SEMINAR REPORT ON

OPTIMIZATION METHODS FOR PID CONTROLLER

TUNING



Submitted By
Saiman Shetty
(4 NM 09 EE 050)

Under the guidance of
Mr. Pradeep Kumar
Assistant Professor

Seminar Report Submitted to Department of Electrical and Electronics,
NMAM Institute of Technology, Nitte






DEPARTMENT OF ELECTRICAL AND ELECTRONICS

1

1. Introduction
2. Literature Review
3. Functional partitioning of the Seminar
3.1 Brief overview of P, PI and PID control techniques.
3.2 Explanation of the Ziegler-Nichols Open loop and Closed loop tuning methods.
3.3 Discussion of problems such as Integral windup and mitigation methods.

4. The PID Control Method
4.1 The Proportional Control Mode
4.2 The Proportional + Integral Control Mode
4.3 Proportional + Integral + Derivative Control Mode
4.4 Necessity of Controller Tuning
4.5 Ultimate Cycling Method
4.6 Process reaction curve method
4.7 Damped Oscillation Method
4.8 Bumpless Transfer
4.9 Integration windup and mitigation methods

5. Conclusions
6. Technical References





2

1. Introduction
Although many innovative methodologies have been devised in the past 50 years to handle more
complex control problems and to achieve better performances, the great majority of industrial
processes are still governed by means of simple proportional-integral-derivative (PID) controllers [1].
This seems to be because PID controllers, despite their simple structure, assure acceptable
performances for a wide range of industrial plants and their usage (the tuning of their parameters) is
well known among industrial operators. However, optimum operation of the PID controller can only be
attained by proper adjustment of the gains by a process called tuning. The objective of this report is to
discuss the types of PID controller tuning and its optimization.

2. Literature Review
A heuristic approach to PID controller tuning was first explained in a 1942 publication titled Optimum
Settings for Automatic Controllers [2] by J.G. Ziegler and N.B. Nichols. This paper was arrived at a
result of experimental majorly intended to convey parameter tuning methods that have proven to be
very useful for a controls technician, setting aside the complex mathematics that the process involves.
Several improvements in the controller tuning algorithms have been developed since then.
A practical approach to tuning the process controllers is explained by Jacques F. Smuts in his book
"Process control for practitioners"[4], where elaborate procedures of various types of PID Control are
described, out of which the Z-N method is adopted for inclusion in this report.

3. Functional partitioning of the Seminar
3.1 Brief overview of P, PI and PID control techniques.
3.2 Explanation of the Ziegler-Nichols Open loop and Closed loop tuning methods.
3.3 Discussion of problems such as Integral windup and mitigation methods.

4. The PID Control Method
A PID controller has a Set Point (SP) that the operator can set to the desired value. The Controllers
Output (CO) sets the position of the control valve. And the output measurement, called the Process
Variable (PV) gives the controller its much-needed feedback. Upon proper implementation, the PID
controller must compare the process variable to its set point and calculate the difference between the
two signals, also called the Error (E).
Then, based on the Error and the PID controllers tuning constants, the controller calculates an
appropriate controller output which corrects the error and stabilizes the controller at a steady state.
PID controllers have three control modes:
3

Proportional Control
Integral Control
Derivative Control
Each of the three modes reacts differently to the error. The amount of response produced by each
control mode is adjustable by changing the controllers tuning settings.

4.1 The Proportional Control Mode
The proportional control mode is in most cases the main driving force in a controller. It changes the
output in proportion to the error . This implies if the error is large, more control action is required to
correct the error, which is intuitively true.





is called the controller gain.
However, in a proportional controller, the main drawback is the steady-state offset that cannot be
eliminated by proportional controller alone. There should be an external bias or a "reset" mechanism
to eliminate this problem.

4.2 The Proportional + Integral Control Mode
In order to nullify the offset created in the proportional control mode, and to improve the speed and
stability of the response, an additional 'memory' of all the previous errors has to be introduced. This is
done by including a weighted integrating action in the control loop. The 'weight' is called

, the
integral gain.

]

For a given error, the speed of the integral action is set by the controllers integral time setting (

). A
large value of

results in a slow integral action, and a small value of

results in a fast integral action.

