136.

00-3

Instrumentation & Control - Course 136

CONTROL MODE SETTINGS

The quality of control obtained from a particular system

depends largely on the adj ustments made to the various mode
settings.

Many control loop components change their phase shift or

gain under different load conditions. Some practical
examples to demonstrate this concept of nonlinearity would
include:

(a) level control in a non uniform tank,

(b) flow control from a square root scale,

(c) variable valve ~p as flow increases.

As a result of this nonlinearity, controller adjustments

which provide good stability at' one load may produce severe
cycling at an increased or decreased load.

One solution to this problem is to compromise between

stability and recovery time. Control adjustments are made to
provide slight cyclic recovery under the most unstable load
and set point conditions. This will provide less than opti-
mum recovery under other load and set point combinations but
ensure stable response.

Two general methods of controller tuning to provide

optimum control of a given process are:

1. Ziegler and Nichols Method (ultimate sensitivity).

2. Process Reaction Curve ~1ethod.

Ultimate Sensitivity Method

This is a method of determining the optimum controller

settings by referring to a set of equations developed by
Ziegler and Nichols. The ultimate sensitivity method was
derived empirically by examining the general characteristics
of many industrial control loops.

November 1981 - 1 -
 136.00-3

The fundamental information required is that proportion-

al band value which will cause constant amplitu~e process
cycling and the reBul ting period of such a cycle. In order
to obtain this cycling, the process is first brought to the
set point on manual control with reset and derivative modes
eliminated. (0 RPM. 0 min.) control is then switched to
automatic and the proportional band is narrowed until the
process just cycles with constant amplitude following an
applied. disturbance. (Set point shift.) The process is now
in a state of marginal stability so that the loop gain is
known to be one and the phase lag is 180 degrees. The appro-
priate Ziegler and Nichols (Z/N) formula can then be used to
find the optimum controller settings.

Ziegler and Nichols Formula

Ultimate Proportiona1 Band (PB u ): That proportional

band setting which results in constant process cycling
following a disturbance.

Ultimate Period (p): The period of the constant ampli-

tude process cycle.

Proportional Onl~

-%PB = 2 PB u

Proportional Plus Reset (PI)

%PB = 2.2 PB u
P
R ~ (MPR)
1.2

proportional Plus Reset Plus Derivative (PID)

%PB = 1.6 PB u

R ~ ~ (MPR)

D = 8P (MIN)

These general formulas can be related to the theoretical

control modes developed in lesson 136.00-2.

Consider a process cycling under straight proportional

control. The proportional band setting must be the ultimate
proportional band and the loop gain is one.

- 2 -
 136.00-3

In order to achieve the quarter decay ratio from con-

stant cycling, each half cycle must have one half the magni-
tude of the previous half cycle. (Visualize that the con-
troller response to the positive deviation results in the
subsequent negative deviation). The loop gain must be reduc-
ed from one to one half. The effect of the control response
to a positive deviation will be a negative deviation of one
half the magnitude. The resulting positive deviation will
now be one half the negative deviation so that the current
positive deviation is a quarter of the original quarter
decay ratio.

The loop gain is halved by halving the controller gain

which would require doubling the ultimate proportlonalband.
100
Recall the general relation: %PB =
Gain

Halving the control gain: 100 100

%PB = = 2 x
.5 Gain Gain

Straight Proportional: %PB = 2 PB u

The operating proportional band must be widened when

reset is apded to correct for' the destablizing effect of
reset. (Increased phase lag.)

Notice that the operating proportional band will be 10%

wider for PI control than for straight proportional. (PB =
2. 2PB u )' The proportional band can be narrowed when deriva-
tive mode is added since closer control will be possible
without danger of instability. (Rate introduces a phase
lead. ) Notice that the proportional band setting for a three
mode controller will be approximately 80% of the setting
required for straight proportional. (PB = l.6PB u ')

From empirical studies it has been found that the maxi-

mum lead from a proportional plus derivative controller will
have a practical limit of 40 degrees. This phase angle can
be used to calculate a general derivative time compared to
the cycle period.

Proportional Plus Derivative

2nD
Tan ( ¢) = P

2nD
Tan ( 40 ) = P

2nD
.839 = p-
.839 P
D =
~ = 7.49

- 3 -
 136.00-3

For ease of recall and to ensure stability, the more

conservative time of PIS coula be selected. This would
correspond to a phase lead of 38 degrees which is still
reasonably close to the practical limit.
The maximum lag that a proportional plus reset control-
ler should contribute at cyclic conditions is known to be 11
degrees. The general reset time can be calculated using the
following phase relationship.

