dominant complex poles. (2) Our
graphical design procedure is simple and direct in
I. INTRODUCTION determining the stabilizing regions of PIO controller,
avoiding the complicated mathematical calculation of the
IO controllers are widely used in industrial process
P control fields because of their simple structure and control
analytical methods.

stability in many practical processes [I], [2]. Traditional

II. RANGE OF PROPORTIONAL-GAIN
investigation of PIO controller is focused on the parameters
tuning, e.g., the well-known Ziegler-Nichols rules [3] for the The SISO feedback configuration considered in this article
processes with S-shape reaction curve [4]. Recent research is depicted in Fig.l, and G(s) and C(s) represent the
trend of PIO controller has turned to the determination of all
transfer functions of the controlled process and the controller,
stabilizing parameters since the authors in [5] investigated the respectively, which are given by
first-order plants with time-delay using a version of
[6] co2n
Hermite-Biehler Theorem derived by Pontrayagin G(s) =
2 e-Ls (1)
applicable to quasi-polynomial. Using the classical Nyquist S + 2scon s + co;
[7] obtained the same results
stability criterion, the author in
K'
as those in [5]. [5] was generalized to the
The method used in C(s) = Kp + - + KdS (2)
second-order integrating processes with time-delay [8]. For s

�
the first-order plus dead-time unstable processes [9],
r(t) ....-----.
- y( t)

T
D-partition technique was applied to characterize the stability
domain in the space of the process parameters and controller
C(s) 1-1 �·I G (s)
parameters, respectively. The author in [10] studied the
second-order time-delay plants with two real time constants
Fig. l. Unity feedback system
by means of a graphical approach to draw the stability region
of PIO controllers. In [II], the stability region in the process
parameter plane was detailed for a second-order time-delay
where S is the damping ratio of the second-order
plant whose transfer function has one zero. In more recent
works, several authors investigated the arbitrary order process, con the undamped natural frequency, L the
time-delay plants controlled by PIO controllers by different
time-delay , and K p' Kj and Kd are the gain-parameters

of PIO controller. It is known that when -1 < S <1,

Manuscript received December 20, 2009. This work was supported in part
by the National Nature Science Foundation of PR China under Grant G( s) has a pair of complex poles in s-plane with negative
60874028.
De-lin Wang is with the School of Electronic Information and Automation, (0 < S < 1 ) or positive ( -1 < S < 0) real part, and when
Tianjin University of Science and Technology, Tianjin, 300222, PR China,
(phone: 86-22-60273295, email: [email protected]).
S ::; -lor S ;::: 1, G( s) has tow real poles in the left or
liang-Hui Zhang is with the School of Electronic Information and tight half s-plane.
Automation, Tianjin University of Science and Technology, Tianjin, 300222,
PR China(e-mail: [email protected]).

978-0-9555293-3-7/10/$25.00 112
The problem of interest in the present article is to Proof Along the similar line as in [10] (Lemma 1 and
determine the stabilizing gain-parameter ranges of Remark2 therein, and letting (J' � 0) the condition (10) can
(Kp'K; , Kd) such that closed-loop system shown in Fig. 1 be derived based on the graphical observation.
Remark 1 From condition (8), it is clear that (10) is
is stable.
equivalent to
To this end, the closed-loop characteristic quasi-
polynomial is first computed as 4Lt;m n+L2m; + 2 >0 (11)

Ll(s) = S(S2+ 2t;OJns+OJ;) Kp + 1 >0 (12)

(3)
+OJ;(K; +K s+Kds2)e-Ls 0 The inequality (11) exhibits the parameters constraint of the
p =

second-order plant (1) which can be stabilized by a PID

Multiplying both sides of (3) by eLs, one gets
controller.
Il:(s) = S(S2 + 2t;mn s + m;)eLS Now, we are ready to derive the stabilizing range of Kp
(4)
+m;(K; +Kps+KdS2) =0 values.

