Computational Intelligence in Stochastic Reconstruction of Porous Microstructures for Image-Based...

Area of Science:

• Materials Science
• Computational Modeling
• Engineering Mechanics

Background:

• Understanding microstructure-property relationships (MPRs) in random porous media is crucial.
• Image-based modeling and numerical simulations are key tools for MPR investigation.
• Stochastic nature of porous media requires large datasets of 3D digital microstructures.

Purpose of the Study:

• To review stochastic microstructure reconstruction methodologies.
• To emphasize computational intelligence-based approaches like generative adversarial networks.
• To discuss limitations of traditional methods and future trends.

Main Methods:

• Review of traditional algorithm-based reconstruction methods.
• Examination of emerging computational intelligence (CI)-based approaches.
• Focus on generative adversarial networks (GANs) and other AI techniques.

Main Results:

• Stochastic reconstruction efficiently generates virtual microstructures from limited data.
• CI-based methods enhance reconstruction fidelity, efficiency, and scalability.
• Enables robust statistical analysis of MPRs via methods like Monte Carlo simulations.

Conclusions:

• Significant progress in AI-driven microstructure reconstruction.
• Challenges remain in data scarcity, computational cost, and interpretability for complex systems.
• Physics-aware ML and hybrid AI offer future solutions for MPR research.