Resin Property Estimation Using Key Additive Variables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately estimate the elongation or tensile strength of resin compositions used in coating materials for electric wires, particularly when incorporating multiple additives like flame retardants and antioxidants, leading to insufficient estimation accuracy.
Innovation Solution
A physical property estimation device and method utilizing a regression model built through machine learning to correlate compound amounts of base polymer, flame retardant, and flame retardant auxiliary with initial elongation or tensile strength, excluding less influential additives like antioxidants and lubricants to improve estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If compound amounts of all materials including antioxidants, copper inhibitors, and lubricants are used as explanatory variables, then the model considers more factors, but the estimation accuracy of initial elongation and initial tensile strength does not improve satisfactorily
Solution Approach 1:
The patent extracts and removes unnecessary explanatory variables (antioxidants, copper inhibitors, and lubricants) from the machine learning model. By identifying that these materials do not significantly influence initial elongation and initial tensile strength, the invention eliminates them from the compound amount data, thereby simplifying the model while improving estimation accuracy for these specific physical properties.
Solution Approach 2:
The patent segments the explanatory variables into different groups based on their influence on specific physical properties. It separates materials into those that significantly affect initial elongation and initial tensile strength (base polymer, flame retardant, flame retardant auxiliary) versus those that do not (antioxidants, copper inhibitors, lubricants), allowing selective use of variables for different estimation purposes.
2Measurement precision
If compound amounts of base polymer, flame retardant, and flame retardant auxiliary are used as explanatory variables, then the estimation accuracy of initial elongation and initial tensile strength improves, but the model cannot estimate other physical properties influenced by antioxidants and copper inhibitors
Solution Approach 1:
The patent implements a dynamic variable selection mechanism where the set of explanatory variables changes based on the target physical property to be estimated. When estimating initial elongation or initial tensile strength, it uses a simplified variable set (base polymer, flame retardant, flame retardant auxiliary). When estimating other physical properties like heat resistance or oxidation resistance, it dynamically includes additional variables (antioxidants, copper inhibitors) as needed.
Solution Approach 2:
The patent changes the parameters (explanatory variables) of the machine learning model according to the specific estimation task. It adjusts the compound amount data inputs based on which physical property is being estimated, optimizing the model's performance for each specific property while maintaining overall versatility through conditional variable inclusion.
3Productivity
If machine learning is applied to estimate physical properties of resin compositions, then the research and development period is shortened and cost is reduced, but the estimation accuracy of physical properties is not satisfactorily obtained
Solution Approach 1:
The patent performs preliminary analysis to identify which materials significantly influence specific physical properties before building the machine learning model. By pre-determining the relevant explanatory variables through material property analysis and experimental data review, it prepares an optimized dataset that enables accurate estimation while maintaining fast processing speeds for research and development applications.
Data Source
AI summary
A physical property estimation device for a resin composition, which estimates a physical property of a resin composition that is manufactured by using a base polymer, a flame retardant, a flame retardant auxiliary, and other materials. The physical property estimation device is provided with a regression model building processing unit that learns by machine learning a relationship between explanatory variable data and physical property data including physical property data as an estimation target, and builds a regression model that shows a correlation between the explanatory variable data and the physical property data, and a physical property estimation processing unit that estimates the physical property of an estimation target by using the regression model. The explanatory variable data includes compound amount data of the base polymer, the flame retardant, and the flame retardant auxiliary respectively, when a physical property to be estimated by the physical property estimation processing unit is either initial elongation or initial tensile strength of the estimation target.