Machine-Learning-Assisted Multi-Element Optimization of Mechanical Properties in Spinel Refractory Materials

Area of Science:

• Materials Science
• Ceramics Engineering
• Computational Materials Science

Background:

• Spinel refractories are crucial in high-temperature industrial applications.
• Optimizing spinel performance requires understanding complex compositional effects.
• Existing methods for material optimization are often time-consuming and empirical.

Purpose of the Study:

• To leverage machine learning for optimizing multi-element compositions in spinel refractories.
• To identify spinel compositions with superior hardness and flexural strength.
• To elucidate the microstructural mechanisms behind enhanced refractory properties.

Main Methods:

• Fabrication of 1120 spinel samples at 1600 °C.
• Construction of an experimental database with 112 data points.
• Application of machine learning models for high-throughput performance prediction and experimental verification.

Main Results:

• Identified (Al2Fe0.25Zn0.25Mg0.25Mn0.25)O4 with highest hardness (1770.6 ± 79.1 HV1).
• Identified (Al2Cr0.5Zn0.1Mg0.2Mn0.2)O4 with highest flexural strength (161.2 ± 9.7 MPa).
• Determined mechanisms including solid solution strengthening, layered structures, and grain boundary reinforcement.

Conclusions:

• Multi-element doping significantly enhances spinel refractory hardness and strength.
• Microstructural analysis reveals key strengthening mechanisms.
• This machine learning-driven approach offers a promising strategy for optimizing refractory materials.