Control Systems and Adaptive

Process . Regulators and

Communication

1
 PID controllers

• The PID controllers represent an efficient solution for

many control problems. Currently around 95% of
controllers are of the type PID, having survived to most
innovative technological elements.
• In this kind of controllers, the control action is built as an
addition of three type of actions: proportional, integral
and derivative. The control signal c(t) could be expressed
by the following equation, where Kp is the value of the
proportional gain, Ti is the integral time constant and Td is
the differential time constant, applied to the error signal
e(t). K t de(t )
c ( t )  K p e( t )   e(t )dt  K
p
T
p d
Ti 0
dt
2
 PID controllers

• This same signal, in the Laplace domain, would have

the following expression:
 1 1   1  K ( s  z1 )( s  z2 )
C ( s )  K p 1   TD s  E ( s )   K p  K I  K D s  E ( s )  D E ( s)
 TI s   s  s

• Its main advantages are that the proportional action

produces a control signal proportional to the error
signal, so it introduces a major correction if the error is
greater, the derivative action provides some advance
on the response of the system and the integral term
allows to eliminate the steady state error.
3
 PID controllers
• Kp will determine the value of the proportional action. If Kp is small
also the proportional action will be and vice versa. This action is
easy to tune and only depends on a parameter, and since the
correction is proportional to the error made can reduce this,
although not eliminate on steady state.
• If there is only proportional action:
K p ( s)
Y ( s)  E ( s)
1  K p ( s)

• This implies that there is always error, which decreases as Kp

increases, but if Kp increases greatly over impulse or instability can
appear.

4
 PID controllers
• Ti is the time required for that the integral action contribute to
output of the controller in an amount equal to the proportional
action. If Ti is small, the integral action will be great. This action
compensates for disturbances and maintains the controlled variable
around the point consigned. Being an integral action eliminates
stationary errors. In contrast, if Ti decreases much can destabilize
the system.

• Td is the time required for that the proportional action contribute to

the output of the controller in an amount equal to the derivative
action. If Td is small the derivative action will be small. The
derivative error anticipates the effect of the proportional action
estimating the error that will occur later, stabilizing quickly the
controlled variable after any disturbance.

5
 PID controllers
comparative representation of the response to a step signal of a plant for
different types of controllers (P, PI, PD, PID).

6
 PID controllers
The code in Scilab for this example would be:
//Examples of controllers P, PD, PI, PID
//
clear
s=poly(0,'s');
//time vector
t=0:0.05:50;
//
//transference function of the plan (order 3)
gp=1/((4*s+1)*(3*s+1)*(s+1));
//
//controllers parameters
Kc=3;Ti=8;Td=1.5;
//
//window
xset('window',1)
xname('Control System with several controllers PID')
//---------------------------------
//P control
gc=Kc;
Mr=gc*gp/(1+gc*gp);
Mrs=syslin('c',Mr);
yp=csim('step',t,Mrs);
7
//
 PID controllers
//PI control
gc=Kc*(1+1/(Ti*s));
Mr=gc*gp/(1+gc*gp);
Mrs=syslin('c',Mr);
ypi=csim('step',t,Mrs);
//
//PD control
gc=Kc*(1+Td*s);
Mr=gc*gp/(1+gc*gp);
Mrs=syslin('c',Mr);
ypd=csim('step',t,Mrs);
//
//PID control
gc=Kc*(1+1/(Ti*s)+Td*s);
Mr=gc*gp/(1+gc*gp);
Mrs=syslin('c',Mr);
ypid=csim('step',t,Mrs);
//
//step response
//graphic
subplot(1,2,1);xset('font',2,3)
plot2d(t',[yp; ypi; ypd; ypid]',style=[2,3,5,6]),xgrid(4)
//títulos y leyendas
xtitle('Step response','time','y(t)');
legends(['P control';'PI control';'PD control';'PID control'],[2,3,5,6],opt=1)
xstring(22,0.6,'Gp=1/((4*s+1)*(3*s+1)*(s+1)), Kc= 3, Ti= 8, Td= 1.5')
//---------------------------------
8
 PID controllers
Tuning rules
• PID controllers are usually adjusted in situ, in order to encompass all the
features of the plant. There are methods of adjustment or tuning both
analytical and experimental, even automatic. Analytical methods require
the knowledge of the desired transfer function, so they are commonly
used experimental methods.
• When designing a controller may use a controller of which complexity
coincide with the complexity of the process being [Link], for
obvious reasons, it is necessary to using a controller whose complexity is
more restricted, in which case you may either simplify the process model
to approximate it to a PID controller or design a controller for a complex
model and approximate it by a PID controller . In both cases it is necessary
a tuning or adjustment of the controller to achieve to fulfil its mission in
the most satisfactory mode. This need gave rise to the appearance of
different methods for tuning these controllers.
• The variation of any parameter may affect the operation of the controller,
so it is necessary to pursue a particular adjustment process. In the
following figure can be seen the effects of the decrease (left) or increase
(right) of the independent variables, keeping unchanged the other.
9
 PID controllers
Tuning rules
Example: effect of the decrease (left) and increase (right) of the Ti value
between 4 and 12 (the correct value for Ti could be 8).

