Material Design Assistance for Predicting New Raw Material Properties
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing material design assistance devices are unable to predict properties of materials containing raw materials not included in their training datasets, limiting their design assistance capabilities.
Innovation Solution
A design assistance device that includes a raw material information storage, a registration reception unit, and a feature generation unit to manage and generate features for new raw materials, using a machine learning model to predict properties of materials with new raw materials, while preventing duplicate registrations based on molecular structure, use, or name matches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is trained on a fixed training dataset, then prediction accuracy for materials in the training dataset is improved, but the ability to predict properties of materials containing new raw materials not in the training dataset deteriorates
Solution Approach 1:
The system segments the raw material information into distinct components (molecular structure, physical properties, chemical properties) and processes each separately. The molecular structure is encoded into features that can be independently evaluated, allowing the model to generalize to new raw materials by combining known structural patterns with new inputs, rather than requiring entire material formulations to be in the training dataset.
Solution Approach 2:
The system transforms raw material molecular structures into numerical feature representations through chemical structure encoding. This parameter transformation allows the machine learning model to work with standardized numerical inputs rather than raw chemical data, enabling it to generalize predictions to new raw materials by recognizing patterns in the encoded structural features even when specific materials weren't in the training set.
2Adaptability or versatility
If the system allows free registration of new raw materials, then adaptability to new materials is improved, but data quality and reliability deteriorate due to duplicate or invalid registrations
Solution Approach 1:
The system implements feedback mechanisms where registration information is validated against existing databases and training datasets before acceptance. Duplicate detection algorithms compare new raw material registrations against stored molecular structures and properties, providing feedback to prevent redundant or erroneous data entry. This feedback loop maintains data quality while allowing legitimate new materials to be registered.
Solution Approach 2:
The system performs preliminary validation and verification of raw material registration information before it is added to the database. Molecular structures are checked for validity, duplicates are detected in advance, and consistency with existing data is verified. This preliminary action ensures that only high-quality, non-duplicate data is registered, maintaining reliability while enabling adaptability.
Data Source
AI summary
A design assistance device is configured to assist design of a material in which a plurality of raw materials are to be formulated, by using a machine learning model that has learned a correspondence relationship between design condition information of a material, in which a plurality of raw materials are to be formulated, and property information of the material. The design assistance device includes a memory; and a processor connected to the memory and configured to store information of names and attributes of a plurality of registered raw materials, receive an input of information of a name and an attribute of a raw material to be newly registered, and store the received information, and generate a feature of the raw material inputtable to the machine learning model, from information that is selected as the design condition information and is information of attributes of a plurality of raw materials stored.