CN109308484A - Aero-engine multiclass failure minimum risk diagnostic method and device - Google Patents
Aero-engine multiclass failure minimum risk diagnostic method and device Download PDFInfo
- Publication number
- CN109308484A CN109308484A CN201810860268.0A CN201810860268A CN109308484A CN 109308484 A CN109308484 A CN 109308484A CN 201810860268 A CN201810860268 A CN 201810860268A CN 109308484 A CN109308484 A CN 109308484A
- Authority
- CN
- China
- Prior art keywords
- engine
- aero
- failure
- kinds
- maximum
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The invention discloses a kind of aero-engine multiclass failure minimum risk diagnostic method and devices, wherein method includes: to constitute original training set；Nondimensionalization processing is carried out to original training set using maximum-Returning to one for minimum value method, and dimension-reduction treatment is carried out using linear discriminant analysis；The posterior probability of failures all kinds of in training set is modeled, to obtain the softmax failure probability model of engine；In-service aero-engine sensor measures parameters are acquired, and carry out maximum-Returning to one for minimum value and linear discriminant analysis dimension-reduction treatment；All kinds of probabilities of malfunction of in-service aero-engine are estimated using softmax failure probability model；The expected loss for being determined as all kinds of failures is calculated according to the empirical loss meter of all kinds of failure modes mistakes, and Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss.This method can be diagnosed to be aero-engine failure with lower expected loss, provide reasonable decision for the maintenance of subsequent aero-engine.
Description
Technical field
The present invention relates to aero-engine health control technical field, in particular to a kind of aero-engine multiclass failure is most
Small diagnosis of risk method and device.
Background technique
Aero-engine is the important subsystem of aircraft, its operating status directly affects flight status.History
On, many aviation accidents are generated by engine failure.Meanwhile the mechanical system as a kind of complexity, aeroplane engine
Machine itself possesses high cost.Therefore, the maintenance support of aero-engine is to the use effect for guaranteeing flight safety, improving aircraft
It can be extremely important with usage economy.
With the development of the social economy, airplane flight is more and more, it is also increasing to the demand of aero-engine.With
This simultaneously, the demand with people to aeroplane performance is higher and higher, and the thrust ratio of engine is increasing, complexity and information
Change level is also being continuously improved.Traditional maintenance policy, such as correction maintenance based on failure and time-based periodic maintenance
It has been difficult to adapt to the demand of aircraft engine maintenance guarantee.(Conditioned-Based Maintenance is based on shape to CBM
The condition maintenarnce of state) it has been to be concerned by more and more people.
Based on actual monitoring state of the CBM according to equipment individual, potential failure is predicted using certain theoretical method
And remaining life, to provide reference and foundation for maintenance decision.CBM can more accurately weigh safety and economy
Contradiction, reduce maintenance risk and maintenance cost to the maximum extent, improve the reliability of equipment.Upgrade on the basis of CBM and spreads out
The aero-engine PHM (Prognostic and Health Management, prognostics and health management) born has become
One emerging integrated technology of aviation field, to the maintenance support system in aviation field produces major transformation.Boat
Empty engine Incipient Fault Diagnosis is an important composition content of aero-engine PHM, the health control to aero-engine
It is most important.It can be diagnosed to be in time which kind of failure is aero-engine occur, and provide reference for maintenance decision.
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.
Present inventor has found that aero-engine is in whole life cycle, gas circuit by a large amount of creative works
Component can generate degeneration since some original state, be embodied in flow, efficiency and pressure ratio and generate dull variation, in turn
The overall performance of engine is influenced, while can also be reflected on sensor measures parameters.This degenerative process proceeds to some threshold
When value, it is believed that the component produces failure.This threshold value can be generally defined as engine complete machine by artificially defined
Some secure threshold is arrived in energy parameter decline.The flow of engine overall performance parameter and all parts, efficiency and pressure ratio are all
Can not be measured directly, therefore the target of Fault Diagnosis of Aeroengines is, is joined according to the measurable sensor of aero-engine
Number, determining engine at this time is in malfunction or normal condition；If it is malfunction, which component further determined
Produce failure.
