Anomaly Detection in Vehicle Subsystems via Sensor Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting degradation and isolating faults in complex systems, such as aircraft, is challenging due to limited sensor data, varying operating conditions, and the need to characterize nominal behavior under dynamic conditions.
Innovation Solution
A method that identifies a chain of operations for complex systems, clusters operation phases based on operating conditions using sensor data, computes statistical values for parameters across these phases, identifies outlier parameters, and excludes them to compute nominal values for sub-systems, thereby detecting anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is collected under varied operating conditions to characterize nominal behavior, then detection accuracy improves, but data variability and complexity increase
Solution Approach 1:
The patent segments the complex sensor data by dividing operation phases into clusters based on operating conditions. This segmentation transforms the complex varied data into manageable groups, allowing nominal behavior to be characterized within each cluster while maintaining overall detection accuracy across diverse operating conditions.
Solution Approach 2:
The patent performs preliminary clustering of operation phases before anomaly detection. By pre-organizing sensor data into clusters based on operating conditions, the system establishes a structured foundation that simplifies subsequent anomaly detection while preserving the benefits of varied operating condition data.
2Measurement precision
If outlier parameters are included in sensor data, then data completeness is maintained, but anomaly detection accuracy deteriorates
Solution Approach 1:
The patent extracts and removes outlier parameters from the sensor data after identifying them through statistical analysis. This extraction process eliminates data points that would degrade anomaly detection accuracy while preserving the majority of informative sensor data, thus maintaining data completeness for valid measurements.
Solution Approach 2:
The patent introduces statistical values as an intermediary between raw sensor data and anomaly detection. These statistical values serve as a mediator to identify and filter outlier parameters, allowing the system to distinguish between valid operational variations and true anomalies while maintaining overall data integrity.
3Reliability
If statistical analysis is performed on all sensor data, then comprehensive coverage is achieved, but computational efficiency decreases
Solution Approach 1:
The patent segments sensor data into clustered operation phases before performing statistical analysis. This segmentation reduces the computational scope by analyzing data in smaller, condition-specific groups rather than processing all sensor data uniformly, thereby improving computational efficiency while maintaining detection reliability through targeted analysis.
Solution Approach 2:
The patent performs preliminary clustering of operation phases before conducting statistical analysis. This preliminary organization of data into meaningful groups reduces the computational burden of subsequent statistical operations while ensuring that analysis is performed on appropriately grouped data, maintaining both reliability and efficiency.
Data Source
AI summary
Techniques for anomaly detection are disclosed. These techniques include identifying a chain of operations for one or more systems, including a plurality of consecutive operations, and clustering operation phases in the chain of operations based on one or more operating conditions, using sensor data. The techniques further include computing statistical values for one or more parameters across the clustered operation phases, identifying one or more outlier parameters in the sensor data based on the computed statistical values, and excluding the one or more outlier parameters from the sensor data. The techniques further include computing one or more nominal values for one or more parameters of a first sub-system, of a plurality of sub-systems in the one or more systems, using the sensor data with the one or more outlier parameters excluded, and detecting an anomaly in the first sub-system based on the computed one or more nominal values.