4.3 Proportional + Integral + Derivative Control Mode
The PID control mode produces additional to the PI mode, an output based on the rate of change of
the error. The derivative mode produces more control action if the error changes at a faster rate. This
4

is done by including a weighted differentiating action in the control loop. The 'weight' is called

, the
derivative gain.

()

()

]

The derivative mode has an adjustable setting called Derivative Time (

). The larger the derivative

time setting, the more derivative action is produced.

Figure 1. PID Control block Diagram


Figure 2. Stability



5

4.4 Necessity of Controller Tuning
The basic objective of controller tuning is
Fast Response
Acceptable Stability
For most practical processes being controlled with PID controllers, these two objectives cannot
completely be achieved simultaneously. Faster the response, worse the stability and better the
stability, slower the responses. Generally, it is more important to have a acceptably stable system
which responds quickly enough as shown in Figure 2. It is seen that for acceptable stability, the
undershoot that follows the overshoot for step actuated system is smaller than the overshoot.
To get the desired stable behavior of the system, it is required to adjust the controller parameter(s)

and

in the right proportion by a process called tuning. There are several known methods for
tuning, the most basic ones being the Ziegler-Nichols tuning methods, which are discussed in this
report.

4.5 Ultimate Cycling Method
The Ziegler-Nichols methods are heuristic tuning procedures derived from elaborate experiments
conducted to arrive at a general tuning rule for all controllers as per [2]. The Ultimate Cycling method
is used to tune closed loop PID controllers.
The following steps are followed to apply the Ziegler-Nichols closed loop method [4]:
i. The process is stabilized and it is ensured that no process changes (eg. product change, grade
change, load changes) are scheduled.
ii. If the loop is oscillating, it is ensured that the Process Variable stops oscillating when the
controller is placed in Manual mode.
iii. The Integral action is turned off by either setting integral gain

to zero or setting integral time

to a very large value.

iv. The derivative action is turned off by setting the derivative parameter to zero.
v. The controller is placed in automatic mode.
vi. A set point change is made and the result is monitored.
vii. A freely oscillating process variable with constant magnitude is got by varying the controller
gain

. If the oscillations are decaying,

is increased by 50%. If the oscillations are growing,

is decreased by 50%.

is decreased by 50% if the gain hits the upper or lower limits.

viii. If the Process Variable oscillates with constant amplitude, and neither the Process Variable nor
the Controller Output hits their limits, the ultimate controller gain

and the period of the

oscillation

is noted.
6

ix. The new controller gains are calculated as per the chart in Table 1 and entered into the
controller.
These values are for interactive PID algorithms only. The controller parameter values should be
correspondingly calculated for non-interactive and parallel algorithms.

Table 1. Z-N Tuning Parameters
Controller

P 0.5

-- --
PI 0.45

--
PID 0.6

4.6 Process reaction curve method
Although the original method proposed in [2] for self-regulating and integrating processes, a modified
version of the reaction curve method is widely used on self-regulating processes but more accurately
so [4]. This method uses three process characteristics: dead time

, process gain

, and time
constant , which are determined by the following steps:
i. The controller is placed in manual and the process is allowed to settle.
ii. A step change of a few percent is made in the controller output (CO) and the process variable
(PV) is allowed to settle out at a new value. The size of this step should be large enough that
the process variable moves well clear of the process noise/disturbance level.
iii. The total change obtained in PV is converted to a percentage of the span of the measuring
device.
iv. The process gain

is calculated as follows:

=



v. A line tangential to the PV response curve is drawn through the point of inflection and
extended to intersect with the original level of the PV (before the step-change in CO). Take
note of the time value at this intersection.
vi. The dead time is measured

as shown in Figure 3.
vii. The value of the PV at 63% of its total change is noted from the curve and the time constant
is determined as shown in Figure 3.
7

viii. A few more trials of steps i through vii are done and the average of the parameters are taken
for the calculation of settings for Controller Gain (Kc), Integral Time (Ti), and Derivative Time
(Td), using the Ziegler-Nichols tuning rules viz.
For P control:

For PI control:

For PID control:

These rules produce a quarter amplitude damping and the controller gains must be divided by 2 to
ensure stability.


