U.S. patent application number 09/751420 was filed with the patent office on 2002-07-04 for prognostics monitor for systems that are subject to failure.
Invention is credited to Johnson, Daniel P..
Application Number | 20020087258 09/751420 |
Document ID | / |
Family ID | 25021902 |
Filed Date | 2002-07-04 |
United States Patent
Application |
20020087258 |
Kind Code |
A1 |
Johnson, Daniel P. |
July 4, 2002 |
PROGNOSTICS MONITOR FOR SYSTEMS THAT ARE SUBJECT TO FAILURE
Abstract
Methods and devices for detecting and predicting parameter
deviations and isolating failure modes in systems that are subject
to failure. In a preferred embodiment, methods are provided for use
with engines, including aircraft, automobile, and industrial
combustion engines. However, numerous other applications are
contemplated. Such engines may be described as having monitor
points having current parameter values, where the monitor points
may correspond to single physical sensors or to virtual or inferred
monitor points having parameter values derived from multiple
sensors. Acceptable ranges, limits, and values for each of the
monitor point parameters may be provided for use with the present
invention. Parameters lying outside of the acceptable ranges may be
said to be in deviation. Ambiguity groups, including one or more
failure modes or physical causes of the parameter deviations may
also be provided. Parameter deviations, after optional filtering,
may generate deviation signals which may be followed by analysis of
the ambiguity groups to isolate the failure mode or modes causing
the deviation. Courses of engine operation ameliorating the failure
mode may be suggested. Methods are also provided for projecting
current trends into the future to predict deviations and isolate
failure modes early, prior to actual occurrence. One preferred use
for the methods is early detection and isolation of faults in
aircraft engines, leading to corrective action including early
preventative maintenance.
Inventors: |
Johnson, Daniel P.;
(Fridley, MN) |
Correspondence
Address: |
HONEYWELL INTERNATIONAL INC.
101 COLUMBIA ROAD
P O BOX 2245
MORRISTOWN
NJ
07962-2245
US
|
Family ID: |
25021902 |
Appl. No.: |
09/751420 |
Filed: |
December 29, 2000 |
Current U.S.
Class: |
701/114 ;
123/479 |
Current CPC
Class: |
F02D 2200/0404 20130101;
F02D 2200/0614 20130101; F02D 2200/503 20130101; F02D 41/1454
20130101; F02D 2041/1432 20130101; F02D 2200/0408 20130101; Y02T
10/40 20130101; F02D 2200/0606 20130101; F02B 77/08 20130101; F02D
35/025 20130101; F02D 41/22 20130101 |
Class at
Publication: |
701/114 ;
123/479 |
International
Class: |
G06G 007/70; G06F
019/00; F02M 051/00 |
Claims
What is claimed is:
1. A method for detecting system failure modes, wherein the system
has a plurality of monitor points, each point having an actual
parameter value, wherein the system has a plurality of failure
modes, the method comprising the steps of: providing a range of
desired parameter values for each monitor point; providing a
monitor point comparator function for said monitor points, wherein
said monitor point comparator function compares said actual
parameter value with said desired parameter value range and
generates a parameter deviation signal if said actual parameter
value is outside of said desired parameter value range; providing a
plurality of ambiguity groups, said ambiguity groups having at
least one failure mode associated with a monitor point deviation
signal; obtaining said actual parameter value; comparing said
actual parameter value with said desired parameter value for each
monitor point using said monitor point comparator function and
generating a parameter deviation signal if said parameter actual
value is outside of said desired parameter value range; and for
each generated parameter deviation signal, searching said plurality
of ambiguity groups for at least one ambiguity group that includes
said generated parameter deviation signal, and identifying at least
one failure mode associated with said generated deviation
signal.
2. A method according to claim 1 wherein the system is a combustion
engine.
3. A method for detecting combustion engine failure modes as in
claim 2, wherein said identifying step includes selecting a group
of ambiguity groups having said parameter deviation signal, taking
the intersection of the failure modes in said selected group of
ambiguity groups, and outputting said intersecting failure
modes.
4. A method for detecting combustion engine failure modes as in
claim 3, wherein said identifying step includes selecting only
those failure modes present in all ambiguity modes having said
parameter deviation signal.
5. A method for detecting combustion engine failure modes as in
claim 2, further comprising outputting suggested engine operating
parameters to ameliorate bad effects of said detected engine
faults.
