U.S. patent number 3,845,370 [Application Number 05/387,267] was granted by the patent office on 1974-10-29 for credibility testing in sampled-data systems.
This patent grant is currently assigned to International Business Machines Corporation. Invention is credited to John P. Mantey.
United States Patent |
3,845,370 |
Mantey |
October 29, 1974 |
CREDIBILITY TESTING IN SAMPLED-DATA SYSTEMS
Abstract
A closed-loop sampled-data servomechanism wherein each sample of
the aperiodic feedback signal from a digital transducer is tested
to determine if it is credible or valid feedback information. If it
is credible, the feedback sample is used as feedback information by
the servomechanism. In addition, the valid sample is stored in a
memory. Thus, the memory always contains the most recent valid
feedback sample. If a particular feedback sample is not valid, the
most recent prior valid feedback sample is obtained from the memory
and is used by the servomechanism to construct a substitute for the
current invalid feedback sample.
Inventors: |
Mantey; John P. (Boulder,
CO) |
Assignee: |
International Business Machines
Corporation (Armonk, NY)
|
Family
ID: |
23529162 |
Appl.
No.: |
05/387,267 |
Filed: |
August 10, 1973 |
Current U.S.
Class: |
318/271 |
Current CPC
Class: |
H02P
7/2855 (20130101); G05B 9/02 (20130101); G05B
2219/42293 (20130101) |
Current International
Class: |
G05B
9/02 (20060101); H02P 7/285 (20060101); H02P
7/18 (20060101); H02p 005/16 () |
Field of
Search: |
;318/314,318,326-328 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Schaefer; Robert K.
Assistant Examiner: Langer; Thomas
Attorney, Agent or Firm: Sirr; Francis A.
Claims
1. A closed-loop speed control servomechanism operable to variably
energize a motor in a start-run-stop sequence to achieve a desired
acceleration, constant speed run and deceleration profile,
comprising:
digital tachometer means driven by said motor and operable to
provide sampled-data period feedback data in accordance with actual
motor speed, said feedback data being aperiodic during acceleration
and deceleration, and being substantially periodic during constant
speed run,
speed memory means containing desired period data and permissible
period window data which define the motor's desired acceleration,
constant speed run and deceleration profiles,
means providing motor start and stop commands,
means including clock means connecting said motor start and stop
commands in controlling relation to said memory means to
periodically interrogate said memory means,
period testing network means receiving as inputs said period
feedback, said desired period data and permissible period window
data, and operating thereon to test said feedback data against a
permissible period window data to determine if said feedback data
is valid or invalid,
hold network means controlled by said testing network means when
said feedback data is valid and operable to hold said valid
feedback data, and
means including said hold network means operable to selectively
supply current valid feedback data or to substitute therefor
feedback data which is a function of prior valid feedback data for
use by said servomechanism, in accordance with said current
feedback data passing or failing said
2. The servomechanism as defined in claim 1 including period
comparison network means controlled by said period testing network
and receiving as inputs said period feedback data and said desired
period data, and operable to supply only valid feedback data to
said hold network means, and means connecting the contents of said
hold network for use by said
3. The servomechanism as defined in claim 2 including command speed
network means, and means summing the output of said command speed
network means
4. The servomechanism as defined in claim 3 wherein the output of
said command speed network means is modified by the contents of
said hold network means only when a period error exists.
Description
BACKGROUND AND SUMMARY OF THE INVENTION
The present invention pertains to the field of closed-loop
sampled-data servomechanisms for use in controlling a process.
More particularly, the present invention pertains to closed-loop
motor speed servomechanisms which derive aperiodic speed feedback
samples from a digital transducer driven by the motor being
controlled.
An inherent characteristic of a sampled-data servomechanism is that
a control decision is made at one sample time and, as a result of
this decision, a mode of process control is instituted, and then
maintained, in essentially open loop fashion, until a subsequent
sample time. At this subsequent sample time, new feedback
information is available and the mode of process control is
updated.
