Novel machine learning driven design strategy for high strength Zn Alloys optimization with multiple constraints

Abstract

Zinc (Zn) alloys offer advantages such as abundant resources and low cost. Nevertheless, their current mechanical properties limit application in more advanced fields. Due to the lack of clear compositional design methods, the development of high-performance Zn alloys is urgently needed. To this end, this work proposes a fast and effective design strategy for Zn alloys based on machine learning (ML). The prediction models for the ultimate tensile strength, elongation, and hardness were successfully developed, with accuracies exceeding 90%. Interpretability analysis of the models was performed using the SHAP method with particle swarm optimization (PSO). Furthermore, a ML-based Zn alloy composition design system (ZACDS) was proposed by integrating the Bayesian optimization algorithm. A novel high-strength Zn alloy was successfully designed using ZACDS, demonstrating good agreement between predicted and experimental mechanical properties. This approach offers a new strategy for Zn alloy design under different compositional constraints and performance requirements.