6. A method for detecting combustion engine failure modes as in
claim 2, wherein said parameters are selected from the group
consisting of: air filter pressure switch; air throttle vale
position; alternator inoperative sensor; alternator output voltage;
backup control select; battery level; cylinder head temperature
(CHD); exhaust gas oxygen (EGO); exhaust gas temperature (EGT);
engine crankshaft position; fuel filter pressure switch; fuel flow
transducer; fuel injector pulse width command; fuel inlet pressure;
fuel pressure; fuel pump outlet pressure; fuel to air ratio; fuel
to air ratio trim; ignition and timing command; knock sensor;
manifold air pressure (MAP); manifold air temperature (MAT); oil
chip detector; oil cooler delta pressure switch; oil filter delta
pressure; oil fuel content; oil pressure; oil quantity; oil
temperature; revolutions per minute (RPM); sparkplug energy pulse;
turbine inlet temperature; turbocharger compressor discharge
pressure; turbocharger compressor seal delta pressure; vibration;
and wastegate valve position, and combinations thereof.
7. A method for detecting combustion engine failure modes as in
claim 2, wherein said identifying step includes selecting a group
of ambiguity groups having said parameter deviation signal by
searching said ambiguity groups in order of probability of fault
mode, taking the intersection of the failure modes in said selected
group of ambiguity groups, and outputting said intersecting failure
modes.
8. A method for detecting combustion engine failure modes as in
claim 2, wherein said providing range of desired parameter values
step includes providing a range of continuous values having a high
limit and a low limit, and wherein said comparator function
generates a deviation signal if said actual value lies outside of
said high and low values.
9. A method for detecting combustion engine failure modes as in
claim 2, wherein said providing range of desired parameter values
includes providing a range of continuous values having only a high
limit and said comparator function generates a deviation signal if
said actual value is greater than said high value.
10. A method for detecting combustion engine failure modes as in
claim 2, wherein said providing range of desired parameter values
includes providing a range of continuous values having only a low
limit and said comparator function generates a deviation signal if
said actual value is less than said low value.
11. A method for detecting combustion engine failure modes as in
claim 2, wherein said providing range of desired parameter values
includes providing binary values having two states, and said
comparator function generates a deviation signal if said actual
value is one state rather than another state.
12. A method for detecting combustion engine failure modes as in
claim 2, further comprising storing a history of said actual
parameter values over time, and saving said stored history prior to
said deviation signal generation for later output.
13. A method for detecting combustion engine failure modes as in
claim 2, wherein said actual parameters values are inferred values
from at least two physical parameter measurements.
14. A method for predicting system failure modes, wherein the
system has a plurality of monitor points, each point having an
actual parameter value, wherein the system has a plurality of
failure modes, the method comprising the steps of: providing a
range of desired parameter values for each monitor point; providing
a monitor point comparator function for said monitor points,
wherein said monitor point comparator function compares said actual
parameter value with said desired parameter value range and
generates a parameter deviation signal if said actual parameter
value is outside of said desired parameter value range; providing a
plurality of ambiguity groups, said ambiguity groups having at
least one failure mode associated with a monitor point deviation
signal; obtaining said actual parameter value history over a
plurality of time points; extrapolating said plurality of time
points into the future; predicting if said extrapolated time points
will generate a deviation signal from said comparator; and, if said
deviation signal will be generated, generating a predicted
deviation signal, predicted time of deviation and outputting said
predicted deviation time; and for each generated predicted
parameter deviation signal, searching said plurality of ambiguity
groups for at least one ambiguity group that includes said
generated predicted parameter deviation signal, and identifying at
least one failure mode associated with said predicted generated
deviation signal.
15. A method according to claim 14 wherein the system is a
combustion engine.
16. A method for predicting combustion engine failure modes as in
claim 15, wherein said identifying step includes selecting a group
of ambiguity groups having said parameter deviation signal, taking
the intersection of the failure modes in said selected group of
ambiguity groups, and outputting said intersecting failure
modes.
17. A method for predicting combustion engine failures as in claim
16, wherein said identifying step includes selecting only those
failure modes present in all ambiguity modes having said parameter
deviation signal.
18. A method for predicting combustion engine failures as in claim
15, further comprising outputting suggested engine operating
parameters to ameliorate bad effects of said detected engine
faults.
19. A method for predicting combustion engine failures as in claim
15, wherein said parameters are selected from the group consisting
of: air filter pressure switch; air throttle vale position;
alternator inoperative sensor; alternator output voltage; backup
control select; battery level; cylinder head temperature (CHD);
exhaust gas oxygen (EGO); exhaust gas temperature (EGT); engine
crankshaft position; fuel filter pressure switch; fuel flow
transducer; fuel injector pulse width command; fuel inlet pressure;
fuel pressure; fuel pump outlet pressure; fuel to air ratio; fuel
to air ratio trim; ignition and timing command; knock sensor;
manifold air pressure (MAP); manifold air temperature (MAT); oil
chip detector; oil cooler delta pressure switch; oil filter delta
pressure; oil fuel content; oil pressure; oil quantity; oil
temperature; revolutions per minute (RPM); sparkplug energy pulse;
turbine inlet temperature; turbocharger compressor discharge
pressure; turbocharger compressor seal delta pressure; vibration;
and wastegate valve position, and combinations thereof.
