Maintenance Prediction Model for Winding Body Defect Ratios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current maintenance systems in production facilities often fail to detect abnormalities before they occur, leading to unnecessary downtime and operational disruptions, as they typically notify facility personnel only after an abnormality has been detected, rather than proactively identifying potential issues.
Innovation Solution
A learned model generation method that uses sensor data from a production apparatus to identify defects in components such as supply reels, bonding rollers, and winding cores by analyzing the intersection of end surface positions of electrode sheets, and generates a model to predict maintenance needs based on defect ratios before and after maintenance, allowing for proactive maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If maintenance systems notify facility personnel only after an abnormality is detected, then the system structure remains simple, but production downtime increases and operational efficiency deteriorates
Solution Approach 1:
The system performs preliminary actions by continuously monitoring sensor data and using machine learning models to predict potential abnormalities before they occur. The defect detection model analyzes patterns in sensor readings from supply reels, bonding rollers, and winding cores to identify early signs of defects, enabling maintenance to be scheduled proactively rather than reactively, thus preventing production downtime while maintaining manageable system complexity
Solution Approach 2:
The system implements feedback mechanisms by continuously collecting sensor data from the production apparatus, comparing it against learned patterns from historical data, and providing real-time predictions about potential defects. This closed-loop feedback enables the system to adapt to changing conditions and improve prediction accuracy over time, achieving high productivity through proactive maintenance without requiring overly complex manual intervention systems
2Measurement precision
If continuous monitoring of sensor data is implemented to detect abnormalities early, then detection accuracy improves, but data processing requirements and computational load increase
Solution Approach 1:
The system extracts only the most relevant features from sensor data for analysis, rather than processing all raw data equally. The machine learning model focuses on specific patterns in sensor readings that are most indicative of defects in supply reels, bonding rollers, and winding cores, filtering out redundant information to reduce computational energy consumption while maintaining high defect detection accuracy
Solution Approach 2:
The system transforms raw sensor data into meaningful parameters and features that are more efficient for processing. By converting continuous sensor readings into defect probability scores and trend indicators, the system achieves accurate defect detection while reducing the computational burden of analyzing raw high-frequency sensor data, thereby lowering energy consumption
3Reliability
If a learned model is generated using all available sensor data including pre-maintenance data, then model comprehensiveness improves, but model accuracy deteriorates when maintenance does not significantly reduce defects
Solution Approach 1:
The system applies partial action by selectively using only the most valuable data subsets for model generation. When maintenance does not significantly reduce defects, the system focuses on learning from post-maintenance data patterns that are most indicative of actual defect conditions, rather than forcing integration of all available data. This selective approach maintains model accuracy while preserving the flexibility to adapt data usage strategies based on maintenance effectiveness
Solution Approach 2:
The system applies different quality standards to different data subsets based on their relevance and reliability. Pre-maintenance data is used with different weighting and validation criteria compared to post-maintenance data, allowing the model to learn from high-quality predictive patterns while filtering out noisy or less relevant information. This local quality approach ensures model accuracy is maintained while preserving adaptability in how different data sources are utilized
Data Source
AI summary
A shape data group that is shape data of each of a predetermined number or more of winding bodies correlated with time information is acquired, a first defect ratio before maintenance is calculated on the basis of each shape data group of the predetermined number of winding bodies read before the maintenance, a second defect ratio after the maintenance is calculated on the basis of each shape data group of the predetermined number of winding bodies read after the maintenance, and a facility state diagnosis model is generated by using each shape data group of the predetermined number of winding bodies read before the maintenance in a case where a difference between the first defect ratio and the second defect ratio is greater than or equal to a predetermined value.