The present invention operates upon the principle that whatever the
mode of process control, the range of possible subsequent feedback
samples can be predicted. If the subsequent feedback sample does
not fall within this range, the particular sample is not credible
and is classified as invalid. Rather than use the invalid sample to
effect a change in the mode of process control, hoping that the
invalid sample at least shows the trend in a process variable and
any resulting overcorrection will appear in subsequent valid
samples, the present invention discards the invalid sample and
substitutes a sample based upon the most recent prior valid sample
or samples.
More specifically, the present invention provides a testing network
or window which tests each feedback sample. Since the validity of a
feedback sample is a function of the sample rate and may be
variable with other process parameters, the window of acceptable
parameter limits is variable with time and may be variable with
other process parameters. In one embodiment, the arrival of each
valid feedback sample is effective to reset the testing network, in
anticipation of the need to test the subsequent feedback
sample.
While prior art devices have recognized that failures may occur in
servomechanisms, prior art solutions have been in the nature of
redundant systems, majority-voting techniques, or the imposition of
the requirement that a given feedback condition must be sensed
twice or more before corrective action is taken.
The present invention provides the simple control technique of
testing the feedback information for validity, and then using the
information if it passes the test, or, alternatively, making a
substitution based upon the most recent valid feedback if the
current feedback fails to pass the test.
The foregoing and other features and advantages of the invention
will be apparent from the following more particular description of
preferred embodiments of the invention, as illustrated in the
accompanying drawing.
BRIEF DESCRIPTION OF THE DRAWING
FIG. 1 is a schematic showing of the invention applied to process
control;
FIG. 2 is a graph showing the operation of FIG. 1's process
variable window;
FIG. 3 is a schematic showing of the invention applied to a speed
control process for a direct current motor;
FIG. 4 is a graph showing the contents of FIG. 3's speed memory or
store, whose contents define the desired tachometer period, i.e.,
motor speed, for a start-run-stop sequence, and also showing the
acceptable period window, i.e., valid feedback information, which
will be accepted and used in the closed-loop feedback system;
and
FIG. 5 is a graph showing the contents of FIG. 4's speed command
network.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
In FIG. 1 process 10 is generally shown as having a process
variable controller 11, a sampled-data first process variable
sensor 12, and a second process variable sensor 17. This general
showing of a controllable process is meant to embody a large
variety of processes, such as the speed control process of FIG. 3,
or, for example, a temperature control process wherein controller
11 compares the output of sensor 12 to a set point temperature and
controls the temperature of the process to the magnitude defined by
this set point.
The control set point or command for process 10 is received by
controller 11 by way of conductor 13. Actual process variable
information, corresponding to the set point characteristic, is
received by controller 11 by way of conductor 14. Controller 11 is
constructed and arranged to compare the command variable magnitude
on conductor 13 with the actual variable magnitude on conductor 14
and to institute corrective action within process 10 in order to
bring the magnitude of the process variable, for example
temperature, to the command magnitude.
Sensor 12 operates to periodically or aperiodically sample the
actual magnitude of the process variable and to supply sampled-data
feedback on condutctor 15 to process variable window network
16.
Network 16 operates to compare the magnitude of each individual
feedback sample to a reference magnitude and to make a decision as
to whether or not the particular feedback sample is a valid
feedback sample or an invalid feedback sample. By definition, a
valid feedback sample is a feedback sample which falls within the
range of expected values. This range is defined by the extent to
which controller 11 can achieve change in the process variable.
This change is always dependent upon the time between feedback
samples, and may additionally be responsive to a second process
variable. When the expected range of feedback magnitudes is in fact
responsive to and variable with a second process variable, a second
process variable sensor 17 is associated with process 10 and
provides an output on conductor 18 to control the testing parameter
utilized by network 16 to test feedback 15.
If a particular feedback sample passes the required test, this
sample is provided as valid feedback information on conductor 19.
