U.S. patent application number 11/608076 was filed with the patent office on 2008-06-12 for system and method for equipment life estimation.
This patent application is currently assigned to GENERAL ELECTRIC COMPANY. Invention is credited to Piero Patrone Bonissone, Neil Holger White Eklund, Kai Frank Goebel, Hai Qiu, Feng Xue, Weizhong Yan.
Application Number | 20080140352 11/608076 |
Document ID | / |
Family ID | 39499290 |
Filed Date | 2008-06-12 |
United States Patent
Application |
20080140352 |
Kind Code |
A1 |
Goebel; Kai Frank ; et
al. |
June 12, 2008 |
SYSTEM AND METHOD FOR EQUIPMENT LIFE ESTIMATION
Abstract
A method to predict equipment life is disclosed. The method
includes making available a set of input parameters, and defining a
model of a health of the equipment as a function of the set of
input parameters. The method continues with receiving at least one
signal representative of a respective one of an actual sensor
output relating to an actual operation attribute margin of the
equipment, predicting a remaining useful equipment life based upon
a sequence of outputs of the model of the health of the equipment,
and generating a signal corresponding to the remaining useful
equipment life.
Inventors: |
Goebel; Kai Frank; (Mountain
View, CA) ; Bonissone; Piero Patrone; (Schenectady,
NY) ; Yan; Weizhong; (Clifton Park, NY) ;
Eklund; Neil Holger White; (Schenectady, NY) ; Xue;
Feng; (Clifton Park, NY) ; Qiu; Hai; (Clifton
Park, NY) |
Correspondence
Address: |
GENERAL ELECTRIC COMPANY;GLOBAL RESEARCH
PATENT DOCKET RM. BLDG. K1-4A59
NISKAYUNA
NY
12309
US
|
Assignee: |
GENERAL ELECTRIC COMPANY
Schenectady
NY
|
Family ID: |
39499290 |
Appl. No.: |
11/608076 |
Filed: |
December 7, 2006 |
Current U.S.
Class: |
702/183 |
Current CPC
Class: |
G05B 2219/32234
20130101; Y02P 90/02 20151101; G05B 19/4184 20130101; G05B
2219/32369 20130101; Y02P 90/26 20151101; G05B 2219/32235 20130101;
Y02P 90/14 20151101 |
Class at
Publication: |
702/183 |
International
Class: |
G06F 15/00 20060101
G06F015/00 |
Claims
1. A method to predict equipment life comprising: making available
a set of input parameters, wherein the making available comprises
making available an operating condition, a degraded abnormal health
condition, and a deterioration condition; executing a computational
model with the set of input parameters to define at least one
modeled operation attribute margin; comparing the modeled operation
attribute margin defined using the set of input parameters absent
the degraded abnormal health condition with a corresponding modeled
operation attribute margin defined using the set of input
parameters comprising the degraded abnormal health condition to
develop a normalized operation attribute margin; defining a health
index based upon the normalized operation attribute margin;
defining a model of a health of the equipment as a function of the
set of input parameters; receiving at least one signal
representative of a respective one of an actual sensor output
relating to an actual operation attribute margin of the equipment;
predicting a remaining useful equipment life based upon a sequence
of outputs of the model of the health of the equipment; and
generating a signal corresponding to the remaining useful equipment
life.
2. The method of claim 1, wherein the predicting comprises:
assessing a plurality of operational data prior to an end of
equipment useful life.
3. The method of claim 1, wherein the predicting comprises:
extrapolating a trajectory of the sequence of outputs of the model
of the health of the equipment.
4. The method of claim 3, wherein the extrapolating comprises:
extrapolating the trajectory of the sequence of outputs of the
model of the health of the equipment using an exponential curve
fit.
5. (canceled)
6. The method of claim 1, further comprising: defining the health
of the equipment by a most limiting one of the at least one modeled
operation attribute margin.
7. The method of claim 1, wherein: the making available the set of
input parameters comprises supplying the set of input parameters of
a gas turbine engine; and the executing the computational model
with the set of input parameters to define the modeled operation
attribute margin comprises defining the modeled operation attribute
margin to comprise at least one of booster stall, high pressure
compressor stall, high pressure compressor pressure ratio, low
pressure turbine clearance, high pressure turbine inlet
temperature, high pressure turbine clearance, high pressure turbine
exit temperature, and core speed.
8. The method of claim 7, wherein the making available comprises:
making available a degraded abnormal health condition comprising
efficiency and flow.
9. (canceled)
10. The method of claim 1, further comprising: executing a
computational model with the set of input parameters to define at
least one modeled sensor output; and predicting a deterioration
magnitude via a deterioration model using the operating condition
and the modeled sensor output.