Substituting s = jm into (4) gives rise to Theorem 1 The necessary condition for the range of Kp
*
I::,. (jm) = jm(_012 + j 2t;mnm + 01;)eiLm values for which the plant, with transfer function G(s) as in
(1), can be stabilized by a PID controller is that if the
+m;(K; + jKpm-Kdm2)=0 parameters constraint (11) is satisfied and the equation
*
Separate 1::,. (j 01) into its real and imaginary parts, one has 2(1 + Lt;m n )
tana = a (1 3)
*
1::,. (jm) = I::,.r (01) + jl::,.; (01) a2 _ L2m2n -2 L?'m0:,' n

where has a solution al in the interval (0,7Z") , then

I::,.r(m) =m;K; -m[m;Kdm + 2t;mnmcos(Lm) 2(; . al2
(5) -1<Kp <Kpu =
: --al smal + (- - -1)cosal
+(01; -m2)sin(Lm)] Lmn L2mn2
Ll;(OJ) OJ[OJ;K -2t;OJnOJsin(LOJ) (14)
p
=

(6) For Kp values outside this range, there are no stabilizing

+ (OJ; -OJ2)cos(LOJ)]
PID controllers. In the following, we give two propositions to
Next, we consider the delay-free stability condition for the
be used to prove Theorem 1.
purpose of giving the stabilizing range of Kp value because
Proposition 1 Let I::,. (z) = P(z,eZ), where P(z,w) is a
a minimal requirement for any delayed system is that the
delay-free closed-loop characteristic polynomial be stable.
polynomial with a principal term. Suppose I::,. ( jy), y E R is
When there exists no delay factor, the closed-loop separated into its real and imaginary parts:
characteristic quasi-polynomial (3) reduces to I::,. (jy) = I::,.r ( Y) + jl::,.; (y). If all zeros of I::,. (z) have

S3 +mn(2t; +mn Kd ) S2 + 01;(1 + Kp )s +m;K; = 0 negative real parts, then the zeros of I::,.r (y) and 1::,.; (y) are
Using the Hurwitz Stability Criterion yields the following real and simple.
stability conditions Proposition 2 Let p and q denote the highest powers of z and
K; >0 (7) eZ, respectively, in I::,. (z) . Let & be an appropriate constant
K p >-1 (8) such that the coefficients of terms of the highest degree in

2t; I::,.r (y) and 1::,.; (y) do not vanish at y=& . Then, for the
Kd >-- (9)
mn equation I::,.r (y) = 0 or 1::,.; (y) = 0 to have real roots, it is
Before giving the main result of the stabilizing range of necessary and sufficient that in the
proportional-gain, the following lemma can be proved, which intervals -2hZ" + & ::::;; y ::::;; 217l" + &, I = 1,2" .., I::,.r (y)
gives the parameters constraint for the existence of a
or 1::,.; (y) has exactly N!!. = 41q + p real roots starting
stabilizing PID controller.
Lemma 1 If the following condition holds with a sufficiently large I.
4Lt;m n + L2m; + 2 Proof of Theorem 1 Obviously, from Lemma 1 and Remark
O< <00 1, the parameters constraint (11) of the controlled plant (1)
2mn2(K p +l) should be satisfied for the existence of a stabilizing PID
(10)
controller.
Then, there exists a stabilizing PID controller for the
second-order plant (1).

113
With the change of variables Z = Lm, �J m) in (6) I > 0, �i(Z) has exactly 41 + 3 roots. Hence, from

can be expressed as Proposition 2, it follows that for -1 < K p < Kpu' �i(z)
2;mn Z2
�i(Z) =z[ m;Kp ___ zsinz+( m; --)cos z]/ L has only real roots.
L L Case 2. Kp � Kp u. For K p =Kpu' the two curves are
(15)
tangent to each other as shown in Fig. 2(b) 0 The roots
To claim (14), we first prove the assertion that the

necessary and sufficient condition for �i(z) to have only

distribution contradicts Proposition 2. For K p > Kpu' the

real and simple roots is given by (14). root Zl in Fig. 2(b) will vanish, which contradicts
Defming the change of variable as z = Ls in (4), one has Proposition 2, too. The combination of Case 1 and Case 2

that the highest power of z and eZ are p=3, and q=l, shows that the upper bound of Kp values is given by K pu.
respectively. From Proposition 1, the zeros of �i(z) must

be real and simple for the requirement of the existence of a

stabilizing controller. Next, from Proposition 2, for

I sufficiently large, it follows that �i(z) =0 should have

exactly41q + P = 41 + 3 real roots in the intervals

-211l + & ::; z ::; 211l + & . It is noted that the roots of

�i(z) =0 are simple (see Fig. 2).

