Aging Diagnosis Learning Update for Low-Maintenance Equipment Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional anomaly diagnosis methods for electric equipment facilities are environmentally dependent, requiring complex threshold parameter adjustments and frequent professional maintenance, leading to high man-hour requirements for aging diagnosis.
Innovation Solution
An aging diagnosis system that updates a determiner using initial learning data from accelerated aging tests and operational measurement data, with teacher aging degree labels, to select and learn additional data, enhancing the learning effect and accuracy of aging determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional anomaly diagnosis methods use temperature, sound and vibration measurements with predetermined thresholds, then anomaly detection can be performed, but environmental dependence increases and requires enormous amount of complex threshold parameters per diagnosis target facility
Solution Approach 1:
The patent replaces the conventional mechanical threshold-based diagnosis system with a machine learning-based determiner system. Instead of using predetermined thresholds for temperature, sound and vibration measurements, the system uses a determiner created through accelerated aging tests and additional learning to automatically determine aging degrees, thereby eliminating the need for complex threshold parameter adjustments per facility
Solution Approach 2:
The patent changes the parameter representation from fixed threshold values to dynamic aging degree labels generated by the determiner. The system transforms physical measurement parameters into learned feature representations that capture aging patterns, allowing the determiner to adapt to different facilities through additional learning without requiring manual threshold reconfiguration
2Measurement precision
If predetermined threshold parameters are continuously fine adjusted to weaken environmental dependence, then diagnosis accuracy improves, but regular maintenance by professional becomes necessary and great amount of man hours is required
Solution Approach 1:
The patent implements self-service by enabling the determiner to automatically adapt to different facilities through additional learning using their own operational data. The system performs self-adjustment without requiring professional intervention, eliminating the need for regular maintenance and manual parameter tuning while maintaining high diagnosis accuracy
Solution Approach 2:
The patent performs preliminary action by conducting accelerated aging tests beforehand to create the initial determiner. This pre-training phase captures aging patterns that would otherwise require extensive manual adjustment during operation, allowing the system to start with high accuracy and maintain it through automatic additional learning rather than professional maintenance
3Measurement precision
If determiner created by accelerated aging test performs additional learning with all learning data, then learning effect increases, but unnecessary learning data increases processing burden
Solution Approach 1:
The patent applies partial action by selectively including only necessary learning data in the additional learning process. The system determines data inclusion based on whether the data contributes to improving aging determination accuracy, excluding redundant or unnecessary data points. This selective approach maintains high learning effectiveness while reducing the processing burden associated with handling excessive learning data
Data Source
AI summary
A determiner which learns acceleration measurement data which has been obtained by an accelerated aging test and indicates that a facility changes from a normal state to an aged state, and advance label data which is obtained by giving a label to data indicating characteristics of aging in the acceleration measurement data. Measurement data of aging diagnosis is obtained from the facility which is operating, teacher aging degree label data is found from a record of maintenance of the facility, and additional data is obtained from the measurement data and the teacher aging degree label data. When a difference between predicted aging degree label data and teacher aging degree label data is greater than a predetermined value, learning data is selected as additional learning data. The additional learning data is learned to update the determiner.