Self-Organizing Map for Image Device State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for estimating the state of image forming devices, such as printers, are limited in providing comprehensive state information for self-repair functions and require improvement in selecting appropriate machine learning methods for accurate state estimation.
Innovation Solution
An image formation system utilizing a self-organizing map generated by a learning processor, which classifies the state of the device using both chronological learning data sets to update the map based on Euclidean distances and hit counts, enabling efficient state estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single state quantity (lifetime counter value) is used to estimate photoconductor lifetime, then the device complexity is reduced and ease of operation is improved, but the measurement precision and reliability of state estimation are insufficient for comprehensive self-repair functions
Solution Approach 1:
The patent segments the state estimation problem into multiple independent state quantities (lifetime counter value, rotation speed, temperature, etc.), each measuring a specific aspect of photoconductor condition. This segmentation allows comprehensive state assessment without requiring a single complex measurement system, thereby improving measurement precision while managing device complexity through modular measurement approaches.
Solution Approach 2:
The patent creates a universal state estimation system that processes multiple different types of state quantities (electrical, mechanical, thermal parameters) through a unified machine learning model (self-organizing map). This multi-functional approach enables comprehensive photoconductor state assessment using diverse measurement data, improving estimation accuracy while maintaining system manageability through a single integrated evaluation framework.
2Reliability
If machine learning with multiple state quantities is used to estimate photoconductor state, then the measurement precision and reliability are improved, but the device complexity and difficulty of detecting and measuring increase due to insufficient consideration of appropriate AI selection and application methods
Solution Approach 1:
The patent transforms the complex machine learning implementation problem into manageable parameter changes by systematically varying key parameters: selecting specific state quantities to measure, choosing appropriate measurement intervals, adjusting the self-organizing map configuration (number of nodes, learning rate), and optimizing the batch processing parameters. This parameter-based approach makes AI implementation more controllable and less difficult while maintaining high reliability.
Solution Approach 2:
The patent implements feedback mechanisms where the state estimation results are continuously monitored and used to adjust the measurement and processing parameters. The system feeds back the classification outcomes and estimation reliability metrics to optimize future measurements and model updates, reducing the difficulty of AI implementation through iterative refinement based on actual performance data.
3Measurement precision
If batch processing with chronological learning data sets is used to update the self-organizing map, then the adaptability and measurement precision are improved, but the processing time and loss of time increase
Solution Approach 1:
The patent implements periodic batch processing where the self-organizing map is updated at predetermined intervals using chronological learning data sets rather than continuously. This periodic update mechanism maintains high classification accuracy by regularly incorporating new data while avoiding the excessive processing time of continuous updates, thus reducing time loss while preserving measurement precision.
Solution Approach 2:
The patent applies partial action by selectively processing only the necessary portion of chronological data sets for model updates. Instead of reprocessing all historical data, the system uses incremental batch processing with recently acquired data sets, achieving sufficient classification accuracy improvement without the excessive processing time required for complete data retraining, thereby optimizing the time-precision tradeoff.
Data Source
AI summary
An image formation system includes a state measuring device that measures a plurality of state quantities of an image forming device, a learning processor that generates a self-organizing map, and a state estimator that estimates a state of the image forming device. The learning processor uses a first learning data set acquired chronologically previously and a second learning data set acquired chronologically subsequently, performs acquisition of first hit counts and generation of a first self-organizing map, and performs acquisition of second hit counts and update of the first self-organizing map to generate a second self-organizing map. In the update, an amount of the update is adjusted based on a product of a Euclidean distance between the first self-organizing map and a self-organizing map to be updated and the first hit count and a product of a Euclidean distance between the second learning data set and the self-organizing map to be updated and the second hit count.