Predicting failures in an aircraft

Patent metadata

Field	Value
Publication number	US-10360741-B2
Application number	US-201715606917-A
Country	US
Kind code	B2
Filing date	May 26, 2017
Priority date	Jun 2, 2016
Publication date	Jul 23, 2019
Grant date	Jul 23, 2019

Abstract

The disclosure herein proposes identifying precursors to all the phenomena which can have an impact on the service use of an aircraft. It relates to a system for predicting failures in an aircraft, including a processor configured to analyze a current behavior of at least one flight parameter of the aircraft to detect any deviation of the current behavior relative to a predetermined behavior model of the parameter, the behavior model being determined from a plurality of series of learning data relating to the parameter collected during the flights of a set of aircraft.

First claim

The invention claimed is:

1. A method for predicting failures in an aircraft, the method comprising analyzing a current behavior of at least one flight parameter of the aircraft to detect any deviation of the current behavior relative to a preconstructed behavior model of the flight parameter, the behavior model being constructed in a preliminary learning phase from a plurality of series of learning data relating to the flight parameter that have been collected during flights of a set of aircraft, wherein detection of any deviation of the current behavior of the flight parameter relative to the preconstructed behavior model of the flight parameter comprises:
collecting, on each current flight, a current series of data from measurements relating to the flight parameter;
segmenting the current series of data into a predetermined number of segments to transform it into a current string of segments, the predetermined number of segments being defined by the preconstructed behavior model;
aligning the segments of the current string of segments relative to a reference string of segments defined by the preconstructed behavior model;
assigning predetermined characteristic measurements to the current string of segments to generate a current measurement vector; and
computing a distance between the current measurement vector and a reference measurement vector defined by the preconstructed behavior model, a value of the distance being representative of a normal or deviant behavior of the flight parameter.

2. The method of claim 1 , further comprising computing a score of abnormal behavior of the flight parameter.

3. The method of claim 2 , further comprising identifying a set of flights of the aircraft exhibiting a deviant behavior and of maintenance actions performed during the set of flights.

4. The method of claim 1 , further comprising identifying a set of flights of the aircraft exhibiting a deviant behavior and of maintenance actions performed during the set of flights.

5. The method of claim 1 , wherein each series of learning data or current series of data comprises temporal data from a flight data recorder and/or temporally indexed failure messages from a centralized supervision system.

6. The method of claim 1 , wherein construction of the behavior model relating to a given flight parameter comprises:
collecting, on each of the flights of the set of the aircraft, at least one series of learning data from measurements relating to the flight parameter thus forming, during the flights of the set of the aircraft, the plurality of series of learning data;
segmenting each series of learning data into an optimal number of segments to transform it into a string of learning segments, the optimal number of segments defining the predetermined number of segments associated with the flight parameter;
transforming the set of series of learning data into a corresponding set of strings of learning segments;
selecting a central string of segments from the set of strings of learning segments, the central string of segments defining the reference string of segments;
aligning the segments of the set of strings of learning segments by being identified relative to the reference string of segments;
generating a learning measurement vector for each string of learning segments by assigning predetermined characteristic measurements to each learning segment and/or to the set of segments of each string of learning segments;
computing a learning distance between each learning measurement vector and a reference measurement vector associated with the reference string of segments; and
partitioning the set of the learning measurement vectors according to consecutive intervals as a function of the learning distances thus defining a normal behavior interval and atypical behavior intervals.

7. The method of claim 6 , wherein the central string of segments corresponds to the medoid of the set of strings of segments, the method comprising:
saving the reference string of segments, the distances between the measurement vectors and the reference measurement vector, and the partitioning of the measurement vectors; and
displaying a graph representative of the behavior model.

8. The method of claim 1 , comprising displaying a result illustrating a behavior deviation of one or more flight parameters.

9. The method of claim 6 , further comprising:
saving the reference string of segments, the distances between the measurement vectors and the reference measurement vector, and the partitioning of the measurement vectors; and
displaying a graph representative of the behavior model.

10. A system for predicting failures in an aircraft, comprising a processor configured to analyze a current behavior of at least one flight parameter of the aircraft to detect any deviation of the current behavior relative to a predetermined behavior model of the flight parameter, the behavior model being determined from a plurality of series of learning data relating to the flight parameter collected during flights of a set of aircraft and detection of any deviation of behavior of a flight parameter comprising:
collecting, on each current flight, a current series of data from measurements relating to the flight parameter;
segmenting the current series of data into a predetermined number of segments to transform it into a current string of segments, the number of segments being defined by the behavior model;
aligning the segments of the current string of segments relative to a reference string of segments defined by the behavior model;
assigning predetermined characteristic measurements to the current string of segments to generate a current measurement vector; and
computing a distance between the current measurement vector and a reference measurement vector defined by the behavior model, a value of the distance being representative of a normal or deviant behavior of the flight parameter.

11. A monitoring system comprising maintenance and management avionics systems, and further comprising the system for predicting failures of claim 10 .

12. The system of claim 10 , wherein the system is configured to compute a score of abnormal behavior of the flight parameter.

13. The system of claim 12 , wherein the system is configured to identify a set of flights of the aircraft exhibiting a deviant behavior and of maintenance actions performed during the set of flights.

14. The system of claim 10 , wherein the system is configured to identify a set of flights of the aircraft exhibiting a deviant behavior and of maintenance actions performed during the set of flights.

15. The system of claim 10 , wherein each series of learning data or current series of data comprises temporal data from a flight data recorder and/or temporally indexed failure messages from a centralized supervision system.

16. The system of claim 10 , wherein construction of the behavior model relating to a given flight parameter comprises:
collecting, on each of the flights of the set of the aircraft, at least one series of learning data from measurements relating to the flight parameter thus forming, during the flights of the set of the aircraft, the plurality of series of learning data;
segmenting each series of learning data into an optimal number of segments to transform it into a string of learning segments, the optimal number of segments defining the predetermined number of segments associated with the flight parameter;
transforming the set of series of learning data into a corresponding set of strings of learning segments;
selecting a central string of segments from the set of string

Assignees

• Airbus Operations Sas
• Airbus Sas

Inventors

• Huet Jean-Max
• Duluc Franck
• Benhamou Eric
• Maillard Bruno

Classifications

• G07C5/0808 Primary

  Diagnosing performance data (testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005) · CPC title

• G05B23/0221

  Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods · CPC title

• G06F11/008

  Reliability or availability analysis · CPC title

• B64D45/00

  Aircraft indicators or protectors not otherwise provided for · CPC title

• G06N5/04

  Inference or reasoning models · CPC title