To address the limited availability of systematic design studies for a concurrent improvement in strength and ductility of long-period stacking ordered (LPSO)-containing Mg-RE-Zn alloys at engineering dimensions, an interpretable machine learning design framework is established in this study. By integrating computer vision with literature and our previous work, micrographs are automatically processed to extract quantitative descriptors of blocky LPSO phases, including volume fraction, size, morphology and dispersion. These LPSO-related descriptors are combined with RE/Zn contents, grain size, texture intensity and fabrication route as input features to develop predictive models for yield strength, ultimate tensile strength and elongation. Shapley additive explanations (SHAP) analysis reveals that grain size and LPSO volume fraction are the dominant factors governing strength, while elongation is most sensitive to the total RE content. Excessive RE content or LPSO volume fraction deteriorates ductility, whereas a uniform dispersion of blocky LPSO phases is beneficial for achieving a favorable strength and ductility synergy. Based on the proposed interpretable optimization framework, a promising composition region was identified for achieving improved strength and ductility synergy in Mg-RE-Zn alloys. Guided by the model predictions and further considering prior processing experience and casting feasibility, an alloy with nominal composition Mg-9.09Gd-3.12Y-2.10Zn was fabricated for experimental validation. After hot extrusion and peak ageing, industrial-scale plates with good formability and no obvious defects are successfully produced, exhibiting a yield strength of 351 ± 1 MPa, an ultimate tensile strength of 435 ± 2 MPa and an elongation of 14.3 ± 0.4%. This work demonstrates an effective integration of quantitative LPSO microstructural characterization with data-driven alloy design, providing a promising framework for guiding the optimization of strength and overall mechanical performance in industrial-scale Mg-RE-Zn alloys.
Wang et al. (Fri,) studied this question.