Experimental Value Optimization Under Allowable Synthesis Conditions

The patent applies preliminary action by performing multi-objective optimization before actual experiments to generate candidate experimental values that satisfy allowable conditions. The system calculates multiple objective functions (including synthesis probability, conductivity, and allowable condition satisfaction) in advance, and selects candidates that meet all conditions before experimentation begins, thereby preventing failed experiments due to condition violations.

The patent changes parameters by introducing multiple objective functions with different weights to balance competing requirements. The system adjusts the weights of different objective functions (synthesis probability, conductivity prediction, allowable condition satisfaction) to generate optimal experimental value candidates that satisfy all conditions while maximizing productivity and reliability.

The patent replaces the mechanical experimentation system with an information processing system that uses multi-objective optimization algorithms. Instead of physically testing multiple composition ratios through synthesis and measurement, the system calculates optimal candidates using trained models (synthesis probability model and conductivity prediction model) and selects the best candidates computationally, thereby reducing experimentation time while maintaining or improving optimization accuracy.

The patent uses copying by creating virtual models of the synthesis process and conductivity measurement through trained machine learning models. The synthesis probability model and conductivity prediction model serve as virtual copies that simulate the outcomes of actual experiments, allowing the system to evaluate multiple composition ratios computationally before performing physical experiments, thus reducing the number of actual experiments needed.