Prediction of Microstructural Representativity From A Single Image.

Area of Science:

• Materials Science
• Computational Materials Science
• Image Analysis

Background:

• Traditional methods for material representativity analysis require extensive datasets and statistical analysis.
• Estimating the Integral Range is crucial for determining microstructural property variance but is data-intensive.
• Limited microstructural data poses a challenge for accurate representativity assessment.

Purpose of the Study:

• To develop a novel method for predicting the representativity of phase fraction from a single 2D or 3D material image.
• To overcome the limitations of traditional methods by reducing data requirements.
• To provide a practical and accessible tool for material scientists and engineers.

Main Methods:

• Leveraging the Two-Point Correlation function to directly estimate variance from a single image.
• Enabling phase fraction prediction with associated confidence levels.
• Validation using open-source datasets across diverse microstructures.

Main Results:

• Successful prediction of phase fraction representativity from individual images.
• Significant reduction in data requirements for representativity analysis.
• Demonstrated efficacy across various microstructural types.

Conclusions:

• The developed method offers a practical solution for representativity analysis with limited data.
• The Two-Point Correlation function provides an efficient way to estimate variance.
• An accessible web-application (https://www.imagerep.io) is available for practical use.