Combined data-driven model for the prediction of thermal properties of Ni-based amorphous alloys

Abstract

Ni-based amorphous alloys are a unique class of materials that are attracting attention in biomass plants because of their outstanding physical properties at high temperatures. Several studies have investigated and designed the relationships between the input and target properties of alloys using machine learning algorithms. The extensive use of these models has a limitation in that the required composition is yet to be determined. To address this issue, we trained four machine learning algorithms to design Ni-based amorphous alloys and predict their thermal properties. The machine learning algorithms were trained using only the compositions of Ni-based amorphous alloys obtained from the relevant literature as the input feature data. Random forest regression was selected to predict and design the Ni-based amorphous alloys. We applied this algorithm to design amorphous alloys with the desired thermal properties and an optimal composition determined via particle swarm optimization. A melt spinner was used to fabricate the alloy. X-ray diffraction and differential thermal analyses were used to evaluate the specimens. Empirical equations were proposed for use in industrial fields.

Keywords

• Machine learning
• Particle swarm optimization
• Empirical equation
• Thermal property
• Ni-based amorphous alloy