Machine Learning for Mesh-to-CAD Structural Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing structural optimization techniques produce discrete representations of mechanical parts as mesh structures that are difficult to edit and cannot be directly input into manufacturing systems like machining, molding, or 3D printing, or 3D printing systems, or 3D scanning systems, as they lack a parameterized representation suitable for these systems.

Innovation Solution

A machine-learning method that converts discrete mesh structures from structural optimization into parameterized 3D CAD models using a function trained on a dataset of 2D polyline profiles associated with primitive parametric curve classes, enabling automatic conversion and reducing overfitting issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If structural optimization produces discrete mesh structures, then optimal mechanical properties are achieved, but the results cannot be directly input into manufacturing systems and are difficult to edit

Engineering Contradiction:
Improveoptimal mechanical propertiesVSAvoidconvertibility to manufacturing systems
Core Design Contradiction:
ReliabilityVSEase of manufacture

Solution Approach 1:

The patent introduces an intermediary conversion process that transforms discrete mesh structures into parameterized 3D CAD models. This intermediary representation serves as a bridge between structural optimization results and manufacturing systems, enabling both to communicate effectively while preserving the optimal mechanical properties from the optimization stage.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The conversion process changes the parameterization state of the model from discrete mesh parameters to continuous parametric CAD parameters. This parameter transformation enables the model to be edited and manipulated in a manner compatible with manufacturing systems while maintaining the geometric fidelity of the optimized structure.

Inventive Principle:
Principle #35Parameter changes

2Ease of manufacture

If existing conversion methods are used to transform mesh structures to 3D CAD models, then conversion is achieved, but the process is tedious and requires many manual interventions

Engineering Contradiction:
Improveconvertibility to 3D CAD modelsVSAvoidmanual intervention time
Core Design Contradiction:
Ease of manufactureVSLoss of time

Solution Approach 1:

The conversion system performs self-service by automatically transforming mesh structures into 3D CAD models without requiring manual engineering interventions. The algorithm autonomously handles the complex conversion process, including feature recognition, parameter extraction, and model reconstruction, thereby eliminating tedious manual operations.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces manual mechanical conversion processes with an automated computational algorithm. Instead of engineers manually manipulating models through CAD tools, the system uses automated image processing and geometric reconstruction algorithms to perform the conversion, significantly reducing time and human effort.

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

3Ease of manufacture

If existing conversion methods are used, then 3D CAD models are obtained, but the models comprise too many parameters leading to overfitting issues

Engineering Contradiction:
Improve3D CAD model generationVSAvoidnumber of parameters
Core Design Contradiction:
Ease of manufactureVSDevice complexity

Solution Approach 1:

The conversion algorithm extracts only the essential geometric parameters needed to define the part geometry, excluding redundant or unnecessary parameters from the final 3D CAD model. This selective extraction reduces the parameter count while maintaining the model's ability to represent the optimized structure accurately.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of allowing the conversion process to generate maximum detail with many parameters, the system inverts the approach by starting with a simplified parametric representation and selectively adding only those features necessary to capture the essential geometry of the optimized structure, thereby avoiding overfitting.

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentEP4629124A1Machine-learning in structural optimization
Publication Date: 2025.10.08 DASSAULT SYSTEMES SA
  • EP4629124A1 patent drawingFigure 1~2
  • EP4629124A1 patent drawingFigure 3
  • EP4629124A1 patent drawingFigure 4

AI summary

The disclosure notably relates to a computer-implemented method for machine-learning a function. The method comprises obtaining a dataset including 2D polyline profiles each representing respectively a portion of a manufacturing contour in a result of a structural optimization that represents a mechanical part, each 2D polyline profile being associated in the dataset with a respective primitive parametric curve class among a predetermined set of primitive parametric curve classes. T he method further comprises training the function based on the dataset. The function is configured to take an input 2D polyline profile and to provide an output primitive parametric curve class. Such a method forms an improved solution for processing a result of a structural optimization that represents a mechanical part.