For this purpose, an object of the present invention is to provide a kind of aero-engine multiclass failure minimum risk diagnostic method,
This method can be diagnosed to be aero-engine failure with lower expected loss, provide reasonably for the maintenance of subsequent aero-engine
Decision.
It is another object of the present invention to propose a kind of aero-engine multiclass failure minimum risk diagnostic device.
In order to achieve the above objectives, one aspect of the present invention embodiment proposes a kind of aero-engine multiclass failure minimum risk
Diagnostic method, comprising the following steps: step S1: acquisition aero-engine includes all kinds of failures and trouble-free sensor measurement
Data, to constitute original training set；Step S2: nothing is carried out to the original training set using maximum-Returning to one for minimum value method
Dimensionization processing, and dimension-reduction treatment is carried out using linear discriminant analysis；Step S3: using the method for Maximum-likelihood estimation, and benefit
It is modeled with posterior probability of the softmax function to all kinds of failures described in training set, to obtain the engine
Softmax failure probability model；Step S4: acquiring in-service aero-engine sensor measures parameters, and carries out maximum-minimum
Value normalization and linear discriminant analysis dimension-reduction treatment；Step S5: using the softmax failure probability model estimate it is described
Use as a servant all kinds of probabilities of malfunction of aero-engine；Step S6: it is calculated and is determined according to the empirical loss meter of all kinds of failure modes mistakes
For the expected loss of all kinds of failures, and Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss.
The aero-engine multiclass failure minimum risk diagnostic method of the embodiment of the present invention can measure according to aero-engine
Sensor parameters, determining engine at this time can be diagnosed with lower expected loss in malfunction or normal condition
Aviation engine failure out is aero-engine health so as to provide reasonable reference to aircraft engine maintenance decision
Administrative skill lays the foundation.
In addition, aero-engine multiclass failure minimum risk diagnostic method according to the above embodiment of the present invention can also have
There is following additional technical characteristic:
Further, in one embodiment of the invention, in the step S3, using the Maximum-likelihood estimation
Method, and modeled using posterior probability of the softmax function to all kinds of failures in the training set, wherein pole
The optimization problem of maximum-likelihood estimation is convex problem, and acquires global minimum using gradient descent method.
Further, in one embodiment of the invention, in the step S6, the decision rule of the fault diagnosis
To calculate the expected loss that the state of present engine is determined as to failure of all categories according to the empirical loss meter, according to the phase
The criterion of loss reduction is hoped to carry out Fault Tree Diagnosis Decision, so that the expected loss of diagnostic result is minimum.
Further, in one embodiment of the invention, in the step S2 and the step S4, to data set into
When row dimensionality reduction, using the dimension reduction method for having supervision --- linear discriminant analysis, to efficiently use classification information.
Further, in one embodiment of the invention, in the step S3, using the softmax function pair
Posterior probability P (the Ck|x(i) ；θ) model,Formula is:
Wherein, x(i)A sample point is represented, θ is a K × (d+1) parameter matrix.
In order to achieve the above objectives, another aspect of the present invention embodiment proposes a kind of aero-engine multiclass failure minimum wind
Dangerous diagnostic device, comprising: the first acquisition module includes all kinds of failures and trouble-free sensor for acquiring aero-engine
Measurement data, to constitute original training set；Processing module, for utilizing maximum-Returning to one for minimum value method to the original instruction
Practice collection and carry out nondimensionalization processing, and dimension-reduction treatment is carried out using linear discriminant analysis；Modeling module, for using maximum likelihood
The method of estimation, and modeled using posterior probability of the softmax function to all kinds of failures described in training set, to obtain
State the softmax failure probability model of engine；Second acquisition module, for acquiring in-service aero-engine sensor measurement
Parameter, and carry out maximum-Returning to one for minimum value and linear discriminant analysis dimension-reduction treatment；Estimation module, described in utilizing
Softmax failure probability model estimates all kinds of probabilities of malfunction of the in-service aero-engine；Determination module, for according to each
The empirical loss meter of class failure modes mistake calculates the expected loss for being determined as all kinds of failures, and the smallest according to expected loss
Criterion carries out Fault Tree Diagnosis Decision.