P
Tan ( 8) =
2TIR

P
Tan (11) = 2TIR = .194

P
R = = _P- (MPR)
2TI(.194) 1. 22

The lead contribution of a three mode controller should

be approxima tely 25 degrees at. process eyel10 cond i tions.
The physical lead limit of 40 degrees would still apply to
the derivative component and reset rate can be increased
until the total lead of the combination is reduced to 25
degrees. Notice that more reset effect is possible with PID
control than with PI control due to the corrective action of
derivative mode.

Tan (25 ) = (h D P
P 2TIR)

2rr(P/8) P
.466 = - 2rrR
P

2TI P
.466 =
8
- 2TIR

P
2 TIR
= .319

R = P (MPR)
2

- 4 -
 136.00-3

Example
Determine the optimum three mode settings usi.ng the Z/N
formula for a process which cycles uniformly 3 times in 27
minutes with a proportional band setting of 65%.

%PB = 1. 6PB u = 1.6(65) = 104%

Ultimate Period = 27 = 9 minutes

"3

P
R = 2" = 2"9 = 4.5 MPR ( • 222 RPM)

P
D = "8 = "89 = 1.125 minutes

Reaction Curve Method

A mul ticapaci ty system response to a step disturbance
can be represented in general "as a dead time and a first
order response. The larger time constant is able to dominate
the response, and the delay <Bue to small time constants will
approach dead time.

PERCENT
PROCESS
/ FIRST ORDER RESPONSE

r-~-:('-
o

- 5 -
 136.00-3

The procedure requires that the control loop be opened

at the final actuator. This can be achieved by placing the
controller on manual. A step change input to the process can
now be achieved by changing the regulated pressure applied to
the valve actuator a discrete amount.

PROCESS e -_ _........
INFLOW

CONDENSATE
STEAM ~---DI':::J--1
PROCESS
PRODUCT

A!S

The resulting process change can be noted with respect

to time, or a trend recorder could be activated to record the
process reaction to the step change. The reaction curve can
then be analyzed to find the effective dead time (L) and the
maximum slope or maximum reaction rate in percent change of
process per minute (%C). In order to appreciate the reaction
curve method of controller tuning it would be worth-while to
review the concepts of capacitance and dead time.

Capacitance

The capacitance of a system is the ratio of the system

capac i ty to some reference variable. Cons ider level control
in a tank, the capacitance of the tank can be found by exam-
ining volume wrt level.

A large capacitance system will not be very sensitve to

small changes in input. A given step change in valve posi-
tion in a large capacitance system will result in a relative-
ly small ampl i tude process response. Such a process can be
said to be a low gain application. Consider the simplifica-
tion that the loop gain must be less than one (at the 180
degrees lag frequency) to ensure stability. (Control gain X
process gain <1). A relatively narrow band control could be
utilized on a low gain process control system.

- 6 -
 136.00-3

If the process was a low capacitance system it would be

sensitive to small changes in input. A given step change in
valve position in a small capacitance system will result in a
relatively large amplitude process response. The process can
be said to be a high gain application. To ensure stability a
wider proportional band must be used.
The capacitance of a system can be estimated by testing
the system with a step change in valve position. A response
factor could be determined to relate percent change in pro-
cess to a particular change in valve pesi tion. The larger
this response factor is, the smaller the capacitance of the
system must be and accordingly the wider the operating pro-
portional band must be to ensure stable control.

%PB a (Response Factor)

General Rule: increase process capaci tance, narrow the

%PB.

Dead Time

Consider a temperature control system that seems to work

adequately with a loop gain of one. The mixed water tempera-
ture is being controlled to the desired set point in a stable
fashion by regulating the inflow of hot water.

HOT WATER

COLD
-0+--------------+---..,..
WATER 0-..........
f. d.1
MIXED
WATER

- 7 -
 136.00-3

The location of the temperature sensor introduces a dead

time element into the control system.
ie, the time to transport the mixed water- from the
mixing tee to the detector.

Assume a slug of cold water enters the system and pulls

the mixed water temperature down.

This slug passes by the sensor causing a decrease in

transmitter signal applied to the controller. Note at this
time that the water at the mixing tee is back to normal temp-
era ture, the cold s1 ug is loea ted at the detector. The con-
troller responds by opening the control valve more and adding
more hot water to the system raising the mixed water tempera-
ture until the cold slug has passed by the sensor. The
control system has caused a slug of hot water to be admitted
to the system which in turn will flow past the sensor. The
controller will respond to the hot slug by decreasing the hot
water inflow to the system causing a second cold slug to be
introduced.

