Abnormality Detection Model Segmentation for Failure and Aging Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting abnormalities in plant systems, such as chemical or oil refinery plants, cannot distinguish between failures and aged deterioration, leading to unclear maintenance intervals and increased costs.
Innovation Solution
An information processing apparatus and method that generates a comparison model when an abnormality is detected, allowing for the determination of whether the abnormality is caused by a failure or aged deterioration by comparing the relationship among metrics indicated by both models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If invariant relation analysis is used to detect system abnormalities, then abnormality detection capability is improved, but the ability to distinguish between failure and aged deterioration deteriorates
Solution Approach 1:
The patent segments the abnormality analysis into two distinct parts: (1) detection of abnormality using invariant relation analysis, and (2) classification of abnormality cause by comparing metric distribution patterns. This segmentation allows the system to maintain high abnormality detection capability while adding cause identification without compromising the original detection function.
Solution Approach 2:
The patent introduces metric distribution patterns as an intermediary element between abnormality detection and cause identification. By analyzing the distribution characteristics of metrics (normal vs. abnormal states), the system can determine whether an abnormality is caused by failure or aged deterioration, thus recovering the lost cause identification information.
2Ease of operation
If maintenance intervals are determined by empirical rules, then ease of operation is improved, but manufacturing precision of maintenance timing deteriorates
Solution Approach 1:
The system enables self-service maintenance scheduling by automatically analyzing metric distributions and determining whether abnormalities indicate failure or aged deterioration. This eliminates the need for manual empirical judgment while providing precise, data-driven maintenance timing recommendations.
Solution Approach 2:
The patent implements feedback by continuously monitoring metric distributions and using the results to adjust maintenance timing. The system provides feedback on the actual system state (failure vs. aged deterioration) which informs optimal maintenance scheduling, thereby improving maintenance timing accuracy while maintaining operational simplicity.
Data Source
AI summary
An abnormality of a system is caused by a failure or aged deterioration is determined. A monitoring apparatus includes a model storage unit, a model generation unit, and a determination unit. The model storage unit stores a monitoring model that is a model for one point in time. The model indicates a relationship among a plurality of metrics of a system. The model generation unit generates a comparison model that is the model for a point in time at which the relationship among the plurality of metrics does not conform to the monitoring model. The determination unit determines that the relationship among metrics in the system has changed in a case where the comparison model indicates the relationship among metrics among which the monitoring model indicates the relationship, and outputs a result of the determination.