11. The method of claim 10, further comprising: comparing the
predicted deterioration magnitude with the deterioration condition
to define a deterioration estimation error; and changing the
deterioration model to reduce the deterioration estimation
error.
12. (canceled)
13. The method of claim 1, wherein: the comparing comprises
comparing a plurality of modeled operation attribute margins
defined using the set of input parameters absent the degraded
abnormal health level with a corresponding plurality of modeled
operation attribute margins defined using the set of input
parameters comprising the degraded abnormal health level to develop
a plurality of normalized operation attribute margins; the method
further comprising defining a limiting normalized operation
attribute margin as a normalized operation attribute margin of the
plurality of normalized operation attributes having a minimum value
at a specific degraded abnormal health level; and the defining the
health index comprises defining the health index based upon the
limiting normalized operation attribute margin.
14. The method of claim 1, further comprising: executing the
computational model with the set of input parameters to define at
least one modeled sensor output; and developing a transfer function
that makes available a predicted health index using the modeled
sensor output and a predicted deterioration condition.
15. The method of claim 14, further comprising: comparing the
predicted health index to the defined health index to define a
health estimation error; and changing the transfer function to
reduce the health estimation error.
16. The method of claim 14, further comprising: determining a
change in the signal representative of the actual sensor output;
and predicting an expected change in the health index via the
transfer function using the change in the signal representative of
the actual sensor output; wherein the predicting the remaining
useful equipment life comprises predicting the remaining useful
equipment life based upon the expected change in the health
index.
17. The method of claim 16, further comprising: detecting a
degraded abnormal health condition; wherein the determining the
change occurs subsequent to the detecting the degraded abnormal
health condition.
18. The method of claim 1, further comprising: estimating a set of
confidence intervals for the predicted remaining useful equipment
life via a statistical technique.
19. The method of claim 18, wherein the estimating comprises:
estimating the set of confidence intervals for the predicted
remaining useful equipment live life via bootstrapping.
20. A program storage device readable by a computer, the device
embodying a program or instructions executable by the computer to
perform the method of claim 1.
21. A prediction system for predicting life of equipment, the
system comprising: a database comprising a set of input parameters;
a processor in signal communication with the database; a
computational model application for executing on the processor, the
computational model performing a method, the method comprising:
executing a computational model application with the set of input
parameters to define at least one modeled operation attribute
margin wherein the set of input parameters comprise an operating
condition, a degraded abnormal health condition, and a
deterioration comparing the modeled operation attribute margin
defined using the set of input parameters absent the degraded
abnormal health condition with a corresponding modeled operation
attribute margin defined using the set of input parameters
comprising the degraded abnormal health condition to develop a
normalized operation attribute; defining a health index based upon
the normalized operation attribute margin; defining a model of a
health of the equipment as a function of the set of input
parameters; receiving at least one signal representative of a
respective one of an actual sensor output relating to an actual
operation attribute margin of the equipment; predicting a remaining
useful equipment life based upon a sequence of outputs of the model
of the health of the equipment; and generating a signal
corresponding to the remaining useful equipment life.
22. (canceled)
23. The system of claim 21, wherein: the equipment comprises gas
turbine engine equipment; and the modeled operation attribute
margin comprises at least one of booster stall, high pressure
compressor stall, high pressure compressor pressure ratio, low
pressure turbine clearance, high pressure turbine inlet
temperature, high pressure turbine clearance, high pressure turbine
exit temperature, and core speed.
24. (canceled)
25. The system of claim 21, wherein the computational model
application further performs: executing the computational model
application with the set of input parameters to define at least one
modeled sensor output; predicting a deterioration magnitude via a
deterioration model using the operating condition and the modeled
sensor output; comparing the predicted deterioration magnitude with
the deterioration condition to define a deterioration estimation
error; and changing the deterioration model to reduce the
deterioration estimation error.
26. (canceled)
27. The system of claim 21, wherein the computational model
application further performs: executing the computational model
with the set of input parameters to define at least one modeled
sensor output; developing a transfer function that makes available
a predicted health index using the modeled sensor output and a
predicted deterioration condition; comparing the predicted health
index to the defined health index to define a health estimation
error; and changing the transfer function to reduce the health
estimation error.
28. The system of claim 27, wherein the computational model
application further performs: determining a change in the signal
representative of the actual sensor output; and predicting an
expected change in the health index via the transfer function using
the change in the signal representative of the actual sensor
output; wherein the predicting the remaining useful equipment life
comprises predicting the remaining useful equipment life based upon
the expected change in the health index.
Description
BACKGROUND OF THE INVENTION
[0001] The present disclosure relates generally to life estimation,
and particularly to equipment subsystem life estimation.
