Accelerated Design of Fe-based Soft Magnetic Materials Using Machine Learning and Stochastic Optimization

Machine learning was utilized to efficiently boost the development of softmagnetic materials. The design process includes building a database composed ofpublished experimental results, applying machine learning methods on thedatabase, identifying the trends of magnetic properties in soft magneticmaterials, and accelerating the design of next-generation soft magneticnanocrystalline materials through the use of numerical optimization. Machinelearning regression models were trained to predict magnetic saturation (B_S),coercivity (H_C) and magnetostriction (λ), with a stochasticoptimization framework being used to further optimize the correspondingmagnetic properties. To verify the feasibility of the machine learning model,several optimized soft magnetic materials – specified in terms of compositionsand thermomechanical treatments – have been predicted and then prepared andtested, showing good agreement between predictions and experiments, proving thereliability of the designed model. Two rounds of optimization-testingiterations were conducted to search for better properties.