Learned Defect Models for Winding Core Fault Attribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing maintenance systems in production facilities often require maintenance to be performed after an abnormality occurs, leading to facility downtime. There is a need for a system that can detect signs of abnormalities before they cause failures, allowing for proactive maintenance.
Innovation Solution
A learned model generation method and apparatus that utilize sensors to acquire data on the positional relationships of electrode sheets wound on winding cores. This data is used to generate learned models that identify defects in the winding process and attribute them to specific components, such as winding cores, allowing for targeted maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If maintenance is performed after an abnormality occurs, then the facility can be restored to normal operation, but facility downtime increases and production efficiency decreases
Solution Approach 1:
The system performs preliminary actions by continuously monitoring winding body parameters and generating learned models that predict potential defects before they occur. The model analyzes historical data and current sensor readings to identify trends indicating future failures, enabling maintenance to be scheduled proactively rather than reactively, thus preventing facility downtime while maintaining high production efficiency
Solution Approach 2:
The system implements feedback mechanisms by continuously collecting sensor data from the winding process, comparing it against learned models, and adjusting maintenance predictions in real-time. The model receives feedback from actual defect occurrences and refines its predictions, creating a closed-loop system that improves reliability while optimizing production scheduling to minimize downtime
2Reliability
If general maintenance procedures are performed, then facility reliability improves, but maintenance time and resource allocation increase
Solution Approach 1:
The system applies local quality by identifying specific components or parameters that exhibit abnormal characteristics through the learned model analysis. Instead of performing blanket maintenance on the entire winding apparatus, the system pinpoints exact locations or components requiring attention based on localized defect patterns detected in sensor data, thereby reducing overall maintenance time while maintaining reliability
Solution Approach 2:
The maintenance approach is segmented by dividing the winding apparatus into monitorable components and analyzing each independently through the learned model. The system segments maintenance tasks based on predicted defect locations and severity, allowing prioritized and targeted maintenance actions rather than comprehensive shutdowns, thus reducing maintenance time while preserving reliability
3Measurement precision
If comprehensive monitoring of all winding parameters is implemented, then defect detection accuracy improves, but system complexity and measurement requirements increase
Solution Approach 1:
The system extracts only the most critical parameters from the full set of available sensor data that are most strongly correlated with winding body defects, as identified by the learned model analysis. By selecting and monitoring only these key parameters rather than all possible measurements, the system achieves high defect detection accuracy while minimizing sensor requirements and system complexity
Solution Approach 2:
The system applies partial action by implementing monitoring at selective critical points in the winding process rather than continuous comprehensive monitoring of all parameters. The learned model identifies which parameter measurements provide the most value for defect prediction, enabling the system to achieve high detection accuracy with a subset of measurements, thereby reducing complexity while maintaining precision
Data Source
AI summary
A facility state diagnosis model generator generates a shape data group in which any of a plurality of winding cores is correlated with any of a plurality of pieces of group data, and generates a replacement data group in which correspondence relationships between the winding cores and the group data are replaced with each other in all combinations of the plurality of winding cores and the plurality of pieces of group data, and generates or updates a plurality of learned models indicating that a cause of a defect is any of the plurality of winding cores by using the replacement data group.