20. A method for predicting combustion engine failures as in claim
15, wherein said identifying step includes selecting a group of
ambiguity groups having said parameter deviation signal by
searching said ambiguity groups in order of probability of fault
mode, taking the intersection of the failure modes in said selected
group of ambiguity groups, and outputting said intersecting failure
modes.
21. A computer program for executing a method for detecting system
failure modes, wherein the system has a plurality of monitor
points, each point having an actual parameter value, wherein the
system has a plurality of failure modes, the method comprising the
steps of: providing a range of desired parameter values for each
monitor point; providing a monitor point comparator function for
said monitor points, wherein said monitor point comparator function
compares said actual parameter value with said desired parameter
value range and generates a parameter deviation signal if said
actual parameter value is outside of said desired parameter value
range; providing a plurality of ambiguity groups, said ambiguity
groups having at least one failure mode associated with a monitor
point deviation signal; obtaining said actual parameter value;
comparing said actual parameter value with said desired parameter
value for each monitor point using said monitor point comparator
function and generating a parameter deviation signal if said
parameter actual value is outside of said desired parameter value
range; and for each generated parameter deviation signal, searching
said plurality of ambiguity groups for at least one ambiguity group
that includes said generated parameter deviation signal, and
identifying at least one failure mode associated with said
generated deviation signal.
22. A method according to claim 21 wherein the system is a
combustion engine.
23. A computer program for executing a method for predicting system
failure modes, wherein the system has a plurality of monitor
points, each point having an actual parameter value, wherein the
system has a plurality of failure modes, the method comprising the
steps of: providing a range of desired parameter values for each
monitor point; providing a monitor point comparator function for
said monitor points, wherein said monitor point comparator function
compares said actual parameter value with said desired parameter
value range and generates a parameter deviation signal if said
actual parameter value is outside of said desired parameter value
range; providing a plurality of ambiguity groups, said ambiguity
groups having at least one failure mode associated with a monitor
point deviation signal; obtaining said actual parameter value
history over a plurality of time points; extrapolating said
plurality of time points into the future; predicting if said
extrapolated time points will generate a deviation signal from said
comparator; and, if said deviation signal will be generated,
generating a predicted deviation signal, predicted time of
deviation and outputting said predicted deviation time; and for
each generated predicted parameter deviation signal, searching said
plurality of ambiguity groups for at least one ambiguity group that
includes said generated predicted parameter deviation signal, and
identifying at least one failure mode associated with said
predicted generated deviation signal.
24. A method according to claim 23 wherein the system is a
combustion engine.
25. A computer monitoring system for coupling to a system, said
computer monitoring system running a program for executing a method
for detecting system failure modes, wherein the system has a
plurality of monitor points, each point having an actual parameter
value, wherein the system has a plurality of failure modes, wherein
the computer monitoring system is coupled to said monitor points
and able to read said monitor points, the method comprising the
steps of: providing a range of desired parameter values for each
monitor point; providing a monitor point comparator function for
said monitor points, wherein said monitor point comparator function
compares said actual parameter value with said desired parameter
value range and generates a parameter deviation signal if said
actual parameter value is outside of said desired parameter value
range; providing a plurality of ambiguity groups, said ambiguity
groups having at least one failure mode associated with a monitor
point deviation signal; obtaining said actual parameter value;
comparing said actual parameter value with said desired parameter
value for each monitor point using said monitor point comparator
function and generating a parameter deviation signal if said
parameter actual value is outside of said desired parameter value
range; and for each generated parameter deviation signal, searching
said plurality of ambiguity groups for at least one ambiguity group
that includes said generated parameter deviation signal, and
identifying at least one failure mode associated with said
generated deviation signal.
26. A method according to claim 25 wherein the system is a
combustion engine.
