Data-driven multi-objective design of reliable and cost-effective Co-Cr-Fe-Ni high-entropy alloys

To address the cost-reliability trade-off in non-equiatomic Co-Cr-Fe-Ni high-entropy alloys (HEAs), this study establishes a multi-objective design framework integrating atomistic simulations and uncertainty-aware machine learning. Validation confirms that Molecular Dynamics (MD) results align well with fine-grained and laser directed energy deposition (L-DED) manufactured alloys, establishing their process applicability. Heteroscedastic Gaussian Process Regression (HGPR) analysis reveals significant compositional robustness of yield strength. Furthermore, Pareto optimization identifies an optimal Fe-rich composition ( ) that achieves a 25% cost reduction while maintaining a high strength of 0.535 GPa and excellent process stability. This work demonstrates the industrial efficacy of shifting from strength maximization to a “reliability-cost” dual-optimization paradigm. • Systematic MD simulations fill the data gap for non-equiatomic Co-Cr-Fe-Ni HEAs. • MD yield strengths align well with fine-grained and laser directed energy deposition (L-DED) experimental data. • Heteroscedastic Gaussian process regression (HGPR) reveals significant compositional robustness of strength. • Reliability-based design identifies a stable process window for alloy manufacturing. • Pareto optimization screens out low-cost Fe-rich alloys with high strength.