+
o
ORIGINAL COLD CONTROL INDUCED CONTROL INDUCED
SLUG HOT SLUG COLO SLUG

This process response will smooth out to approximate a

sine wave. The net result will be that the control system
will maintain the disturbance introduced cycle because the
dead time has caused the correction to be made well after the
upset occurred.

This constant amplitude oscillation can be attenuated by

decreasing the controller gain. The system will now provide
a controlled underdamped response with the wider proportional
band.

- 8 -
 136.00-3

General Rule: increase system dead time, widen the %PB.

From examining capaci tance, it was found that the %FB

setting must be proportional to the system response factor.
Including dead time in the consideration would require that
the %PB setting be proportional to the dead time interval.
%PB ~ (Response Factor x Dead Time)

140

TEMPERATURE
°c

130

MNUTE
o l.l 2.4
TIME

The percent change in valve position applied to the

system to cause a step response mus-t be noted. (%b.V).

The reSUlting process reaction curve can be examined to

determine the effective dead time (L), and the maximum slope
of the reaction curve. A straight line is drawn tangent to
the point of inflection on the reaction curve, a second
straight line is drawn horizontally from the initial process
value. The intersect of these two straight lines will define
the effective dead time (L) wrt the step initiation.

- 9 -
 136.00-3

The slope of the tangent line, which is the maximum

slope of the reaction curve will provide the process response
rate factor.

Once the dead time approximation, the percent change in

valve position, and the process reaction rate is known; then
the applicable reaction rate formula can be used to determine
the optimum mode settings.

Reaction Curve Formula

Proportional

100( %C)( L)
%PB =
HV

Proportional Plus Reset

1l0(%C) (L)
%PB =
%AV

L
R = -:3 (MPR)

Proportional plus Reset Plus Derivative

83(%C)(L)
%PB =
MV

R = 2L (MPR)

o = ~ (min)

Example

Refer to the previous reaction curve sketch in the

notes. This process response was achieved wi th a 12kPa
increase to a 20 - 100 kPag valve actuator. The controlled
range of the process is 100 - 200 degrees C. Determine the
three mode controller settings using the reaction curve
method.

Calculate the percent change in valve position:

12
MV = 80 x 100 = 15%

- 10 -
 136.00-3

Calculate the percent change in process per minute:

%C = 10
2.4 - 1.1
=~
1.3
= 7.69 %/rnin.

Determine the effective dead time:

from the curve, L = 1.1

%PB =
(83)(%C) (L) = 83(7.69)(1.1) = 46.8% = 47%
%AV 15

R = 2L = 2(1.1) = 2.2 MPR

L
D =
"2 = 1;1 = .55 minutes.

Zielgler and Nichols devised the reaction curve method

also as a result of empirical studies on a wide variety of
systems. The ultim~te sensitivity and reaction curve methods
will give very similar results for a given application. The
period of oscillation in a system will be approximately four
times the effective de\3.d time of the reaction curve for tha~
system. If the dead time is much greater than the first
order time constant (say six times), then only PI control
should be used since derivative mode would compound the con-
trol problem. If the dead ~ime is zero, then the %PB can be
set to zero (on/off control) since the system can be cha~act­
erized as pure first order and will not cycle.

ASSIGNMENT

1. A three mode controller is to be tuned to a particular

process which cycles 5 times in one half hour under
cri tical proportional control. The PB u = 60%. Deter-
mine the control mode settings required using the
Ziegler/Nichols method.

2. Repeat Question 1 with a process cycle time of 0.5 min-

utes. State the two mode combination which WQuld seem
to provide adequate control.

3. Repeat Question 2 for a process cycle time of 60

minutes.

- 11 -
 136.00-3

4. A process is analyzed by the reaction curve method and

the effective dead time is found to be zero minutes.
What will be the proportional band setting?

5. Would a large capacitance, slow response process be con-

trolled with a wide band or a narrow band controller
setting?

6. A control system is operating adequately under PI con-

trol with a 20%PB and 3MPR settings. A substantial cap-
aci tance increase results from new constructon addition
to the system but the control settings are inadvertently
left unchanged.

When this system was subjected to a slight operational

disturbance the process cycled for a prolonged time
(with a 15 minute period) approaching marginal stab-
ility. What recommendations would you make concerning
this system problem?

7. Analyze the following reaction curve sketch to determine

the correct 3-mode control setting. A 2% valve change
was made to initiate the process response.

PROCE55
5
PERCENT

L MINUTES

o 5 10 15 20 25 30
TIME

M. J. MacBeth

- 12 -