Design of PID Controllers
by the Maximum Stability Degree Criterion
Ion Fiodorov, Bartolomeu Izvoreanu, Irina Cojuhari
Automatics and Informational Technologies Department
Technical University of Moldova
Chisinau, Republic of Moldova
fiodorov_ion@yahoo.com; izvor@mail.utm.md; cojuhari_irina@mail.utm.md
Abstract - The new approach for determining and imposing the
maximum stability degree of the system in function of the control
object parameters and the number of tuning parameters of PID
controller is proposed in this work. The elaborated method of
synthesizing the typical controllers represents a simple analytical
procedure, that requires the reduced volume of calculation and
not impose the restrictions on the complexity of control object.
The control systems synthesized by this criterion can be
characterized by the high speed, a low overshot and good
robustness.
Keywords - control object; control algorithm; the PID
controller; design of controllers; automatic control system; the
maximum stability degree
I. INTRODUCTION
One of the most important problem in the automatic
control, from the theoretical and the practical point, is the
problem to synthesis of controllers. The synthesis of
controllers involves the designing or the choosing a control
algorithm (the structure of controller) and determination its
optimal parameters, such as to be achieved and maintained the
objective of controlling and satisfied the performances
imposed to the automatic control system as: stability, dynamic
and static properties, sensitivity, disturbance rejection,
robustness at the parameter object variation etc. [1, 2].
The practice of the automation of the technological
processes demonstrates that about 90-95% of the controllers
actually used in the industrial installations use the controllers
that functioned on the basis of the PID algorithm, and its
variations. This is explained by the fact that in the limits of the
automatic control structure, the PID control algorithm is
optimal, if we refer to the specific functions of the controller.
The failure function of the automatic control system can be
appeared not from reason of non-satisfactory action of PID
algorithm (with condition that it is tuned optimal), but from
the reason of uncertainty information about the control object
and the degree of approximation of the mathematical model of
process [3, 4].
It should be noted that the use of PID control algorithm
does not guarantee priori the optimal control of the system and
either its stability. And if the controller is tuned incorrectly,
then the performance of the automatic control system
decrease, which leads, for example to reduction in the quality
of the output final production, low energy efficiency and as a
result increases in operating costs and harmful emissions into
the environment [4]. The providing of the high performances
depends on the choosing the corresponding values of the
controller tuning parameters. The determination of tuning
parameters, in dependency of the desired performances,
represents the complex and difficult procedure, for which
there is no single approach and that continues to be one of the
most actually problem in the automatic control [1, 2, 4].
One of the criteria used to design the automatic control
systems is the maximum stability degree criterion. This
criterion assumes the maximum displacement in front of the
imaginary axe of the localization domain in the complex plan
of the nearest roots. This fact involves the reducing of the
duration of the transient response and ensures the high
stability degree of the system, which leads to the satisfactory
performances of the control system even in the most
unfavorable variations of the object parameters from their
nominal values [5, 6, 9].
In this work is proposed the new approach for determining
and imposing the maximum stability degree of the system in
function of the control object parameters and number of tuning
parameters of the controller.
II. DESIGN OF THE PID CONTROLLERS FOR THE CONTROL
SYSTEMS WITH MAXIMUM STABILITY DEGREE
A. Formulation and solving of the problem to synthesis the
control system with maximum stability degree
The problem of synthesis the control system with
maximum stability degree can be formulated as follows. Is
considered that is given the control system (Fig.1) that
includes the control object with known parameters
k exp( −τ s )
H F (s) = , (1)
a 0 s r + a1 s r −1 + ... + a r −1 s + a r
and selected controller
m
H R (s) = ∑b s
j =1
j
( j −1) , (2)
where k is the transfer coefficient of the control object; d1
J = ηm = , (4)
τ - time delay; ai , i = 0, ..., r - the model object parameters; nd 0
r - object order; b j - the tuning parameters of the controller;
when the each real parts of the all roots of the characteristic
m - the number of tuning parameters. equation are equal α1 = α 2 = ... = α n = J .