10
 PID controllers
Tuning rules
• There are several methods for tuning, each best
suited to each type of plant, the type of controller
and/or specifications. Some of them are:
– Step response Ziegler-Nichols method
– Frequency response Ziegler-Nichols method
– Chien, Hrones, and Reswick method (CHR)
– Cohen-Coon method
– Empirical tuning based on rules

11
 PID controllers
Tuning rules
• Step response Ziegler-Nichols method: is based on a registered
information of the open loop step response of the process. The step
response only need two parameters (figure). First we need to determine
the point where the slope of the tangent straight line to unitary step
response has the maximum value.
The intersections between the tangent
and the axis give the values for a and L.

• Ziegler-Nichols have given the parameters for the

controller in function of the values of a and L as
shown in the next table:

12
 PID controllers
Tuning rules
• Frequency response Ziegler-Nichols method: is based in the knowing of
the point where the Nyquist plot of the process intersects the negative
real axis (this point is characterized by parameters Ku and Tu). This method
represent that a point of the Nyquist plot can be moved by changing the
PID controller parameters. The process is as shown:
• To carry out the adjustment is necessary first to set the integral time
constant Ti to its maximum value (∞) and differential time constant Td to
minimum (0). Then, starting with a small value, would adjust the value of
Kp until the process starts to oscillate, which occurs for a value of Kp = Ku
being the oscillation period Tu. The values calculates by this method are
the indicated in the next table:

13
 PID controllers
Tuning rules
• Experimentally, this method can be carried out as follows:

1. Proportional gain setting: To perform the adjustment first it’s necessary

to adjust the integral time constant Ti to his maximum value and the
differential time constant Td to his minimum value. Following Kp will be
adjusted until to obtain the desired response values.
2. Integral action setting: After the proportional gain has been adjusted
would proceed to reduce the integral time constant Ti to reduce the
steady state error (objective of this action). An important swaying could
appear. It would reduce the gain slightly and repeat the process until to
obtain the desired response values​​.
3. Derivative or differential action setting: Keeping the previously set
values ​would increase the value of Td until to obtain a faster response. If
necessary would increase slightly the gain value.

14
 PID controllers
Tuning rules
• Chien, Hrones, and Reswick method (CHR): for tuning a PID controller by
this method, the parameters a and L will be determined in the same mode
than Ziegler-Nichols method, and the controller parameters will be
obtained in function of these values.
• This method determines various settings in function of the percentage of
overshoot and according if are obtained from the response to load
disturbances or variations in the set point. The values of the controller
parameters ​obtained from the response to load disturbances are the
indicated in the following table:

15
 PID controllers
Tuning rules
• The controller parameters obtained from the response to variations in the
set point are the indicated in the following table:

16
 PID controllers
Tuning rules
• Cohen-Coon method: this method is based on a model of the process
such that: K p sL
P( s )  e
1  sT
The main design criterion is the rejection of load disturbances, and also
provides a table of values ​based on analytical and numerical calculations.
considering: KpL L
a 
T LT