Figure 3. Reaction Curve

4.7 Damped Oscillation Method
In many cases, plants are not supposed to undergo through sustained oscillations, as is the case
when Ultimate cycling method is used. Hence, damped oscillation method is preferred for these cases.
8


Figure 4. Damped oscillation Method

The controller gain

is kept low, the integral and differential times

and

are set to zero. Now,

is increased slowly till a decay ratio

of

is obtained for step response. The period of oscillation

is noted. If

is the proportional gain settling for obtaining

decay ratio, the controller

parameters are measured as follows:

4.8 Bumpless Transfer
It is a normal practice to set up some processes using manual control initially, and once the process is
close to normal operating point, the control is transferred to automatic mode through auto/manual
switch. In such cases, in order to avoid any jerk in the process, the controller output immediately after
the changeover should be identical to the output set in the manual mode. This can be achieved by
forcing the integral output at the instant of transfer to balance the proportional and derivative outputs
against the previous manual output; i.e.
Integral output = {(previous manual) (proportional + derivative) output}.
Similarly, for automatic to manual transfer, initially the manual output is set equal to the controller
output and the difference is gradually reduced by incrementing or decrementing the manual output to
the final value of the manual signal and thus affecting a changeover.
Another way to transfer from Auto to Manual mode in a bump less manner, the set point may be made
equal to the present value of the process variable and then slowly changing the set point to its desired
value. These features can be easily be implemented if a digital computer is used as a controller. This
provision eliminates the chance of the process receiving sudden jolt during transfer.

9

4.9 Integration windup and mitigation methods
A significant problem with integral action is that when the error signal is large for a significant period of
time. This can occur every time when there is large change in set point. If there is a sudden large
change in set point, the error will be large and the integrator output in a P-I-D control will build up with
time [3]. As a result, the controller output may exceed the saturation limit of the actuator. This windup,
unless prevented may cause continuous oscillation of the process, which is not desirable.

There exist several methods through which integral windup can be prevented, two of which are
discussed in this report. The characteristics of majority of actuators are similar to that of an amplifier,
in that the output varies linearly with the input till the input is within a particular range, beyond which
the output becomes constant at either the upper or lower limit of the actuator.

The first method uses a switch to break the integral action, whenever the actuator goes to saturation.
This can be illustrated by Figure 5.



Figure 5. Scheme for anti-integration windup

When the switch S is closed, transfer function of the controller can be obtained as:

]

So when the switch is closed, the controller acts as a P-I controller. On the other hand, if the switch is
open, it is a simple P- controller. The switch is activated by the position of the actuator. If the actuator
is operating in the linear range, the switch is closed, and the controller is in P-I mode. But whenever
the actuator is in the saturation mode, the switch is automatically opened; the controller becomes a P-
controller. As a result, any windup due to the presence of integral mode is avoided.

10



Figure 6. Alternative scheme for anti-integration windup action

Another technique for anti-windup action is illustrated in Figure 6. It is assumed that the slope of the
actuator in the linear range is unity. As a result, when the actuator is operating in the linear range the
error

is zero, and the controller acts as a PI controller. But when the actuator is in saturation mode,
the error

is negative for a positive . This will reduce the integral action in the overall control loop,
hence preventing integration windup.

5. Conclusions
The Ziegler-Nichols method is an empirical method and is obtained from extensive experimentation
with different processes. There is no firm mathematical basis behind these selections but an intuitive
one. This method has proven to be very useful without sufficient a priori knowledge about the system.
Despite simplicity, this method lacks robustness that is offered by other techniques of tuning available.
However, the other methods of tuning require initial values of tuning parameters, which can most
suitably be determined using the Z-N method in the absence of a system model. Further modifications
to the controller designs can be used to reduce basic errors that occur in the PID control system.

6. Technical References
[1] Visioli, "Tuning of PID controllers with fuzzy logic", IEE Process-Control Theory Application, Vol.
148, No. 1, January 2001, Pg. 1 to 8.
[2] Ziegler, Nichols, Optimum Settings for Automatic Controllers", American Society for Mechanical
Engineers transaction, 1942, Pg. 759 to 765.
[3] D.P. Eckman, "Automatic process Control", Wiley Eastern, New Delhi, 1958,
ASIN: B0000CK1WV
[4] Jacques F. Smuts, "Process control for practitioners", OptiControls Inc., 2011, ISBN-13:
9780983843818