Sensor Anomaly Classification With Cause Sensor Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods in time-series data from sensors face challenges due to inappropriate labeling of data with the same label containing different sensor behaviors, making it difficult to accurately classify anomalies and identify their causes.
Innovation Solution
An analysis apparatus and method that classifies anomalies based on time-series sensing data using a learned model and identifies the sensor responsible for the anomaly by comparing feature values with past data, allowing for the update of anomaly models and more accurate labeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning with manual labeling is performed, then anomaly classification can be achieved, but labeling accuracy deteriorates due to human error and subjective judgment
Solution Approach 1:
The system enables sensors to automatically label their own anomaly data by comparing current sensor behaviors against historically learned normal behaviors. The sensor that detects the anomaly also identifies itself as the source, eliminating the need for manual human labeling and the associated errors.
Solution Approach 2:
The system uses feedback from past sensing data to automatically update and refine anomaly classification. By continuously comparing current sensor readings with historical data and updating the learned model, the system improves labeling accuracy over time without manual intervention.
2Quantity of substance
If data with the same label contains different sensor behaviors, then data volume increases, but classification reliability deteriorates due to inappropriate labeling
Solution Approach 1:
The system applies local quality by identifying which specific sensor detected the anomaly and using that sensor's individual characteristics and historical behavior patterns for classification. Instead of treating all sensors uniformly, the classification is tailored to the specific sensor's local behavior patterns, improving reliability.
3Ease of operation
If manual labeling is performed by humans, then anomaly detection can be implemented, but time consumption increases and productivity decreases
Solution Approach 1:
The system eliminates manual labeling operations by enabling sensors to automatically identify and label their own anomaly detections. The sensor that detects an anomaly automatically becomes the source of the labeled data, completely removing the time-consuming manual labeling process and significantly improving productivity.
Data Source
AI summary
An analysis apparatus includes: an anomaly classifying unit that classifies the type of an anomaly having occurred based on time-series sensing data in occurrence of anomaly received from a plurality of sensors and a learned model learned in advance; and an identifying unit that identifies a sensor having detected information corresponding to the cause of the anomaly classified by the anomaly classifying unit based on information corresponding to the sensing data and information corresponding to past data that is past sensing data stored in advance.