From Affirmation 1 can be concluded that by choosing the
r(t) e(t) u(t) y(t) tuning parameters b j ( j = 1,..., m) of the controller it can be
HR(s) HPF(s)
- created the situation when the real parts of roots from
characteristic equation are equal with each other, then the
calculated parameters b j will be optimal in sense of ensure
Figure 1. The structure scheme of control system. the maximum stability degree of the designed system.
For estimation of the tuning parameters b j ( j = 1,..., m) it
The controller (2) represents the general form of PID
controller and its variation, which is characterized by the is formulated the following affirmation.
following transfer function Affirmation 2. If the n order characteristic equation (3) of
the stable system, where the number of tuning parameters
ki ki + k p s + k d s 2 m ≤ (n − 1) , and characteristic equation transcribed by its
H R ( s) = k p + + kd s = ,
s s decomposition in n linear factors with real parts of the roots
α1 = α 2 = ... = α n = J , represent the characteristic equations
where k p , k i , k d are the tuning parameters of the controller. of the same system (1), (2), then the optimal values of tuning
The characteristic equation of the system (1), (2) is parameters b j ( j = 1, ..., m), in the sense of ensure the
presented in the following form maximum stability degree J , are determined by equalizing
the coefficients of these equations with indexes
n m
i = ( n, ( n − 1), ..., ( n − ( m − 1)))
A( p, b j ) = ∑ i =0
ai p n−i + k ∑b p
j =1
j
j −1 = d0 pn +
(3)
+ d1 p n −1 + ... + d n −m pm + ... + d n−1 p + d n = 0, d i (k , ai , b j ) = qi (a0 , J , ω k ),
(5)
i = (n, (n − 1), ..., (n − (m − 1))).
where n is the order of designed system; d i = f i (ai ) , for the
Demonstration. The roots of the characteristic equation are
i = 0,..., (n − m); d i = f i (k , ai , b j ), for the the pairs of conjugate complex roots
i = ((n − (m − 1)),..., n) . For simplicity is considered initially
that time delay τ = 0 . pi , i +1 = −α i , i +1 ± jωk , (6)
It is necessary to determine a vector of tuning parameters α i > 0, i = 1, 2, ..., n; ωk ≥ 0, k = 1, ..., l
b j of the controller (2), so as for the known parameters of the
object (1) to be ensured the maximum value of the stability where α i is the real part, ωk - imaginary part, k - the order
degree of the system (1), (2)
number of the conjugate complex roots’ pairs, l - the number
of conjugate complex roots’ pairs.
η m = maxη (b j ) j = 1, ..., m ,
According to the Affirmation 1 it is introduced the notion
of the maximum stability degree J , and in the roots (6) of the
where η m is the maximum stability degree (MSD); η - the characteristic equation are substituted α i = J , i = 1, ..., n
stability degree of control system.
The number of tuning parameters m of the controller (2) pi = − J ± jω k , k = 1, ..., n / 2 , (7,a)
are chosen from the consideration of satisfaction the Hurwitz
if n is the even number,
stability conditions and it must be satisfied the condition
m ≤ n − 1 [5]. pi = − J ± jω k , k = 1, ..., (n − 1) / 2 , pn = − J , (7,b)
On the basis of solving the formulated problem is put the if n is the odd number.
following affirmation.
Affirmation 1. [6] If the coefficients d 0 and d1 from the For imposing to the system the maximum stability degree,
characteristic equation (3) are fixed, the stability degree η of taking into account the roots pi , the characteristic equation
the control system will obtain the maximum possible value, (3) is transcribed by its decomposition in n linear factors
equal with
 A( p ) = d 0 ( p − p1 )( p − p 2 ) × ... × ( p − p n ) =
l z
= d0 ∏[( p − jω
k =1
k + J )( p + jω k + J )] ∏( p + J) =
i =1
(8)
= q0 p n + q1 p n−1 + ... + q n −m p m + ... + q n−1 p + q n = 0,
where l is the number of the conjugate complex roots’ pairs;
z - the number of real roots; n = 2l + z - the degree of the
characteristic equation of the designed control system;
q0 = d 0 = a0 ; qi = f i (a0 , J , ωk ), i = (1,..., n) .
