Fleet Device Anomaly Detection Using Peer Reference Databases
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting anomalies in devices within a fleet, such as motor vehicles, rely on empirical threshold comparisons that are not precise and do not effectively utilize data from similar devices for accurate anomaly detection.
Innovation Solution
A method involving the construction of a main database from collected data, extraction of reference and analysis databases, and statistical comparison of data from a device of interest with similar devices to identify deviations beyond predetermined thresholds, using decorrelated quantities and descriptive features like model, software version, and operating state.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data is compared to empirically set thresholds, then anomaly detection can be performed, but the detection precision is insufficient and does not account for behavior of similar devices
Solution Approach 1:
The system performs preliminary actions by constructing a main database from historical data of multiple devices and pre-processing it to extract reference databases for different device types. This preparation work is done in advance so that when anomaly detection is needed, the comparison can be made against pre-organized reference data rather than raw historical data, improving both precision and efficiency.
Solution Approach 2:
The system segments the overall database into multiple reference databases, each dedicated to a specific device type or category. This segmentation allows for more precise comparisons by matching devices only with their同类 counterparts, thereby improving detection precision while managing complexity through organized modular structures.
2Reliability
If data is compared with thresholds set based on professional experience, then anomaly detection is possible, but the method lacks objectivity and consistency
Solution Approach 1:
The system implements feedback by continuously comparing device data against reference databases constructed from historical performance data. The statistical analysis provides objective feedback on whether deviations are within normal ranges or represent anomalies, replacing subjective professional judgment with data-driven objective criteria that improve reliability and consistency.
Solution Approach 2:
The system replaces the mechanical/manual process of professional threshold setting with an automated statistical analysis system. Instead of experts manually determining thresholds based on experience, the system automatically computes statistical parameters from historical data, eliminating subjectivity and improving reliability through consistent automated processing.
3Measurement precision
If statistical analysis is performed comparing device data with reference database, then detection precision is improved, but data processing time and computational resources increase
Solution Approach 1:
The system performs preliminary processing of historical data to construct reference databases with pre-computed statistical parameters (means, standard deviations, etc.) for each device type. This advance preparation stores processed reference information that can be quickly accessed during anomaly detection, avoiding the need to perform complex statistical calculations on raw historical data each time an anomaly check is performed.
Solution Approach 2:
The system dynamically adapts the reference databases to different device types and operating conditions. By organizing reference data according to device categories and maintaining statistical parameters specific to each type, the system optimizes the matching process to compare devices only with relevant counterparts, reducing unnecessary computational overhead while maintaining high detection precision.
Data Source
AI summary
A method for detecting an anomaly in a device of interest belonging to a fleet of devices, includes: acquiring a new dataset, with each data item representing the value of a respective quantity of the device of interest; updating a main database, for adding the new dataset thereto; weighting the datasets of the main database; extracting, from the main database, a reference database consolidating datasets associated with devices of the fleet of devices, excluding datasets associated with the device of interest; extracting, from the main database, an analysis database consolidating datasets associated with the device of interest; comparing, for at least one quantity, corresponding data of the analysis and reference databases; and detecting an anomaly in the device of interest, when, for at least one of the quantities, the corresponding data of the analysis database deviates beyond a predetermined threshold relative to the corresponding data of the reference database.