This valid feedback sample is stored in last valid feedback memory
network 20 and is additionally applied to conductor 14 to be used
as current feedback information by controller 11. In addition,
valid feedback information is effective to reset window network 16,
by way of conductor 21, preparing this network for a subsequent
feedback sample. Since the ability of network 16 to test a given
feedback sample is time dependent, it is necessary to reset this
network to a zero time state such that its testing parameter may
again begin its time dependent variation, awaiting a subsequent
feedback sample.
By way of example, network 20 may be a register which holds the
last valid feedback in binary form. In addition, the output of
network 20, which will be used only if subsequent feedback samples
fail to pass the test imposed by network 16, may be a substitution
which is based upon a most recent prior valid sample or samples.
For example, this substituted signal may be a valid sample whose
magnitude decays with time.
Assume for the moment that the next feedback sample provided by
conductor 15 is in fact invalid feedback, as occasioned, for
example, by momentary failure of sensor 12. In this case, this
particular feedback sample fails to pass the test imposed by
network 16. As a result, conductor 22 becomes active and enables
memory network 20 to supply a substitute signal based upon the last
valid feedback information contained therein to conductor 14 by way
of conductor 23. As a result, process controller 11 continues to
control process 10 by using substitute signals based on the last
valid feedback information received from sensor 12.
This mode of process control continues until a valid feedback
sample is obtained from sensor 12. When this valid feedback sample
is obtained, it is stored in memory network 20, it is supplied to
conductor 14 for use by controller, and window network 16 is reset
to its zero time state, awaiting the next feedback sample.
FIG. 2 is a graph showing schematically the operation of window
network 16. In this graph a nominal expected feedback magnitude is
represented by broken line 30. At time t0 window network 16 is
reset to its zero time state wherein the window test to be
subsequently instituted by network 16 is defined by solid lines 31
and 32. By way of example, curve 31 defines the magnitude of the
process variable sensed by sensor 12 if controller 11 institutes a
100 percent ON control of the condition changing portion of process
10 at time t0. This in a temperature control process, for example,
would constitute maximum energization of an associated heater, or
in a speed control process, would institute full energization of a
motor. Curve 32 defines the expected feedback which can occur with
variation in time if controller 11 institutes a 100 percent OFF
condition of the condition changing portion of process 10 at time
t0. It is known that valid feedback information must exist
somewhere between curves 31 and 32. Any feedback sample which is
either above curve 31 or below curve 32, is invalid feedback
information.
Assume for the moment that the next feedback sample is obtained at
time t1, and that this feedback sample is valid. This feedback
sample is effective to reset window network 16 such that its
testing parameters are now defined by broken lines 33 and 34. The
testing window defined by these broken lines is identical to the
testing window defined by curves 31 and 32, and is time
displaced.
FIG. 3 is a schematic showing of the present invention as applied
to a process involving the speed control of direct current motor
40. As an exemplary environment of such a process, motor 40 is
directly connected to tape drive capstan 41, associated with
magnetic recording tape 42. In this environment it is preferable
that motor 40 be a high torque - low inertia motor of the type
described in U.S. Pat. No. 3,490,672, issued to G. A. Fisher and H.
E. Van Winkle.
The feedback sensor in this process is a digital tachometer 43. As
is well known to those of skill in the art, such a tachometer
provides a feedback pulse on conductor 44 for each unit of rotation
of motor 40. By way of example, tachometer 43 may provide 500
pulses for each revolution of motor 40.
The speed of motor 40 is measured by period measurement network 45.
This network is effective to count the number of cycles of high
frequency clock 46 which occur between adjacent output pulses from
tachometer 43. Network 45 utilizes these two inputs to derive a
tachometer period measurement, T, on conductor 46.
Capstan motor 40 operates in a start/stop mode and the apparatus of
FIG. 3 is operable to control the speed of this motor during
acceleration, constant speed run and deceleration of the motor.