The characteristic equation’s roots (8), location in the
complex plane are presented in the Fig. 2.
synthesized control system.
The Affirmation 2 is demonstrated.
It is noted, that the tuning parameters b j from (9), the
expressions for its calculation are obtained by the proposed
method, are dependent of the known control object parameters
(1) and of the variable J that represents the stability degree of
the system. To determinate the value of the maximum stability
degree it is formulated the Affirmation 3.
Affirmation 3. If the stable system (1), (2) is described by
n order characteristic equation (3), where the tuning
parameters number are m ≤ (n − 1) , then the maximum
stability degree η max (b j ) is determinate by the geometric
mean root of the algebraic equation An −m ( p) = An( m) ( p) = 0 ,
equals with

d n −m
J = n −m , (10)
c n− m d 0
n(n − 1)(n − 2)...(n − m + 1)
cn −m = cnm = ,
m!

where An − m ( p) = An( m) ( p) = 0 represents m order derivative

Figure 2. The characteristic equation’s roots,
location in the complex plane.
Assume that the equations (3) and (8) satisfy the
conditions m ≤ (n − 1), α1 = α 2 = ... = α n = J . If the
respectively equations represent the characteristic equations of
the same control system then they are equivalent. From this
reason the coefficients attached to the variables with
respectively order from both equations are equal to each other.
Coefficients from the characteristic equation (3), with
indexes i = (n, (n − 1), ..., (n − (m − 1))) include the tuning
parameters b j ( j = 1, ..., m) of the respectively controller.
Therefore, equaling the coefficients from equation (3) with
coefficients with the same indexes from the equation (8)
d i (k , ai , b j ) = qi (a 0 , J , ω k ),
i = (n, (n − 1), ..., (n − (m − 1))),
of the characteristic equation (3). This geometric mean root is
obtained by equalizing the characteristic equation’s
coefficients (3), (8) with indexes i = (n − m)
d n − m (ai ) = qn − m (a0 , J , ωk ), ω k = 0 . (11)
Demonstration. Initial is considered the case when the
characteristic equation (3) of the control system (1), (2) satisfy
the condition of ensure the maximum stability degree and has
n real roots pi = −α i = − J , i = (1, ..., n) .
In [5] is stated that if the n order characteristic equation
(3) of the designed system (1), (2) satisfy the Hurwitz stability
conditions and the number of tuning parameters m in the
tuning algorithm is chosen so that m ≤ (n − 1) , then its m
order derivative
An−m ( p) = An( m) ( p) = [n(n − 1)(n − 2)...(n − m + 1)]d 0 p n−m +
+ [(n − 1)(n − 2)...(n − m)]d1 p n−( m+1) + ... (12)
+ (m + 1)!d n−m−1 p + m!d n−m = 0

and after some transformation, are obtained the expressions
for the calculation of the tuning parameters b j of the
respectively tuning algorithm
b j = f j (k , a0 , ai , J , ω k ), j = (1,..., m) ,
where J is the maximum stability degree, and the free
parameters ωk represent the imaginary parts of the complex
conjugate roots’ pairs of the characteristic equation of the
also will satisfy the Hurwitz conditions and its roots will be
placed in the same circular domain as the roots of the initial
polynomial (3).
In the same time is known that the multiple root with n
multiplicity of the n degree polynomial will be the root of its
(9) (2.9)
derivative with order 1, 2, ..., (n − 1) [7].
From this can be concluded that expression (12) has the
same multiple roots pi = − J , i = (1, ..., n − m) as equation (3).
The coefficients of the algebraic equation (12) are
dependent just of the known parameters of the control object
(1), so in this way the roots of this equation can be ensured the maximum stability degree for the respectively
determined. cases.
The generalized characteristic of the numerical values of The first part of the Affirmation 3 is demonstrated.
the characteristic equation’s coefficients and its roots Further, taking into account the definition of maximum
stability degree, the characteristic equation (8), which was
D (λ ) = a0 λ n + a1λ n −1+ ... + an −1λ + an = 0 obtained by decomposing in n linear factors of the
characteristic equation (3), is transcribed for
can be the parameter [8] pi = − J , i = (1, ..., n)

n
λ mg = n λ1 ⋅ λ2 ... λ n = n a n / a 0 , n!
