Candidate Formulation Evaluation for Extrapolation Risk

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional design support systems fail to evaluate the degree of extrapolation of candidate designs for target substances, leading to significant errors between predicted and experimental values, thereby undermining the reliability and motivation for experimentation.

Innovation Solution

A design evaluation device and method that determine a reference design, assess the degree of extrapolation of candidate designs using a classification based on formulation differences, and output the results, incorporating user-defined criteria to quantify the newness and challenge of candidate designs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If machine learning is used to predict material formulations, then development efficiency is improved, but prediction accuracy deteriorates in extrapolation regions

Engineering Contradiction:

Improvedevelopment efficiencyVSAvoidprediction accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system performs preliminary evaluation of the degree of extrapolation for candidate designs before experimentation. By calculating the degree of extrapolation in advance and comparing it against threshold values, the system identifies high-risk candidate designs that may yield inaccurate predictions. This preliminary assessment allows users to adjust predictions or select alternative candidates before committing to expensive experiments, thereby maintaining high development efficiency while improving prediction accuracy.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If candidate designs are selected from extrapolation regions, then design novelty is improved, but reliability of prediction deteriorates

Engineering Contradiction:

Improvedesign noveltyVSAvoidprediction reliability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system provides feedback on the degree of extrapolation for each candidate design by calculating it against existing experimental data and comparing with threshold values. This feedback mechanism allows users to understand the reliability level of each candidate design's prediction. Users can then adjust their selection criteria or modify candidate designs based on this feedback, enabling them to pursue novel designs while maintaining acceptable prediction reliability through informed decision-making.

Inventive Principle:

Principle #23Feedback

Data Source

PatentEP4641577A1Design evaluation device, design evaluation method, and program

Publication Date: 2025.10.29 RESONAC CORP

Figure 1
Figure 2
Figure 3

AI summary

A design evaluation device evaluates a degree of extrapolation of a candidate design of a target substance. The design evaluation device includes a reference design determination unit configured to determine a reference setting indicating a formulation of a target substance to be produced using two or more material substances, a candidate design determination unit configured to determine a candidate design indicating the formulation that is different from the reference design, a class determination unit configured to determine a class indicating a degree of extrapolation of the candidate design, based on a classification indicating a difference in the formulation between the candidate design and the reference design, and a result output unit configured to output the candidate design and the class in association with each other.