DETECTING FAULT STATES OF AN AIRCRAFT

• US 20190333294A1
• Filed: 04/30/2018
• Published: 10/31/2019
• Est. Priority Date: 04/30/2018
• Status: Active Grant

• Current Assignee: The Boeing Co.
• Original Assignee: The Boeing Co.
• Inventors: Sundareswara, Rashmi, Lu, Tsai-Ching, Betz, Franz David
• Granted Patent: US 10,964,132 B2
• Time in Patent Office: Days
• CPC Class Codes: B64D 2045/0085 Devices for aircraft health monitoring, e.g. monitoring flutter or vibration
• CPC Class Codes: B64D 45/00 Aircraft indicators or protectors not otherwise provided for camouflage F41H3/00
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/045 Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G07C 5/008 communicating information to a remotely located station transmission systems for measured values G08C
• CPC Class Codes: G07C 5/0808 Diagnosing performance data testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005

Abstract

An apparatus for detecting fault states of an aircraft is provided. The apparatus receives training data including operational parameters of the aircraft operating over a plurality of training flight legs and applies a first clustering algorithm separately to the training data to produce first clustered data. The apparatus applies a second clustering algorithm to the first clustered data to produce second clustered data that indicates a plurality of states describing behavior of the aircraft operating over the training legs and applies a third clustering algorithm to identify one or more fault states of the aircraft from the plurality of states based on the training data and the second clustered data. The apparatus receives messages carrying data from the aircraft over a digital datalink, detects a fault state of the one or more fault states from the data, and generates an alert of the fault state.

Claims

1. An apparatus for detecting fault states of an aircraft, the apparatus comprising a processor and a memory storing executable instructions that, in response to execution by the processor, cause the apparatus to at least:

receive training data including operational parameters of an aircraft operating over a plurality of training flight legs;

apply a first clustering algorithm separately to the training data for each of the plurality of training flight legs to produce first clustered data, the first clustering algorithm being configured to cluster the operational parameters into one or more respective clusters for each of the plurality of training flight legs;

apply a second clustering algorithm to the first clustered data to produce second clustered data that indicates a plurality of states describing behavior of the aircraft operating over the plurality of training flight legs;

apply a third clustering algorithm to the second clustered data to identify one or more fault states of the aircraft from the plurality of states based on the training data and the second clustered data;

establish a digital datalink with the aircraft;

receive messages carrying data from the aircraft over the digital datalink;

detect a fault state of the one or more fault states from the data; and

in response,

generate an alert.

2. The apparatus of claim 1, wherein the apparatus being caused to apply the third clustering algorithm includes being caused to:

identify each of the plurality of training flight legs as a first type of flight leg in which the training data describes the aircraft in a first set of the plurality of states, or a second type of flight leg in which the training data describes the aircraft in a second set of the plurality of states, based on the training data and the second clustered data; and

identify the one or more fault states of the aircraft from the plurality of states based on a difference in states between the first set of the plurality of states of the first type of flight leg and the second set of the plurality of states of the second type of flight leg.

3. The apparatus of claim 1, wherein the apparatus being caused to apply the first clustering algorithm includes being caused to apply the first clustering algorithm separately to the training data of each of a plurality of flight phases for each of the plurality of training flight legs.

4. The apparatus of claim 1, wherein the apparatus being caused to apply the second clustering algorithm includes being caused to produce a library including the second clustered data that indicates the plurality of states, and the apparatus being caused to apply the third clustering algorithm includes being caused to identify the one or more fault states based on the training data and the library.

5. The apparatus of claim 1, wherein the apparatus being caused to apply the first clustering algorithm includes being caused to apply a Dirichlet Process Gaussian Mixture Model (DPGMM), and the apparatus being caused to apply the second clustering algorithm includes being caused to apply a DPGMM or a Density Based Spatial Clustering of Applications with Noise (DBSCAN).

6. The apparatus of claim 1, wherein the apparatus being caused to apply the second clustering algorithm includes being caused to assign the training data for each of a plurality of points in time for each of the plurality of training flight legs separately to one of the plurality of states.

7. The apparatus of claim 1, wherein the apparatus being caused to detect the fault state includes being caused to at least:

assign the data for each of a plurality of points in time separately to one of the plurality of states; and

detect that the data for one of the plurality of points in time is assigned to the fault state.

8. The apparatus of claim 1, wherein the apparatus being caused to apply the third clustering algorithm includes being caused to apply the third clustering algorithm to a histogram of the second clustered data for each of the plurality of training flight legs.

9. A method of detecting fault states of an aircraft, the method comprising:

receiving training data including operational parameters of an aircraft operating over a plurality of training flight legs;

applying a first clustering algorithm separately to the training data for each of the plurality of training flight legs to produce first clustered data, the first clustering algorithm being configured to cluster the operational parameters into one or more respective clusters for each of the plurality of training flight legs;

applying a second clustering algorithm to the first clustered data to produce second clustered data that indicates a plurality of states describing behavior of the aircraft operating over the plurality of training flight legs;

applying a third clustering algorithm to the second clustered data to identify one or more fault states of the aircraft from the plurality of states based on the training data and the second clustered data;

establishing a digital datalink with the aircraft;

receiving messages carrying data from the aircraft over the digital datalink;

detecting a fault state of the one or more fault states from the data; and

in response,

generating an alert.