[0002] Estimating a remaining useful life, also herein referred to
as a remaining life, of a subsystem is known in the art as
prognostics. Remaining life estimates provide valuable information
for operation of modern complex equipment. Remaining life estimates
provide decision making aids that allow operators to change
operational characteristics (such as load), which in turn may
prolong a life of the subsystem. Remaining life estimates also
allow planners to account for upcoming maintenance and set in
motion a logistics process that supports a smooth transition from
faulted to fully functioning equipment. Predicting remaining life
is not straightforward because, ordinarily, remaining life is
dependent upon future usage conditions, such as load and speed, for
example. In addition, an understanding of the underlying physics
that govern remaining life is hard to come by in particular for
complex machinery where numerous fault modes can potentially be the
driver for remaining life. Examples of equipment that may benefit
from use of remaining life estimates are aircraft engines (both
military and commercial), medical equipment, and power plants.
[0003] A common approach to prognostics is to employ a model of
damage propagation contingent on future use. Such a model is often
times based on detailed materials knowledge and makes use of finite
element modeling. Because such models are extremely costly to
develop, they are limited to a few important parts of a subsystem,
but are rarely applied to a full subsystem.
[0004] Another approach is a data-driven approach to take advantage
of time series data where equipment behavior has been tracked via
sensor measurements during normal operation all the way to an end
of equipment useful life. The end of equipment useful life can
represent a totally non-functioning state of the equipment for
example, equipment failure. The end of equipment useful life can
also represent a state of the equipment wherein the equipment no
longer provides expected results. When a reasonably-sized set of
these observations exists, pattern recognition algorithms can be
employed to recognize these trends and predict remaining life.
These predictions are often made under an assumption of
near-constant future load conditions. However, such run to end of
equipment useful life data are often not available because, when an
observed system is complex, expensive, and, safety is important,
such as aircraft engines, for example, faults will be repaired
before they lead to the end of equipment useful life. This deprives
the data-driven approach from information necessary for its proper
application.
[0005] Accordingly, there is a need in the art for a life
estimation arrangement that overcomes these limitations.
BRIEF DESCRIPTION OF THE INVENTION
[0006] An embodiment of the invention includes a method to predict
equipment life. The method includes making available a set of input
parameters, and defining a model of a health of the equipment as a
function of the set of input parameters. The method continues with
receiving at least one signal representative of a respective one of
an actual sensor output relating to an actual operation attribute
margin of the equipment, predicting a remaining useful equipment
life based upon a sequence of outputs of the model of the health of
the equipment, and generating a signal corresponding to the
remaining useful equipment life.
[0007] Another embodiment of the invention includes a prediction
system for predicting life of equipment. The prediction system
includes a database comprising a set of input parameters, a
processor in signal communication with the database, and a
computational model application for executing on the processor, the
computational model application performing a method. The method
includes defining a model of a health of the equipment as a
function of the set of input parameters, receiving at least one
signal representative of a respective one of an actual sensor
output relating to an actual operation attribute margin of the
equipment, predicting a remaining useful equipment life based upon
a sequence of outputs of the model of the health of the equipment,
and generating a signal corresponding to the remaining useful
equipment life.
[0008] The system includes a database including a set of input
parameters, a processor in signal communication with the database,
and a computational model application for executing on the
processor. The computational model performs a method including
executing the computational model with the set of input parameters
to define at least one modeled sensor output, and defining a model
of a health of the equipment as a function of the set of input
parameters. The method continues with receiving at least one signal
representative of a respective one of an actual sensor output
relating to an actual operation attribute margin of the equipment
and predicting a remaining useful equipment life based upon a
sequence of outputs of the model of the health of the
equipment.
BRIEF DESCRIPTION OF THE DRAWINGS
[0009] Referring to the exemplary drawings wherein like elements
are numbered alike in the accompanying Figures:
[0010] FIG. 1 depicts a schematic flowchart in accordance with an
embodiment of the invention;
[0011] FIG. 2 depicts two charts depicting a plurality of
normalized margins and a health index in accordance with an
embodiment of the invention;
[0012] FIG. 3 depicts a chart depicting a health index trajectory
in accordance with an embodiment of the invention;
[0013] FIG. 4 depicts a chart depicting a health index trajectory
applied to collected data in accordance with an embodiment of the
invention;
[0014] FIG. 5 depicts a schematic diagram of a prediction system in
accordance with an embodiment of the invention; and
[0015] FIG. 6 depicts a flowchart of a method for estimating
equipment life in accordance with an embodiment of the
invention.
DETAILED DESCRIPTION OF THE INVENTION
[0016] An embodiment of the invention will provide a subsystem
level prognostics process over an entire subsystem without needing
to assess the model of damage propagation mechanics at the
materials level. In an embodiment, the process will operate in an
absence of complete run to end of equipment useful life time series
data.