27. A computer monitoring system for coupling to a system, said
computer monitoring system running a program for executing a method
for predicting system failures, wherein the system has a plurality
of monitor points, each point having an actual parameter value,
wherein the system has a plurality of failure modes, wherein the
computer monitoring system is coupled to said monitor points and
able to read said monitor points, the method comprising the steps
of: providing a range of desired parameter values for each monitor
point; providing a monitor point comparator function for said
monitor points, wherein said monitor point comparator function
compares said actual parameter value with said desired parameter
value range and generates a parameter deviation signal if said
actual parameter value is outside of said desired parameter value
range; providing a plurality of ambiguity groups, said ambiguity
groups having at least one failure mode associated with a monitor
point deviation signal; obtaining said actual parameter value
history over a plurality of time points; extrapolating said
plurality of time points into the future; predicting if said
extrapolated time points will generate a deviation signal from said
comparator; and, if said deviation signal will be generated,
generating a predicted deviation signal, predicted time of
deviation and outputting said predicted deviation time; and for
each generated predicted parameter deviation signal, searching said
plurality of ambiguity groups for at least one ambiguity group that
includes said generated predicted parameter deviation signal, and
identifying at least one failure mode associated with said
predicted generated deviation signal.
28. A method according to claim 27 wherein the system is a
combustion engine.
Description
FIELD OF THE INVENTION
[0001] The present invention is related generally to monitoring
systems, and more specifically, to devices and methods for
monitoring system performance or parameters to detect and isolate
present and future predicted failures.
BACKGROUND OF THE INVENTION
[0002] Numerous systems or components in use today are subject to
physical failure. Many of these systems have a monitoring mechanism
for monitor the operation of the system. The monitoring mechanisms
often have one or more monitoring points. One illustrative system
is a combustion engine. Inevitably, combustion engines fail. Some
engine failures may be prevented through replacement of parts,
standard service procedures, and/or major overhauls. Engine
failures or component failures may cause more harm than necessary
due to a lack of real time information about the nature of the
component failure and the lack of a real time suggestion as to a
course of action which could ameliorate the component failure. The
prevention of engine failure may not be carried out due to the lack
of knowledge of current engine parameters, lack of knowledge of
trends in engine parameters, and lack of knowledge of current
component failures.
[0003] In combustion engines used in automobiles, unexpected or
puzzling engine failures are at best inconvenient, at worst
presenting safety issues. Failure in combustion engines used in
industrial applications may cause unplanned down time and loss of
production. Combustion engines used in aircraft may cause major
loss of life upon unexpected failure.
[0004] What would be desirable, therefore, are systems and methods
for better detecting and isolating system or component failures.
Systems and methods for predicting future system or component
failures before they occur would also be advantageous.
SUMMARY OF THE INVENTION
[0005] The present invention includes methods and devices for
detecting faults or failure modes in systems that are subject to
failure. The present invention is described with reference to an
engine, but numerous other applications are contemplated. An engine
may have numerous monitor points, with each point having an actual
parameter value, and the engine having numerous failure modes. The
monitor points may be measurable values which can include, for
example, continuous values such as engine oil pressure, or binary
state values such as ON or OFF, HI ALARM or NOT HI ALARM.
[0006] The present invention may use a range of acceptable values
for each monitor point parameter. Values outside of this range may
generate a deviation result or signal. Some ranges are binary
values, such as OFF, which have no acceptable values other than the
desired binary value. Other ranges are continuous values, such as
two endpoints for a range, or a midpoint together with a margin
about the midpoint. Another range is a high or low limit for a
value, with any value exceeding the limit being cause for a
deviation condition. In some embodiments, exceeding different
limits for the same measured quantity can result in different
deviation outputs.
[0007] Numerous ambiguity groups may be created by a user to work
with the present invention, or may be included with a product
according to the present invention. The ambiguity groups can
include a number of failure modes, or faults that can be a cause,
or the cause, of out of range values or deviations emanating from
the monitor point parameter values. One class of failure modes is
mechanical failures, for example, a ruptured fuel line. In one
embodiment, an ambiguity group includes one or more failure modes,
the monitor point parameter deviation that would result from the
failure mode or combination of failure modes, and the probability
of the combination of failure modes occurring together to produce
the deviation signal. The ambiguity groups can be clustered
together about common deviation signals than can be caused by the
failure modes. In one example, all ambiguity groups that could
result in lower than expected RPM could be grouped together in
vectors or collections of ambiguity groups. In one embodiment, the
ambiguity groups are collected together in a table.
[0008] In operation, the engine failure detector can collect data
from sensor values, and operate upon the sensor values to form
parameter values for monitor points. Some monitor points are
directly coupled to sensors while other monitor points are
synthesized from multiple sensor values to form virtual or inferred
monitor points.
[0009] The parameter value may be obtained, and compared with the
desired parameter value range by a monitor point comparator
function. If the comparator function decides that the actual
parameter value exceeds the acceptable range, a deviation signal
may be generated. In some embodiments, a presumptive deviation
signal is filtered prior to outputting the deviation signal. The
filtering is often used to reduce the number of false alarms.
