ATI
Q45D-ODO Optical D.O. System
Part 7 –PID Controller
O & M Manual
Revision C (7/09)
55 -
D
Derivative gain. The addition of derivative control can be problematic in
many applications, because it greatly contributes to oscillatory behavior.
In inherently slow chemical control process’, differential control is
generally added in very small amounts to suppress erratic actions in the
process that are non-continuous, such as pumps and valves clicking on
and off. However, as a starting point for chemical process control, its best
to leave the “D” term set to 0.
Based on these descriptions, the focus on tuning for chemical applications really
only involves adjustment of “P” and “I” in most cases. However, increasing both
increases the response of the controller. The difference is in the time of recovery.
Although combinations of high “P’s” and low “I” will appear to operate the same
as combinations of low “P’s” and high “I’s”, there will be a difference in rate of
recovery and stability. Because of the way the algorithm is structured, large “P’s”
can have a larger impact to instability, because the proportional gain term
impacts all the other terms directly. Therefore, keep proportional gain lower to
start and increase integral gain to achieve the effect required.
Many of the classical tuning techniques have the user start with all values at 0,
and then increase the P term until oscillations occur. The P value is then
reduced to ½ of the oscillatory value, and the I term is increased to give the
desired response. This can be done with the Q45D controller, with the exception
that the I term should start no lower than 1.0.
If it appears that even large amounts of integral gain (>20) don’t appreciably
increase the desired response, drop I back to about 1.0, and increase P by 1.00,
and start increasing I again. In most chemical control schemes, I will be
approximately 3 times the value of P.
7.3
Classical PID Tuning
Unlike many high speed position applications where PID loops are commonly
used, the chemical feed application employed by this instrument does not require
intense mathematical exercise to determine tuning parameters for the PID. In
fact, the risk of instability is far greater with overly tuned PID control schemes. In
addition, many of the classical mathematical exercises can be damaging or
wasteful in the use of chemicals when the process is bumped with large amounts
of input error to seek a response curve. Because of this, the general adjustment
guidelines described in section 8.2 are sufficient for almost all application tuning
for this instrument. Beyond this, many sources are available for classical tuning
methods.