[0017] An embodiment of the invention will utilize a hybrid
model-based and data-driven approach to address particular
challenges of a low-sampling rate of operational data available and
to capitalize on thermodynamic engine models. An embodiment will
take advantage of fault signature instances obtained by the model
while observing partial fault trajectories in real data. An
embodiment will transform engine observations into a space in which
end-of-life margins can be defined. The equivalent of inverse
damage propagation lines are then fit to the transformed
observations, which allows remaining useful life to be estimated.
In an embodiment, different sources of uncertainty are quantified.
To that end, uncertainties such as engine model uncertainty, fault
modeling uncertainty, sensor noise, variations in time of fault
identification, variations in duration of fault condition,
variations in fault propagation assumptions, and variations in
transfer function parameters, for example, are considered.
[0018] As used herein, the term prognostics shall refer to the
estimation of remaining useful subsystem life. The remaining useful
life (RUL) estimates are in units of time or cycles. A time
estimate typically has associated uncertainty that is described as
a probability density curve around an actual estimate. Operators
can choose a confidence that allows them to incorporate a risk
level into their decision making. Typically, a confidence level on
RUL estimates increases as a prediction horizon decreases, such as
toward an approach of an end of component life, for example.
[0019] Prognostics is closely linked to diagnostics. As used
herein, the term diagnostics shall refer to a detection of a fault
condition, or an observed change in an operational state that is
related to a verifiable event. A fault is a first sign of a
potential end of equipment useful life at some future time. An
example of such a fault is an increase in engine fuel consumption
resulting from a cracked turbine blade. A direct cost of the end of
equipment useful life is unavoidable: ultimately, the equipment
must be replaced. Moreover, there are indirect costs to the end of
equipment useful life that are in many cases far greater than the
direct cost of the repair. One source of indirect costs is
secondary damage, for example, an end of the useful life of a
component in a compressor stage of a gas turbine often causes
damage to the rear stages. Another indirect cost is unscheduled
maintenance. It is often less expensive to replace a faulty
component during scheduled maintenance before it has reached the
end of its useful life than to have a component reach the end of
its useful life in the field which may result in unscheduled
maintenance and possibly operational disruption.
[0020] In the absence of any evidence of damage or faulted
condition, prognostics reverts to statistical estimation of
fleet-wide life, such as Weibull curves or other suitable
mechanisms. It is more common to employ condition-based prognostics
in a presence of an indication of abnormal wear, faults, or other
non-normal situation. It is therefore important to include accurate
diagnostics soon after a fault to provide a trigger point for
prognostic algorithms to operate.
[0021] Condition-based prediction systems depend on reliable fault
diagnostics to initiate prognostic algorithms. If diagnostics
recognizes a start point of damage too late, the damage propagation
models may lag reality and underestimate a magnitude of damage. If
prognostic algorithms are kicked off when there is no real damage,
a benefit of remaining life estimation is reduced. Accordingly,
presence of an accurate diagnostic fault detection algorithm will
be assumed as a basis for an embodiment of a prognostic RUL
prediction.
[0022] An embodiment of the process is broken down into an off-line
training process to develop models used for RUL estimation, and an
on-line monitoring process to utilize the developed models for
estimating the RUL.
[0023] Referring now to FIG. 1, a schematic flowchart 100 that
flows left to right, top to bottom, of an embodiment of the
off-line training process is depicted.
[0024] In an embodiment, the process will begin with an input of a
set of input parameters 104 that include equipment operating
conditions 105, various deterioration levels 110 that describe a
state of wear of a subsystem of equipment, and various degraded
abnormal health conditions 115, at varying magnitudes, to a physics
based computational model 125. In an embodiment, the computational
model 125 is a thermo-dynamic component level model of an aircraft
engine, also herein referred to as a cycle deck 125. In an
embodiment, the equipment is an aircraft engine, and the operating
conditions 105 are referred to as points within a flight envelope,
and include conditions such as altitude, throttle position, speed,
and air temperature, for example. In an embodiment, the degraded
abnormal health conditions 115 are defined in terms of combinations
of subsystem efficiency and flow. The degraded abnormal health
conditions 115 can be faults, but particular fault signatures do
not necessarily need to be known at this stage.
[0025] In an embodiment, the process will continue by observing at
least one output of the cycle deck 125 for normal conditions,
deteriorated conditions, and combinations of faults and
deterioration, as defined by the set of input parameters 104. The
at least one output is assessed over all operating conditions 105,
or throughout the entire flight envelope. The at least one output
includes modeled operation attribute margins 130, also herein
referred to as margins, of the subsystem of interest, as well as
expected sensor outputs 135, also herein referred to as modeled
sensor outputs. In an embodiment, typical margins 130 include
booster stall margin, high pressure compressor (HPC) stall margin.
