Machine Learning Device for Print Defect Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Industrial printing using electrophotographic image formation apparatuses faces challenges in adequately adjusting control information, leading to high rates of print defects due to the complexity of adjusting parameters like secondary transfer voltage and toner fixing temperature, which are critical for print quality, and requires manual intervention by specialized engineers, resulting in increased costs and time.

Innovation Solution

A machine learning device and method that generate learned models based on datasets including feature value information, medium information, and control information to optimize secondary transfer voltage and toner fixing temperature for different printing speeds, enabling automated adjustment of control information to improve print quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual adjustment of control parameters is performed by specialized engineers, then print quality can be improved, but time consumption and costs increase

Engineering Contradiction:

Improveprint qualityVSAvoidtime consumption

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system automatically adjusts control parameters by acquiring state variables during printing, comparing them with teaching data, and determining optimal parameters without requiring specialized engineers. The image formation apparatus performs self-diagnosis and self-adjustment, eliminating manual intervention while maintaining print quality.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical adjustment process with an automated information processing system. Control parameters are determined through automated comparison of state variables with teaching data stored in memory, substituting human engineer intervention with computational analysis and automatic parameter determination.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Manufacturing precision

If control parameters are adjusted for different print media, then print quality improves, but device complexity increases

Engineering Contradiction:

Improveprint qualityVSAvoidparameter adjustment complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system uses a universal approach by storing multiple sets of teaching data corresponding to different print media types in a single memory unit. The same image formation apparatus can handle various media (ordinary paper, thick paper, transparent sheets, etc.) by automatically selecting the appropriate teaching data set, eliminating the need for separate adjustment mechanisms for each media type.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system manages parameter complexity by changing the state of stored teaching data according to the detected print media type. Instead of physically reconfiguring the apparatus, the control unit changes which set of control parameters is active by comparing state variables with the appropriate teaching data, simplifying the adjustment process while maintaining precision.

Inventive Principle:

Principle #35Parameter changes

3Manufacturing precision

If multiple control parameters are adjusted simultaneously, then print quality improves, but difficulty of adjustment increases

Engineering Contradiction:

Improveprint qualityVSAvoidparameter adjustment difficulty

Core Design Contradiction:

Manufacturing precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent segments the adjustment process by first determining the print media type, then selecting the corresponding teaching data set, and finally comparing state variables with that specific set. This stepwise segmentation breaks down the complex simultaneous adjustment of multiple parameters into manageable sequential steps, reducing adjustment difficulty while maintaining quality.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system introduces teaching data as an intermediary between the state variables and the control parameters. Instead of directly adjusting multiple parameters simultaneously, the control unit uses teaching data as a reference mediator to determine optimal parameters, simplifying the adjustment process by providing a clear comparison target for each state variable.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11381690B2Machine learning device including a state variable acquisition unit, a teaching data acquisition unit and a learned model generation unit that generates a first learned model by performing machine learning based on a first state variable dataset and first teaching data and generates a second learned model by performing machine learning based on a second state variable dataset and second teaching data

Publication Date: 2022.07.05 OKI ELECTRIC INDUSTRY CO LTD

AI summary

A machine learning device according to an embodiment may include: a state variable acquisition unit that acquires first and second state variable datasets including print results performed at first and second printing speeds respectively by an image formation apparatus; a teaching data acquisition unit that acquires first and second teaching data corresponding to the first and second printing speeds; and a learned model generation unit that generates a first learned model by performing machine learning based on the first state variable dataset and the first teaching data, and generates a second learned model by performing machine learning based on the second state variable dataset and the second teaching data. Each of the first and second state variable datasets includes: feature value information; medium information; and first control information. Each of the first and second teaching data includes: second control information; and a print defect value.