Data-Driven Anomaly Detection for Aircraft Subsystems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern aircraft systems face challenges in anticipating and predicting flight deck effects, which can lead to unplanned maintenance and business losses due to the complexity of monitoring multiple subsystems effectively.
Innovation Solution
A data-driven anomaly detection method that involves monitoring parameters, collecting data, generating monitoring quantities, and using statistical analysis to identify anomalies before they cause flight deck effects, utilizing baseline models to detect deviations and generate alerts, and contribution plots to determine the contributing factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional monitoring methods are used to track multiple aircraft subsystems, then system coverage is comprehensive, but the complexity of monitoring and analysis increases significantly
Solution Approach 1:
The patent segments the complex monitoring task by dividing it into multiple specialized modules: data collection module, baseline model construction module, anomaly detection module, and contribution analysis module. Each module handles a specific aspect of the monitoring process, making the overall system more manageable and less complex while maintaining comprehensive coverage of multiple aircraft subsystems
Solution Approach 2:
The patent introduces baseline models as intermediary representations that capture normal system behavior patterns. These baseline models serve as mediators between raw sensor data and anomaly detection, simplifying the comparison process by providing a reference framework against which current system states can be evaluated without directly analyzing all raw data
2Measurement precision
If comprehensive parameter monitoring is implemented across all subsystems, then anomaly detection capability is improved, but data processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary action by constructing baseline models during periods when the system is known to be operating normally. This pre-computation of reference patterns allows the anomaly detection phase to proceed more quickly, as the heavy computational work of establishing what constitutes normal behavior is done in advance, enabling real-time or near-real-time anomaly detection
Solution Approach 2:
The patent extracts only the most relevant features and parameters from the comprehensive sensor data that are necessary for anomaly detection. By selecting and extracting key indicators rather than processing all available data, the system maintains high anomaly detection capability while reducing the computational burden and processing time
3Measurement precision
If statistical analysis is performed on all sensor data to detect anomalies, then detection accuracy is improved, but system resource consumption increases
Solution Approach 1:
The patent applies local quality by performing statistical analysis selectively rather than uniformly across all data. The contribution plot technique identifies which specific parameters or sensors are contributing most to detected anomalies, allowing the system to focus computational resources on analyzing only those local areas of the system that are exhibiting abnormal behavior, thereby reducing overall resource consumption while maintaining detection accuracy
Data Source
AI summary
A method for data-driven anomaly detection may include monitoring a plurality of parameters associated with a plurality of subsystems of a system. The method may also include collecting data corresponding to each of the plurality of parameters from the plurality of subsystems and generating monitoring quantities based on the data. The method may also include determining if any quantities in the monitoring quantities exceed a predetermined limit. A contribution plot may be generated corresponding to each of the parameters in response to any of the quantities exceeding the predetermined limit. The method may further include determining which parameter is likely to cause an effect based on the contribution plot.