Aircraft System Anomaly Detection Using CDF Area Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods in aircraft systems, particularly in valves of air intake lip de-icing systems, fail to accurately distinguish between normal large, short-lived variations and actual anomalies due to information loss from data compression, leading to false positive detections.
Innovation Solution
Employing a learning system with pre-trained encoder/decoder neural networks, specifically recurrent neural networks with short-term and long-term memory, to reconstruct measurement series during normal operation, and comparing cumulative distribution functions to detect anomalies based on area separation rather than threshold-based detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If encoder/decoder neural network is used to compress and reconstruct measurement data, then information processing efficiency is improved, but measurement precision deteriorates due to information loss from compression
Solution Approach 1:
The patent changes the detection parameter from point-wise anomaly scores to cumulative distribution function area. This parameter transformation allows the system to tolerate local reconstruction errors from compression while maintaining sensitivity to true anomalies, effectively resolving the contradiction between compression efficiency and measurement precision.
Solution Approach 2:
The patent transitions from one-dimensional anomaly score comparison to two-dimensional cumulative distribution function area comparison. This dimensional change provides a more robust metric that is less sensitive to the information loss inherent in data compression, thereby maintaining measurement precision while preserving processing efficiency.
2Reliability
If detection threshold is set to maximize F-score, then anomaly detection sensitivity is improved, but false positive rate increases due to short-lived normal variations
Solution Approach 1:
The patent performs preliminary aggregation of anomaly information through cumulative distribution function calculation before making the detection decision. This preliminary action smooths out the effect of short-lived normal variations that would otherwise trigger false positives, while preserving sensitivity to sustained true anomalies.
Solution Approach 2:
The cumulative distribution function area serves as an intermediary metric between the raw anomaly scores and the final detection decision. This intermediary transformation reduces the impact of transient normal variations on the detection outcome, thereby reducing false positives while maintaining detection sensitivity.
3Loss of time
If encoder compression is applied to measurement data, then data processing speed is improved, but information completeness deteriorates leading to reconstruction errors
Solution Approach 1:
The patent changes the evaluation parameter from exact reconstruction fidelity to cumulative distribution function area comparison. This parameter change makes the system robust to information loss from compression, as the cumulative area metric is less sensitive to local reconstruction errors while still capturing true anomalies.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for detecting an anomaly in a system (110) of an aircraft (100), comprising: - obtaining a current series of measurements of one or more physical quantities (PT1, PT2) of the system (110), during a time period when the system (110) is in operation; - on the basis of the current series of measurements, providing, by means of an encoder-decoder (126), a current reconstructed series; and - comparing the reconstructed current series with the current series of measurements in order to obtain a current series of anomalies. The method further comprises: - computing a current distribution function of the current series of anomalies; - computing an area separating the current distribution function from a reference distribution function; and - comparing the area with a predefined threshold.