MATAI: A Generalist Machine Learning Framework for Property Prediction and Inverse Design of Advanced Alloys

2025-11-13

AI Summary

Advanced alloy design faces challenges arising from high-dimensional compositional spaces, conflicting multi-objective performance requirements, and stringent manufacturability constraints. Method: We propose a physics-informed multitask learning framework integrated with a bilevel optimization strategy—combining local search and symbolic constraint programming—to jointly enable accurate property prediction and constraint-aware inverse design. The approach unifies a curated alloy database, an end-to-end deep neural network (directly mapping composition to density, strength, and ductility), a manufacturability-constrained optimization engine, and an AI–experiment closed-loop feedback mechanism. Results: Applied to titanium-based systems, the method discovered a novel alloy—exhibiting low density (<4.45 g/cm³), high strength (>1000 MPa), and superior ductility (>5%)—within only seven iterative cycles; experimental validation confirmed its overall performance surpasses that of TC4. This work establishes the first inverse alloy design paradigm that is physics-guided, constraint-embedded, and experimentally closed-loop driven.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

The discovery of advanced metallic alloys is hindered by vast composition spaces, competing property objectives, and real-world constraints on manufacturability. Here we introduce MATAI, a generalist machine learning framework for property prediction and inverse design of as-cast alloys. MATAI integrates a curated alloy database, deep neural network-based property predictors, a constraint-aware optimization engine, and an iterative AI-experiment feedback loop. The framework estimates key mechanical propertie, sincluding density, yield strength, ultimate tensile strength, and elongation, directly from composition, using multi-task learning and physics-informed inductive biases. Alloy design is framed as a constrained optimization problem and solved using a bi-level approach that combines local search with symbolic constraint programming. We demonstrate MATAI's capabilities on the Ti-based alloy system, a canonical class of lightweight structural materials, where it rapidly identifies candidates that simultaneously achieve lower density (<4.45 g/cm3), higher strength (>1000 MPa) and appreciable ductility (>5%) through only seven iterations. Experimental validation confirms that MATAI-designed alloys outperform commercial references such as TC4, highlighting the framework's potential to accelerate the discovery of lightweight, high-performance materials under real-world design constraints.

Problem

• Predicts mechanical properties of alloys from composition using machine learning
• Solves constrained optimization for alloy design under manufacturability limitations
• Accelerates discovery of lightweight high-strength alloys through AI-experiment feedback

Innovation

• Integrates alloy database with deep neural networks
• Uses multi-task learning for mechanical property prediction
• Combines local search with symbolic constraint programming