24.10 Empirical optimization
With missing distance data can empirically be optimized by means of the self-optimization or in manual attempts.With
the attempts for empirical optimization the following is to be considered:
w
It is to be guaranteed that correcting variable and controlled variable take never forbidden values!!!
w
The conditions for the attempts should be always alike, in order to win comparable statements.
w
The test sequence must be oriented at the goal of the optimization: Leadership- or interference behaviour.
w
The operating point of the controller must be alike with the attempts.
The control parameters are to be set as follows with their first use:
Xp maximum: to the largest adjustable value,
Tv relatively large: time max., which the controlled system needs for distinct beginning of the reaction.
Tn large:time max., which the controlled system needs for the entire process.
The time requirement for an empirical optimization is large. In order to achieve an useful result in relatively short time ,
the following is recommended for appropriate procedure results:
Adjust Tn=Tv=0 and Xp largest possible(p-controller). The Xp is reduced from attempt to attempt, as long as the control
is sufficiently stable. If it becomes too unstable, then the Xp is to be increased and next step is *.
Measure lasting offset: If it is sufficiently small, then the optimization is successfully terminated (P). If it is too large,
then the controlled system is better regulated with PD (adjust Tv relatively large and next step is Ö ).
Reduce Xp from attempt to attempt, as long as the control is sufficiently stable. If it becomes too unstable, then the
next step is £.
Tv is to be made smaller and determined whether the regulation can be sufficiently stabilized again. If, then it the next
step is ¢, if not, then Xp is to increase and the next step is ¤ .
Determine whether with the procedures Ö and ä the Xp was substantially made smaller. If, then the next step is <,
if not, then the controlled system better is pi-regulated (Tv set to 0 and the next step is >).
Measure lasting offset. If it is sufficiently small, then the optimization is successfully terminated (PD). If it is too large,
then the controlled system is better PID-regulated (no longer change Xp and Tv and the next step is >).
Tn is adjusted largely and reduced from attempt to attempt, as long as the control is sufficiently stable. If it becomes
too unstable, then the Xp is to be increased, and the optimization is successfully terminated (PID or pi).
g
For the controlled variable (process value X) the empirical optimization is substantially improved with a
writer (or trend function of the engineering tool) in time requirement and quality, and evaluation of the
test results is clearly simplified.
g
The procedure mentioned can only with restrictions be generalized and does not lead to a clear
improvement of the behavior with all controlled systems.
g
Changes of the operating point (Y0), the switching point distance (Xsh) and the lasting switching
period (Tp1 and Tp2) lead to results, which can be better or worse. With 3 - Point - step controllers
TM must be adjusted to the real running time of the conncted actuator.
Controller 9499 040 50611
Empirical optimization 190
To Next Page
Loading...
