Inkjet Nozzle Side-Skipping Detection Using Multi-Nozzle Image Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting nozzle ejection failures in inkjet printing apparatuses require significant parameter setting efforts and fail to accurately detect side skipping due to neglecting the relationship between nozzle ejection positions.
Innovation Solution
A method using machine learning to generate a trained model from learning images, allowing for the detection of ejection failures, particularly side skipping, by analyzing the relative positional relationships among nozzles without requiring extensive parameter setting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional numerical analysis methods are used for detecting ejection failures, then detection can be performed based on imaging data, but significant parameter setting efforts are required and side skipping cannot be accurately detected
Solution Approach 1:
The patent replaces conventional numerical analysis methods with machine learning-based image processing. A trained model automatically analyzes imaging data to detect ejection failures, eliminating the need for manual parameter setting while accurately detecting side skipping and other nozzle abnormalities.
Solution Approach 2:
The system employs a self-learning mechanism where the trained model automatically adapts to detect various types of ejection failures without requiring manual intervention for parameter adjustment. The model independently processes imaging data and identifies anomalies based on learned patterns.
2Measurement precision
If conventional detection methods focus on individual nozzle evaluation, then simple ejection failures can be detected, but the relationship between nozzle ejection positions is neglected causing side skipping to be undetected
Solution Approach 1:
The patent merges the evaluation of multiple nozzles by analyzing their relative positional relationships simultaneously. The trained model processes imaging data to detect not only individual nozzle failures but also side skipping by considering the spatial relationships between adjacent nozzles' ejection positions.
Solution Approach 2:
The system transitions from evaluating nozzles in isolation to analyzing them in a multi-dimensional spatial context. By incorporating positional relationship information as an additional dimension of analysis, the model can detect side skipping that occurs when ink droplets are ejected at incorrect positions relative to adjacent nozzles.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Regarding an inkjet printing apparatus, it is possible to detect side skipping, which is one of types of ejection failures of a nozzle, without requiring a large amount of time and effort for setting a parameter. A trained model is generated by performing machine learning by using a plurality of learning images each formed by ejecting ink from N nozzles, the N being an integer of 2 or more (S20). A plurality of partial images each corresponding to the N nozzles are cut out from a captured image of a print image of a test pattern (S50). The partial image is given, as input data, to the trained model (S60), and a mean square error between the input data and output data is calculated (S70). Based on a result of comparing the mean square error with a threshold, it is determined whether or not an ejection failure nozzle is included in the N nozzles corresponding to the input data (S80).