[0010] Once a deviation signal is generated, a fault isolator
functionality can operate on the provided deviation signal or
signals. In one embodiment, all the ambiguity groups having a
particular deviation signal are collected together. Given the
collected ambiguity groups, one or more of the ambiguity groups can
be selected as the most likely ambiguity groups to have generated
the deviation signal. In one embodiment, the ambiguity group having
the highest probability is selected. In another embodiment, the
possible ambiguity groups are analyzed to determine intersecting
failure modes, if any. It is contemplated that the ambiguity groups
may be Boolean ANDed together to determine common failure
modes.
[0011] The failure mode or modes selected as most likely to have
caused the deviation signal may be output to a human operator, and
may be fed to an automatic control function which may determine a
course of action to alert or ameliorate the condition caused by the
failure mode.
[0012] Some embodiments of the invention have failure predictors as
well. A history of values for monitor point parameters may be
collected by a data historian functionality. Some embodiments
record the data in a sliding window. The historical data can be
extrapolated or projected out into the future, and a prediction
made as to whether the projected monitor point parameter value
trends will exceed the range of acceptable values for that monitor
point parameter. If a deviation condition is predicted, a projected
deviation signal and the projected time for that deviation may be
output for use by a human operator and/or fed back to an engine
control system to ameliorate or even postpone the predicted
deviation signal condition. In many embodiments, the predicted
deviation signal is analyzed to isolate a fault in a manner similar
to the fault isolation performed for a present deviation signal. In
some embodiments, the early deviation signal predicted is used to
suggest preventative maintenance earlier than otherwise
planned.
BRIEF DESCRIPTION OF THE DRAWINGS
[0013] FIG. 1 is a schematic diagram of a combustion engine and
engine monitoring system;
[0014] FIG. 2 is a simplified data flow diagram of a fault
isolating method according to the present invention;
[0015] FIG. 3 is a schematic diagram of a table of ambiguity
groups;
[0016] FIG. 4 is a high level flow chart of a method for detecting
and isolating engine faults; and
[0017] FIG. 5 is a high level flow chart of a method for predicting
engine faults.
DETAILED DESCRIPTION OF THE INVENTION
[0018] FIG. 1 illustrates an engine fault monitoring system 20
coupled to a combustion engine 22 and having a computer 24.
Combustion engine 22 can be any combustion engine, including
automobile engines and aircraft engines. In a preferred embodiment,
engine 22 is an aircraft engine. Computer 24 can be any suitable
computing device, including general purpose computers and dedicated
single purpose computing devices, which can include single board,
microprocessor based devices.
[0019] Engine 22 may be seen to have a number monitor points 30,
coupled to computer 24 through a number of data communication lines
or channels 40. Monitor points 30, are associated with parameters,
often the direct or filtered output of physical sensors. Examples
of sensors include oil pressure sensors and oil temperature
sensors. Examples of corresponding monitor points can be oil
pressure and oil temperature. For purposes of understanding the
present invention, monitor points may be considered to be
associated with physical entities, concepts, values, or states,
while monitor point parameters may be considered to be the data
associated with the monitor point.
[0020] Examples of some corresponding monitor point parameters can
be hi-oil pressure, hi-hi oil pressure, low oil pressure, low-low
oil pressure, hi-oil temperature, hi-hi-oil temperature, low oil
temperature, low-low oil temperature. Parameters may be selected
from the group consisting of: air filter pressure switch; air
throttle valve position; alternator inoperative sensor; alternator
output voltage; backup control select; battery level; cylinder head
temperature (CHD); exhaust gas oxygen (EGO); exhaust gas
temperature (EGT); engine crankshaft position; fuel filter pressure
switch; fuel flow transducer; fuel injector pulse width command;
fuel inlet pressure; fuel pressure; fuel pump outlet pressure; fuel
to air ratio; fuel to air ratio trim; ignition and timing command;
knock sensor; manifold air pressure (MAP); manifold air temperature
(MAT); oil chip detector; oil cooler delta pressure switch; oil
filter delta pressure; oil fuel content; oil pressure; oil
quantity; oil temperature; revolutions per minute (RPM); sparkplug
energy pulse; turbine inlet temperature; turbocharger compressor
discharge pressure; turbocharger compressor seal delta pressure;
vibration; and wastegate valve position, and combinations
thereof.
[0021] Monitor points may be virtual, synthesized, or inferred
monitor points, where the monitor point is created using
information from more than one sensor. One example of a virtual
monitor point would lower the expected RPM, given the fuel and air
flow going to the engine. Engine efficiency would be another
example of a virtual monitor point. A virtual monitor point 39 is
illustrated in FIG. 1, using a series of monitor points 30 as
inputs.
