Stacking ensemble learning assisted design of Al-Nb-Ti-V-Zr lightweight high-entropy alloys with high hardness

• Introduces a stacking ensembled ML model with optimized generalization performance for hardness prediction of HEAs. • The top three features that significantly affect the hardness are first ionization energy, metal radii and mixing enthalpy. • Trained a classification ML model which accurately forecasts alloy phase structures with Macro_F1 score = 0.95. • Experimental validation of two samples confirmed the model’s reliability with less than 8% error in hardness prediction. To improve the accuracy and efficiency of machine learning models in predicting and designing the mechanical properties and designing of lightweight high-entropy alloys, we have trained multi-classification machine learning models using stacking ensemble method. This ensembled model achieves high prediction accuracy of 0.9457 and good anti-overfitting performance. Two candidate high-entropy alloys with high hardness from the predicted results (Al 0.38 Ti 0.36 V 0.05 Zr 0.16 Nb 0.05 and Al 0.51 Ti 0.28 V 0.04 Zr 0.16 Nb 0.01 ) were selected to prepare bulk samples using arc melting method. The experimentally measured micro Vickers hardness of two samples were 723.7 HV and 691.0 HV respectively, and only slightly lower than the hardness values predicted by the model, with an error of less than 8 %. The phase structure of the samples, which is a mixture of HCP and FCC, also agrees well with the predicted results. This indicates that our machine learning approaches is highly effective in predicting the hardness of high-entropy alloys, with accuracy that has been experimentally verified, thereby significantly enhancing the efficiency of designing new lightweight high-hardness high-entropy alloys.