Speed memory network 47 contains a stored memory of the desired
tachometer period, Tr, which should occur during these three phases
of a motor cycle. In addition, this network contains a stored
memory of the acceptable deviation from this desired tachometer
period, this deviation being defined as the acceptable period
window .DELTA.Tr.
When a start command is received, by way of conductor 48, the
accelerate and subsequent constant run contents of memory 47 are
interrogated in a continuous fashion, as controlled by clock 49.
Subsequently, when a stop command is received on conductor 50, the
deceleration portion of memory 47 is interrogated, again as
controlled by clock 49.
FIG. 4 shows a representative content of memory 47 wherein solid
line 62 represents the desired tachometer period Tr, or command set
point for the motor's accelerate, constant speed run, and
decelerate intervals. The .DELTA.T range defined by broken lines 63
and 64 defines the acceptable period window .DELTA.Tr. At any given
time, the actual period measurement, T, supplied by network 45,
must fall within the bounds of broken lines 63 and 64 in order to
be classified as valid feedback information.
Period comparison network 51 and period testing network 68 are
effective to compare the period feedback information T on conductor
46 with the desired tachometer period information Tr received from
memory 47 by way of conductor 52, and with the acceptable period
window information .DELTA.Tr received from memory 47 by way of
conductor 53 to determine if the feedback information is valid and,
if valid, to compute the period error Te for use to control the
speed of motor 40.
More specifically, network 68 compares the actual period T to the
desired period Tr to determine if the difference between these two
periods is less than the acceptable period window. Network 68 may,
for example, be an arithmetic and logic unit controlled to perform
the test .vertline.T-Tr.vertline.<.DELTA.Tr. In other words,
equation .vertline.T-Tr.vertline.<.DELTA.Tr must be satisfied.
If this test is passed, network 51 is enabled by way of conductor
69. Network 51, which may be a similar arithmetic and logic unit,
now computes the period error Te by way of the equation Te=K(T-Tr),
where k is a constant. The computed period error is supplied as
valid feedback to conductor 54, where it is stored in hold or
memory network 55.
This hold network may provide a steady magnitude signal to output
conductor 61. By way of example, network 55 may include a register
holding the binary value of Te, the register driving a bipolar
digital to analog converter whose output comprises conductor 61.
Alternatively, this signal magnitude or output may exponentially
decay toward the expected subsequent period error by providing the
predictor network described in U.S. Pat. No. 3,758,757, by O. R.
Buhler and J. T. Cutter.
Assuming for the moment that the next sample provided by network 45
fails to pass the test .vertline.T-Tr.vertline.<.DELTA.Tr, a new
calculation of the period error Te is not made, and the contents of
hold network 55 is retained for use by the servomechanism, the
invalid feedback information being disregarded.
The apparatus of FIG. 3 includes a speed command network 56 which
is controlled by a start command on conductor 57 and a stop command
on conductor 58 to provide a signal to power amplifier 59. This
signal is selected to normally energize motor 40 to cause the motor
to follow a desired acceleration, constant speed run and
deceleration profile. The output of network 56 appears on conductor
60 and is summed with the output of network 55 to variably control
the energization of motor 40. For example, if the particular
feedback sample derived from network 45 indicates that the motor is
slow, the output of network 61 is effective to add to the signal
provided by conductor 60 and to thereby increase the energization
of motor 40. On the other hand, should the motor be fast, the
output of network 55 is effective to subtract from the output of
network 56 and thereby decrease the energization of motor 40.
FIG. 5 shows the contents of speed command network 56 for the
accelerate, constant speed run, and decelerate intervals of the
motor. The magnitude of curve 65 is representative of the degree of
motor energization which will usually achieve the desired speed
profile for a start/stop cycle of motor 40.
While the invention has been particularly shown and described with
reference to preferred embodiments thereof, it will be understood
by those skilled in the art that various changes in form and
details may be made therein without departing from the spirit and
scope of the invention.
* * * * *