Machine Learning Material Structure Generation and Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Generating new material structures is a time-consuming and expensive process typically performed manually through experimentation and finite element analysis, making it difficult to quickly develop and test new materials.

Innovation Solution

A material generation system using machine learning models to generate and analyze new material structures, determining stress and strain values, and updating models based on finite element analysis results to improve accuracy and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual experimentation and finite element analysis are used to generate new material structures, then accuracy and reliability are improved, but time consumption and cost increase significantly

Engineering Contradiction:

Improveaccuracy of material structure generationVSAvoidtime to generate and test new materials

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent creates virtual copies of material structures through machine learning models that replicate the behavior and properties of real materials. These digital twins allow for rapid simulation and testing without physical experimentation, significantly reducing time while maintaining accuracy through iterative refinement against experimental data

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent performs preliminary screening and analysis of potential material structures using machine learning models before committing to full-scale experimentation. This preliminary action identifies promising candidates early in the design process, reducing the overall time required by focusing experimental resources on the most promising options

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If manual experimentation and finite element analysis are used to generate new material structures, then measurement precision is improved, but productivity decreases

Engineering Contradiction:

Improveprecision of stress and strain analysisVSAvoidnumber of material structures generated per unit time

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces traditional mechanical finite element analysis systems with machine learning-based computational models. These ML models achieve comparable measurement precision for stress and strain analysis but execute orders of magnitude faster, enabling high-throughput screening of material structures

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the analysis approach by changing from deterministic numerical methods to probabilistic machine learning models trained on experimental data. This parameter change allows the system to maintain measurement precision while dramatically increasing productivity through parallel processing and optimized inference

Inventive Principle:

Principle #35Parameter changes

3Reliability

If traditional manual methods are used for material structure generation, then model accuracy is improved through iterative refinement, but device complexity and operational difficulty increase

Engineering Contradiction:

Improveaccuracy of stress and strain determinationVSAvoidcomplexity of analysis system

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the material structure generation and stress-strain analysis functions into a single integrated machine learning model. This unified approach eliminates the need for separate finite element analysis software and manual modeling steps, reducing system complexity while maintaining accuracy through joint training on coupled datasets

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20240220683A1Generating and analyzing material structures based on material parameters and machine learning models

Publication Date: 2024.07.04 VOLKSWAGEN AG

AI summary

A method, apparatus and system are provided to generate and analyze material structures. A first machine learning model may generate material structures and a second machine learning model may determine stress values and strain values for the generated material structures. The material structures are generated based on material parameters.