3D Print Layer Configuration Prediction for Target Texture

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current 3D printing technologies face challenges in predicting the layer configuration of building materials to achieve desired textures, often requiring a trial-and-error process that is time-consuming and labor-intensive, especially when multiple ink colors are used to create complex patterns.

Innovation Solution

A method involving machine learning to predict the layer configuration by producing specimens with varying layer configurations, measuring their textures, and using these data to train a computer to output the corresponding layer patterns for desired textures, reducing the need for repetitive testing and increasing efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If a trial-and-error process is used to determine layer configuration, then desired texture can be obtained, but the time and work required increase significantly

Engineering Contradiction:
Improvetexture qualityVSAvoidtime for determining layer configuration
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training a machine learning model with specimen data before actual 3D printing. Multiple specimens with different layer configurations are created and measured in advance to build a training dataset. The ML model learns the relationship between layer configurations and textures during this preliminary phase, so that when actual printing occurs, the optimal layer configuration can be predicted immediately without trial-and-error testing.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If multiple specimens with different layer configurations are produced and measured, then accurate texture data is acquired, but the workload for specimen production and measurement increases

Engineering Contradiction:
Improvetexture measurement accuracyVSAvoidwork efficiency in specimen production
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent uses copying by creating multiple specimens that replicate different layer configuration scenarios. These specimens serve as copies or representations of potential 3D print outcomes. By measuring these specimen copies instead of repeatedly printing and testing full-scale objects, the system achieves accurate texture data collection while significantly reducing the workload and time required compared to full production cycles.

Inventive Principle:
Principle #26Copying

3Manufacturing precision

If machine learning is used to predict layer configuration, then the prediction accuracy improves, but the complexity of the system increases

Engineering Contradiction:
Improvelayer configuration prediction accuracyVSAvoidsystem complexity for ML implementation
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary element - a machine learning model - that mediates between the input parameters and the desired texture output. Instead of directly computing the complex relationship between layer configurations and textures through traditional methods, the ML model serves as an intermediary that has been pre-trained to understand these relationships. This intermediary approach simplifies the actual prediction process while maintaining high accuracy, as the complex learning has already been performed during the training phase with specimen data.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3716603B1Layer configuration prediction method and layer configuration prediction apparatus
Publication Date: 2022.12.14 MIMAKI ENGINEERING CO LTD
  • EP3716603B1 patent drawingFigure 1(a)~1(c)
  • EP3716603B1 patent drawingFigure 2(a)~2(c)
  • EP3716603B1 patent drawingFigure 3(a)~3(c)

AI summary

A layer layering pattern corresponding to desired texture is appropriately predicted. A layer configuration prediction method of predicting a layering pattern of layers of a building material to express texture set to a CG image representing an object includes: a specimen production step (S102) of producing a plurality of specimens by depositing layers of the material in different configurations; a specimen measurement step (SI04) of performing, on each specimen, measurement to acquire a texture parameter corresponding to the texture; a learning step (S106) of causing a computer to perform machine learning of the relation between each of the specimens and the texture parameter; a setting parameter calculation step (S204) of calculating a setting parameter corresponding to the texture set to the CG image; and a layer configuration acquisition step (S206) of providing the setting parameter as an input to the computer having been caused to perform the machine learning, and acquiring an output representing the layering pattern of layers of the material corresponding to the setting parameter.