The aero-engine multiclass failure minimum risk diagnostic device of the embodiment of the present invention can measure according to aero-engine
Sensor parameters, determining engine at this time can be diagnosed with lower expected loss in malfunction or normal condition
Aviation engine failure out is aero-engine health so as to provide reasonable reference to aircraft engine maintenance decision
Administrative skill lays the foundation.
In addition, aero-engine multiclass failure minimum risk diagnostic device according to the above embodiment of the present invention can also have
There is following additional technical characteristic:
Further, in one embodiment of the invention, the modeling module be further used for using it is described greatly seemingly
The method so estimated, and modeled using posterior probability of the softmax function to all kinds of failures in the training set,
Wherein, the optimization problem of Maximum-likelihood estimation is convex problem, and acquires global minimum using gradient descent method.
Further, in one embodiment of the invention, the decision rule of the fault diagnosis is according to the experience
Loss meter calculates the expected loss that the state of present engine is determined as to failure of all categories, according to the smallest standard of expected loss
Fault Tree Diagnosis Decision is then carried out, so that the expected loss of diagnostic result is minimum.
Further, in one embodiment of the invention, when carrying out dimensionality reduction to data set, using the dimensionality reduction side for having supervision
Method --- linear discriminant analysis, to efficiently use classification information.
Further, in one embodiment of the invention, in the step S3, using the softmax function pair
Posterior probability P (the Ck|x(i) ；θ) model,Formula is:
Wherein, x(i)A sample point is represented, θ is a K × (d+1) parameter matrix.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, in which:
Fig. 1 is the process according to the aero-engine multiclass failure minimum risk diagnostic method of one embodiment of the invention
Figure；
Fig. 2 is to be shown according to the structure of the aero-engine multiclass failure minimum risk diagnostic device of one embodiment of the invention
It is intended to.
Specific embodiment
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to is used to explain the present invention, and is not considered as limiting the invention.
The aero-engine multiclass failure minimum risk proposed according to embodiments of the present invention diagnosis is described with reference to the accompanying drawings
Method and device describes the aero-engine multiclass failure minimum risk proposed according to embodiments of the present invention with reference to the accompanying drawings first
Diagnostic method.
Fig. 1 is the flow chart of the aero-engine multiclass failure minimum risk diagnostic method of one embodiment of the invention.
As shown in Figure 1, the aero-engine multiclass failure minimum risk diagnostic method the following steps are included:
Step S1: acquisition aero-engine includes all kinds of failures and trouble-free sensor measurement data, to constitute original
Beginning training set.
It is understood that firstly, the embodiment of the present invention acquires the normal sensing data of engine and includes each gas
The sensing data of circuit unit failure constructs training set.
Specifically, being directed to certain type fanjet, acquiring it in normal state and separately including four kinds of gas path components
The operation data of (compressor, fan, high-pressure turbine, low-pressure turbine) failure.Operation data includes the measurement number of 13 sensors
According to, be fuel oil respectively with respect to reduced discharge, twin shaft revolving speed, the total temperature of fan inlet and outlet and stagnation pressure, compressor inlet and outlet total temperature
The total temperature and stagnation pressure, low-pressure turbine of the total temperature and stagnation pressure, high-pressure turbine inlet and outlet imported and exported with stagnation pressure, combustion chamber are imported and exported total
Mild stagnation pressure.By above-mentioned operation data Ji Gouchengxunlianji.
Step S2: nondimensionalization processing is carried out to original training set using maximum-Returning to one for minimum value method, and uses line
Property discriminant analysis carry out dimension-reduction treatment.