[0022] Computer 24, executing a computer program, can analyze the
data provided by the monitor point parameters, trend the monitor
point parameters historically, and store the acceptable deviation
ranges for the points. The appropriate data can be output through
data communication line or channel 44 to a computer output or
input/output device 26. In many embodiments, computer output device
26 is optimized for human input and output. In one example,
deviation signals or alarms and suggested courses of action are
output to device 26. In some embodiments, device 26 accepts human
operator inputs to modify the operation of engine 22. In some
embodiments, either computer or human generated changes in engine
operation are fed back to engine 22 through a command interface
line or channel 42.
[0023] FIG. 2 illustrates a simplified data flow diagram 60 of the
data flow which can occur in computer 24 or other comparable
controller. Data flow diagram 60 includes a comparator
functionality or comparator 52 and a data historian/predictor
functionality 54. An acceptable deviation range 50 is also present,
coupled to historian/predictor 54 through data line or channel 51
and to comparator 52 through data line or channel 53. The result of
historian/predictor 54 is coupled to comparator 52 through data
communication line or channel 55. The result of historian/predictor
54 can also be coupled to output or input/output device 26 through
a data communication line or channel 62.
[0024] Data historian/predictor 54 can trend data for the monitor
points over time, where the sample interval can vary and be
appropriate to the time constant or period for the point being
monitored. In one embodiment, the historical data is stored in a
sliding window or circular file, for a fixed period of time, before
being overwritten with new data. The historical data may also be
filtered prior to being stored and may also be compressed.
[0025] The historian/predictor 54 preferably has an extrapolation
or projection functionality, which takes the historical data and
extrapolates or projects the historical trend out into the future.
Using the acceptable deviation range provided by range 50, and the
projection provided to comparator 52 by predictor 54 through data
line 55, the comparator can determine whether any of the projected
points will fall outside of the acceptable ranges. If the projected
data falls outside of the acceptable ranges, the comparator can
output a predicted deviation signal through output line 44 to
device 26. In a preferred embodiment, the projected time of first
deviation is also output from comparator 52 to device 26. In some
embodiments, the historical data leading to the predicted deviation
is also output to device 26 through data line or channel 62. The
observation of the historical trend leading to the conclusion may
be useful in some applications to allow for a human analysis which
may indicate that the prediction is questionable. In some
embodiments, the historical data preceding a predicted deviation is
copied to another location and stored, or marked as data to not be
written over, in order to preserve the data for future
analysis.
[0026] The projection made by historian/predictor 54 is a straight
line extrapolation of a least squares fit data in some embodiments.
In other embodiments, a more sophisticated extrapolation method is
used. One method utilizes a weighted least squares form of
estimator, with the degree of the polynomial dependent upon the
parameter being tested. Once the predicted deviation is output,
further analysis may be performed in some embodiments of the
invention.
[0027] FIG. 3 illustrates ambiguity groups which are themselves
grouped into groups within a table 100. Table 100 is presented for
purposes of illustration, as the actual implementation of the
invention may take many forms. Table 100 is a sparsely populated
table having rows and columns, with the rows containing ambiguity
groups such as ambiguity groups 102, 104, 106, 112, 114, 120, 122,
123 and 125. The columns in an example ambiguity group include a
probability column or element 129, and a series of failure mode
columns 130, 131, through 139. Failure mode columns are numbered 00
through 24, and may be referred to by column number for convenience
and clarity below.
[0028] Table 100 also includes a deviation signal column or element
140, which contains a particular deviation signal, including the
types of deviation signals previously discussed. For purposes of
illustration, ambiguity groups in Table 100 are collected into
groups having the same deviation signal. For example, ambiguity
groups 102, 104 and 106 are grouped together into ambiguity group
collections 108, where collections 108 have the same value in
deviation signal column 140. Ambiguity group collections 116, 124,
and 126 are also shown. In one example, deviation signal column 140
may contain the deviation signal for hi-oil pressure for all of the
ambiguity groups within collection 108. Probability column 129 may
contain the estimated probability of each of the ambiguity groups
occurring.
[0029] An ambiguity group may thus consist of a series of failure
modes that may result in the occurrence of a particular deviation
signal. In the example in Table 100, failure modes 00 through 24
are shown. Each failure mode may correspond to a physical event or
state that may lead to a deviation signal. In one example, a
failure mode may be a ruptured fuel line while a deviation signal
may be a lower than expected RPM and a loss of fuel pressure.
