Data-Driven Microstructure Optimization of Spark Plasma Sintered Ceramics

Spark plasma sintering (SPS) is a field-assisted sintering technique that enables rapid densification of powders through Joule heating, achieved by the simultaneous application of an electric current and uniaxial pressure. The key advantages of SPS include reduced sintering time and temperature, which minimize grain growth and promote a fine-grained microstructure. Due to these advantages, SPS has found applications in several key sectors, including aerospace, automotive, electronics, and biomedical industries.
Despite its benefits, several studies have reported non-uniform microstructures in ceramic samples processed by SPS. The simultaneous application of pressure and temperature can lead to pronounced thermal and stress gradients within the sample.
In this work, decoupling thermal and mechanical loading during SPS is proposed as a strategy to improve microstructural homogeneity and tailor densification mechanisms in ceramic systems. This is achieved using a specialized die configuration in which the electrode connections are decoupled from the high-pressure punches, enabling independent pressure application at defined stages of the sintering process. However, this tooling concept has received limited research to date.
The primary objective of this study is to understand the sintering mechanisms under the modified SPS setup and to optimize processing conditions for improved microstructural control and enhanced densification. To establish the process–property–microstructure relationship, a combination of simulations, experiments, and data-driven methods is employed. Since direct measurement of the sample temperature during SPS is not feasible, simulations are used to predict temperature gradients within the specimen. These models are validated through experiments conducted at varying processing parameters. The resulting microstructures are analyzed for grain size distribution and porosity using AI-driven image analysis, which enables automated, objective, and reproducible evaluation of large datasets while extracting comprehensive microstructural information. Machine learning techniques are subsequently applied to optimize sintering parameters and achieve performances better than that of conventional SPS configurations.