Aircraft Component Anomaly Detection Using Time-Sequence Approximation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting operating anomalies in aircraft components, such as the oil circulation system, are inadequate as they only cover pre-identified anomalies during specific flight phases and fail to detect anomalies not accounted for by existing algorithms.
Innovation Solution
A method and device using an approximation module to determine an approximated time sequence from acquired values, calculating a global anomaly score based on differences between acquired and approximated values, and detecting anomalies by comparing the score to a threshold, utilizing machine learning algorithms like autoencoders to minimize approximation errors and distinguish normal from abnormal operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If prior art methods use pre-identified anomaly detection algorithms for specific flight phases, then detection coverage for known anomalies is improved, but detection capability for unknown or unaccounted anomalies deteriorates
Solution Approach 1:
The system performs preliminary actions by acquiring reference time sequences from multiple flights during normal operations before actual anomaly detection. These reference sequences are stored and used to train the approximation module in advance, enabling the system to recognize normal operation patterns and detect deviations without requiring pre-programmed anomaly knowledge.
Solution Approach 2:
The approximation module performs self-service by automatically learning normal operation patterns from reference data and autonomously detecting anomalies in acquired time sequences. The system improves its detection capability through continuous learning from stored reference sequences without requiring external reprogramming or manual anomaly definition.
2Ease of operation
If traditional methods rely on instantaneous snapshot measurements at specific flight phases, then detection simplicity is improved, but detection accuracy for continuous anomalies deteriorates
Solution Approach 1:
The system implements continuous monitoring by acquiring time sequences of status indicator values throughout the entire operation of aircraft components, rather than taking isolated snapshots. This continuous acquisition captures the evolution of component behavior over time, enabling detection of anomalies that develop gradually or occur between discrete measurement points.
Solution Approach 2:
The system acquires more data than traditional methods by collecting continuous time sequences of multiple status indicators throughout operation. This excessive data acquisition includes redundant information that, when processed by the approximation module, provides more robust anomaly detection capability compared to minimal snapshot measurements.
3Reliability
If approximation module is trained to minimize errors for normal operation reference sequences, then false positive reduction is improved, but detection sensitivity for subtle anomalies may deteriorate
Solution Approach 1:
The system segments the anomaly detection process into two distinct phases: training the approximation module on normal operation reference sequences to establish a baseline, and then using this trained module to detect deviations in acquired sequences. This segmentation allows the system to achieve high reliability by learning what normal operation looks like across multiple flights, while maintaining sensitivity to subtle anomalies through continuous comparison against this established baseline.
Data Source
AI summary
Method for detecting an operating anomaly of an aircraft component associated with at least one status indicator, said method—for at least one acquired time sequence (Seq1, Seq2) comprising acquired values of said at least one indicator—comprising steps of:determining (E20) an approximated time sequence (Seq1*, Seq2*) of approximated values of said acquired values, by an approximation module (AE1, AE2);determining (E30) a global anomaly score (A1, A2) for said acquired time sequence (Seq1, Seq2) from differences between said acquired values (Seq1, Seq2) and approximated values; anddetecting (E40) an operating anomaly of said component as a function of a comparison of said global anomaly score with a first threshold,said approximation module being configured to minimize approximation errors between time sequences of reference values and time sequences of approximated values.