The industry-relevant fabrication of supports in fuel-electrode supported Solid Oxide Cells (SOCs) by tape casting typically involves a multi-stage process, demanding precise control over tape thickness and density. However, conventional SOC manufacturing processes are resource-intensive and often rely on industry/R&D unpublished knowledge and trial-and-error practices to achieve the target properties of the resulting tape. Hence, machine learning (ML) was employed for predicting the thickness and density across three distinct stages of the fabrication process: tape casting, sintering, and NiO-reduction process. Our developed ML models (e.g., Extra Trees and Ridge Regressions) demonstrate exceptional accuracy ( R 2 > 0 . 9 ) for each specific prediction task. Concurrently, experimental data analysis was conducted to elucidate the impact of the manufacturing parameters on the tape properties. Our data-driven ML approach offers a pathway towards achieving precise tape property control and advancing more efficient SOC support manufacturing.
Le-Dinh et al. (Thu,) studied this question.