Flight Data Anomaly Detection via Kernel Entropy Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing flight data analysis systems require predefined detection rules to identify abnormal flights, which can lead to missed detection of singular events and non-compliance with safety standards.
Innovation Solution
A method that uses Gaussian kernel entropy component analysis to automatically detect abnormal flights by grouping flight data into signature vectors, classifying them based on distance to a normal flight zone, and determining an abnormality score to identify abnormal flights without predefined rules, along with generating a phantom flight for comparison.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If predefined detection rules are used to identify abnormal flights, then the system can detect common abnormal patterns, but it fails to detect singular events and unexpected issues
Solution Approach 1:
The patent transforms the detection approach by changing from fixed threshold parameters to dynamic statistical parameters. It uses kernel density estimation to compute probability density functions for flight parameters, allowing the system to adapt to different flight conditions and detect anomalies based on statistical deviations rather than predetermined thresholds. This enables detection of singular events that deviate from normal statistical patterns.
Solution Approach 2:
The system performs self-learning by automatically establishing baseline statistics from historical flight data without requiring expert-defined rules. The kernel density estimation algorithm automatically learns the normal distribution patterns of flight parameters, and the anomaly detection mechanism automatically identifies deviations from these learned patterns, making the system self-adapting to new flight scenarios.
2Measurement precision
If expert analysis is used to detect abnormal flights, then detailed analysis of predefined events is possible, but the process requires manual intervention and is time-consuming
Solution Approach 1:
The patent replaces the mechanical expert analysis process with an automated computational system. Instead of relying on human experts to manually review flight data and detect anomalies, the system uses kernel density estimation algorithms and statistical computing to automatically identify abnormal flights. This substitution maintains high detection precision while eliminating the time loss associated with manual expert review.
Solution Approach 2:
The patent introduces statistical probability density functions as an intermediary between raw flight data and anomaly detection. Rather than directly comparing flight parameters against fixed thresholds or requiring expert interpretation, the system uses kernel density estimation to create a probabilistic model that automatically mediates the detection process, enabling both precision and automation.
3Reliability
If traditional flight data analysis systems are used, then predefined events can be detected, but the system complexity increases with the need for rule definition and maintenance
Solution Approach 1:
The patent extracts the complexity of rule definition from the system by removing the need for predefined detection rules entirely. Instead of requiring experts to manually define thresholds and detection criteria, the system extracts statistical patterns directly from flight data using kernel density estimation. This extraction eliminates the configuration complexity while maintaining detection reliability.
Solution Approach 2:
The patent creates a universal detection mechanism that can handle multiple types of flight anomalies without requiring separate rules for each event type. The kernel density estimation approach provides a single, unified framework that automatically adapts to different flight parameters and anomaly types, making the system multi-functional without increasing complexity.
Data Source
Figure 1a~1b
Figure 2
AI summary
The invention relates to a method of analysing flight data recorded during N flights of at least one aircraft, by means of a flight data recorder of an aircraft, the data being grouped together by flight i in a signature vector of the flight X i of size d, the components of which correspond to data recorded during said flight i of the aircraft, a flight i being thus defined by the signature vector X i , the method comprising the following steps : Gaussian kernel entropy component analysis of the flight signatures X i to obtain a zone of normal flights and classifying the flight signatures X i with respect to their distance to said zone; determining, for each flight i, an abnormality score z i defined by the distance of a flight signature X i with respect to the zone of normal flights; detecting, as a function of the abnormality score z i , at least one abnormal flight.