Metric Model Combination for Plant System Abnormality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In plant systems, the timing at which metric relations change varies across different pairs and even within the same state, leading to errors in correlation models used for monitoring, making it difficult to accurately detect system abnormalities.
Innovation Solution
An information processing apparatus that stores models for each metric set and determines system states by comparing combinations of models conforming to specific states, allowing for accurate monitoring even when relation changes occur differently across metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a correlation model is generated for each period associated with a state of the system, then the system can detect abnormalities by comparing metric values to the correlation model, but the correlation model becomes inaccurate when the timing of relation changes differs across metric pairs
Solution Approach 1:
The patent segments the correlation model into multiple metric-specific correlation models, each tailored to the specific timing characteristics of its metric pair. Instead of using a single unified correlation model for all metrics, the system generates individual correlation models for each metric pair, allowing each to accurately reflect its unique temporal relationship patterns without being constrained by other metrics' timing variations.
2Device complexity
If a single correlation model is used for all metric pairs, then the system structure remains simple, but the model cannot accurately represent different timing characteristics of various metric pairs
Solution Approach 1:
The patent applies local quality by making each correlation model specific to its metric pair's local characteristics. Each metric pair receives a customized correlation model that reflects its unique temporal relationship patterns, rather than forcing all metrics to conform to a single global model structure. This localized approach ensures that each metric pair's specific timing characteristics are accurately captured.
Data Source
AI summary
Monitoring of a state of a system can be accurately performed even when a timing at which a relation changes is different for each set of the metrics. A monitoring apparatus 100 includes a model storage unit 122 and a determination unit 115. The model storage unit 122 stores, for each of a plurality of metric sets in a system, a model representing a relation among metrics included in the corresponding metric set. The determination unit 115 determines and outputs whether the system is in one state, by comparing a combination of models to which the plurality of metric sets conform respectively when the system is in the one state and a combination of models to which the plurality of metric sets conform respectively when the system is in a state to be determined.