RU2011129298A

RU2011129298A - IDENTIFICATION OF FAILURES IN THE AIRCRAFT ENGINE

Info

Publication number: RU2011129298A
Application number: RU2011129298/08A
Authority: RU
Inventors: Жером ЛАКАЙ
Original assignee: Снекма
Priority date: 2008-12-15
Filing date: 2009-12-14
Publication date: 2013-01-20

Claims

1. A method for identifying failures in an aircraft engine (1), the method is characterized in that it comprises the following steps, in which:

using sensors (3a-3f) to collect time series measurements from said aircraft engine (1) and its environment;

using processor means (5) for calculating, from said measurements of a time series, indicators that are characteristic of elements of said aircraft engine (1);

use processor means (5) to determine, from the said characteristic indicators, a set of normalized indicators that are representative of the operation of the aforementioned aircraft engine (1);

using processor means (5) for constructing an anomaly vector representing the behavior of said engine (1) as a function of said set of normalized indicators;

using processor means (5), in the event of a failure detected by the anomaly vector, to select a subset of the basis vectors having directions belonging to a certain neighborhood of the direction of the anomaly vector, the mentioned subset of the basis vectors is selected from the set of basis vectors associated with the failures of the aircraft engine (1) and defined using criteria established by experts; and

using processor means (5) to identify failures associated with said subset of basis vectors;

and the fact that the selection of said subset of the basis vectors contains the steps in which:

using processor means (5) for calculating the geodetic distances between the projection of the above-mentioned anomaly vector and the projections of the mentioned basic vectors on a sphere in a dimension space equal to the number of indicators in the said set of normalized indicators minus the number of linear relationships between the indicators;

using processor means (5) for comparing said geodetic distances in pairs;

using processor means (5) for classifying basis vectors in increasing order of geodesic distances with respect to said anomaly vector; and

use processor means (5) to form said subset of the basis vectors from the first basis vectors having a classification order less than a certain rank.

2. The method according to claim 1 , characterized in that said sphere has a radius of 1.

3. The method according to claim 1, characterized in that it includes the following steps, in which:

use processor means (5) to determine, for each basis vector, the a priori probability of occurrence based on criteria established by experts; and

use processor means (5) to calculate, for each basis vector, the posterior probability of occurrence as a function of the a priori probability of occurrence and the mentioned geodesic distances.

4. The method according to claim 1, characterized in that said set of normalized indicators

contains indicators

identified by means (5) of the processor using criteria established by experts.

5. The method according to claim 4, characterized in that said set of normalized indicators

additionally contains dynamic indicators built by means (5) of the processor as a function of indicators at the present and past moments

, and representing the behavior of said aircraft engine over time.

6. The method according to claim 1, characterized in that the construction of the above-mentioned vector of anomalies consists in that:

use processor means (5) to form a vector

indicators from said set of indicators; and

use processor means (5) to construct said anomaly vector z by re-normalizing said vector

indicators using the following formula:

Where

- the average value of the indicator vectors, and

- square root of the pseudo-inverse signal

covariance matrix

.

7. The method according to claim 6, characterized in that it includes the following steps, in which:

use the processor tool (5) to calculate the norm of the above-mentioned anomaly vector using the Mahalanobis distance:

; and

using processor means (5) for detecting anomalies of said aircraft engine using a trigger threshold defined as a function of the statistical distribution of said norm of the anomaly vector.

8. The method according to claim 1, characterized in that the said set of basis vectors is constructed in accordance with the imitating behavior of indicators in case of anomalies.

9. The method according to claim 3, characterized in that it further consists in the fact that:

use processor tool (5) to establish a grid of solutions when applying criteria established by experts;

using processor means (5) for applying Bayes rules to derive component-wise failure probabilities from said posterior probabilities of occurrence and from said decision grid; and

using processor means (5) for detecting faulty physical components that are responsible for said failures when using said component-wise failure probabilities.

10. The method according to claim 9, characterized in that said decision grid is formed from a matrix of conditional probabilities, that the component is faulty, knowing that the failure has been noticed, and from a sequence of coefficients corresponding to the a priori probability of failure of each component.

11. The method according to claim 9, characterized in that the said grid of solutions is complemented by machine learning.

12. System for identifying failures in an aircraft engine (1), the system is characterized in that it contains:

sensors (3a-3f) for collecting time series measurements from said aircraft engine (1) and its environment;

means (5) for calculating, from said measurements of a time series, indicators that are characteristic of elements of said aircraft engine (1);

means (5) for using said characteristic indicators to determine a set of normalized indicators representing the operation of said aircraft engine (1);

means (5) for constructing an anomaly vector representing the behavior of said engine (1) as a function of said set of normalized indicators;

means (5) for selecting, in the case of an anomaly detected by said anomaly vector, a subset of basis vectors having directions belonging to a certain neighborhood of the direction of said anomaly vector, said subset of basis vectors being selected from a set of basis vectors associated with the failures of said aircraft engine, and defined using criteria established by experts;

means (5) for identifying failures associated with said subset of basis vectors;

and the fact that the means for selecting said subset of basis vectors contains:

means (5) for calculating the geodesic distances between the projection of the anomaly vector mentioned and the projections of the mentioned basis vectors on a sphere in a space of dimension equal to the number of indicators from the said set of normalized indicators minus the number of linear relationships between the said indicators;

means (5) for comparing said geodetic distances in pairs;

means (5) for classifying basis vectors in increasing order of their geodesic distances with respect to said anomaly vector; and

means (5) for forming said subset of the basis vectors from the first basis vectors having a classification order less than a certain rank.

13. The system according to p. 12, characterized in that it includes:

means (5) for determining, for each basis vector, the a priori probability of occurrence when applying the criteria established by experts; and

means (5) for calculating, for each basis vector, the posterior probability of occurrence as a function of the a priori probability of occurrence and the mentioned geodesic distances.

14. The system according to p. 12, characterized in that it further includes:

means (5) for establishing a grid of decisions when applying criteria established by experts;

means (5) for using Bayes rules to derive componentwise failure probabilities from posterior probabilities of occurrence and from the mentioned decision grid; and

means (5) for detecting faulty physical components that are responsible for said failures, according to said component-wise failure probabilities.

15. A computer program including instructions for implementing a method for identifying failures according to claim 1, when executed by a processor means.