17
 PID controllers
Tuning rules
• Empirical tuning based on rules: the methods described above are approximate
methods that require a posterior manual tuning. This manual tuning is done on the
closed loop response, introducing a disturbance (change in the set point, change in
the control variable, etc..), analyzing and correcting the response of the controller
parameters. These adjustments are based on simple rules, developed from a process
of experimentation. These rules are:
- Increasing the proportional gain decreases the stability.
- The error decays faster if we decrease the integration time constant.
- Decreasing the integration time constant decreases the stability.
- Increasing the derivative time constant increases the stability.
• As can be appreciated, changing a parameter in one direction or another affects the
tuning differently. It is therefore common to use tuning maps which objective is to
show, intuitively, how the changes in the controller parameters affect to the
behaviour of the closed loop system. In this mode may be defined boundaries within
which can move parameter values or limit values for which the system becomes
unstable. These tuning rules were implemented in automatic tuning process.
18
 PID controllers
Controllers design
• Basic design: for a basic design, first we will choose the
system topology, that is, where to locate the controller: serial,
parallel. After selecting the controller settings, we must
choose a type of controller that meets the required
specifications, the most widely used is the PID controller. Thus
the complexity of the
controller is restricted.
You can also use different
compensators (lead, lag or
lag-lead).

19
 PID controllers
Controllers design
• After selecting the controller, and according to the
required specifications, we must to determine its
parameters, for which we will choose the analysis
method most appropriate to be used, according to
specifications: root locus approach or frequency
response analysis.
• Finally we should check that the designed system meets
the required function and, if necessary, adjust the
parameters to make it so.

20
 PID controllers
Controllers design
• Improved design: sometimes it is desirable to move some of
the actions of the controller to the feedback loop. For
example, if a PID controller has a step input, the derivative
action makes that appears a pulse in the control. Therefore
other possible configurations are used to avoid the problems
that can occur.

21
 PID controllers
Controllers design
• PI-D Controller: avoids the phenomenon of the set point
reaction. Are avoided aggressive control actions that may
damage the actuators. The action becomes slower and
reduces the over shoot.

22
 PID controllers
Controllers design
• I-PD controller: the proportional and derivative actions are
only on the feedback loop. With this kind of controller, with a
step signal do not produce an abrupt change in the control
signal, which may not be suitable for some types of actuators.

23
 PID controllers
Controllers design
• Integral control with state feedback configuration: allows a
finest control if increases the order of the plant.
• PI-PD, PID-PD controllers: The PI-PD controller provides an
excellent four-parameters controller for control of integrating,
unstable and resonant processes.
• Feed-forward control: allows to measure the disturbances and
perform a corrective action as soon as the disturbance
appears.

24
 PID controllers
Controllers design
• With PID controllers can appear some problems, such as:
- The tuning: the choice of values ​of the parameters Kp, Ti and Td
can become complicated, in fact, it was observed that according to
the tuning method chosen these values ​may be different.
- Integrator windup effect: this effect appears when the control
action increases so much that produce the saturation in the
actuators. This problem breaks the control loop because the
actuator will remain at its limit value regardless of the control
signal. The control will not be effective until the control signal falls
below the saturation level.
To eliminate this effect there are many controller configurations
that allow to override the problem.

25
 PID controllers
Controllers design
• Robust design: the first step in designing a control system is to
design an equivalent mathematical model of the physical
plant. Sometimes this model may not be linear or too
complex. A complex model is not very useful because
complicates excessively the design process. For this reason
normally the design is done by simple models but which
reflect the intrinsic characteristics of the physical system.
Obviously this solution generates an uncertainty whether the
designed control will be suitable for the required function.
Uncertainty can be of different types depending on the
features that are deemed (stability, gain margin, phase
margin, etc..). Uncertainty is added to the nominal model and
may be treated in different ways.

26
 PID controllers
Controllers design
• In robust control theory we expect to approximate the plant
model by a linear model with constant coefficients, knowing it
will be incurred an error that is intended to be bounded. In
this way we can design control techniques valid for
multivariable systems, to ensure at least the system stability.
• In a robust control the system must to be stable for the entire
plant situations and the performance must meet the design
specifications for all possible situations. that is, in the
presence of uncertainty.