Another example of a failure mode may be a blocked fuel filter,
which may lead to deviation signals including high fuel pressure in
one fuel line segment, low pressure in another segment, a high
delta pressure across the fuel filter, and lower than expected RPM
for a given throttle setting.
[0030] As can be seen in Table 100, some ambiguity groups have more
than one failure node checked, indicating that more than one
failure mode is required to result in the occurrence of the
deviation signal for that ambiguity group. In one example,
ambiguity group 102 requires failure modes 13, 16 and 18 to occur,
while ambiguity group 104 only requires failure mode 10 to occur.
The probability column 129 may contain an estimated probability for
the ambiguity group. In one example, the occurrence of several
failure modes together may be expected, and may be common, having a
high probability. In another example, the occurrence of the
multiple failure modes may be unlikely to occur at the same time,
and the ambiguity group has a low probability. In yet another
example, a single failure mode leading to a deviation signal may be
estimated to be either highly probable or highly unlikely, and the
value may be estimated by an engineer to populate the table.
[0031] Upon the occurrence of one or more deviation signals, the
computer may indicate to the human operator the occurrence of the
deviation signal, and may also attempt to provide an explanation of
a possible cause or causes. In one example, a failure mode in
column 05 in Table 100 may trigger a deviation signal corresponding
to both ambiguity group 123 and 125. In some instances, the
probability column may indicate that failure mode 05 is the likely
cause, for either deviation signal standing alone. In other
instances, the combination of both deviation signals may increase
the probability that failure mode 05 is the cause or one cause of
the deviation signal. In one embodiment, the probabilities of the
ambiguity groups having an active deviation signal are combined. In
one method, an intersection or Boolean AND operation is performed
on all ambiguity groups having an active deviation signal. In some
embodiments, failure modes are given negative weights or other
indicia that they are required or strongly suggested to be NOT
present in order for the particular ambiguity group to be the cause
of a deviation signal.
[0032] In some modes of operation, it may occur that more than one
failure mode may be the cause, and the failure modes may be
unrelated, resulting in true ambiguity being conveyed to the human
operator as the cause or causes of the deviation signal or signals.
In some embodiments, where true ambiguity means that different
failure modes could lead to identical deviation signals, the most
serious failure mode may be selected and output. In some
embodiments, in some situations, no single failure mode may readily
be ascertainable, as for example, when an unexpected fault mode may
be the cause. In this case, in some methods, the union of the
ambiguity groups may be taken and the most serious of the failure
modes selected for output. When a detection test indicates failure,
each failure mode in the ambiguity group is said to be "indicted."
When multiple detection tests show failure, the failure mode with
the highest number of indictments may be indicated as the most
likely fault. In a preferred embodiment, all indicated failure
modes are made available for human and/or machine analysis. Failure
modes adjudged to be the possible or most likely causes of the
active deviation signal or signals may be output through device 26
for human consumption.
[0033] Computer 26 may also have active programs executing to
ameliorate the perceived failure mode caused by deterioration of
engine operation. The computer may provide suggested changes in
engine operation. The suggested changes in operation may be output
to a human operator and/or fed back to the engine controls
automatically.
[0034] FIG. 4 illustrates a method which may be incorporated in a
computer program executing within computer 26 or other computer.
The computer program may be any suitable compiled or interpreted
program, including languages such as Basic, Java, C, C++, ladder
logic, or sequential function chart programming. In step 202, each
monitor point parameter may be updated to obtain a current value.
Monitor point parameters may be updated by directly reading analog
or discrete sensors, or by performing calculations on sensor values
to obtain derived or inferred monitor point parameter values.
[0035] The obtained, current monitor point parameter values may be
compared in step 204 by a comparator function against an acceptable
range of monitor point values. Some ranges may have no breadth,
such as binary or discrete values, for example ON/OFF or HI/NOT HI
switch values. Some ranges may have endpoints denoting a range of
acceptable values. Other ranges may be acceptable only below a
value, at or below a value, above a value, or at or above a value.
If the comparator detects an out of range value for a monitor
point, a presumptive deviation signal may be generated.
[0036] The presumptive deviation signal may be filtered to inhibit
false alarms in step 206. Some filtering may include a timer or
counter, requiring the presumptive deviation to persist for a
period of time or a number of cycles before a deviation signal is
generated. The filter may require that a discrete deviation remain
unchanged for a period of time or number of cycles, to eliminate
momentary false alarms or bounce.
[0037] In step 210, for each generated parameter deviation signal,
the ambiguity groups may be searched to find those groups having
the deviation signal being generated. In step 212, the ambiguity
groups having the matching deviation signals may be collected. In
some embodiments, the ambiguity groups are presorted by deviation
signal, and the searching and collection is made easier.