HPC pressure ratio, low pressure turbine (LPT) clearance margin,
high pressure turbine (HPT) inlet temperature margin, HPT clearance
margin, HPT exit temperature margin, and core speed margin, for
example. In an embodiment, some margins 130 are capable to be
directly measured as sensor outputs, while other margins 130, such
as clearance and stall margin for example, are not capable to be
directly measured in the field with current technology. By taking
advantage of the cycle deck 125, those margins 130 that cannot be
directly measured can be calculated and included into a health
assessment process, to be described further below. Therefore, a
complete and systematic health estimation, which takes into account
all possible margins 130, can be achieved. Sensor outputs may also
be referred to as remote monitoring (RM) parameters.
[0026] In an embodiment, the process proceeds by normalizing each
margin 130. An example stall margin 140 and a set of corresponding
sensor outputs 135 for a variety of inputs parameters 104 are
depicted. Each margin 130 is normalized by taking a ratio
(expressed as a percentage) of the margin 130 provided by the cycle
deck 125 that results from input parameters 104 including degraded
abnormal health conditions (or faults) and deterioration conditions
at a particular point in the flight envelope to the same,
corresponding margin 130 resulting from input parameters 104 that
include an undeteriorated and fault free engine at the same flight
envelope point. It will be appreciated that in an embodiment, the
normalization will be performed for many margins 130, at many
degraded abnormal health levels (or faults) 115 and deterioration
110 levels, and for many flight envelope 105 points.
[0027] Referring now to FIG. 2, an illustrative example of a
plurality of normalized HPT margins 210 in the presence of an HPT
fault is depicted. It will be appreciated that as a value of
normalized margins 210, also herein referred to as normalized
operation attribute margins, approach 0%, a stronger probability of
failure exists, with a 0% normalized margin value defining an
existence of failure. As used herein, the term "failure" refers to
at least one of the normalized margins 210 violating a specified
operational limit, such as a temperature, speed, or clearance
limit, for example. Note that some margins 130, depending upon a
type of fault, may increase in a presence of at least one of
deterioration and fault, and so, when normalized, achieve values
over 100%. It will be further appreciated that a normalized margin
211 that reaches the value of 0% with a minimum degraded abnormal
health level (or fault magnitude) 215 will be defined as a limiting
value, also herein referred to as a limiting normalized operation
attribute margin, thereby defining a maximum tolerable degraded
abnormal health level (or fault magnitude) 220. A health index (HI)
250 is defined as the limiting normalized operation attribute
margin 211, or the most limiting margin 211 for a given degraded
abnormal health level (or fault magnitude) 115 and deterioration
level 110 at a particular point in the flight envelope 105. For
example, the health index 250 may be represented as HI=min(m.sub.1,
m.sub.2, . . . m.sub.n), where m.sub.1, m.sub.2, m.sub.n represent
the plurality of normalized margins 210.
[0028] Referring back now to FIG. 1, an embodiment of the process
proceeds by building a deterioration model 150 to provide a
deterioration magnitude estimate 111 based upon the expected sensor
outputs 135, via a suitable algorithm. In an embodiment, a neural
network is used for building and refining the accuracy of the
deterioration model 150. Operating conditions 105 or flight
envelope points, and the equipment sensor outputs 135, provided by
the cycle deck 125, are used as input to the deterioration model
150. The deterioration magnitude estimate 111, output by the
deterioration model 150, is used as a training target to ascertain
an accuracy of the deterioration model 150. Because the
deterioration level 110 is used as an input parameter 104 to the
cycle deck 125 to generate the expected sensor outputs 135, it will
be used as a basis against which to compare the deterioration
magnitude estimate 111 provided by the deterioration model 150.
Accordingly, the suitable algorithm will train or refine the model
150 to reduce the differences between the deterioration magnitude
estimate 111, and the deterioration level 110 input to the cycle
deck 125.
[0029] An embodiment of the process will continue by building a
transfer function (TF) 155. The TF 155 is a model that defines a
relationship between sensor variables (RM) 135 and the health index
250. Accordingly, the TF 155 is configured to provide a predicted
health index 251 using the expected sensor outputs 135 and the
deterioration estimate 111 as input. The TF 155 is built and
trained via a suitable algorithm in a fashion similar to the
deterioration model 150 described above. In an embodiment, the
algorithm to build the TF 155 is a neural network. The health index
250 is used as a training target, to ascertain the accuracy of the
TF 155, with the expected sensor outputs 135 and the deterioration
estimate 111, provided by the deterioration model 150, used as the
training input to develop the TF 155. An output of the TF 155, or a
health assessment of the equipment, is depicted graphically as
chart 156.
