A Closed-Loop Framework for Inverse Design: Dynamic Training and Intelligent Optimization for Heterostructured...

Area of Science:

• Materials Science
• Computational Materials Science
• Machine Learning

Background:

• Heterostructured materials exhibit desirable mechanical and physical properties, driving significant research interest.
• Designing these materials faces challenges due to the strength-toughness trade-off, requiring extensive experimental and simulation efforts.
• Current design approaches are time-consuming and costly, limiting the exploration of the vast material design space.

Purpose of the Study:

• To develop a comprehensive scientific machine learning framework for efficient and accelerated design of heterostructured materials.
• To integrate a deep learning model for accurate forward prediction and an optimization algorithm for inverse design.
• To demonstrate the framework's effectiveness using metal matrix composites as a representative material system.

Main Methods:

• Implemented a Back-Propagation Neural Network with Continual Learning (BPNN-CL) for predicting material properties.
• Utilized a Non-dominated Sorting Genetic Algorithm II with Partition Monitoring and Chaotic Perturbation (NSGA-II-PMCP) for inverse design and optimization.
• Validated the framework's performance on metal matrix composites, comparing prediction accuracy and optimization efficiency against conventional methods.

Main Results:

• The BPNN-CL model achieved prediction errors within 10% for most experimental data.
• BPNN-CL reduced the Mean Absolute Percentage Error for elastic modulus prediction by over 14.9% compared to six conventional machine learning methods.
• The NSGA-II-PMCP algorithm demonstrated enhanced generalization and high efficiency in inverse design, outperforming existing optimization techniques.

Conclusions:

• The proposed scientific machine learning framework provides a unified platform for heterostructured material design.
• This approach significantly accelerates the discovery and development of advanced materials.
• Integration with manufacturing techniques like 3D printing offers a scalable pathway for future material innovation.