Screened Cryogenic Alloys using Machine Learning

The attached files include cryogenic alloys screened using machine learning pipeline. 1. CryoAlloy_CandidatePool_AllTemps_Scored.csv (4,456 rows × 67 columns) Complete scored candidate pool containing 1,114 unique alloy compositions evaluated at four cryogenic temperatures (4, 20, 77, 120 K). Each row includes composition (37 elements, at.%), thermodynamic descriptors, predicted mechanical properties (YS, UTS, elongation), conformal uncertainty half-widths, risk-adjusted scores, physical plausibility flags, and weighted final scores.2. CryoAlloy_Shortlist_ALL_T_uncertainty_gated.csv (200 rows × 67 columns) Top 50 uncertainty-gated model-prioritised candidates at each of the four target temperatures (4, 20, 77, 120 K), ranked by weighted final score. Candidates pass physical plausibility filters (UTS ≥ YS, elongation ≥ 5%) and are scored using risk-adjusted predictions (pred − conformal half-width).3. CryoAlloy_Pareto3D_4K.csv (55 rows × 67 columns) Three-objective Pareto-optimal candidate set at 4 K, containing alloys that are non-dominated in risk-adjusted yield strength, UTS, and elongation simultaneously. These represent the fundamental strength–ductility trade-off surface for the superconducting temperature regime.4. CryoAlloy_Pareto3D_20K.csv (54 rows × 67 columns) Three-objective Pareto-optimal candidate set at 20 K (liquid hydrogen regime). Non-dominated alloys in risk-adjusted yield strength, UTS, and elongation, representing the trade-off surface for liquid-hydrogen applications.5. CryoAlloy_Pareto3D_77K.csv (66 rows × 67 columns) Three-objective Pareto-optimal candidate set at 77 K (liquid nitrogen regime). The largest Pareto front across the four temperatures, reflecting greater compositional diversity among non-dominated candidates at this well-characterised temperature.6. CryoAlloy_Pareto3D_120K.csv (57 rows × 67 columns) Three-objective Pareto-optimal candidate set at 120 K (LNG regime). Non-dominated alloys in risk-adjusted yield strength, UTS, and elongation for liquefied natural gas storage and transport applications.7. CryoAlloy_Pareto2D_4K.csv (5 rows × 67 columns) Two-objective Pareto front at 4 K, containing candidates non-dominated in risk-adjusted yield strength versus UTS. A smaller subset than the 3D front due to the stricter dominance criterion without the elongation axis.8. CryoAlloy_Pareto2D_20K.csv (5 rows × 67 columns) Two-objective Pareto front at 20 K (risk-adjusted yield strength vs UTS). Identifies the strength-optimised candidates at liquid-hydrogen temperature without the elongation trade-off dimension.9. CryoAlloy_Pareto2D_77K.csv (5 rows × 67 columns) Two-objective Pareto front at 77 K (risk-adjusted yield strength vs UTS). Candidates representing the best achievable strength combinations at liquid-nitrogen temperature.10. CryoAlloy_Pareto2D_120K.csv (4 rows × 67 columns) Two-objective Pareto front at 120 K (risk-adjusted yield strength vs UTS). The smallest 2D front, reflecting tighter convergence among top-strength candidates at the upper cryogenic boundary.11. Holdout_DOI_Benchmark_Summary.csv (3 rows × 9 columns) Summary results of the hold-out DOI benchmark evaluation for all three targets (yield strength, UTS, elongation). Reports the number of withheld DOIs, train/test sizes, and six performance metrics (R², MAE, RMSE, Spearman ρ, Top-10% hit rate) under zero-DOI-overlap conditions.