[0030] In an embodiment, the method will continue by using a
historical time series of signals representative of actual sensor
data 136 from engines that are known to have experienced serious
faults as input to the TF 155. Output from the TF 155 are HI
trajectories 255. Referring now to FIG. 3, an exemplary HI
trajectory 255 is plotted as a line among many individual HI 250
points. Under normal operating conditions, represented by a dashed
line within a first portion 310, a typical HI trajectory 255 will
drop at a steady pace, or a small slope that is reflective of
normal equipment deterioration. However, subsequent to an
occurrence of a fault, as represented by a solid line within a
second portion 320, the slope of the HI trajectory 255 will
decrease dramatically, and the HI trajectory 255 will drop
significantly downward. HI trajectories 255 are recorded to the
point where maintenance was performed, prior to an end of equipment
useful life. In an embodiment, a curve fitting extrapolator 160
(shown in FIG. 1) will fit an exponentially decaying curve 330 for
the time period subsequent to fault initiation, within the second
portion 320. The curve 330 can be fit to several different
functions. In an embodiment, the curve function is y=b+m
(1-e.sup..alpha.x.sup.n). In an embodiment, b is the value at the
fault initiation point, m is a multiplier, such as m=2, for
example, .alpha. is modifier, and n is an exponent, such as n=2.5,
for example.
[0031] An embodiment of the invention continues by finding an
intersection of the fitted curve 330 and the constant HI=0,
depicted graphically by a chart 165 in FIG. 1, which represents a
mean of the estimated time of failure. A RUL 170 estimation is a
difference between the estimated time of failure and a current time
step. A statistical technique to estimate a variance of the data,
such as bootstrapping, for example, is used to estimate confidence
intervals for the RUL 170 estimation.
[0032] In an embodiment, a noise distribution obtained during
normal operation, that is prior to the fault initiation, as
depicted by the first portion 310 in FIG. 3, is used to superimpose
noise onto the RUL 170 estimate. In an embodiment, the data are
then ready for training of other remaining life estimators.
[0033] In an embodiment, the on-line monitoring process is intended
to utilize actual data collected to provide the RUL 170 estimate
following a discovery of a fault. In an embodiment, the on-line
monitoring process will utilize a bank of models, based on
diagnostic information developed by the off-line training process
described above. In an embodiment, the bank of models can be
executed selectively or in parallel. In an embodiment, there are
dedicated models for each of at least one sub-system within a
system of interest. In an embodiment, the models may be further
refined by root cause. In an embodiment, the on-line monitoring
process can operate in several modes.
[0034] In an embodiment, one mode of the on-line monitoring process
will calculate the HI 250 via the TF 155 at all times, such as for
every sample, or flight, for example. In an embodiment, this mode
will proceed independently of fault triggers from diagnostics, and
provide additional information about the RUL 170 only when a fault
indication is given.
[0035] In another embodiment, a mode of the on-line monitoring
process is such that the prognostics module remains dormant, and
"awakes" when prompted by a diagnostic flag, such as the indication
of a fault, for example. In response to the diagnostic flag, data
prior to the diagnostic flag are retrieved to calculate the HI 250
data.
[0036] In an embodiment, both of the on-line monitoring processes
described will fit an exponentially decaying curve to the HI 250
trajectory 255 following the diagnostic flag. The curve can be fit
to several different functions. An embodiment includes the function
y=b+m (1-e.sup..alpha.x.sup.n) where b is the value at the fault
initiation point, m is a multiplier such as m=2, for example,
.alpha. is modifier, and n is an exponent, such as n=2.5, for
example.
[0037] A mean of the RUL 170 estimate at a time step of interest is
the difference between the current time and the intersection of the
fitted exponentially decaying line and the constant HI=0. In an
embodiment, a statistical technique to estimate a variance of the
data, such as bootstrapping, for example, is used to estimate
confidence intervals for the RUL 170 estimation. In an embodiment,
the curve fitting and RUL 170 estimate are performed dynamically,
that is, for every point for which a HI 250 is available.
[0038] The RUL estimation process has been tested in experiments,
using actual data relating to an aircraft engine, as depicted in
FIG. 4. It will be appreciated that application of the prediction
process disclosed herein to the actual data depicts the HI
trajectory 255. It can be seen that the HI trajectory 255 includes
the first portion 310 represented by the dashed line therein. The
dashed line depicts a drop at a steady pace, reflective of normal
engine deterioration. However, subsequent to the occurrence of the
fault, the second portion 320, represented by the solid line, has
the slope of the HI trajectory 255 that decreases dramatically.