Specifically, for training set X=x (1),x(2),…,x(m), wherein x(i)A sample point is represented, is a d
Dimensional vector, i.e. x=[x1,x2,…,xd], vector represents a sensor, d=13 herein per one-dimensional.For every one-dimensional j,
Calculate the maximum value of all sample points in the dimensionAnd minimum valueLine is carried out to the dimension by following formula
Property transformation:
Then dimensionality reduction is carried out using linear discriminant analysis.Note normal condition is C1Class, what four gas path components broke down
State is respectively C2,C3,C4,C5.For convenience's sake, remember that total classification number is K, K=5 herein, remember Call={ C1,C2,…,
CK}。
(1) each classification C in raw data set X is calculatedkMean vector mk
(2) within class scatter matrix S is calculatedWWith inter _ class relationship matrix SB, and then calculateSB
(3) it calculatesSBFeature vector (e1,e2,…,ed) and corresponding characteristic value (λ1,λ2,…,λd)
(4) maximum d' characteristic value is selected, the matrix W that their corresponding feature vectors are formed a d × d' is (each
Column are a feature vectors)
(5) projected to original sample in new subspace using W: X'=X × W, X' are the samples of n × d' after projection
Subspace.Wherein, X' is exactly the training set after dimensionality reduction.
Step S3: using the method for Maximum-likelihood estimation, and using softmax function to failures all kinds of in training set after
It tests probability to be modeled, to obtain the softmax failure probability model of engine.
In one embodiment of the invention, in step s3, it using the method for Maximum-likelihood estimation, and utilizes
Softmax function models the posterior probability of all kinds of failures in training set, wherein the optimization of Maximum-likelihood estimation is asked
Entitled convex problem, and global minimum is acquired using gradient descent method.
Further, in one embodiment of the invention, in step s3, using softmax function to posterior probability P
(Ck|x(i)；It θ) models, formula are as follows:
Wherein, x(i)A sample point is represented, θ is a K × (d+1) parameter matrix.
Specifically, note training set is X={ x(1),x(2),…,x(m), it is equipped with sample observations Y={ y(1),y(2),…,y(m), the probability of the observation out is denoted as L (θ), wherein θ is unknown parameter:
L (θ)=P (Y1=y(1),…,Ym=y(m)；θ)
Assuming that be between sample it is independent identically distributed, L (θ) can further be write as:
Wherein 1 (y(i)=Cj) it is indicative function, work as y(i)=CjWhen the expression formula value be 1, be otherwise 0.In order to succinctly rise
See, classification C can be enabledk=k。Using soFtmax function comes to P (Ck|x(i) ；θ) model:
Wherein θ is a K × (d+1) parameter matrix:
For x(i)D dimensional feature,K=1,2 ..., K are corresponding weight,
It is intercept item, should also introduces thus
Then, likelihood function L (θ) can be write as:
Logarithm is taken to L (θ), obtains log-likelihood function l (θ):
Maximum-likelihood estimation seeks to minimize following cost function:
J (θ) is minimized using gradient descent method, the gradient formula of J (θ) is as follows:
The more new formula of iteration each time:J=1,2 ..., K.
Step S4: in-service aero-engine sensor measures parameters are acquired, and carry out maximum-Returning to one for minimum value and line
Property discriminant analysis dimension-reduction treatment.
In one embodiment of the invention, in step S2 and step S4, when carrying out dimensionality reduction to data set, using there is prison
The dimension reduction method superintended and directed --- linear discriminant analysis, to efficiently use classification information.
It is understood that when the embodiment of the present invention carries out dimensionality reduction to data set, using a kind of dimensionality reduction side for having supervision
Method --- linear discriminant analysis can efficiently use classification information.
Specifically, when carrying out maximum-Returning to one for minimum value, maximum valueAnd minimum valueIt should be in step S2
Value, matrix W when dimensionality reduction should be the W in step S2.
Step S5: all kinds of probabilities of malfunction of in-service aero-engine are estimated using softmax failure probability model.
It is understood that the embodiment of the present invention is estimated using softmax failure probability model obtained in step S3
The probability of in-service all kinds of failures of aero-engine.
Specifically, the embodiment of the present invention is by each data point x in the data set of in-service aero-engine(i)Substitute into step
Softmax failure probability model obtained in rapid S3, can be obtained x(i)Belong to the probability P (C of each statek|x(i)；θ).
