SINTONÍA DE CONTROLADORES PID:

Profesora: Silvia Ochoa Cáceres

Universidad de Antioquia
01-2021

Control de Procesos 1
 PIDs Tuning
• Procedure for adjusting the parameters of a
controller.
• In the case of PID controllers, the tuning
procedure allow us to find the de KC, τI and TD
values that provide acceptable performance
and for which the system is stable (in closed
loop).

2
 PIDs Tuning

• Involves the selection of the values for Kc,

τI, and τD.
• Affects the performance of the control
system
• Impacts the closed loop stability

3
 PIDs Tuning
We look for parameters that:
• Reduce the process variability
• Allows the process to remain stable,
even for the worst disturbance
scenario.
• Avoid excessive variation in the
manipulated variable

4
 In summary tuning means:

Finding the best set of controller parameters

for ...
AFFECTING the DYNAMIC
PERFORMANCE as desired.
5
 Stages in Controllers Tuning

Find/calculate an starting value for the parameters

Implement and Perform

Fine tuning

Monitor the performance in the

Controlled Variable and the Manipulated
6 variable Behavior
 ¿Why fine tuning?
• Performance can be improved by
making small changes to the tuning
parameters, when:
– The model has errors (it is not the same
as the real plant)
– Most tuning methods do not perform
optimization.
– Some processes require more "exact"
control.
7
 TUNING based on …
Trial and error Stability

Z-N Tyreus-
Process Reaction Closed loop Luyben
Curve
(1/4 decay ratio) Performance
Z-N
Cohen – Optimization Correlations
Open loop
Coon 8
 Tuning by trial and error
Con el controlador en manual, eliminar el efecto de
las acciones integral y derivativa.

Ajustar la ganancia en su valor mínimo (Kc≅0.005)

Colocar el controlador en Automático

Realizar una pequeña perturbación y observar

la respuesta de la variable controlada

Duplicar la ganancia y perturbar nuevamente

NO
Es el sistema
sub amortiguado
y oscilatorio?
SI
La ganancia es la 9ganancia última.
 Disminuya la ganancia a la mitad de su valor último

Introduzca la acción integral reduciendo TI en factores de 2,

y perturbe el sistema para ver el efecto.

Encuentre el valor de TI para el cual el sistema se comporta

subamortiguado, y ajuste TI al doble de este valor.

Introduzca la acción derivativa incrementando TD,

perturbe el sistema y encuentre el valor de TD
que suavice la respuesta sin amplificar el ruido.

Aumente KC hasta lograr un equilibrio entre

el tiempo de respuesta, el sobrepaso y la oscilación.

10
 Tuning by Ziegler and Nichols Open Loop
Method (Step response)
The objective of this
method is to provide y(t)
tuning parameters that
produce a decay ratio
(razón de asentamiento)
of 1/4 in the oscillation ysp
of the controlled
variable.

11
 Tuning by Ziegler and Nichols Open Loop
Method (Step response)
• The tuning parameters are based on the
parameters of a FOPDT model (from a reaction
curve) of the process in open loop.
• The transfer function of a first order model plus
dead time is given by:

Kp −t d s
G p (s) = e
τs + 1
12
 Tuning by Ziegler and Nichols Open Loop
Method (Step response)
• However, it is important to keep in mind two important points
1. For tuning purposes, the transfer function of the process to be
controlled is considered composed of Gp, Gm and Gf.
2. Despite what was mentioned in point 1, several of the tuning
methods (ZN-OL, Cohen-Coon, Ciancone-Marlin, Lopez-Rovira,
among others) require that the transfer function of the process to
be controlled be of order one.
• The transfer function of the process to be controlled is then
defined as:
' Debe ser
K p KmK f K Adimensional!
G′p ( s ) ≈ −t d s
≈ −t d s
p
e e
τs + 1 13
τs + 1
 5
=
3 +1