Now, let �i(Z) =0, according to (15), one arrives at

either z=O or

2;m n Z2
m; K p - __ Zsin z +(m; -- ) cos Z=0 (16)
L L2 3
Z
4

i.e., we have one root at the origin and other roots are given
Fig. 2. Plots of the curves on both sides of(l7) for different Kp
by equation (16). The equation (16) is a nonlinear one, a
graphical method is employed in the following to get the roots
Next, one proves that the upper bound of Kp values is
distribution. From (16 ), one has
given by (14). From Fig. 2(b), it follows that the two curves
Kp 2; . 1 Z2
-=--sm z+-( -I)cos z -- (17) are tangent to each other at al =Zl in the interval (0,1l),
z Lm n z L2m2n i.e., one has
d
and the plots of the functions on both sides of (17) are shown
- Kp
in Fig. 2 for different Kp values, where the intersection, [-L=a,
d
Z Z
denoted by zj ,j = 1,2,3 , arranged in increasing order of d
2; . 1 Z2
magnitude, of the two curves represent the positive real roots =-[--sm Z+-( -1) cos zL a ,
-
--

of (17).
dz Lmn z L2m2n
Now, one chooses & =0 to satisfy the requirement that

the coefficients of terms of highest degree in �i(z) do not

vanish at Z = & = 0 , i.e., cos(&) # 0 . In the following, we (18)

consider two different cases for Kp values. a2I L2m2n·
smal
_

Case 1. -1 < Kp < Kpu . In this case, the two curves are L2mn2al

shown in Fig. 2(a). one observes that in the interval [0,21l] , On the other hand, from (17), for z =aI E (0,1l), one

has
�i(z) has 4 roots, including the one at origin. Since the two
Kp 2; . 1 al2
functions on both sides of (17) are odd functions, one -=--smaI +-( -I)cosaI -- (19)
concludes that in the interval [-21l,21l] , �i(z) has
al L mn al L2m n2

41 + 3 = 4 x 1 + 3 = 7 roots. Further observation shows Eliminating Kp between equations (18) and (19) concludes

that, in the interval [-211l,211l] , for any positive integer that al is the solution of the equation (13) in the interval

114
(0, Jl') . Once al is determined, the upper bound K pu in (14) B=-20"y('mn -0")+ y(m; - l +0"2 -2 'mn O")
can be obtained using (19). Thus, we have proven the Suppose that, for fixed Kp and 0" , one has found, in one
assertion that the necessary and sufficient condition for

IJ.j(z) to have only real and simple roots is given by (14).

way or another, a point (y, K? , K�, Kp'0") on the

Finally, we prove the conclusion of the Theorem. From imaginary axis, x=O, such that

Proposition 1, the necessary condition for the existence of a IJ. r(y,KjO ,K�,K p ,O") = 0
stabilizing PIO controller is that the zeros of IJ.j(z) should (24)
IJ.j(y,KjO ,K�,K p ,O") =0
be real and simple, which is equivalent to (14) according to
i.e., there is a root on the imaginary axis in z-plane. Then,
the above assertion. This completes the proof the Theorem.
according to the Implicit Function Theorem, if the Jacobi
Remark 2 Theorem 1 only gives a necessary condition for
Matrix
the stabilizing range of Kp values. It can be shown, using
8IJ.r 8IJ.r
the algorithm developed in the next section, that when Kp
8Kj 8Kd
values are inside the range given by (19) or (22), the (25)
8IJ.j 8IJ.j
stabilizing region in the (K pK d) -plane would be
8KI 8Kd (y,K;0 ,Kd,K
0
p'u)
non-empty.
is nonsingular, the equation (24) has a unique local solution
III. GRAPHICAL STABILITY CRITERION
curve ( Kj(y), K d (y) ), and the following Proposition
In this section, the plotting of the stabilizing region in
holds [15].
( Kj ,K d)-plane are discussed for fixed Kp values in the Proposition 3 The critical roots are in the right-half plane in
the parameter space to the left of the curve
stabilizing range given by Theorem 1, using a graphical
stability criterion applicable to time-delay systems. To this (Kj(y), K d (y)) when we follow this curve in the
end, defme the change of varlables.
direction of increasing y , whenever detJ < 0 and to the
s=z -O",z=x+ jy,O" >0 (20)
right when detJ >0 . Here J is the Jacobi matrix defined in
i.e., one shifts the imaginary axis in s-plane to the vertical (25).
line z = 0" in z-plane, and substitute (20) into (4), the From (22). (23)and (25), one obtains
following characteristic quasi-polynomial in z is obtained
detJ =-2m:oy < 0,VO" >O,y >0 (26)

