High-throughput characterization methods for Ni-based superalloys and phase prediction via deep learning

Abstract

The strengthening phases characteristics in the alloy determine the mechanical properties of the alloy, but it is a hard task to predict the precipitation of complex alloys. In this work, we quickly detected 33,484 groups of Ni-based superalloys composition information and microstructure image by integrating high-throughput experiment and a nested UNet 3+ architecture for image recognition, and established a database of γ′ precipitation. Based on the database, a high-confidence prediction model was established, which could accurately predict the volume fraction, average size and size distribution of γ′ prediction in different alloys. Compared with the traditional methods, the proposed approach has a remarkable advantage in acquiring and analyzing the experimental data, which can also be applied to other multi-component alloys.

Keywords

• Superalloy
• Diffusion-multiple
• Deep learning
• High-throughput
• Powder metallurgy