10
= ′=
+1 3 +1
3
=
1.5 + 1
 Tuning by Ziegler and Nichols Open Loop
Method (Step response)
• The tuning parameters are calculated based
on:
Controlador KC TI TD
P  1  τ 
  *  
 K 'P   td 
PI  0.9   τ 
  *   3.3 * t d
 K 'P   td 
PID  1.2   τ 
  *   2 * td 0.5 * t d
 K 'P   td 
15
 Tuning by Ziegler and Nichols Open Loop
Method (Step response)
• It is necessary to take into account that these
formulas are empirical, and they should only
be applied when :

td
0.1 < <1
τ

16
 Tuning by Cohen-Coon Method

Controlador KC TI TD
P  1  τ   td 
  *   * 1 + 
 K 'P   td   3τ 

 3t 
PI  1  τ  
t d  30 + d 
τ 
 *  0.9 + d  
t
  * 
 K 'P   td   12τ  t 
9 + 20 d 
τ 

PID 
+
td   6t 
 1  τ   16 3  t D  32 + d  4t d
  *   *  τ   τ  t 
 K ' P   t d   12  t  11 + 2 d 
 13 + 8 d  17 τ 
 
τ 
Ejemplo: Determine los parámetros de sintonía
para un controlador para el IC
mv
Vapor TC 2 ,
= = ,
4 +1
Corriente de TT
Proceso
2.45 , 0,4
Tin T =
= 0,5 + 1
4 +1

Kp=[°C/(kg/min)] Km=[mA/°C]
kd=[°C/°C] Kf=(kg/min)/%

Parámetros/Método ZN-OL Cohen-Coon

Kc
τI
Ejemplo: Determine los parámetros de sintonía para un controlador para el IC

1. Determinar la Función de transferencia: K p Km K f

G′p ( s ) ≈ e −t d s
τs + 1
Es importante recordar que se debe garantizar una Kp’ adimensional!
Adimensional?

K p' = 2.45 * 0.4 * 0.2 * 100

16

[ ]  °C
K =
'   mA   (kg / min) 
 *  *  *
%
 (kg / min)   °C  
p
%  mA
Por lo tanto:
1.225 −0.8 s
G′p ( s ) = e
4s + 1
Ejemplo: Determine los parámetros de sintonía para un
controlador para el IC
2. Aplicar las Correlaciones de Sintonía
Para ZN-OL

 0.9   τ  τ I = 3.3* t d = 2.64 min

Kc =   *   = 3.67
 K 'P   td 

Para Cohen - Coon

 3t 
t D  30 + d 
τ 
 1  τ  
 *  0.9 + d  = 3.74
t τI =  = 1.88 min
Kc =   *  t 
 K 'P   td   12τ  9 + 20 d 
τ 
 Ejemplo: Determine los parámetros de sintonía para un
controlador para el IC
3. Observar el Comportamiento dinámico
T (°C)

%OP
 Ejemplo: Determine los parámetros de sintonía para un
controlador para el IC
4. Cuantificar el Desempeño

ISE(°C )
2

Conclusión?
 TUNING based on …
Trial and error Stability

Z-N Tyreus-
Process Reaction Closed loop Luyben
Curve
(1/4 decay ratio) Performance
Z-N
Cohen – Optimization Correlations
Open loop
Coon 23
 TUNING BY DYNAMIC PERFORMANCE

By Optimization Correlations

24
 PERFORMANCE

• It is the ability of the control system to achieve

the desired dynamic response, measured in
terms of the error in the controlled variable.
• Recall that the error is the deviation between
the real value of the controlled variable at
time t, and the set point.

Error= ysp -y
25
 PERFORMANCE

• The performance of the control

system must take into account two
main factors:
–The performance of the controlled
variable.
–The behavior of the manipulated
variable.
26
 PERFORMANCE OF THE CONTROLLED
VARIABLE
• To measure the performance of the controlled
variable, performance indices based on the
integral of the error are frequently used:
1. Integral of the Absolute Value of the Error (IAE):

tf

IAE =  ySP (t ) − y (t ) dt
0
Where tf is the simulation time.

27
DESEMPEÑO DE LA VARIABLE CONTROLADA

– Integral of Square of Error (ISE):

tf

ISE =  [ ySP (t ) − y (t )] dt
2

– Integral time absolute error (ITAE):

tf

ITAE =  t ySP (t ) − y (t ) dt
0

28
DESEMPEÑO DE LA VARIABLE CONTROLADA

– Integral Error (IE):

tf

IE =  [ ySP (t ) − y (t ) ]dt
0

– These performance indices are only comparative

for the same integration time.