[0038] In step 214, the collected ambiguity groups can be analyzed
and one or more failure modes selected for output. In some
embodiments, the failure modes are selected based on probability of
the ambiguity groups occurring. In other embodiments, the failure
modes are selected based on the intersection or Boolean AND of the
ambiguity groups having an active deviation signal. The selected
deviation signals and failure modes can be output for human and
machine consumption is step 216.
[0039] In some embodiments, the deviation signals and failure modes
are analyzed as in step 218 to determine a suggested mode of
operation to ameliorate the problem. In one example, if the engine
torque output fails to perform as designed, and the inlet manifold
pressure also indicates a failure, but the turbocharger pressure
output indicates no failure, then there may be a problem in the air
throttle valve, for example, valve sticking or a leak in the
manifold. A suggested course of action to ameliorate the problem
could be to maintain the engine power through the fuel/air ratio
through varying the fuel. On the other hand, if the engine torque
output fails, but the inlet manifold pressure indicates no failure,
then the resulting ambiguity group should include the fueling
system, and a suggested reconfiguration strategy that includes
manipulating air or spark would be indicated.
[0040] FIG. 5 illustrates a method 300 for predicting deviations
and isolating failure modes early, prior to their actual
occurrence. Method 300 may be implemented in an executable computer
language, as discussed with respect to method 200 in FIG. 4. Method
300 may be used to generate a predicted deviation to warn early of
an impending problem, prior to the time at which the problem
actually occurs. Many of the steps of method 300 are similar or
identical to those of method 200, previously discussed, and are so
noted. Method 300 may be run concurrently with method 200, with may
of the steps being the same or combined.
[0041] In step 302, values may be obtained for each monitor point,
to obtain a parameter value, as discussed with respect to step 202.
In step 304, current values may be added to historical values
previously collected. In some embodiments, the current values are
sampled only at certain time intervals, for example, with data only
being stored every 10 cycles, or every 20 seconds. In some
embodiments, the current values are filtered, and the filtered
values rather than instantaneous values are recorded in the data
historian function. In some embodiments, the data is stored in a
moving window or circular queue, so that older values are over
written by newer values.
[0042] In step 306, the set of stored historical values may be
extrapolated or projected out into the future. In one embodiment, a
weighted least squares form of estimator is used, with the degree
of the polynomial being dependent upon the parameter being tested.
In some embodiments, the rate of change of the value is taken into
account, with more recent values effectively being given more
weight. In some embodiments, the noise or deviation of the signal
within a time window is analyzed, and a deviation signal set to
denote the excessive noise and variability rather than a projected
absolute value crossing a limit.
[0043] In step 308, the projected parameter value is compared with
the acceptable range for that parameter, and an early presumptive
deviation alarm may be generated. The comparator may be the same or
similar to that discussed with respect to step 204 in method 200.
The presumptive early deviation signal preferably has an estimated
time of actual deviation associated with the signal. In step 310,
the presumptive early alarm may be filtered, using methods similar
to those discussed with respect to step 206 in method 200. After
filtering, the presumptive early deviation may be termed the early
deviation signal, which may be processed by step 312. In step 312,
the ambiguity groups may be searched as discussed with respect to
step 210 in method 200. The matching ambiguity groups may be
collected in step 314, as discussed with respect to step 212 in
method 200.
[0044] In step 316, at least one failure mode may be selected,
using methods the same or similar to the methods discussed with
respect to step 214 or method 200. The selected failure mode or
modes may be output for human and/or machine consumption in step
318, as discussed with respect to 216 of method 200. Step 320 may
be handled somewhat differently than step 218 of method 200. In the
case of the early deviation signal, the recommended course of
action may be carried out over a longer time frame and be less
urgent. The output to the human operator may well be an advisory
signal rather than an alarm. In one example, the recommended course
of action may be to recommend preventative maintenance earlier than
otherwise indicated. In a preferred method, the logic and
historical data that led to the projected early deviation signal is
presented for human and/or machine consumption.
[0045] While the present invention is described with reference to
an engine, numerous other applications are contemplated. It is
contemplated that the present invention may be applied to any
system that is subject to failure so long as adequate sensor data
can be collected.
[0046] Numerous advantages of the invention covered by this
document have been set forth in the foregoing description. It will
be understood, however, that this disclosure is, in many respects,
only illustrative. Changes may be made in details, particularly in
matters of shape, size, and arrangement of parts without exceeding
the scope of the invention. The invention's scope is, of course,
defined in the language in which the appended claims are
expressed.
* * * * *