One-Class Classifier Anomaly Evaluation for Machine Condition Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
One-class classifiers in machine condition monitoring systems provide limited information, making it difficult to assess the severity of anomalies detected, as they only indicate whether a measurement is within or outside the normal operating range, without providing insights into how different the anomaly is from the normal distribution.
Innovation Solution
The method evaluates anomalies by projecting them onto the closest boundary of the decision region and calculating the distance from the anomaly to this boundary, using techniques like Euclidean distance and k-mean algorithms to determine the extent of deviation from the normal operating range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a one-class classifier is used to detect anomalies, then anomaly detection capability is provided, but the information provided is too limited for sophisticated fault diagnostics
Solution Approach 1:
The patent transforms the one-class classifier output from a single binary decision (anomaly/normal) into multiple dimensions by calculating distances to different boundaries of the decision region. This dimensional expansion provides richer information about anomaly characteristics, enabling sophisticated fault diagnostics while maintaining the simplicity of the one-class classifier approach.
2Ease of operation
If only binary decision output is used from one-class classifier, then classification simplicity is maintained, but the ability to evaluate anomaly severity is lost
Solution Approach 1:
The patent segments the anomaly evaluation process by calculating distances to multiple different boundaries of the decision region. Each boundary represents a different aspect of anomaly severity, allowing the system to maintain simple binary classification while providing detailed severity evaluation through multiple distance measurements.
3Measurement precision
If distance calculation to decision region boundary is added, then anomaly evaluation capability is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-establishing the decision region boundaries during the training phase. This allows the system to maintain simple binary classification while providing detailed severity evaluation through multiple distance measurements, without adding significant computational complexity during real-time operation.
Data Source
AI summary
A method for monitoring machine conditions provides additional information using a one-class classifier in which an evaluation function is learned. In the method, a distance is determined from an anomaly measurement x to a boundary of a region R1 containing all acceptable measurements. The distance is used as a measure of the extent of the anomaly. The distance is found by searching along a line from the anomaly to a closest acceptable measurement within the region R1.