29
 TUNING BY DYNAMIC PERFORMANCE

• The best tuning parameters will be those for

which a lower performance index is found
(ISE, IAE, etc.) and that also that satisfies the
limits imposed on the manipulated variables.
• When evaluating the performance of the
control system, it must be taken into account
that the manipulated variable does not vary
excessively or aggressively.

30
 TUNING BY DYNAMIC PERFORMANCE:
CORRELATIONS
• There are several types of correlations used to
tune PID controllers which accounts for the
dynamic performance; among which are:
– Ciancone correlations (regulation and tracking).
– Rovira correlations (tracking).
– López correlations (regulation).

31
 TUNING BY DYNAMIC PERFORMANCE:
CORRELATIONS
• These types of correlations are applied based
on first order models with dead time (open
loop determined), including the sensor-
transmitter and the EFC.
'
K p KmK f K
G′p ( s ) ≈ −t d s
≈ −t d s
p
e e
τs + 1 τs + 1

32
 TUNING BY DYNAMIC PERFORMANCE:
CIANCONE CORRELATIONS
• The Ciancone correlations are presented graphically
depending on the control algorithm to be used (PID,
PI, etc).
• The method is based on the dimensionless fraction
dead time :
td θ
Fraction Dead Time = =
td + τ θ + τ

33
 CIANCONE – MARLIN
Correlations for PID:
For regulation and tracking control

34
 CIANCONE –
MARLIN
Correlations for PI:
For regulation and
tracking control

35
 LÓPEZ CORRELATIONS :
• The López correlations (for regulation) based
on the IAE are given by:

KC TI TD
−0.985
P  0.902   t d 
  *  
 K 'P   τ 
−0.986
 0.984   t d   τ   td 
0.707
PI
  *    * 
 K 'P   τ   0.608   τ 
τ
−0.921 0.749 1.137
PID  1.435   t d     d
t  d
t
  *      
* 0. 482 * τ *  
 K 'P   τ   0.878   τ  τ 
36
 CORRELACIONES DE Rovira:
• The Rovira correlations (for tracking control)
based on the IAE are given by:

KC TI TD
−0.861 τ
PI  0.758   t d 
  *   t 
1.02 − 0.323 d 
 K 'P   τ  τ 
−0.869 τ
 
PID  1.086  *  td  0.914
 td   td 
 K'  τ 0.740 − 0.130  0.348 *τ *  
 P    τ  τ 

37
Ejemplo: Determine los parámetros de sintonía
para un controlador para el IC
mv
Vapor TC 2 ,
= = ,
4 +1
Corriente de TT
Proceso
2.45 , 0,4
Tin T =
= 0,5 + 1
4 +1

Kp=[°C/(kg/min)] Km=[mA/°C]
kd=[°C/°C] Kf=(kg/min)/%

Parámetros/Método Ciancone López

Kc
τI
Ejemplo: Determine los parámetros de sintonía para un
controlador para el IC
1. Determinar la Función de transferencia:
1.225 −0.8 s
G′p ( s ) = e
4s + 1
2. Aplicar las Correlaciones de Sintonía
Ciancone-Marlin: dado que el ejercicio planteado es control por
regulación, y previamente se definió usar un PI, se usarán las
gráficas a continuación:
td θ 0.8
Fraction Dead Time = = = = 0.17
t d + τ θ + τ 0.8 + 4
 De la Figura:

K c K p' = 1.7
1.7 1.7
Kc = ' = =1.39
K p 1.225

De la Figura:
τI
= 0.35
td + τ
τ I = 0.35 * 4.8 = 1.68 min
 Ejemplo: Determine los parámetros de sintonía para un
controlador para el IC

2. Aplicar las Correlaciones de Sintonía

López:

K c = 3.93 ; τ I = 2.11 min

 Ejemplo: Determine los parámetros de sintonía para
un controlador para el IC
3. Observar el Comportamiento dinámico
T (°C)

