System and method for detecting aircraft energy anomaly using neural network model

Patent metadata

Field	Value
Publication number	US-12409947-B2
Application number	US-202318349321-A
Country	US
Kind code	B2
Filing date	Jul 10, 2023
Priority date	Sep 22, 2022
Publication date	Sep 9, 2025
Grant date	Sep 9, 2025

Abstract

The present disclosure relates to a system and method for detecting an aircraft energy anomaly using an artificial neural network learning model. A system for detecting an aircraft energy anomaly using an artificial neural network learning model according to the present disclosure includes an input interface device for receiving ADS-B (Automatic Dependent Surveillance-Broadcast) data, a memory storing a program which generates a specific energy feature for energy state analysis by extending the ADS-B data through a preprocessing, and a processor for executing the program, wherein the processor generates an energy distribution model by the use of the specific energy feature, and performs artificial neural network-based energy anomaly learning.

First claim

What is claimed is:

1. A system for detecting an aircraft energy anomaly using an artificial neural network learning model, the system comprising:
an input interface device for receiving ADS-B (Automatic Dependent Surveillance-Broadcast) data;
a memory storing a program which generates a specific energy feature for energy state analysis by extending the ADS-B data through a preprocessing; and
a processor for executing the program,
wherein the specific energy feature relates to whether aircraft energy is too high or too low during an approach stage of an aircraft to a runway, and
wherein the processor generates an energy distribution model by the use of the specific energy feature, and performs artificial neural network-based energy anomaly learning.

2. The system of claim 1 , wherein the processor extracts the specific energy feature by the use of location information and motion information of the ADS-B data, and generates the energy distribution model.

3. The system of claim 2 , wherein the processor manages energy distribution data for each arrival airport runway.

4. The system of claim 3 , wherein the processor predicts an arrival runway by the use of runway heading information and latitude and longitude information of the runway threshold of the arrival airport.

5. The system of claim 1 , wherein the processor predicts whether an aircraft energy anomaly exists in real time by the use of a trained learning model and test data.

6. The system of claim 5 , wherein the processor calculates a specific energy value at a certain point in the ADS-B data of a flight which has arrived at a corresponding runway, for a remaining distance to the corresponding runway to generate a quantile value, and determines an energy state of a certain flight by comparing the quantile value with a quantile model of a runway of an arrival airport.

7. The system of claim 5 , wherein the processor analyzes an energy state at a point of the ADS-B data, which is time series data, and utilizes an analysis result as an input of a long short-term memory (LSTM) to predict an energy state at a next time point.

8. A method for detecting an aircraft energy anomaly using an artificial neural network learning model, the method comprising:
(a) generating a specific energy feature for energy state analysis by extending ADS-B data; and
(b) generating an energy distribution model by the use of the specific energy feature, and performing artificial neural network-based energy anomaly learning,
wherein the specific energy feature relates to whether aircraft energy is too high or too low during an approach stage of an aircraft to a runway.

9. The method of claim 8 , wherein the step (a) includes extracting the specific energy feature by the use of location information and motion information of the ADS-B data.

10. The method of claim 9 , wherein the step (a) includes managing energy distribution data for each arrival airport runway.

11. The method of claim 10 , wherein the step (a) includes predicting an arrival runway by the use of runway heading information and latitude and longitude information of the runway threshold of an arrival airport.

12. The method of claim 10 , wherein the step (b) includes learning energy data for each distance class of a runway, wherein time-series energy anomaly learning is performed by the use of an autoencoder learning result for each distance class of a runway as an input to an LSTM model.

13. The method of claim 8 , further comprising (c) predicting whether or not there is an aircraft energy anomaly in real time by the use of test data and a learning model which has been trained and distributed in the step (b).

14. The method of claim 13 , wherein the step (c) includes predicting an energy anomaly by the use of a hybrid model in which an autoencoder and an LSTM are combined.

Assignees

• Electronics & Telecommunications Res Inst

Inventors

• Park Noh-Sam
• Kim Ji Yeon
• Jang Jong Hyun

Classifications

• G08G5/54

  for approach or landing · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• B64D2045/0085

  Devices for aircraft health monitoring, e.g. monitoring flutter or vibration · CPC title

• G08G5/26

  Transmission of traffic-related information between aircraft and ground stations · CPC title

• G08G5/727

  from a ground station · CPC title