IJ. *(z,KpKd,Kp'O") =(z -O")[(z -O"i Then, solving equations (24) for Kj and Kd in terms of

+ 2 'mn(z-0") +m;]eL(z-u) (21)

K p'0" and y, one gets

+m;[Kj +Kp(z-O")+KAz-O"i] =0 Kd = !
2mn O"y
[Asin(Ly) + Bcos(Ly)
Let z = jy , i.e., along the imaginary axis in z- plane, and

partition the corresponding IJ. *(jy, Kj , K d , Kp , 0") into

+ Kpm;y]
(27)
its real and imaginary components yields

IJ.*(jy,KpKd,Kp'O") =IJ.r(y, Kj,Kd,Kp'O")

�
Kj=- [Acos(Ly) -Bsin(Ly)
mn
+ jIJ.j(y,KpKd,Kp'O") -Kpm;O" + Kdm;(0"2 _ y2)]
where As the coefficients of the characteristic quasi-polynomial

IJ.r (y,K j ,Kd,Kp'O") =e-LU[Acos(Ly) (21) and L are real, if jy is a root of (21), so is the complex

conjugate of it. Hence, it is sufficient to consider

- Bsin(Ly)] + Kjm; (22)
y E[0,+(0).
-Kpm20"+
n Kdm2(0"2
n _ y2) When y = O(x = 0), (21) gives a linear relationship

IJ.j(y,Kj, Kd, Kp'O") =e-LU[Asin(Ly) between Kj and Kd•

(23)
+ Bcos(Ly)] + Kpm;y-2Kdm;O"y
A=-O"(m; - l +0"2 -2 'mn O") -2 l('mn -0")

115
aJ
K d =- � Kj +![aJ;(Kp + 1) (a)

.
0=0.5

i. .
a a (28)

� 1
� 1 ---+- --------- -------r----------------------:---------------------
+a(a -2saJn )] o - - �- -.
------------------ - ------------------ - _--------------------

which gives a piece of stability boundaries. When

2 --·-
-1� � --------------------
� �------ ----------·� --------�
· ---------
0 5 10 15
(b)
y E (0,+00), expression (27) detennines other pieces of
0=0.2

stability boundaries, and the stabilizing region can be

identified in (Kj, K d) - plane using proposition 3 as shown

by examples in the next section. (c) 0=0.0001

';�l
IV. STABILITY REGIONS

In this section, an algorithm is first proposed for

o 5 10 15
detennining the stabilizing parameter set of PID controllers. ki

Then, an illustrating example is given to shown the design

procedure and the shapes of the stabilizing regions. Fig. 3. Stabilizing regions for different a > 0
Algorithm 1 For detennining the stabilizing parameter set of
PID controllers.
Step 1. For a given second-order plant (1), check if the - -
-
-
I
- -
constraint (11) is satisfied. If so, go to next step. Otherwise, - - .J
6 J..-
I
- _ - -- I
the given plant (1) can not be stabilized by PID controller. 5 -

Step 2. Calculate the bound of Kp values according to

Theorem 1.

Step 3. Select a Kp value in the range given by (14), and

draw the stability boundary line in (Kj, K d) - plane using

the linear relationship (28) for small enough a > 0
Step 4. For the same K p and a values as in Step 3, draw
kd ki

the curve given by (27) for big enough y E (0,+00), and

�
Fig. 4. 3-dimentional stabilizing regions for = 0.5
identify stabilizing region based on Proposition 3. For more
details, see the following example. 7"
Example 1 Consider the second-order plant given by (1),
_ - -
.....

L 0. 2,S 0.5,aJn 2, corresponding to a -

and take = = = 1 - -- - :
_ -­
-1-
I

pair of stable poles. For this set of parameters of the plant, it is --

0.5
computed that the parameters constraint in (11) is
JZ
4LSaJn + L2aJ; + 2 = 2.96 > 0 0