Step S6: it is calculated according to the empirical loss meter of all kinds of failure modes mistakes and is determined as that the expectation of all kinds of failures is damaged
It loses, and Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss.
In one embodiment of the invention, in step s 6, the decision rule of fault diagnosis is according to empirical loss table
The expected loss that the state of present engine is determined as to failure of all categories is calculated, is carried out according to the smallest criterion of expected loss
Fault Tree Diagnosis Decision, so that the expected loss of diagnostic result is minimum.
Specifically, in a practical situation, the virtual condition of engine is determined that loss when various states is incomplete
The same, true classification is C by notejSample point x be mistakenly classified as classification CiGenerated loss is λij, then sample x is classified
For CiExpected loss are as follows:
So our target is to find a decision criteria h:X → Y, to minimize the overall loss of sample:
R (h)=Ex[R(h(x)|x)]
Least risk Bayes decision rule are as follows:
In actual task, we are still to estimate posterior probability P (C firstk| x), then according to Least risk Bayes
Decision carries out decision to new samples.
In addition, the aero-engine multiclass failure minimum risk diagnosis side of the embodiment of the present invention returned based on softmax
Other compositions of method and effect be all for a person skilled in the art it is known, in order to reduce redundancy, do not repeat them here
The aero-engine multiclass failure minimum risk diagnostic method proposed according to embodiments of the present invention, according to aeroplane engine
The measurable sensor parameters of machine, determining engine at this time is to be in malfunction or normal condition, can be with lower expectation
Loss is diagnosed to be aero-engine failure, so as to provide reasonable reference to aircraft engine maintenance decision, sends out for aviation
Motivation health control technology lays the foundation.
The aero-engine multiclass failure minimum risk diagnosis proposed according to embodiments of the present invention referring next to attached drawing description
Device.
Fig. 2 is the structural representation of the aero-engine multiclass failure minimum risk diagnostic device of one embodiment of the invention
Figure.
As shown in Fig. 2, the aero-engine multiclass failure minimum risk diagnostic device 10 include: the first acquisition module 100,
Processing module 200, modeling module 300, the second acquisition module 400, estimation module 500 and determination module 600.
Wherein, the first acquisition module 100 includes all kinds of failures and trouble-free sensor for acquiring aero-engine
Measurement data, to constitute original training set.Processing module 200 is used for using maximum-Returning to one for minimum value method to original training
Collection carries out nondimensionalization processing, and carries out dimension-reduction treatment using linear discriminant analysis.Modeling module 300 is used to use maximum likelihood
The method of estimation, and modeled using posterior probability of the softmax function to failures all kinds of in training set, to obtain engine
Softmax failure probability model.Second acquisition module 400 is used to acquire in-service aero-engine sensor measures parameters,
And carry out maximum-Returning to one for minimum value and linear discriminant analysis dimension-reduction treatment.Estimation module 500 is used to utilize softmax failure
Probabilistic model estimates all kinds of probabilities of malfunction of in-service aero-engine.Determination module 600 is used for wrong according to all kinds of failure modes
Empirical loss meter accidentally calculates the expected loss for being determined as all kinds of failures, and carries out failure according to the smallest criterion of expected loss
Diagnose decision.The device 10 of the embodiment of the present invention can be diagnosed to be aero-engine failure with lower expected loss, continue a journey after being
The maintenance of empty engine provides reasonable decision.
Further, in one embodiment of the invention, modeling module 300 is further used for using Maximum-likelihood estimation
Method, and modeled using posterior probability of the softmax function to all kinds of failures in training set, wherein maximum likelihood
The optimization problem of estimation is convex problem, and acquires global minimum using gradient descent method.
Further, in one embodiment of the invention, the decision rule of fault diagnosis is according to empirical loss meter
The expected loss that the state of present engine is determined as to failure of all categories is calculated, event is carried out according to the smallest criterion of expected loss
Barrier diagnosis decision, so that the expected loss of diagnostic result is minimum.
