Anomaly Detection via Sensor Contribution Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current system analyzing techniques struggle to accurately detect anomaly factors when changes in sensor values are small and appear in multiple sensors, leading to difficulties in identifying affected data items and reducing anomaly detection sensitivity.
Innovation Solution
A system analyzing device that collects sensor data, stores a correlation model, and acquires standard contributions to identify the ratio of contribution of each sensor data item to a predicted value, enabling accurate extraction of anomaly factors even when changes are small and widespread.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional anomaly detection methods are used, then the system can detect anomalies, but the sensitivity is reduced when changes in sensor values are small and appear in multiple sensors
Solution Approach 1:
The patent segments the anomaly detection process by evaluating model breakdown for each explanatory variable separately rather than treating all variables uniformly. This segmentation allows the system to identify which specific sensor data items contribute to anomaly detection, preventing information loss when changes are small and widespread across multiple sensors.
2Quantity of substance
If the system monitors all sensor data items, then comprehensive coverage is achieved, but the ability to identify specific anomaly factors is reduced due to information loss
Solution Approach 1:
The patent introduces feedback by calculating contribution amounts that indicate how much each explanatory variable contributes to model breakdown. This feedback mechanism allows the system to trace back from overall anomaly detection to specific sensor data items, preventing information loss even when monitoring a large number of sensor data items.
3Device complexity
If the system uses a single correlation model for all sensor data, then the analysis is simplified, but the accuracy of identifying specific anomaly factors decreases
Solution Approach 1:
The patent applies local quality by evaluating model breakdown contributions locally for each explanatory variable rather than applying a uniform analysis to all variables. This localized evaluation maintains relatively simple system architecture while achieving high precision in identifying specific anomaly factors through contribution amount calculations.
Data Source
AI summary
A system analyzing device according to the present invention includes: a collection unit that collects a plurality of pieces of sensor data of a monitored system; a storage unit that stores a correlation modes based on at least one of a plurality of pieces of sensor data; and a standard contribution acquisition unit that acquires, for a predicted value of an objective variable of a regression equation thereof, a standard contribution indicating a ratio of contribution of each of the data included as explanatory variables.