<Bob Clark (940806.1530 EDT)>
Bill Powers (940801.0900 MDT) Subject: Disturbance; one-way control;
briefly, the meeting
Your excellent and clear discussion of the canonical diagram is not
related to the reason for my restricting my proposed definition of
"disturbance" to "small" changes.
In Bob Clark (940704.1355 EDT) Subject: RKC-MORE PROG CONTROL, I had
suggested a definition of "control:"
*Control is the production of actions in opposition to the acts of an
*external entity. A system demonstrating such actions is a "Control
*System."
This definition was repeated later, Bob Clark (940713.1750 EDT),
*This definition of "Control" is illustrated by "The TEST for the
*Controlled Variable."
Paraphrasing B:CP pp 233-4:
*In this Test, a small change (a "disturbance") is applied to the
*variable of interest and the output of the system is observed for
*action in opposition to the applied change. If the opposing action
*is greater than accounted for by ordinary laws of physical science,
*the Test illustrates "Control."
These definitions require a definition of "disturbance" applicable to
The Test for "control."
For this purpose, I offered the definition in question [Bob Clark
(940713.1750 EDT) Subject: RKC - TERMINOLOGY]:
*A Disturbance is a small change in a variable external to the
*system. Its magnitude is "small" compared to the ordinary
*variations of this variable. Its duration is "small" compared to
*the time used for observing "the production of actions in
*opposition." To be considered a "Disturbance," it must be "small" in
*both respects.
The reason I propose to limit the magnitude of disturbance, is to
facilitate observation of possible "actions in opposition." If too
large a disturbance is used, the output capability of the system
could be over-whelmed. Thus the response could be over-looked.
Regarding the canonical diagram, it is presented in B:CP p 61 and p
274. The "system-environment equations" are included on p 274. They
need not be restated here.
However, the first sentence on p 275 states:
"These equations represent the steady-state conditions of the
"variables under the assumptions that only one such condition exists
"at a time, and that transient effects die out rapidly to zero."
In the steady-state situation, nothing is changing. The disturbance
is, perhaps, opposed, perhaps not. In this situation, without other
data, no conclusion on this point is possible.
For The Test, there must be some way to determine the difference
between "opposition" and "non-opposition." In some situations,
physical laws permit an estimate. But, perhaps better, is to examine
two Cases:
Case I: The disturbance is applied with the system disconnected from
the external variable. There are various ways this can be done. The
change in the external variable is observed as a function of the
disturbance.
Case II: The disturbance is applied with the system connected to the
external variable. The change in the external variable is observed
again as a function of the disturbance.
The difference, if any, in the two observations, indicates the
possible existence of a system controlling the external variable in
question. There could be other interactions and/or side-effects
involved.
My definitions presumed that some means is available for determining
the effects of the disturbance with the system disconnected. Perhaps
that condition should have been included as part of the definition.
Using "disturbances" as tools for identifying controlled variables
and investigating other properties of control systems is very
powerful. By varying not only the magnitude of the disturbance but
also its timing, the dynamic properties of the system can also be
determined.
Regards, Bob Clark
···
Subject: RKC - TERMINOLOGY, where I pointed out that: