Abnormality Pattern Detection Using Multi-Outlier Similarity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality detection methods using outlier detection techniques can only perform binary judgments of normal or abnormality and struggle to identify specific abnormality patterns, especially when failure data is scarce, leading to difficulties in selecting the most suitable detection method and accurately specifying the type of abnormality.
Innovation Solution
An abnormality detection device that calculates deviation scores using multiple outlier detection methods and calculates similarity degrees between deviation sensitivity information and inclination information to detect abnormalities, enabling the identification of specific abnormality patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If outlier detection techniques are used for abnormality detection, then abnormality can be detected when only normal data exists, but only binary information of normal or abnormality can be obtained without specifying abnormality patterns
Solution Approach 1:
The patent segments the abnormality detection process into multiple independent outlier detection methods, each targeting specific abnormality patterns. Instead of using a single general outlier detection method, the system divides detection into multiple specialized methods (e.g., statistical methods, machine learning methods, deep learning methods) that can identify different types of abnormalities, thereby preserving abnormality pattern information while maintaining detection capability.
Solution Approach 2:
The patent creates a universal abnormality detection system that integrates multiple outlier detection methods with different functionalities. Each detection method serves multiple purposes: detecting abnormalities, identifying patterns, and providing diagnostic information. This multi-functional approach allows the system to overcome the limitation of binary classification by incorporating pattern recognition capabilities across different detection methodologies.
2Ease of operation
If a single outlier detection method is applied, then the detection process is simple, but abnormality patterns that do not follow the algorithm cannot be detected
Solution Approach 1:
The patent merges multiple outlier detection methods into a unified detection system that processes data through various algorithms simultaneously. By combining statistical methods, traditional machine learning methods, and deep learning methods, the system achieves comprehensive abnormality pattern coverage while maintaining operational simplicity through integrated processing and unified output interpretation.
Solution Approach 2:
The patent creates a composite detection system that integrates different detection methodologies with complementary strengths. Each detection method acts as a component with specific detection capabilities, and their combination forms a robust system that can detect diverse abnormality patterns. This composite approach ensures that no single algorithm's limitations constrain the overall system's versatility.
3Measurement precision
If multiple outlier detection methods are used to detect various abnormality patterns, then detection accuracy improves, but it becomes difficult to select the most suitable method
Solution Approach 1:
The patent implements feedback mechanisms that automatically evaluate the performance of different outlier detection methods based on detected abnormality patterns and data characteristics. The system provides feedback on which methods are most effective for specific patterns, enabling automatic method selection and reducing complexity. This feedback-driven approach allows the system to adaptively choose suitable methods without requiring manual intervention.
Solution Approach 2:
The detection system performs self-service by automatically selecting and applying the most suitable outlier detection methods based on the characteristics of the input data and observed abnormality patterns. The system autonomously determines which detection methods to employ without external guidance, reducing selection complexity while maintaining high detection accuracy through intelligent method matching.
Data Source
AI summary
A deviation inclination calculation unit (22) calculates a deviation score by using evaluation data obtained from a subject apparatus as an input, in each of a plurality of outlier detection methods specifying data deviated from other data from among subject data, and calculating deviation scores indicating deviation degrees of the data specified, and calculates deviation inclination information from the deviation scores calculated. An abnormality detection unit (23) calculates, for each abnormality pattern, a similarity degree between deviation sensitivity information indicating sensitivity for each of a plurality of abnormality patterns with respect to each of the plurality of outlier detection methods, and the deviation inclination information calculated, and detects an abnormality of the subject apparatus.