Electrostatic Droplet Ejection Using ML for Pattern Uniformity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing electrostatic liquid droplet ejection methods are affected by patterns on the substrate, leading to non-uniform droplet sizes and inconsistent pattern formation.

Innovation Solution

A liquid droplet ejection method and device that utilize machine learning to generate droplet ejection conditions based on pattern recognition, incorporating factors like substrate conductivity, contact angle, and droplet shape, to stabilize droplet ejection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If electrostatic liquid droplet ejection is used, then finer liquid droplets can be ejected, but liquid droplet ejection conditions vary depending on pattern presence on substrate, resulting in non-uniform droplet sizes

Engineering Contradiction:
Improvedroplet size uniformityVSAvoidsensitivity to substrate pattern
Core Design Contradiction:
Manufacturing precisionVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary image capture of the substrate pattern before droplet ejection, and uses machine learning to predict and compensate for pattern-induced ejection variations in advance, ensuring uniform droplet sizes are achieved despite substrate pattern variations

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback by capturing images of the substrate pattern, feeding this information into the machine learning model, and adjusting ejection conditions based on the predicted pattern effects, thereby maintaining droplet uniformity across different substrate regions

Inventive Principle:
Principle #23Feedback

2Reliability

If machine learning-based condition adjustment is implemented, then stable droplet ejection can be achieved, but device complexity increases due to imaging unit and control unit requirements

Engineering Contradiction:
Improveejection stabilityVSAvoidsystem structure
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The machine learning model acts as an intermediary between the substrate pattern recognition and droplet ejection control, processing image data and translating it into adjusted ejection conditions, thereby achieving stable ejection with manageable system complexity

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system replaces complex mechanical adjustment mechanisms with an intelligent control system that uses machine learning algorithms to automatically determine optimal ejection conditions based on substrate pattern analysis, simplifying the overall system architecture

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Stable and uniform droplet ejection is achieved, unaffected by substrate patterns, enabling precise pattern formation and dynamic condition adjustment.

Implementation Method 1

an electrostatic ejection type ink jet head that can eject finer liquid droplets has attracted attention

Methodology Applied
Scientific EffectElectrostatic ejection: Electrostatics

Data Source

PatentEP4711049A1Droplet discharge device and droplet discharge method
Publication Date: 2026.03.18 SIJTECHNOLOGY INC
  • EP4711049A1 patent drawingFigure 1
  • EP4711049A1 patent drawingFigure 2
  • EP4711049A1 patent drawingFigure 3

AI summary

A liquid droplet ejection method is provided including capturing an image of a pattern on a substrate, acquiring first image data corresponding to the pattern, and applying the first image data to a machine learning model to perform machine learning and generate liquid droplet ejection conditions based on an electrostatic method. The liquid droplet ejection method described above may further include generating the machine learning model by performing machine learning on pre-acquired first image data.