∑ i!(n − i)! J
(13)
A( p ) = d 0 ( p + J ) n = d 0 i p n −i =
i =0
that is called the geometric mean root, where λ1 , λ2 , ..., λn are = d 0 ( c 0 p n + c1 Jp n −1 + c 2 J 2 p n − 2 + ...
the roots of the characteristic equation. + c n − m J n − m p m + ... + c n −1 J n −1 p + c n J n ) = (16)
= q 0 p n + q1 p n −1 + q 2 p n − 2 + ...
But, if λ1 = λ2 = ... = λn = η , then λ1 ⋅ λ2 ⋅...⋅ λn = η n , + q n − m p m + ... + q n −1 p + q n = 0 ,
and expression (13) takes the form
where ( p + J ) n = 0 represents the Newton binomial,
λmg = n η n = η . (14) ci , i = (0, ..., n) - the binominal coefficients [7]

From (14) it can be concluded that the geometric mean c0 = 1 ;

root of the n multiple roots of the n order characteristic
equation is equal with absolute value of the multiple root. n ( n − 1)...( n − i + 1)
c i = c ni = =
In the case of equation (12), for which it was established 1 ⋅ 2 ...i
i
(17)
that has the same multiple roots as (3), the expression for
determination of the geometric mean root according to (13)
∏ ( n − l + 1)
l =1 n!
= = , i = (1, ..., n ).
and (14), takes the following form i! i!( n − i )!

p mg = n − m J n − m = J =
m! d n − m d (15)
= n−m = n−m n−m ,
[n( n − 1)(n − 2)...( n − m + 1)] d 0 cn − m d 0
n(n − 1)(n − 2)...(n − m + 1)
where cn − m = , represents the
m!
binominal coefficient with index i = (n − m) [7].
This suggests that the geometric media root of the m order
derivative of the characteristic equation (3) of the control
system with n multiple roots pi = − J is equal with absolute
value of its multiple root and represents the maximum possible
stability degree of the designed control system η m = J (4).
In the case that characteristic equation (3) of the system
(1), (2) does not satisfy the condition of ensure the maximum
stability degree, established by the Affirmation 1, and does not
contain n multiple roots, then and the expression (12), which
represents the m order derivative of the equation (3), also will
not have multiple roots. If α i are not equal with each other
and some α i 0 > η m , then exists other α i < η m and the
stability degree η will be less than the maximum possible
stability degree (4). In this case the value of J , determined in
the conformity with (10), does not represent the maximum
stability degree, just represents the maximum geometric mean
root, that can be imposed to the designed control system. In
this way by the maximization of the geometric mean root is
For the equations (3) and (16), that are equivalent because
represent the characteristic equations of the same system (1),
(2) is applied the equality (11)
d n − m (ai ) = qn − m (a0 , J , ω k ), ω k = 0 ,
where d n − m and qn − m = cn − m d 0 J n − m are the coefficients
attached to the p m from (3) and (16) respectively.
In conformity with value of this coefficients can be
concluded that d n − m = cn − m d 0 J n − m , from that is obtained the
following expression for the calculation of the maximum
stability degree
d n −m
J = n−m . (18)
cn − m d 0
It is observed that the obtain expression (18) is identical
with (15) and (10).
It is demonstrated and the second part of the Affirmation 3.
To verify the veracity of the Affirmation 3 it is presented
an example.
Let be given the control object described by the following
transfer function
 H F (s ) =
k
=
(T1 s + 1)(T2 s + 1)(T2 s + 1)
k
= ,
a 0 s 3 + a1 s 2 + a 2 s + a 3
where k is the transfer coefficient of the object; T1 , T2 , T3 -
the time constants; a 0 = T1T2T3 ; a1 = T1T2 + T1T3 + T2T3 ;
a 2 = T1 + T2 + T3 ; a3 = 1 .
