Equipment State Determination Using Multi-Period Classification Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing equipment monitoring systems face challenges in accurately determining the state of equipment as normal or abnormal, particularly when dealing with changing tendencies in measurement values or noise, as they rely on single classification models learned from multiple periods, which can lead to incorrect classifications.
Innovation Solution
The system employs multiple classification models learned from different periods, using a logical product of their results to determine the equipment's state, and dynamically generates new classification models based on recent data to adapt to changes, while selecting models that prioritize recent learning data for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single classification model learned from multiple periods is used, then the system is simple to operate, but the classification accuracy deteriorates when measurement values have changing tendencies or noise
Solution Approach 1:
The patent divides the classification system into multiple independent classification models, each trained on data from a specific period. Instead of using a single model for all periods, the system segments the modeling task temporally, creating specialized models for different time ranges. This allows each model to capture period-specific characteristics while maintaining overall system accuracy.
Solution Approach 2:
The patent implements a dynamic model selection mechanism where the system automatically selects or weights classification models based on the recency of their training periods. More recent models are given higher weights or selected preferentially, allowing the system to adapt to changing measurement tendencies over time without requiring manual reconfiguration.
2Measurement precision
If multiple classification models learned from different periods are used, then the classification accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent combines multiple classification models into a unified determination system that aggregates their outputs. Rather than maintaining completely separate systems, the models are merged through a determination unit that synthesizes their classification results, achieving improved accuracy while managing complexity through integration.
Solution Approach 2:
The patent introduces weighting parameters that reflect the recency of each model's training period. By dynamically adjusting these parameters, the system optimizes the contribution of each model based on current relevance, allowing accurate classification without requiring an equal number of models for every possible period.
3Adaptability or versatility
If classification models are trained on historical data including abnormal periods, then the system adapts to long-term changes, but the reliability of abnormality detection deteriorates due to noise and changing tendencies
Solution Approach 1:
The patent performs preliminary classification using individual period-specific models before making final abnormality determinations. Each model independently classifies data from its training period, allowing the system to establish baseline expectations for normal variation in each period before aggregating results to detect true abnormalities.
Solution Approach 2:
The determination unit provides feedback by aggregating classification results from multiple models and using this collective information to improve overall detection reliability. The system learns from the patterns across multiple period-specific models, allowing it to distinguish between period-specific variations and true abnormalities more effectively.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is an apparatus including: an acquisition unit which acquires measurement data indicating a state of a target; a supply unit which supplies the measurement data acquired by the acquisition unit to a plurality of classification models respectively learned by learning data, which includes measurement data in a period in which the state of the target is normal, for periods different from each other, the plurality of classification models classifying measurement data as either normal or abnormal in response to the measurement data being input; and a determination unit which determines the state of the target as either normal or abnormal based on a plurality of classification results output from the plurality of classification models. Selected drawing: Fig. 1