A Neural Algorithm for the Detection and Correction of Anomalies: Application to the Landing of an Airplane

Abstract

The location of the plane is key during the landing operation. A set of sensors provides data to get the best estimation of plane localization. However, data can contain anomalies. To guarantee correct behavior of the sensors, anomalies must be detected. Then, either the faulty sensor is isolated or the detected anomaly is filtered. This article presents a new neural algorithm for the detection and correction of anomalies named NADCA. This algorithm uses a compact deep learning prediction model and has been evaluated using real and simulated anomalies in real landing signals. NADCA detects and corrects both fast-changing and slow-moving anomalies; it is robust regardless of the degree of oscillation of the signals and sensors with abnormal behavior do not need to be isolated. NADCA can detect and correct anomalies in real time regardless of sensor accuracy. Likewise, NADCA can deal with simultaneous anomalies in different sensors and avoid possible problems of coupling between signals. From a technical point of view, NADCA uses a new prediction method and a new approach to obtain a smoothed signal in real time. NADCA has been developed to detect and correct anomalies during the landing of an airplane, hence improving the information presented to the pilot. Nevertheless, NADCA is a general-purpose algorithm that could be useful in other contexts. NADCA evaluation has given an average F-score value of 0.97 for anomaly detection and an average root mean square error (RMSE) value of 2.10 for anomaly correction.

Keywords: airplane landing; anomaly correction; anomaly detection; deep learning.

Figure 1

Anomaly detection and correction zone during the landing of an airplane.

Figure 2

Example of simulated time series of the Z coordinate during the landing process: Z^GPS, Z^IRS, Z^ILS, and Z^RA. On the right side, a table relates each sensor to each coordinate. The sensor coordinate cell shows whether or not the signal has oscillations. There is no signal if the cell is empty.

Figure 3

PM^Z prediction. In the lower part, an example of measure prediction for the GPS is explained.

Figure 4

Main elements and basic behavior of NADCA. The red dot at time i + 1 is the measure prediction.

Figure 5

NADCA-L: Generalization of NADCA-B for anomaly detection and correction in signals without oscillations.

Figure 6

Steps of NADCA-L.

Figure 7

NADCA-O: Generalization of NADCA for anomaly detection and correction in signals with oscillations.

Figure 8

GPS and IRS for the Z coordinate (real values of a landing).

Figure 9

ILS, ILS^L, RA, and RA^L for the Z coordinate (real values of a landing).

Figure 10

IRS, GPS, GPS^L, ILS, and ILS^L for the Y coordinate (real values of a landing).

Figure 11

IRS portion as a function of GPS (real values for the X coordinate).

Figure 12

PM^Z (real values).

Figure 13

Evaluation curve for the Z coordinate using real landing values.

Figure 14

PM^Y (real values).

Figure 15

Evaluation curve for the Y coordinate using real landing values.

Figure 16

Predictive model for the X coordinate (real values).

Figure 17

Evaluation curve for the X coordinate using real landing values.

Figure 18

Envelopes for GPS and IRS using real values.

Figure 19

Envelope for ILS using real values.

Figure 20

An envelope for the positive differences and another for the negative ones using real RA values.

Figure 21

Envelope for ILS using real values for the Y coordinate.

Figure 22

Relationship between $\mathsf{\alpha }$ and $C^{*}$ (or $C^{**}$).

Figure 23

Anomaly detected and corrected using NADCA-L. The small anomaly appears in red.

Figure 24

Anomaly detected and corrected using NADCA-O.

Figure 25

Bias and small noise anomalies detected and corrected on a specific landing using NADCA.

Figure 26

Noise anomaly detected and corrected on a specific landing using NADCA.

Figure 27

Noisy bias anomaly detected and corrected on a specific landing using NADCA.

Figure 28

Drift anomaly detected and corrected on a specific landing using NADCA.

Figure 29

Drift anomaly detected and corrected on a specific landing using NADCA.

Figure 30

On the left side, coupling problem between X^IRS and Y^ILS for a drift anomaly on X^IRS. On the right side, the anomaly detection and correction on X^IRS.

Figure 31

Drift anomaly detected and corrected on a specific landing using NADCA.