Further, in one embodiment of the invention, when carrying out dimensionality reduction to data set, using the dimensionality reduction side for having supervision
Method --- linear discriminant analysis, to efficiently use classification information.
Further, in one embodiment of the invention, wherein using softmax function to posterior probability P (Ck|x(i)；It θ) models, formula are as follows:
Wherein, x(i)A sample point is represented, θ is a K × (d+1) parameter matrix.
It should be noted that the aforementioned explanation to aero-engine multiclass failure minimum risk diagnostic method embodiment
It is also applied for the aero-engine multiclass failure minimum risk diagnostic device of the embodiment, details are not described herein again.
The aero-engine multiclass failure minimum risk diagnostic device proposed according to embodiments of the present invention, according to aeroplane engine
The measurable sensor parameters of machine, determining engine at this time is to be in malfunction or normal condition, can be with lower expectation
Loss is diagnosed to be aero-engine failure, so as to provide reasonable reference to aircraft engine maintenance decision, sends out for aviation
Motivation health control technology lays the foundation.
In addition, term " first ", " second " are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of aero-engine multiclass failure minimum risk diagnostic method, which comprises the following steps:
Step S1: acquisition aero-engine includes all kinds of failures and trouble-free sensor measurement data, to constitute original instruction
Practice collection；
Step S2: nondimensionalization processing is carried out to the original training set using maximum-Returning to one for minimum value method, and uses line
Property discriminant analysis carry out dimension-reduction treatment；
Step S3: using the method for Maximum-likelihood estimation, and using softmax function to all kinds of failures described in training set after
It tests probability to be modeled, to obtain the softmax failure probability model of the engine；
Step S4: acquiring in-service aero-engine sensor measures parameters, and carries out maximum-Returning to one for minimum valueization and linearly sentence
It Fen Xi not dimension-reduction treatment；
Step S5: all kinds of probabilities of malfunction of the in-service aero-engine are estimated using the softmax failure probability model；
And
Step S6: calculating the expected loss for being determined as all kinds of failures according to the empirical loss meter of all kinds of failure modes mistakes, and
Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss.
2. aero-engine multiclass failure minimum risk diagnostic method according to claim 1, which is characterized in that described
In step S3, using the method for the Maximum-likelihood estimation, and using the softmax function to all kinds of in the training set
The posterior probability of failure is modeled, wherein the optimization problem of Maximum-likelihood estimation is convex problem, and utilizes gradient descent method
Acquire global minimum.
3. aero-engine multiclass failure minimum risk diagnostic method according to claim 1, which is characterized in that described
In step S6, the decision rule of the fault diagnosis is to be calculated to sentence the state of present engine according to the empirical loss meter
It is set to the expected loss of failure of all categories, Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss, so that diagnostic result
Expected loss it is minimum.
4. aero-engine multiclass failure minimum risk diagnostic method according to claim 1, which is characterized in that described
In step S2 and the step S4, when carrying out dimensionality reduction to data set, using the dimension reduction method for having supervision --- linear discriminant analysis,
To efficiently use classification information.
5. aero-engine multiclass failure minimum risk diagnostic method according to claim 1, which is characterized in that described
In step S3, using the softmax function to the posterior probability P (Ck|x(i)；It θ) models, formula are as follows:
Wherein, x(i)A sample point is represented, θ is a K × (d+1) parameter matrix.
6. a kind of aero-engine multiclass failure minimum risk diagnostic device characterized by comprising
First acquisition module includes all kinds of failures and trouble-free sensor measurement data for acquiring aero-engine, with
Constitute original training set；
Processing module, for carrying out nondimensionalization processing to the original training set using maximum-Returning to one for minimum value method, and
Dimension-reduction treatment is carried out using linear discriminant analysis；
Modeling module, for the method using Maximum-likelihood estimation, and using softmax function to all kinds of events described in training set
The posterior probability of barrier is modeled, to obtain the softmax failure probability model of the engine；
Second acquisition module for acquiring in-service aero-engine sensor measures parameters, and carries out maximum-Returning to one for minimum value
Change and linear discriminant analysis dimension-reduction treatment；
Estimation module, for estimating all kinds of events of the in-service aero-engine using the softmax failure probability model
Hinder probability；And
Determination module is determined as that the expectation of all kinds of failures is damaged for calculating according to the empirical loss meter of all kinds of failure modes mistakes
It loses, and Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss.