10. The method of claim 9, wherein applying the third clustering algorithm includes:

identifying each of the plurality of training flight legs as a first type of flight leg in which the training data describes the aircraft in a first set of the plurality of states, or a second type of flight leg in which the training data describes the aircraft in a second set of the plurality of states, based on the training data and the second clustered data; and

identifying the one or more fault states of the aircraft from the plurality of states based on a difference in states between the first set of the plurality of states of the first type of flight leg and the second set of the plurality of states of the second type of flight leg.

11. The method of claim 9, wherein applying the first clustering algorithm includes applying the first clustering algorithm separately to the training data of each of a plurality of flight phases for each of the plurality of training flight legs.

12. The method of claim 9, wherein applying the second clustering algorithm includes producing a library including the second clustered data that indicates the plurality of states, and wherein applying the third clustering algorithm includes identifying the one or more fault states based on the training data and the library.

13. The method of claim 9, wherein applying the first clustering algorithm includes applying a Dirichlet Process Gaussian Mixture Model (DPGMM), and wherein applying the second clustering algorithm includes applying a DPGMM or a Density Based Spatial Clustering of Applications with Noise (DBSCAN).

14. The method of claim 9, wherein applying the second clustering algorithm includes assigning the training data for each of a plurality of points in time for each of the plurality of training flight legs separately to one of the plurality of states.

15. The method of claim 9, wherein detecting the fault state includes at least:

assigning the data for each of a plurality of points in time separately to one of the plurality of states; and

detecting that the data for one of the plurality of points in time is assigned to the fault state.

16. The method of claim 9, wherein applying the third clustering algorithm includes applying the third clustering algorithm to a histogram of the second clustered data for each of the plurality of training flight legs.

17. A computer-readable storage medium for detecting fault states of an aircraft, the computer-readable storage medium being non-transitory and having computer-readable program code stored therein that in response to execution by a processor, causes an apparatus to at least:

receive training data including operational parameters of an aircraft operating over a plurality of training flight legs;

apply a first clustering algorithm separately to the training data for each of the plurality of training flight legs to produce first clustered data, the first clustering algorithm being configured to cluster the operational parameters into one or more respective clusters for each of the plurality of training flight legs;

apply a second clustering algorithm to the first clustered data to produce second clustered data that indicates a plurality of states describing behavior of the aircraft operating over the plurality of training flight legs;

apply a third clustering algorithm to the second clustered data to identify one or more fault states of the aircraft from the plurality of states based on the training data and the second clustered data;

establish a digital datalink with the aircraft;

receive messages carrying data from the aircraft over the digital datalink;

detect a fault state of the one or more fault states from the data; and

in response,

generate an alert.

18. The computer-readable storage medium of claim 17, wherein the apparatus being caused to apply the third clustering algorithm includes being caused to:

identify each of the plurality of training flight legs as a first type of flight leg in which the training data describes the aircraft in a first set of the plurality of states, or a second type of flight leg in which the training data describes the aircraft in a second set of the plurality of states, based on the training data and the second clustered data; and

identify the one or more fault states of the aircraft from the plurality of states based on a difference in states between the first set of the plurality of states of the first type of flight leg and the second set of the plurality of states of the second type of flight leg.

19. The computer-readable storage medium of claim 17, wherein the apparatus being caused to apply the first clustering algorithm includes being caused to apply the first clustering algorithm separately to the training data of each of a plurality of flight phases for each of the plurality of training flight legs.

20. The computer-readable storage medium of claim 17, wherein the apparatus being caused to apply the second clustering algorithm includes being caused to produce a library including the second clustered data that indicates the plurality of states, and the apparatus being caused to apply the third clustering algorithm includes being caused to identify the one or more fault states based on the training data and the library.

21. The computer-readable storage medium of claim 17, wherein the apparatus being caused to apply the first clustering algorithm includes being caused to apply a Dirichlet Process Gaussian Mixture Model (DPGMM), and the apparatus being caused to apply the second clustering algorithm includes being caused to apply a DPGMM or a Density Based Spatial Clustering of Applications with Noise (DBSCAN).

22. The computer-readable storage medium of claim 17, wherein the apparatus being caused to apply the second clustering algorithm includes being caused to assign the training data for each of a plurality of points in time for each of the plurality of training flight legs separately to one of the plurality of states.

23. The computer-readable storage medium of claim 17, wherein the apparatus being caused to detect the fault state includes being caused to at least:

assign the data for each of a plurality of points in time separately to one of the plurality of states; and

detect that the data for one of the plurality of points in time is assigned to the fault state.

24. The computer-readable storage medium of claim 17, wherein the apparatus being caused to apply the third clustering algorithm includes being caused to apply the third clustering algorithm to a histogram of the second clustered data for each of the plurality of training flight legs.