Material Property Prediction Using Segmented Training Datasets

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for predicting material properties of structural materials are limited by the use of generic training datasets and algorithms that fail to account for material processing variations and specific target properties, leading to unreliable predictions.

Innovation Solution

Classifying structural materials into specialized classes based on chemical composition and processing parameters, generating targeted training datasets and machine learning algorithms that focus on specific material properties, and extracting relevant properties for accurate predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a single generic training dataset and machine learning algorithm are used for all structural materials, then the system can cover a wide range of materials, but the prediction accuracy and reliability for specific materials deteriorates

Engineering Contradiction:
Improvecoverage range of materialsVSAvoidprediction accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the generic training dataset into multiple specialized training datasets based on material classes (e.g., metals, polymers, ceramics) and processing types (e.g., heat treatment, cold working). Each specialized dataset is tailored to specific material properties and processing conditions, enabling the machine learning algorithm to achieve higher prediction accuracy for each material category while maintaining broad adaptability through the modular dataset structure.

Inventive Principle:
Principle #1Segmentation

2Device complexity

If material processing parameters are not considered in the training dataset, then the dataset structure remains simple, but the prediction reliability for processed materials deteriorates

Engineering Contradiction:
Improvedataset structure complexityVSAvoidprediction reliability
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent applies preliminary action by pre-processing and categorizing material processing parameters (such as heat treatment temperatures, cooling rates, deformation amounts) before creating the training datasets. Processing parameters are standardized and integrated into the dataset structure in advance, allowing the machine learning algorithm to directly utilize these pre-organized parameters for accurate predictions without requiring complex real-time processing during prediction.

Inventive Principle:
Principle #10Preliminary action

3Ease of manufacture

If a single machine learning algorithm is trained for all target properties, then the training process is simplified, but the prediction accuracy for specific target properties deteriorates

Engineering Contradiction:
Improvetraining process simplicityVSAvoidprediction accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent applies local quality by creating specialized machine learning models for specific target properties (e.g., yield strength, tensile strength, elongation) within each material class. Each model is trained on specialized training datasets that contain only the relevant features and target properties for that specific application. This targeted approach enables each model to achieve high prediction accuracy for its specific target property while maintaining ease of implementation through a modular, property-specific model structure.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4300504A1System and method for processing material properties of structural materials
Publication Date: 2024.01.03 TOTAL MATERIA AG
  • EP4300504A1 patent drawingFigure 1
  • EP4300504A1 patent drawingFigure 2~3a
  • EP4300504A1 patent drawingFigure 3b

AI summary

A computer implemented method for processing material properties of structural materials, the method comprising: generating training dataset(s) for training a machine learning algorithm for determining values(s) of specific target material property(s) of an assessment material; training a machine learning algorithm for determining value(s) of specific target material property(s) of an assessment material using the training dataset; and determining value(s) of specific target material property(s) of an assessment material using the trained machine learning model.