Assume that the control system is composed from the
control object (19) and the PID controller described by the
transfer function
ki
H R (s) = k p + + kd s .
s
Because the PID controller contains three tuning
parameters k p , ki and k d , then m = 3 . The characteristic
equation in the closed loop of the system (19), (20) is
presented in the following form
A ( p ) = a 0 p 4 + a1 p 3 + ( kk d + a 2 ) p 2 +
+ ( kk p + a 3 ) p + kk i =
= d 0 p 4 + d 1 p 3 + d 2 p 2 + d 3 p + d 4 = 0,
where d 0 = a0 ; d1 = a1 ; d 2 = kk d + a2 ;
d 3 = kk p + a3 ; d 4 = kki .
It is introduced the notion of the maximum stability degree
J and using the substitution pi = − J ± jω k , the characteristic
equation (21) is transcribed by its decomposition in n = 4
linear factors
A( p ) = a 0 ( p − jω1 + J )( p + jω1 + J ) ×
× ( p − jω 2 + J )( p + jω 2 + J ) =
= a 0 [( p + J ) 2 + ω12 ][( p + J ) 2 + ω 22 ] =
= a0 [ p 4 + 4 Jp 3 + (6 J 2 + ω12 + ω 22 ) p 2 +
+ 2 J ( 2 J 2 + ω12 + ω 22 ) p + J 4 + J 2 (ω12 + ω 22 ) +
+ ω12ω 22 ] = q0 p 4 + q1 p 3 + q 2 p 2 + q3 p + q 4 = 0 ,
where q 0 = a 0 ; q1 = 4a 0 J ; q 2 = a 0 (6 J 2 + ω12 + ω 22 );
q 3 = 2a 0 J (2 J 2 + ω12 + ω 22 );
q 4 = a 0 ( J 4 + J 2 (ω12 + ω 22 ) + ω12ω 22 ).
In accordance with (11), equalizing the characteristic
equation’s coefficients (21) and (22) with indexes
i = ( n − m) = 4 − 3 = 1 and considered that ω k = 0 is obtained
d1 = q1 (a0 , J ) ⇒ a1 = 4a0 J .
From this expression the relation for the calculation the
maximum stability degree takes the following form
J = a1 / 4a0 . (23)
(19)
However, following the m = 3 derivative of the
characteristic equation (21), where m is the number of tuning
parameters from the control law (20), is obtained the following
algebraic equations
A4(3) ( p) = A4−3 ( p) = 4a0 p + a1 = 0 ,
where its geometric mean root in accordance with (13) or (15),
is taking the following form
p gm = a1 / 4a0 . (24)
(20)
As it can be observed, the expressions (24) and (23) are
identical and correspond to the expression (4) from the
Affirmation 1.
Therefore, is confirmed again the fact that the maximum
stability degree J, determinate by the Affirmation 3, is
determined by the geometrical mean root of the equation
An−m ( p) = An( m) ( p) = 0, which represents the m order
(21) derivative of the characteristic equation of the designed
system.
B. The synthesing algorithm of controllers for the control
systems with maximum stability degree
For the practice application was elaborated the following
synthesizing algorithm of PID controllers for the control
systems with maximum stability degree [9]:
1. Is given the mathematical model of the control object with
known parameters (1).
2. Is chosen the control algorithm u(t) (2).
3. Is obtained the transfer function of the control system in
the closed loop.
4. Is obtained the characteristic equation of the control
system in the closed loop
(22)
n m
An ( p, b j ) = ∑
i =0
a i p n −i + k ∑b
j =1
j p j −1 =
= d 0 p n + d1 p n−1 + ... + d n −m p m + ... + d n−1 p + d n = 0,
where k is the transfer coefficient of the control object;
ai , i = 0, ..., r - the model object parameters; b j - the
tuning parameters of the controller; m - the number of
tuning parameters; n - the order of designed system.
5. Is verified the condition m ≤ (n − 1) :
o If the condition is fulfilled then is passing to the next
step.
o If the condition is not fulfilled then the procedure of
synthesizing the controller is resuming from the
beginning with another type of control algorithm.
