Inverse design of high-performance Mg-Gd based magnesium alloys by machine learning method

An inverse design framework is developed by employing machine learning methods and a multi-objective co-optimization strategy, which achieves the intelligent design of chemical composition and processing parameters of thermo-mechanical treatments based on the desired mechanical properties of Mg alloys. Based on the database collected for extruded Mg-Gd and Mg-Y based alloys, the inverse design framework is established by integrating the optimized forward model with the non-dominated sorting genetic algorithm. The optimized forward model is constructed by evaluating the performance of different machine learning algorithms, in which the Random Forest algorithm is experimentally validated to accurately describe the relationship between chemical composition and mechanical properties. In addition, the non-dominated sorting genetic algorithm is implemented to achieve the simultaneous optimization of different mechanical properties. Based on the validation from a series of experimental measurements, the established inverse design framework is adopted to develop advanced Mg alloys. With the different desired mechanical properties as inputs, the chemical composition and processing parameters of solid solution and extrusion are efficiently designed for a high-strength Mg-11.5Gd-6.0Y-1.0Zn-0.2Mn (wt.%) alloy and a high-plasticity Mg-2.5Gd-1.0Zn (wt.%) alloy, which exhibits the tensile-strength/elongation of 417 MPa/3.2 % and 223 MPa/34%, respectively. The present advances provide a transparent route for the inverse design of advanced Mg alloys based on the desired mechanical properties.