High-Throughput Screening of Al-Li Alloys via Automated Compositional Optimization and Microstructural Prediction Using Bayesian Neural Networks

**Abstract:** This paper presents a novel framework for accelerating the discovery and optimization of aluminum-lithium (Al-Li) alloys for aerospace applications. Utilizing a closed-loop Bayesian Neural Network (BNN) predictive model, we automate the process of compositional design and microstructural prediction, dramatically reducing the time and resources required for traditional trial-and-error experimentation. The BNN model rapidly predicts the performance characteristics of Al-Li alloys based on their elemental composition, incorporating established thermodynamic and kinetic principles. Coupled with a Bayesian optimization algorithm, the framework iteratively refines alloy compositions to meet targeted mechanical strength and density requirements, circumventing the limitations of traditional alloy development methodologies. This approach promises a ten-fold reduction in experimental iterations, accelerating the introduction of high-performance Al-Li alloys into the aerospace industry.