Control Signal Reliability for Facility Abnormality Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality detection systems for facilities like power generation plants face challenges in accurately determining abnormality due to imbalanced data distribution across different modes, leading to high determination costs and potential false positives.
Innovation Solution
An abnormality detection apparatus that includes a reliability calculation unit to assess the abnormality degree of control signals using a control signal model, and an abnormality diagnosis unit to provide reliability information alongside abnormality detection results, thereby improving the accuracy of abnormality determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If outlier detection is used for abnormality detection in facilities with imbalanced mode data, then abnormality can be detected, but false positives increase when training data for certain modes is extremely low
Solution Approach 1:
The patent segments the abnormality detection process into two independent parts: sensor signal analysis and control signal analysis. Each part has its own model (sensor signal model and control signal model) that processes different data sources separately. This segmentation allows the system to evaluate both sensor abnormalities and control abnormalities independently, then combine the results to reduce false positives caused by imbalanced training data in either domain.
2Measurement precision
If manual data analysis is performed to determine whether training data is sufficient, then determination accuracy improves, but determination cost increases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically evaluate training data sufficiency through the control signal model. The control signal model independently assesses whether sufficient training data exists for the current mode and automatically adjusts the determination process accordingly, eliminating the need for manual data analysis while maintaining high determination accuracy.
3Reliability
If control signal model is introduced to assess training data sufficiency, then false positives are reduced, but device complexity increases
Solution Approach 1:
The patent merges the control signal model with the existing sensor signal model into a unified abnormality detection system. Both models work together in an integrated framework where the control signal model's assessment of training data sufficiency directly influences the sensor signal model's abnormality determination. This merging allows the system to reduce false positives while maintaining manageable complexity through coordinated operation of the two models.
Data Source
AI summary
An abnormality detection apparatus (100) detects an abnormality of a facility. A model generation unit (110) extracts a feature amount of control signal time series data (143) as training data, and generates a control signal model (144) that outputs an abnormality degree for each control signal from the time series data of the control signal based on the training data. A reliability calculation unit (120) calculates the reliability of the control signal in a specific time range by inputting to the control signal model (144), the time series data of the control signal in the specific time range of an abnormality detection result detected in a sensor signal, as verification data. An abnormality diagnosis unit (130) outputs information in which the reliability of the control signal in the specific time range is added to the abnormality detection result in the specific time range, as an abnormality diagnosis result (146).