Plant Learning Data Weighting for Reliable Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing failure prediction models in large plants, such as power generation or chemistry, require high-quality learning data to accurately predict failures, but unskilled personnel may generate erroneous learning data during preprocessing, reducing prediction accuracy and performance.
Innovation Solution
A system for generating learning data that filters plant data based on warning conditions, differentiates weights applied to new and existing data, and combines them to create new learning data, ensuring high reliability and accuracy, even when an unskilled user selects data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If unskilled personnel perform data preprocessing for learning data generation, then the ease of operation is improved, but the manufacturing precision deteriorates due to erroneous data generation
Solution Approach 1:
The system performs automatic filtering of plant data based on warning conditions without requiring manual intervention. The learning data generation part automatically combines new data and existing learning data with appropriate weights, enabling the system to self-correct errors that would normally require skilled operator intervention.
Solution Approach 2:
The data incorporating part acts as an intermediary between raw plant data and the learning data generating part. It automatically filters data based on warning conditions and prepares cleaned data for combination, serving as a mediator that prevents erroneous data from reaching the learning model.
2Quantity of substance
If all plant data is used for learning data generation, then the quantity of substance is improved, but the reliability deteriorates due to inclusion of low-quality data
Solution Approach 1:
The system applies different quality standards to different portions of the data. The data incorporating part identifies and filters out low-quality data segments based on warning conditions, while preserving high-quality data. This creates locally differentiated quality within the overall dataset.
Solution Approach 2:
The learning data generating part changes the weight parameters assigned to different data sources based on their quality. It dynamically adjusts the combination ratio between new data and existing learning data, giving higher weight to reliable data and lower weight to potentially erroneous data.
3Adaptability or versatility
If new data and existing learning data are combined with equal weights, then the adaptability is improved, but the reliability deteriorates due to inappropriate weight allocation
Solution Approach 1:
The system dynamically adjusts the weights assigned to new data and existing learning data based on the specific characteristics and quality of the input data. Rather than using fixed equal weights, the learning data generating part adaptively determines the optimal combination ratio to maintain reliability.
Solution Approach 2:
The system changes the weight parameters based on data quality assessment. The learning data generating part modifies the combination ratio between new and existing data according to their respective reliability, ensuring that high-quality data has greater influence on the learned model.
Data Source
AI summary
A system for generating learning data is provided. The system for generating the learning data includes a data incorporating part configured to generate new data by filtering plant data based on a warning condition to incorporate the plant data for one configuration of a plant into existing learning data and a learning data generating part configured to differentiate a weight applied to the new data and the existing learning data, respectively, by comparing the number of the new data and the number of the existing learning data, and generate new learning data by combining the new data with the existing learning data to which the weight is applied.