%OP
Parámetros/Método Ciancone López ZN-OL Cohen-
Coon
Kc 1,39 3,93 3,67 3,74

τI ISE(°C 2 ) 1,68 2,11 2,64 1,88

 Ejercicio
• Para el sistema del IC, verifique
matemáticamente si el lazo cerrado usando
los parámetros de Ciancone resultarán o no en
un sistema estable en lazo cerrado. Tenga en
cuenta que la función de transferencia del
controlador será:
 1   1 
Gc ( s ) = K c 1 +  = 1.39 1 + 
 τIs  1.68 s 
 TUNING BY DYNAMIC PERFORMANCE:
OPTIMIZATION.
Tuning by optimization is characterized by:
• Uses as an objective function to minimize any of
the performance indices (IAE, ISE, etc).
• It requires more computational effort than other
tuning methods.
• Nonlinear phenomenological models can be
used.

45
 TUNING BY DYNAMIC PERFORMANCE:
OPTIMIZATION
• It is required when the system to be
controlled has complex dynamics (several
coupled dynamics), and when the model
obtained by the reaction curve presents a
large error compared to the real system.
• It can be univariable (adjusting only one
parameter at a time) or multivariable
(adjusting KC, TI, and TD simultaneously).

46
TUNING BY DYNAMIC PERFORMANCE: OPTIMIZATION

Definir parámetros de sintonía iniciales

(curva de rxn, correlaciones, etc)

Evaluar el IAE y determinar en que

dirección Mejora el desempeño
(Algoritmo de Optimización: ej Nelder-Mead)

Determinar nuevos parámetros de

sintonía en la dirección de mejora.

Parámetros
 IAEi+1 - IAEi  <Tol
de Sintonía
47
 COMENTARIOS
• All the Correlations addressed in this talk,
except those of CIANCONE, are applicable
when: tD
0.1 ≤ ≤1
τ
• Ciancone correlations are independent of the
value of the ratio td/τ, thanks to the fact that
they are based on the dimensionless
parameter t d
td + τ
48
 COMENTARIOS
• The Ciancone correlations were developed based on
the IAE, and taking into account changes in the
process parameters up to ± 25% (error in the model).

• By taking into account errors in the model, the

tuning by means of the Ciancone correlations
makes the control system more robust than when it
is tuned using the López correlations, who does not
take into account errors in the model.

49
PROCEDIMIENTO PARA LA SINTONÍA POR DESEMPEÑO
DINÁMICO: CORRELACIONES.
Identificar el modelo de
primer orden con retardo
G(s)=f (K, τ, tD)
Control por
Regulación o
Regulación Control por
Seguimiento
Seguimiento
Correlaciones Correlaciones Correlaciones
de López de Rovira de Ciancone

No %E< 25%?
tD
0.1 ≤ ≤1
τ

No
Si Si Sintonizar por
Obtener los parámetros:
KC, TI y TD. otro Método
50
 FINALMENTE …
• Tuning methods based on performance
indices, either by optimization or by
correlations, ensure that the performance of
the control system will be satisfactory, but
they do not evaluate the stability.

51
 FINALMENTE …
• Correlation tuning methods use approximate models
(first order with dead time) that in some cases can
deviate from the real system, thus affecting the
performance of the control system.
• The correlations seen here have been obtained by
empirical methods, and their validity is restricted to
certain intervals, so it is necessary to evaluate
whether or not they are applicable depending on the
specific case that is being worked on.

52
 Fecha Examen 3
Jueves 17 de Junio.
Fechas para tener en cuenta…
Martes 01.06 PID Charla 15
Jueves 03.06 Sintonía 1 Charla 16
Martes 8.06 Asignación en clase Sintonía 1
Charla 17
Jueves 10.06 Estabilidad 1
Martes 15.06 Estabilidad 2 Charla 18

Jueves 17.06 Examen 3 Se fijó por encuesta

Sintonía y estabilidad
Martes 13.07 Asignación en clase

15/20/22 de Julio Exposiciones Control Avanzado

Jueves 27.07 Examen Acumulativo

Control de Procesos 54