Epoxy Resin Composition Search Using Inverse Property Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods struggle to efficiently search for a thermosetting epoxy resin composition with a well-balanced set of characteristics for applications like CFRP, as these characteristics are often contradictory and difficult to achieve simultaneously.

Innovation Solution

A method involving an information processor that calculates feature values from actual data related to thermosetting epoxy resin compositions, trains predictive models using these values, and performs inverse analysis to find compositions with desired physical properties, utilizing molecular fingerprints and descriptors like the distance between oxygen atoms in the resin's glycidyl group and ether moiety.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If limited specific thermosetting epoxy resin compositions are used as raw materials for composite materials, then manufacturing reliability is maintained, but search efficiency for novel compositions with required characteristics deteriorates

Engineering Contradiction:
Improvemanufacturing reliabilityVSAvoidsearch efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent creates a virtual copy of the thermosetting epoxy resin composition by calculating molecular feature values from structural formulas and using machine learning models to predict physical properties. This virtual modeling allows rapid evaluation of novel compositions without physical experimentation, improving search efficiency while maintaining reliability through validated predictive models

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent performs preliminary calculation of molecular feature values and training of predictive models before actual composition search. By pre-establishing the relationship between molecular structure and physical properties through machine learning, the system can quickly evaluate novel compositions without time-consuming experimental trials

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If multiple contradictory characteristics are pursued simultaneously in thermosetting epoxy resin compositions, then material performance balance is improved, but search complexity increases

Engineering Contradiction:
Improvematerial performance balanceVSAvoidsearch complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent transforms the complex multi-characteristic search problem into a parameter optimization problem by calculating molecular feature values (such as oxygen atom distances, molecular weights, functional group counts) and using machine learning models to predict multiple physical properties simultaneously. This allows systematic adjustment of molecular parameters to achieve balanced performance across multiple contradictory characteristics

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent develops a universal machine learning framework that can simultaneously predict multiple physical properties (strength, heat resistance, flexibility, etc.) from a single set of molecular feature values. This multi-functional approach simplifies the search process by evaluating all desired characteristics through one integrated model rather than separate searches

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4425501B1Search method for thermosetting epoxy resin composition, information processing device, and program
Publication Date: 2025.11.12 DIC CORP
  • EP4425501B1 patent drawingFigure 1
  • EP4425501B1 patent drawingFigure 2
  • EP4425501B1 patent drawingFigure 3

AI summary

To improve a technique of searching for a thermosetting epoxy resin composition. A method of searching for a thermosetting epoxy resin composition executed by an information processor, the method including the steps of: training a plurality of predictive models each corresponding to a target variable using actual data related to a thermosetting epoxy resin composition; and searching for a thermosetting epoxy resin composition with a desired balance of physical properties by inverse analysis using the plurality of predictive models, wherein the actual data include a polymer composition, a structural formula, an epoxy equivalent, and a blending condition related to a thermosetting epoxy resin, and the target variable includes a bending modulus, a bending strain, a bending strength, and a glass transition temperature.