Abnormality Propagation Estimation Using Normal Operation Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality detection and diagnostic systems fail to accurately identify causal abnormality factors and require extensive data collection for unknown abnormality factors, limiting their effectiveness in quickly restoring systems to a normal state.
Innovation Solution
An estimation apparatus comprising a normal index estimation unit and an abnormality propagation information estimation unit that estimate indices indicating normal operation and abnormality propagation, allowing for the identification of causal abnormality factors without relying on abnormal data, thereby reducing the need for extensive data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If abnormal data corresponding to known abnormality factors is collected for specification, then known abnormality factors can be specified, but time and effort are needed for data collection and unknown abnormality factors cannot be specified
Solution Approach 1:
The system performs preliminary learning using only normal operation data to build a baseline model of system behavior. This preliminary action enables the system to detect and specify both known and unknown abnormality factors without requiring pre-collection of abnormal data, thereby eliminating time-consuming data collection processes while maintaining specification accuracy
Solution Approach 2:
The system creates a virtual copy or model of normal system operation through machine learning. This copied normal state serves as a reference framework that can compare against actual operations to identify deviations indicating abnormality factors, eliminating the need to physically collect and store actual abnormal data samples
2Ease of operation
If abnormality factor specification uses correlation between data pieces, then the process is simplified, but causal abnormality factors that do not directly contribute to abnormality determination cannot be specified
Solution Approach 1:
The system replaces simple correlation-based mechanical analysis with machine learning-based causal inference. The learning unit analyzes operational data to identify causal relationships between variables and abnormality factors, substituting the limited correlation method with a more sophisticated approach that maintains operational simplicity while significantly improving causal factor detection accuracy
Solution Approach 2:
The system changes the analytical parameters from simple correlation coefficients to complex causal relationship metrics derived through machine learning. By transforming the analysis parameters to include temporal sequences, variable interactions, and causal pathways, the system achieves both operational simplicity through automated analysis and high accuracy in identifying causal abnormality factors
Data Source
AI summary
An estimation apparatus 1 includes: a normal index estimation unit 2 configured to estimate, using a second variable output by a second component 21 that influences a first variable output by a first component 21, an index A indicating that the first variable is achieved at a normal time; and an abnormality propagation information estimation unit 3 configured to estimate abnormality propagation information expressing an index indicating that an abnormality propagates to a third variable output by a third component 21 influenced by the first component 21, by changing the first variable.