6. Is introduced the notion of the maximum stability degree
J and using the substitution pi = − J ± jω k , the
 characteristic equation of the control system is automatic control systems with maximum stability degree. It is
decomposed in n linear factors argued from theoretical point of view a new approach of
determination the maximum stability degree of the designed
A( p ) = d 0 ( p − p1 )( p − p 2 ) × ... × ( p − p n ) = control system.
l z It is developed a new synthesis method of PID controllers
= d0 ∏ [( p − jω k + J )( p + jω k + J )] ∏( p + J ) = to the models object with inertia with known parameters for
k =1 i =1 the control system with maximum stability degree.
= q0 pn + q1 p n−1 + ... + q n−m p m + ... + q n−1 p + q n = 0,
The elaborated method of synthesis the typical controllers
represents a simple analytical procedure, which requires a
where l is the number of the conjugate complex roots’ reduced volume of calculations, doesn’t have the restrictions
pairs; z - the number of real roots; n = 2l + z - degree of on the complexity of the control object and can be applied to
the characteristic equation of the designed control system. the various types of models of objects: with inertia of arbitrary
7. From (11) d n− m (ai ) = qn −m (a0 , J , ω k ), ω k = 0, is order; inertia and identical elements; inertia and astatism;
obtained the expression for the determination of the inertia and time delay; inertia, astatism and time delay.
maximum stability degree J of the control system The proposed method ensures to designed systems the
maximum stability degree, which offers the high performance
and good robustness of designed control systems.
dn−m
J = n−m , REFERENCES
cn − m d 0
[1] I. Dumitrache, Ingineria reglarii automate. Bucureşti: Politehnica Press,
n(n − 1)(n − 2)...(n − m + 1) 2005. 725 p.
where cn − m = cnm = .
m! [2] C. Lazăr, D. Vrabie, S. Carari, Sisteme automate cu regulatoare PID.
Bucureşti: MATRIX ROM, 2004. 220 p.
[3] M. Bertocco, S. Cappellazzo, A. Flammini, M. Parvis, “A multi-layer
8. Using (5) d i (k , ai , b j ) = qi (a0 , J , ω k ), architecture for distributed data acquisition,” In: Proceedings of the 19th
IEEE Instrumentation and Measurement Technology Conference, 2002.
i = (n − (m − 1),..., n), are obtained the expressions for IMTC/2002, vol. 2, 2002, pp. 1261-1264.
determination of tuning parameters [4] A. O’Dwyer, Handbook of PI and PID controller tuning rules, 3rd
Edition. Imperial College Press, 2009. 608 p.
[5] А. М. Шубладзе, “Способы синтеза систем управления
b j = f j (k , a0 , ai , J , ω k ), j = (1,..., m) , максимальной степени устойчивости (Methods of synthesis of control
systems with maximum stability degree),” In: Автоматика и
телемеханика, №1, 1980, pp. 28-37.
where free parameters ωk represent the imaginary parts
[6] Д. П. Ким, Теория автоматического управления. Т. 1. Линейные
of the complex pairs roots of the characteristic equation. системы (Automatic Control Theory. T.1, Linear systems). Moscow:
9. By the computer simulation is verified the obtained ФИЗМАТЛИТ, 2003. 288 p.
performances of the control system. If the performances [7] А. Г. Курош, Курс высшей алгебры (Course of Higher Algebra).
Моscow: Наука, 1971. 431 p.
are satisfied then the tuning procedure is finished, but if it
[8] И. М. Макаров, Б. М. Менский, Линейные автоматические системы
was obtained the unsatisfied performances then the tuning (Linear automatic systems). М.: Машиностроение, 1982, 504 p.
procedure is resuming from the beginning with another [9] I. Fiodorov, “O metodă de acordare a regulatoarelor la modele de
type of control algorithm. obiecte cu inerţie şi timp mort,” In: Proceedings of the 3-rd International
Conference on Telecommunications, Electronics and Informatics ICTEI-
III. CONCLUSIONS 2010. Chişinău, 2010, v. I, pp. 129-134.
Based on the theoretical investigations, it is formulated and
proposed a new solution of the problem to synthesis the