Polymer Blend Design Using Learned Models for Multi-Property Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing material design systems for polymer materials lack consideration of blending substances and order, making them unsuitable for designing polymers that simultaneously satisfy multiple desired physical properties, and are inefficient due to reliance on trial and error.
Innovation Solution
A material design apparatus and method that utilizes a learned model to correlate monomer blend proportions with polymer properties, allowing for the generation and selection of comprehensive analysis points within specified ranges to efficiently design polymers that meet multiple physical property requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If trial production based on material developer experience is performed, then design conditions can be adjusted, but repeated trials are required which consume considerable time and effort
Solution Approach 1:
The system performs preliminary computation and simulation before actual trial production. The material design support system calculates optimal blending proportions and polymerization conditions using learned models and global search algorithms, preparing design conditions in advance to avoid repeated trials during actual production
Solution Approach 2:
The patent replaces the mechanical trial-and-error process with an information processing system. A computer-based material design support system uses machine learning models and optimization algorithms to predict material properties and determine optimal conditions, substituting physical trial production with computational analysis
2Ease of operation
If local condition search is performed in the vicinity of previous design conditions, then adjustment is simple, but global search for optimal design condition is not achieved
Solution Approach 1:
The material design support system performs multiple functions: it can conduct both local searches around existing conditions and global searches across the entire parameter space. The system uses different search algorithms selectively based on the design stage and requirements, making it adaptable to various search needs
Solution Approach 2:
The search strategy is dynamic and adaptive. The system transitions from global search to local search as the design process progresses, and adjusts search parameters based on feedback from previous trials and learned models, making the search process flexible and responsive
3Extent of automation
If existing material design support systems are used, then inverse problem analysis can be performed, but they do not consider blending substances and order which are essential for polymer design
Solution Approach 1:
The system extends the parameter space of inverse problem analysis to include monomer blending proportions, number of monomers, and polymerization stage assignments. These additional parameters are specifically tailored for polymer material design, making the automated system adaptable to polymer synthesis requirements
Solution Approach 2:
The system handles composite polymer materials consisting of multiple monomer types blended in specific proportions. It considers the combination of different monomers and their sequential polymerization in multiple stages, enabling design of complex polymer compositions that go beyond simple materials
Data Source
AI summary
A material design apparatus includes a learned model that has learned a correspondence between input information about a blend proportion of a monomer and output information about physical property values of a polymer by machine learning. Each unit of the material design apparatus is configured to: receive as input a blend proportion range of at least one monomer; receive required ranges of physical property values of a polymer; generate a comprehensive analysis point of a polymer polymerized from multiple monomers, the multiple monomers including, within the blend proportion range, at least one monomer of which the blend proportion range is input; input the generated comprehensive analysis point into the learned model to calculate physical property values of a polymer, to create a data set, and to store the created data set; and select a polymer within the required ranges of the physical property values from the data set.