7. aero-engine multiclass failure minimum risk diagnostic device according to claim 6, which is characterized in that described to build
Mould module is further used for the method using the Maximum-likelihood estimation, and using the softmax function to the training set
In the posterior probability of all kinds of failures modeled, wherein the optimization problem of Maximum-likelihood estimation is convex problem, and utilizes ladder
Degree descent method acquires global minimum.
8. aero-engine multiclass failure minimum risk diagnostic device according to claim 6, which is characterized in that the event
The decision rule of barrier diagnosis is to be calculated the state of present engine being determined as failure of all categories according to the empirical loss meter
Expected loss, Fault Tree Diagnosis Decision is carried out according to the smallest criterion of expected loss, so that the expected loss of diagnostic result is minimum.
9. aero-engine multiclass failure minimum risk diagnostic device according to claim 6, which is characterized in that data
When collection carries out dimensionality reduction, using the dimension reduction method for having supervision --- linear discriminant analysis, to efficiently use classification information.
10. aero-engine multiclass failure minimum risk diagnostic method according to claim 6, which is characterized in that wherein,
Using the softmax function to the posterior probability P (Ck|x(i)；It θ) models, formula are as follows:
Wherein, x(i)A sample point is represented, θ is a K × (d+1) parameter matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810860268.0A CN109308484A (en) | 2018-08-01 | 2018-08-01 | Aero-engine multiclass failure minimum risk diagnostic method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810860268.0A CN109308484A (en) | 2018-08-01 | 2018-08-01 | Aero-engine multiclass failure minimum risk diagnostic method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109308484A true CN109308484A (en) | 2019-02-05 |
Family
ID=65225760
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810860268.0A Pending CN109308484A (en) | 2018-08-01 | 2018-08-01 | Aero-engine multiclass failure minimum risk diagnostic method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109308484A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111581763A (en) * | 2019-02-15 | 2020-08-25 | 中国航发商用航空发动机有限责任公司 | Method for evaluating diagnosis result of gas circuit fault of aircraft engine |
| CN115130595A (en) * | 2022-07-05 | 2022-09-30 | 重庆电子工程职业学院 | Prediction-based aircraft data analysis and maintenance system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102809413A (en) * | 2012-08-06 | 2012-12-05 | 北京万集科技股份有限公司 | Detecting method and detecting device for rigged dynamic weighing of vehicle |
| CN102980771A (en) * | 2012-12-04 | 2013-03-20 | 南京航空航天大学 | Portable failure detection system and method for aero-engine gas path component |
| CN103489032A (en) * | 2013-09-27 | 2014-01-01 | 南京航空航天大学 | Aero-engine gas path component health diagnosis method based on particle filtering |
| CN105206270A (en) * | 2015-08-20 | 2015-12-30 | 长安大学 | Isolated digit speech recognition classification system and method combining principal component analysis (PCA) with restricted Boltzmann machine (RBM) |
| CN108108765A (en) * | 2017-12-28 | 2018-06-01 | 北京理工大学 | It is a kind of based on probability density than data fusion equipment fault diagnosis method |
| CN108229649A (en) * | 2016-12-22 | 2018-06-29 | 三星电子株式会社 | For the method and apparatus of deep learning training |
-
2018
- 2018-08-01 CN CN201810860268.0A patent/CN109308484A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102809413A (en) * | 2012-08-06 | 2012-12-05 | 北京万集科技股份有限公司 | Detecting method and detecting device for rigged dynamic weighing of vehicle |
| CN102980771A (en) * | 2012-12-04 | 2013-03-20 | 南京航空航天大学 | Portable failure detection system and method for aero-engine gas path component |