[0039] Referring now to FIG. 5, an exemplary embodiment of a
prediction system 500 is depicted. In an embodiment, the prediction
system includes a turbine engine 510, at least one actual sensor
520, a data transfer unit (DTU) 530, a processor 540, an interface
unit 550, a computer 560, and a database 570. The computer 560
further includes a program storage device 580.
[0040] While an embodiment of the process has been described having
a turbine engine, it will be appreciated that the scope of the
invention is not so limited, and that the invention will also apply
to prediction systems including other pieces of equipment, such as
locomotive engines, medical equipment, and rolling mills, for
example.
[0041] In an embodiment, the at least one sensor 520 is disposed
and configured to be responsive to an operating condition of the
engine 510, and to generate a signal representative of the
operating condition of the engine 510. The at least one sensor 520
is in signal communication with the data transfer unit 530, which
makes available to the processor 540 the signal representative of
the operating condition of the engine 510.
[0042] In an embodiment, the processor 540 is in signal
communication with an interface device 550, such as to allow for
the on-line monitoring process, as described herein. The processor
540 is also in signal communication with the computer 560. In an
embodiment, the computer 560 is in signal communication with the
database 570, which is configured to store and make available to
the computer 560 the set of input parameters 104. The computer 560
also includes the program storage device 580 configured to store,
and make available to the computer 560 for execution, the
computational model 125.
[0043] While an embodiment of the invention has been described
having a computer in signal communication with the processor, it
will be appreciated that the scope of the invention is not so
limited, and that the invention will also apply to prediction
systems that have the computer in direct signal communication with
the data transfer unit. It will be further appreciated that an
embodiment of the invention will also include the computer in
signal communication via the data transfer unit via a variety of
communication protocols, such as cellular, wireless internet, and
others, for example, to allow an connection between the computer
and the data transfer unit during use of the equipment, to enable a
remote, on-line monitoring process.
[0044] In view of the foregoing, the prediction system 500 performs
the method of predicting equipment life. Referring now to FIG. 6,
an embodiment of a generalized flowchart 600 of an exemplary method
of predicting equipment life is depicted.
[0045] The method begins in the off-line mode with making available
610 the set of input parameters 104 to the computational model 125.
An embodiment includes executing 620 the computational model with
the set of input parameters 104 to define the at least one modeled
sensor output 135. The method continues with defining 625 the
transfer function 155, or a model of a health of the equipment as
the function of the set of input parameters 104. In the on-line
mode, the method continues with receiving 630 the at least one
signal representative of a respective one of the actual sensor
output 136 relating to the actual operation attribute of the
equipment. The method continues with describing 635 the current
health 156 of the equipment. The method continues with predicting
640 the remaining useful equipment life 170 based upon a sequence
of outputs of the TF 155 of the health of the equipment, and
generating 650 the signal corresponding to the remaining useful
equipment life. In an embodiment, the predicting 640 includes
assessing a plurality of operational data prior to an end of the
equipment useful life. In an embodiment, the signal corresponding
to the remaining useful equipment life will cause the predicted
remaining useful equipment life to be displayed via at least one of
the computer 560 and the interface device 550.
[0046] In an embodiment, the predicting 640 includes extrapolating
the trajectory of the sequence of outputs of the TF 155. In another
embodiment, the extrapolating includes the exponential curve fit.
In an embodiment, the set of input parameters 104 are input
parameters 104 of the gas turbine engine 510. In an embodiment, the
making available 610 the set of input parameters 104 includes the
operating conditions 105, the degraded abnormal health conditions,
or levels 115, and the deterioration conditions, or levels 110.
[0047] In an embodiment, the method further includes executing the
computational model 125 with the set of input parameters 104 to
define the modeled margin 130. In an embodiment, the method further
includes defining the health of the equipment by the most limiting
of the at least one modeled margin 130. In an embodiment, the
modeled margin 130 includes at least one of booster stall, high
pressure compressor stall, high pressure compressor pressure ratio,
low pressure turbine clearance, high pressure turbine inlet
temperature, high pressure turbine clearance, high pressure turbine
exit temperature, and core speed. An embodiment includes making
available the degraded abnormal health condition 115 including
efficiency and flow.
[0048] In an embodiment, the method further includes predicting the
deterioration magnitude estimate 111 via the deterioration model
150 using the operating condition 105 and the modeled sensor output
135. In an embodiment, the method further includes training, or
refining the deterioration model 150 by comparing the predicted, or
estimated deterioration magnitude 111 with the set of input
parameters 104, including the deterioration level 110, to define a
deterioration estimation error, and, changing or refining, via a
suitable algorithm, the deterioration model 150 to reduce the
deterioration estimation error. In an embodiment, the suitable
algorithm is the neural network.
