|
Symptom:
Do you find that using auto-tuning usually does a good job in tuning that
controller, but is some cases, the performance does not reach desired goals?
Probable cause:
Auto-tune controllers must use mathematical equations to calculate the
PID settings. These settings are based on the controller performing a
series of step output bumps to determine the process gain, deadtime, and
time constant. As long as the process is linear and no other upsets occur,
these calculations should give reasonable results. However, in some real
world situations, auto-tune may not give the best results.
Corrective action:
The following situations are examples where other actions may be
appropriate.
|
Auto-tune may not: |
Corrective Action |
|
Account for valve deadband |
Manual tuning with higher gain will usually help
performance. |
|
Account for large changes in installed process gain
|
Characterizing a valve or other component will
enhance control. |
|
Automatically set filtering |
Any filtering needs to be documented and minimized. |
|
Automatically set sample periods |
Sample periods should be a function of process time
constant. |
|
Calculate a predicted settling time |
In some loops, the time for PV to reach and stay on
SP is important. |
|
Function if process is cycling or PV not near SP |
Manual techniques may be needed to stabilize a loop
or bring PV to SP. |
|
Always give the best tuning for interacting loops |
Interacting loops can be decoupled with fast and slow
tuning. |
|
Correctly interpret stick/slip limit cycling |
Correct interpreting the PV cycling may point to
valve maintenance. |
All
of these corrective actions are addressed with proper training of those who
troubleshoot and tune loops.
LOOPtips Archive |
