CONTROL ACTIONS

Two step control action

This can be defined as 'the action of a controller whose output changes from one state to another due to a
variation in its input' One example of this control is that of a float operated filling v/v say for a cistern. In
normal condition the output of the float is nil and no water passes through the valve, should the water
level drop the float detects this and operates the valve to change to its second state which is open and
water flows. When the level reestablishes then the float controls the valve to return to its primary state
which is closed. In this way the float is controlling the water level by changing the valve between two
different states. ! more realistic system is shown below.
The system wor"s as follows# the level drops until the lower float is uncovered, the controller
detects this and opens the filling valve, the filling v/v remains open until the top float is covered and then
the controller shuts the valve
The distance between the floats is termed the 'Overlap' i.e. the distance between the high and
low controlling values $ on some systems this can be altered by altering the high or low set point of the
controller, in the above system this would mean altering the position of the floats %
If there where any delays or lags in the sensing side, say the float switch was a little stic"y or
the filling v/v was slow to fully open then the level would fall below rise above the low and high set points
respectively. This is termed 'Overshoot', it can be seen if the controller 'response to change' time was
speeded up so the overshoot could be reduced.
The system may be represented in bloc" form as follows#
The measuring unit signal $ in this case an electrical on/off % is the Measured value on which
the controller operates. The signal is being 'fed bac"' to the controller hence to measured value is Feed
back for the controller# i.e. the controller can see the direct results of its action.
Feed forward signals are sometimes used on systems which have an inherently high &rocess
'ag# an example of this may be on a (arine )iesel engine *ac"et fresh water cooling system where part of
the control is that the inlet temperature to the engine is monitored and fed forward to the controllers,
should the temperature at inlet rise then conse+uently the outlet temperature must also rise. !s the rise
has already been detected then the controller can start increasing the sea water cooling to the *ac"et
water coolers even though no temperature rise on the outlet from the engine may have been detected.
This type of control, as it ta"es no account of what is happening to the process $ is the engine running and
hence re+uires the extra cooling or is it stopped % is not very accurate and normally $ and as in this case %
re+uired ,eedbac" to improve it.
The actions of controllers having variable output
Proportional control action
This where the change in output signal from a controller is proportional to the change in input signal
The control can be summed up in the following#
Output - .onstant x )eviation
Output this is the output from the controller and goes to the control element $ say the filling v/v on the
previous example i.e. the piece of e+uipment that actually alters the process.
Constant This is the '/ain' of the controller, as the output varies with the deviation, the amount it varies
can be altered.
0ay if the deviation changes by one unit the output changes by one unit, hence the gain is
one. If the output varied by two for the same one change in deviation then the gain would be two.
0imilarly if the change in output was one half a unit for a one unit change in input then the gain would be
half. !nother way used to describe /ain is '&roportional band', here a gain of one is described as a
proportional band $&b%of 1223. ,or a gain of two the &b is 423, and for a gain of a half the &b is 5223,
hence it can be seen that the magnitude of the &b is opposite to the gain.
Deviation This is the difference between the set point of the controller and the measured value. If the set
point was one unit and the measured value was two units the deviation would be one unit.
)eviation - 0et point (easured value
The important thin" to remember is that the narrower the &roportional band the higher the
gain and hence the higher the output varies for a change in deviation, this has the effect of ma"ing the
controller control the process +uic"er by operating the controlling element more for smaller variations
measured value. This has the negative effect as will be seen of ma"ing the system unstable
OFFSET
,or a proportional controller to wor" there must be an deviation, if the deviation is 6ero then the controller
output to the controlling element is 6ero. ,or the example of the tan" and filling v/v obviously this is not
possible, with the water constantly flowing out of the manual outlet v/v then the filling valve $ or
controlling element % must always be some degree open. If the level is at the level of the set point then
the output is nil, the filling v/v is shut and the level drops, deviation occurs and the filling v/v opens. with
this it can be seen that the system is not stable# what would happen in reality is the level would change
$ say the level was low and was now rising% until it reached a point close to the set point where the
deviation multiplied by the gain would give an output signal to the filling v/v such that the flow of water in
to the tan" e+ualled the flow of water out of the tan".
This deviation is called 'offset'
Therefore a proportional only controller when in e+uilibrium must have offset
The amount of offset will be determined by the /ain, for the tan" system if the gain is high the deviation
can be small for a larger output
The offset will increase for increased loads on the system i.e. if the outlet v/v on the example
where to be opened further obviously the filling v/v would have to be opened further, and hence the
deviation $ offset % to give the re+uired output would have to be greater.
,or the system above all the control would be positive as the filling v/v would only be open if
the level was low and hence the offset would always be positive, when the level rose above the set point,
say caused by 'ags leading to Overshoots or the filling v/v lea"ing slightly the deviation would be negative
and the output 6ero.
Proportional action an instabilit! "#unting$
!s the gain increases so the output increases for smaller and smaller changes in deviation, eventually the
response starts to loo" similar to that of a two step controller with the control valve flying from full open
to full shut with the slightest deviation from the set point. This would be o.". if the system was devoid of
all 'ags, with lags however, particularly between the controller and controlling element, there is a
tendency for 'over shoot'.
