Aircraft Engine Anomaly Detection via Temporal Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current aircraft engine monitoring methods are inefficient due to the need for continuous, independent observation of numerous engine elements, leading to high computational demands and potential false anomalies or undetected faults, as they lack continuous monitoring of proper functioning indicators.
Innovation Solution
A method involving the definition of a behavior model using temporal regression, recalculating and monitoring a statistical evolution of the model to detect anomalies in the aircraft engine's control means, specifically for the regulation circuit of stator valves in turbojet engines, by breaking down data into flight regimes and using rational filters to identify deviations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If continuous monitoring of numerous engine elements is performed independently, then detection coverage is improved, but computational time and data processing load increase significantly
Solution Approach 1:
The patent combines multiple independent monitoring streams into a single integrated behavior model that processes engine parameters collectively. Instead of analyzing each engine element separately, the system merges data from various sensors and subsystems into a unified temporal regression model, reducing computational overhead while maintaining comprehensive anomaly detection coverage.
Solution Approach 2:
The behavior model serves multiple functions simultaneously: it characterizes normal engine behavior, detects anomalies, identifies drift conditions, and provides diagnostic information. This multi-functional approach eliminates the need for separate monitoring systems for each function, reducing overall computational load while improving detection reliability.
2Measurement precision
If a detailed model with many measurements is used, then measurement precision is improved, but device complexity and calculation requirements increase
Solution Approach 1:
The patent employs a dynamic behavior model that adapts to changing engine operating conditions through continuous recalibration. The model transitions between different operational states (normal operation, drift, anomaly) and adjusts its parameters accordingly, maintaining high detection accuracy without requiring an excessively complex static model structure.
Solution Approach 2:
The system changes key parameters of the behavior model based on operating conditions and detected states. During normal operation, the model uses baseline parameters; when drift is detected, parameters are recalibrated; during anomaly detection, different threshold parameters are applied. This dynamic parameter adjustment maintains precision while avoiding permanent complexity.
3Ease of operation
If independent discontinuous observation of engine elements is performed, then ease of operation is maintained, but false anomalies and undetected faults increase
Solution Approach 1:
The patent implements continuous monitoring through the ongoing calculation and updating of the behavior model. Rather than periodic discrete checks, the system continuously processes incoming sensor data, continuously recalibrates the model during drift conditions, and continuously evaluates anomaly indicators, eliminating gaps in detection coverage while maintaining operational simplicity.
Solution Approach 2:
The system incorporates feedback mechanisms where detected anomalies and drift conditions trigger model recalibration and threshold adjustments. The behavior model uses feedback from actual engine performance to continuously improve its characterization of normal vs. abnormal states, reducing false positives and ensuring reliable detection without complex manual intervention.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The invention relates to a method and to a system for detecting anomalies in an aircraft engine (1), comprising: - a means (5) for defining a behaviour model for a means (21) for controlling said aircraft engine (1) according to a temporal regression modelling the behaviour of said control means (21) according to a dataset relating to said control means including measurements of past behaviour and control and status measurements of said control means (21); - a means (5) for continuously recalculating said behaviour model for every new dataset; and - a means (5) for monitoring a statistical variation of said behaviour model to detect a behaviour anomaly in said control means reflecting an operating anomaly of said engine (1).