[0049] In an embodiment, the method further includes comparing the
modeled margin 130, defined using the set of input parameters 104
absent the degraded abnormal health condition 115 with the
corresponding modeled operation attribute 130 defined using the set
of input parameters 104 including the degraded abnormal health
condition 115 to develop the plurality of normalized operation
attributes 210. In an embodiment, the method includes defining the
health index 250 based upon the plurality of normalized operation
attributes 210. In an embodiment, the comparing includes comparing
the plurality of modeled operation attributes 210 defined using the
set of input parameters 104 absent the degraded abnormal health
level (or fault magnitude) 215 with the corresponding plurality of
modeled operation attributes 210 defined using the set of input
parameters 104 including the degraded abnormal health level (or
fault magnitude) 215 to develop the plurality of normalized margins
210. The method further includes defining the limiting normalized
margin as the normalized margin having a minimum value at any
given, or specific, degraded abnormal health level (or fault
magnitude) and defining the health index 250 based upon the
limiting normalized margin 211 corresponding to the minimum
degraded abnormal health level (or fault magnitude) 215.
[0050] In an embodiment, the method further includes developing the
transfer function 155 that makes available the predicted health
index 251 using the modeled sensor output 135 and the predicted
deterioration condition, or estimate 111. In an embodiment, the
method further includes comparing the predicted health index 251 to
the defined health index 250 to define a health estimation error,
and changing, or refining via a suitable algorithm, the transfer
function 155 to reduce the health estimation error. In an
embodiment, the suitable algorithm is the neural network.
[0051] An embodiment of the invention further includes determining
the change in the signal representative of the actual sensor output
136, and predicting the expected change, or health index trajectory
255 via the transfer function 155 using the change in signal
representative of the actual sensor output 136. In an embodiment,
predicting the remaining equipment life 170 is based upon the
predicted expected change in the health index 251, or HI trajectory
255, as made available by the curve fitting extrapolator 160. An
embodiment of the invention further includes detecting the degraded
abnormal health condition in the equipment 510 and determining the
change of the actual sensor output 136 subsequent to the detecting
the degraded abnormal health condition.
[0052] An embodiment of the invention may be embodied in the form
of computer-implemented processes and apparatuses for practicing
those processes. Embodiments of the invention may also be embodied
in the form of a computer program product having computer program
code containing instructions embodied in tangible media, such as
floppy diskettes, CD-ROMs, hard drives, USB (universal serial bus)
drives, or any other computer readable storage medium, wherein,
when the computer program code is loaded into and executed by a
computer, the computer becomes an apparatus for practicing the
invention. Embodiments of the invention may also be embodied in the
form of computer program code, for example, whether stored in a
storage medium, loaded into and/or executed by a computer, or
transmitted over some transmission medium, such as over electrical
wiring or cabling, through fiber optics, or via electromagnetic
radiation, wherein when the computer program code is loaded into
and executed by a computer, the computer becomes an apparatus for
practicing the invention. When implemented on a general-purpose
microprocessor, the computer program code segments configure the
microprocessor to create specific logic circuits. A technical
effect of the executable instructions is to predict a remaining
useful life of equipment.
[0053] As disclosed, some embodiments of the invention may include
some of the following advantages: the ability to predict a
remaining useful life of equipment absent detailed materials level
damage propagation models for each part of the equipment; the
ability to predict remaining useful life of equipment absent run to
failure data; the ability to express overall component health as a
function of various operation attributes; the ability to map system
observables to component health; the ability to estimate equipment
deterioration; the ability to extrapolate remaining life estimates
for virtual run-to-failure; and the ability to provide remaining
life estimates during on-line equipment life assessment.
[0054] While the invention has been described with reference to
exemplary embodiments, it will be understood by those skilled in
the art that various changes may be made and equivalents may be
substituted for elements thereof without departing from the scope
of the invention. In addition, many modifications may be made to
adapt a particular situation or material to the teachings of the
invention without departing from the essential scope thereof.
Therefore, it is intended that the invention not be limited to the
particular embodiment disclosed as the best or only mode
contemplated for carrying out this invention, but that the
invention will include all embodiments falling within the scope of
the appended claims. Also, in the drawings and the description,
there have been disclosed exemplary embodiments of the invention
and, although specific terms may have been employed, they are
unless otherwise stated used in a generic and descriptive sense
only and not for purposes of limitation, the scope of the invention
therefore not being so limited. Moreover, the use of the terms
first, second, etc. do not denote any order or importance, but
rather the terms first, second, etc. are used to distinguish one
element from another. Furthermore, the use of the terms a, an, etc.
do not denote a limitation of quantity, but rather denote the
presence of at least one of the referenced item.
* * * * *