This can occur with reduced gain when the process lags are increased, for systems with a very
large lags even small changes in gain can seriously effect the stability of a system and especially its ability
to resist step $ or rapid % load changes.
,or smaller values of gain the system can be set up to have minimum of hunt and be self
stabili6ing .
Split range control" negative an positive offset$
! system could be designed to control both the outlet valves and inlet valves $this is what is seen on the
feed water system level control with the spill and filling being controlled from the one controller% # here
the controller would be set up so that when the level is at the set point its output is mid range $ say for a
controller operating in the 7 to 14 psi range this would be 8 psi%
The control valves would be set up so that one, say the filling v/v would go from close at 8.4
psi to open at 14 psi, and the spill v/v would go from close at 9.4 psi to open at 7 psi. The 1 psi in the
middle is called the 'Deadband' and is there to ensure both v/vs are not open at the same time.$ The v/v
acting to open with increasing input signal is called 'Direct Acting' and the v/v closing with increasing
pressure is called 'Reverse Acting'%
It can be seen that there can now be an offset in the positive i.e. water being used and hence
the ma"e up v/v has to be open and in the negative i.e. there is too much water entering the system and
the spill v/v's have to be opened.
Offset is not a desired result of the control of a system, however for proportional only
controllers this is a direct conse+uence. That is why for all controllers performing important functions#
including the ma"e up/spill system controller above other types of controlling action are added to remove
the offset
Other t!pes of controlling action
Integral action " an the re%oval of offset $
Integral action is defined as the action of a controller whose output signal changes at a rate proportional
to the input$deviation from the set point% signal.
What this means is that if a controller has a constant deviation then the integral action
controller will increase its output continuously until it reaches maximum $often referred to as 'Saturation'%
If the deviation is 6ero the integral action controller is 6ero
If the deviation is small the rate the controller output increases by is small
If the deviation is great then the integral action controller will rapidly increase its output.
ntegral action is included in proportional controllers to re!ove the inherent offset of the
proportional action" the offset is the deviation the integral action re#uires to alter the output
ntegral action ti!e
The amount of integral action, or how fast the integral action increases or decreases the output for one
unit of deviation is expressed as the time ta"en to repeat the proportional action after a stepped change
in input.
:ate of increase of output - )eviation x Integral action time
What this means is say the load changes in the simple filling system example by the manual
v/v being opened and the level suddenly dropping by a foot, the proportional action will see this load
change and give a stepped change in output i.e. if the foot drop in water level e+uals a change in input
signal to the controller of one unit away from the set point, the controller will give a stepped change in
output e+uating to the gain $which is say two % times the deviation $ one unit% which e+uals a change in
output of two units.
Whilst all this has been going on the integral part of the controller has seen the deviation and
has decided to increase the rate of output by an amount e+ual to the deviation multiplied by the integral
action time. The time ta"en for the output to increase by a further two units $ remembering that this was
how much the proportional action changed the output% is the integral action time and is measured in
seconds.
The shorter the integral action time $ less seconds % the more rapidly the integral part of the
controller will increase the output# The longer the integral action time the slower the integral action will
increase the output.
! common way of expressing integral action is in ':epeats per minute', integral action time is
seconds per repeat, hence if the I!T is 12 $ seconds per repeat %, this would e+uate to 12/;2 minutes per
repeat, or more simply 1/; mins. The repeats per minute is therefore ;.
ntegral action and stabilit$
The introduction of integral action into a controller introduces an extra time lag, remembering
the diagram showing that the integral action will ta"e time to increase the output to a stepped load
change, whereas the proportional action will give a stepped change. 'ags introduce instability, hence it
would more difficult to find settings which give a stable output.
ntegral action is alwa$s used with proportional action
&erivative action
The definition of this is the action of a controller whose output is proportional to the rate of change of
input.
That is to say for the filling system if the level was falling slowly the output of the controller
would be small. If the level was flying down at a great rate of "nots then the derivative controller would
give a high output. It is +uite obvious that the derivative action ta"es no account of the deviation from the
set point but is only interested in the rate of change of deviation and hence#
Derivative action b$ itself cannot be used for control%
The purpose of adding derivative action to a controller is to increase the responses that
deviation is removes as +uic"ly as possible. That is to say if the level in our filling system is falling the
proportional action will increase the filling at a the same rate, however as with seen, if there is a lag in the
system particularly between the controller and controlling element# then there is a possibility of instability
and a hunt.
If we where at the point where the water level was *ust starting to fall less rapidly but not at
the point where it was actually starting to rise, all the proportional and integral action see is a large
deviation and so "eep the water v/v wide open, the derivative action, however, sees this slowing down of
the drop in water level, its output is dependent on the rate of change and hence reduces, and so the
output from the controller reduces.
The introduction of derivative action introduces a stabili&ing effect into a control s$ste!
Derivative action ti!e
Output - )erivative action x rate of change of input
Derivative action 'coefficient( This is described as the time the proportional action ta"es to
repeat the derivative action after a ramped $ or constant rate of change% input. The units are seconds.