Combined machine learning and CALPHAD approach for discovering processing-structure relationships in soft magnetic alloys

Abstract

FINEMET alloys have desirable soft magnetic properties due to the presence of Fe3Si nanocrystals with specific size and volume fraction. To guide future design of these alloys, we investigate relationships between select processing parameters (composition, temperature, annealing time) and structural parameters (mean radius and volume fraction) of the Fe3Si domains. We present a combined CALPHAD and machine learning approach leading to well-calibrated metamodels able to predict structural parameters quickly and accurately for any desired inputs. To generate data, we have used a known precipitation model to perform annealing simulations at several temperatures, for varying Fe and Si concentrations. Thereafter, we used the data to develop metamodels for mean radius and volume fraction via the k-Nearest Neighbour algorithm. The metamodels reproduce closely the results from the precipitation model over the entire annealing timescale. Our analysis via parallel coordinate charts shows the effect of composition, temperature, and annealing time, and helps identify combinations thereof that lead to the desired mean radius and volume fraction for nanocrystals. This work contributes to understanding the linkages between processing parameters and microstructural characteristics responsible for achieving targeted properties, and illustrates ways to reduce the time from alloy discovery to deployment.