Machine Learning Screening for Sodium Superionic Conductor Materials

The patent applies preliminary action by performing high-throughput calculations and machine learning predictions before actual experiments. The system pre-screens materials using computational methods to identify promising candidates, so that subsequent experimental measurements are performed only on selected materials, dramatically reducing the number of experiments needed and the associated time loss.

The patent uses copying by creating virtual models of materials through high-throughput calculations and machine learning surrogate models. These computational copies allow researchers to evaluate ionic conductivity and electrochemical stability in silico, reducing the need for physical experimentation and thereby decreasing research time while maintaining measurement accuracy for final validation.

The patent performs preliminary computational screening using high-throughput calculations and machine learning to identify promising material candidates before conducting expensive experimental measurements. This pre-screening reduces the number of materials that require costly synthesis and characterization, thereby reducing overall research costs while maintaining measurement accuracy for the final selected candidates.

The patent creates virtual copies of materials through computational modeling and machine learning surrogate models. These digital twins allow researchers to evaluate material properties in silico, reducing the need for expensive physical synthesis and experimentation. Only the most promising candidates identified through computational copying proceed to actual experimental validation, significantly reducing research costs.

The patent segments the material search process into multiple stages: initial high-speed machine learning screening followed by intermediate high-throughput calculation validation, and final experimental verification. Each stage uses methods appropriate to its purpose, with machine learning providing rapid initial filtering and more accurate but slower methods providing subsequent validation, thereby achieving both speed and accuracy.

The patent introduces high-throughput calculations as an intermediary between machine learning predictions and final experimental verification. The machine learning surrogate models provide rapid initial predictions, which are then refined through more accurate but computationally intensive high-throughput calculations before experimental validation, serving as a bridge that maintains both speed and accuracy.

The patent segments the validation process into multiple layers: machine learning predictions are first filtered, then candidates undergo high-throughput calculation verification for electrochemical stability, and finally experimental validation is performed. This segmented approach ensures that machine learning efficiency is maintained while reliability is progressively verified at each stage through more robust methods.

The patent implements feedback loops where machine learning model predictions are continuously refined based on results from high-throughput calculations and experimental measurements. The system learns from the outcomes of more reliable methods to improve the accuracy and reliability of its machine learning predictions, creating a self-improving system that maintains both efficiency and reliability.