27
 PID controllers
Controllers design
• It is therefore essential to determine the type of uncertainty
(parametric, structured, unstructured) to determine its size
and importance, and to enclose it. To estimate the uncertainty
we can observe the data of the sensors and actuators, to
experiment with the system at different operating points, etc.
• For multivariable systems should choose the closest
description to the effect that causes uncertainty. Robustness
margin is related to the specific type of model uncertainty.
• The control with two degrees of freedom: Allows to adjust the
closed loop and feedback loop characteristics independently,
which improves system response.

28
 PID controllers
Controllers design
• Design by pole placement method: many of the properties of
a system can be expressed by its poles. With this we expect to
design a controller so that the closed loop system has poles at
the desired location. The method requires a complete model
of the process. Is possible to find a controller that gives the
closed-loop poles desired, provided that the controller is
sufficiently complex. For a PID control is necessary to restrict
the complexity of the model using approximate methods,
therefore, the selected poles must be chosen to ensure that
the model is valid.
• The design process will depend on the type of controller (PI,
PID) and the characteristics of the system (number of poles,
order, oscillatory system, etc.)
29
 PID controllers
Controllers design
• Dominant poles design: is a simplification of the above
method. Sometimes is difficult to specify all the closed loop
poles, so that the dominant poles are used to characterize the
system.
• Lambda tuning: is a special case of the poles placement
method. It is a simple method that can give good results in
certain circumstances if the design parameter is chosen
appropriately. The basic method cancels a pole of the process,
which will result in a poor response of the load to
disturbances in process dominated by time constant.

30
 PID controllers
Controllers design
• Algebraic design: is a process in which the transfer function of
the controller is obtained from the specifications by a direct
algebraic method. There are several methods, all of them related
to the allocation of poles.
- Standard forms: It begins by determining a transfer function in
a certain way, calculating her parameters in mode to minimize
the error criterion chosen.
- Haalman method: for systems with a delay L, Haalman
proposed to select a loop transfer function of the form
2  sL
G ( s )  P( s )C ( s )  e
3Ls
depending on the transfer function of the process and applying
this method will be easy to determine the parameters and the
type of controller.
31
 PID controllers
Controllers design
• Internal model control (IMC): its name comes from the fact
that the controller contains internally a model of the process.
This model ( Pm(s) ) is connected in parallel with the process
and applies the inverse of Pim(s) model and a filter (Gf). It is
considered that all disturbances affecting the process are
reduced to an equivalent disturbance d. The controller
obtained can be represented by the following function, from
which is deduced that this type of serial controller cancels the
poles and zeros of the process.

G f Pim
C
1  G f Pim Pm

32
 PID controllers
Controllers design
• Design for disturbance rejection: in the above methods we
have only taken into account the characterization of the
dynamics of the process, without considering the
disturbances directly. For the study we have used a step
disturbance and analyzed the response of the system, but not
all the disturbances have this form. Furthermore we have not
taken into account the amplification of the measurement
noise in the feedback. With a design oriented to interference
rejection we try to obtain a compromise solution between the
attenuation of load disturbances and the amplification of
measurement noise due to the feedback.

33
 PID controllers
Operation conditions
• There are many requirements that are demanded to a control
system. About 95% of the installed controllers are PID type
which shows that these controllers work properly if the
requirements are not too strictest. Usually most stable process
can be controlled using a PI controller, in fact, it is frequent that
the derivative action is not used. This occurs in the majority of
first-order processes.
• The PID controller is usually sufficient for second order
processes and processes with a delay in where the derivative
action may be sufficient to accelerate the system response.
• Not will be sufficient for higher-order processes, in systems with
large delays or in systems with oscillatory modes. For these
cases it is advisable to use more sophisticated control systems
that the PID. 34
 References
Bibliography
• Karl J. Åström. Control PID Avanzado.
• Karl J. Åström. PID Controllers: Theory, Design and Tunning.

Interesting links
• [Link]
• [Link]
• [Link]
• [Link]
• [Link]
• [Link]
• [Link]
ID
• [Link]

35