-0.5
Hence, the stabilizing PID controller exists. According to

Theorem 1, using (13) and (14), one has a) = 1 . 25 -1 -

2

andK pu = 5.7 3 . Pick up K p = 1 .00, for different a > 0, 1.5

15
10

the stabilizing boundary curves in (Kj , K d)-plane are drawn kd 0.5 0

ki
as shown in Fig. 3. The arrows along the curve represent the
direction of increasing y, and according to the sign of Fig_ 5. 3-dimentional stabilizing regions for � = -0.1

detJ in (26), and Proposition 3, the stabilizing region are -1 < Kp < 5.73 , 3-dirnensional stabilizing regions are
identified as the filled areas. Obviously, as a decreases, the
drawn as shown in Fig.4.
stabilizing region increases, and for a sufficiently
small a > O,a = 0.000 1, the complete stabilizing region
Next, take L = 0.2,'; = -O.I,aJn = 2 , corresponding

to a pair of unstable poles. In this case, it is computed that

is recovered as shown in Fig. 3(c) by the filled triangle. In this

way, and by gridding Kp values in the range 4LSaJn + L2aJ; + 2 = 1 . 36 > 0

116
i.e., the stabilizing PID controller exists. The 3-dimensional
stabilizing regions are plotted as shown in Fig.5. In this case,
the stabilizing regions become smaller than the case of
;=0.5.
V. CONCLUSIONS

In this article, the problem of determining all stabilizing

parameters of PID controllers has been discussed for
second-order systems with a pair of complex poles plus
time-delay. The parameters constraint of this kind of plant has
been derived, which guarantees the existence of stabilizing
PID controllers. Then, the stabilizing range of the 3
gain-parameters of PID controller has been determined using
a version of Hermite-Biehler Theorem and a graphical
stability criterion fit for quasi-polynomial. The graphical
method adopted in the article is simple and direct in plotting
the stabilizing regions, avoiding the complicated
mathematical calculations of the stabilizing boundary lines.

REFERENCES
[I] K. Astrom, and T. Hagglund, PID controllers: Theory, Design, and
Tuning. Research Triangle Park, NC: Instrument Society of American ,
1995
[2] G. J. Silva, A. Datta, and S. P. Bhattacharyya, PID Controllers for
Time-Delay Systems. Berlin, Germany: Birkhauser, 2004.
[3] lG. Ziegler, and N. B. Nichols, Optimum setting for automatic
controller. Trans. ASME, vo1.64, pp.759-768, 1942
[4] G .. F. Franklin, J.D. Powell, and A. Emami- Naeini,Yeedback Control
of Dynamic Systems. 3m Edition, New York: Addison-Wesley
Publishing Company, 1995
[5] G. J. Silva, A. Datta, and S.P. Bhattacharyya, New results on the
synthesis of PID controllers. IEEE Tran. Autom. Control, vo1.47, no.2,
pp. 241-252, 2002.
[6] I. S. Pontryagin, On the zeros of some elementary transcendental
function. Amer. Math. Soc. Transl. vol.2, pp. 95-110, 1955.
[7] G. Martelli, Comments on: "New results on the synthesis of PID
controllers. " IEEE Trans. Autom. Control, vo1.50, no.9, pp. 1468-1469,
2005.
[8] L. Ou, W. Zhang, and D. Gu, Sets of stabilizing PID controllers for
second-order Integrating process with time-delay. lEE Pro. Control
Theory & Applications, vo1. l53, no.5, pp. 607-614, 2006.
[9] C. Hwang and lH. Hwang, "Stabilization of fust-order plus dead-time
unstable processes using PID controllers," lEE Proc. Control Theory &
Applications, vo1. l51, no. I, pp.89-94, 2004.
[lO] D. -J. Wang, "Further results on the synthesis of PID controllers. " IEEE
Trans. Autom. Control, vol. 52, no. 6, pp. 1127-1132, 2007.
[II] G. Martelli, Stability of PID-controlled second-order time-delay
feedback systems. Automatica, vo1.45, no. !I, pp. 2718-2722, 2009.
[12] L. Ou, W. lhang, and L'yu, Low-order stabiliziation of LTI systems
with time delay. IEEE Trans. Autom. Control, vo1.54, no.4, pp.774-787,
2009.
[13] V. A. Oliveira, L. V. Cossi, M. C. M. Teixeira and A. M.F. Silva,
Synthesis of PID controllers for a class of time delay systems,
Automatica, vo1.45, no.7, pp. !778-1782, 2009.
[14] N. Hohenbichler, All stabilizing PID controllers for time delay systems,
Automatica, vo1.45, no. !I, pp. 2678-2684, 2009.
[15] O. Diekmann, S. A. van Gils, S. M. Verduyn Lunel and H. -O. Walther,
Delay Equations: Functional-, Complex-, and Nonlinear Analysis,
Applied Mathematical Sciences, Springer-Verlag, 1995.

117