| CN103489032A (en) * | 2013-09-27 | 2014-01-01 | 南京航空航天大学 | Aero-engine gas path component health diagnosis method based on particle filtering |
| CN105206270A (en) * | 2015-08-20 | 2015-12-30 | 长安大学 | Isolated digit speech recognition classification system and method combining principal component analysis (PCA) with restricted Boltzmann machine (RBM) |
| CN108229649A (en) * | 2016-12-22 | 2018-06-29 | 三星电子株式会社 | For the method and apparatus of deep learning training |
| CN108108765A (en) * | 2017-12-28 | 2018-06-01 | 北京理工大学 | It is a kind of based on probability density than data fusion equipment fault diagnosis method |
Non-Patent Citations (3)
| Title |
|---|
| 彭喜元等: "《数据驱动的故障预测》", 31 March 2016, 哈尔滨工业大学出版社 * |
| 王和勇: "《面向大数据的高维数据挖掘技术》", 31 March 2018, 西安电子科技大学出版社 * |
| 王小川等: "《MATLAB神经网络43个案例分析》", 30 August 2013, 北京航空航天大学出版社 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111581763A (en) * | 2019-02-15 | 2020-08-25 | 中国航发商用航空发动机有限责任公司 | Method for evaluating diagnosis result of gas circuit fault of aircraft engine |
| CN111581763B (en) * | 2019-02-15 | 2023-10-17 | 中国航发商用航空发动机有限责任公司 | Method for evaluating air path fault diagnosis result of aero-engine |
| CN115130595A (en) * | 2022-07-05 | 2022-09-30 | 重庆电子工程职业学院 | Prediction-based aircraft data analysis and maintenance system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hanachi et al. | Performance-based gas turbine health monitoring, diagnostics, and prognostics: A survey | |
| US20190392098A1 (en) | Learning Expected Operational Behavior Of Machines From Generic Definitions And Past Behavior | |
| Xu et al. | PHM-oriented integrated fusion prognostics for aircraft engines based on sensor data | |
| US8417410B2 (en) | Operations support systems and methods with power management | |
| CN104756029B (en) | A kind of system of the parts group of monitoring device | |
| US8321118B2 (en) | Operations support systems and methods with power assurance | |
| CN106169001B (en) | A kind of aero-engine complete machine method for predicting residual useful life based on gas circuit performance parameter Graphic Pattern Matching | |
| CN102880170A (en) | System failure early warning method based on baseline model and Bayesian factor | |
| CN110998465B (en) | Turbine diagnostic feature selection system | |
| Simon et al. | Aircraft engine gas path diagnostic methods: public benchmarking results | |
| US7801695B2 (en) | Operations support systems and methods with model-based torque estimates | |
| CN107077134A (en) | For the instrument verified to the system for monitoring aircraft engine | |
| CN103324155A (en) | System monitoring | |
| CN113053171A (en) | Civil aircraft system risk early warning method and system | |
| Armstrong et al. | Implementation of an integrated on-board aircraft engine diagnostic architecture | |
| CN110096801A (en) | A kind of aero-engine high-pressure turbine disk dynamic reliability prediction technique | |
| CN109308484A (en) | Aero-engine multiclass failure minimum risk diagnostic method and device | |
| Zhao et al. | A hierarchical structure built on physical and data-based information for intelligent aero-engine gas path diagnostics | |
| Calderano et al. | An enhanced aircraft engine gas path diagnostic method based on upper and lower singleton type-2 fuzzy logic system | |
| Feng et al. | A multi-rate sensor fusion approach using information filters for estimating aero-engine performance degradation | |
| Ruan et al. | Remaining useful life prediction for aero-engine based on lstm and cnn | |
| Goebel | Conflict resolution using strengthening and weakening operations in decision fusion | |
| Zhou et al. | Fault diagnosis based on relevance vector machine for fuel regulator of aircraft engine | |
| Alberto-Olivares et al. | Remaining useful life prediction for turbofan based on a multilayer perceptron and kalman filter | |
| Shah et al. | Multi-Auto Associative Neural Network based sensor validation and estimation for aero-engine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190205 |