Horide Tomoya, Okumura Shin, Ito Shunta, Yoshida Yutaka
Department of Electrical Engineering, Nagoya University, Furo-cho, Chikusa, Nagoya, Japan.
Commun Eng. 2025 Jun 23;4(1):114. doi: 10.1038/s44172-025-00434-1.
Process engineering of materials determines not only materials properties, but also cost, yield and production capacity. Although process design is generally based on the experience of process engineers, mathematical/data-science modeling is a key challenge for future process optimization. Here we create new opportunities for process optimization in YBaCuO film fabrication through data/model-driven process design. We show integrated modelling of substrate temperature and critical current density in YBaCuO films. Gaussian process regression augmented by transfer learning and physics knowledge was constructed from a small amount of data to show substrate temperature dependence of critical current density. Non-numerical factors such as chamber design and substrate material were included in the transfer learning, and physics-aided techniques extended the model to different magnetic fields. Magnetic field dependence of critical current density was successfully predicted for a given substrate temperature for a five-sample series deposited using different pulsed laser deposition systems. Our integrated process and property modelling strategy enables data/model-driven process design for YBaCuO film fabrication for coated conductor applications.
材料的工艺工程不仅决定材料性能，还决定成本、产量和生产能力。虽然工艺设计通常基于工艺工程师的经验，但数学/数据科学建模是未来工艺优化的关键挑战。在此，我们通过数据/模型驱动的工艺设计为钇钡铜氧（YBaCuO）薄膜制造中的工艺优化创造了新机会。我们展示了YBaCuO薄膜中衬底温度和临界电流密度的集成建模。通过迁移学习和物理知识增强的高斯过程回归，利用少量数据构建模型，以显示临界电流密度对衬底温度的依赖性。迁移学习纳入了腔室设计和衬底材料等非数值因素，物理辅助技术将模型扩展到不同磁场。对于使用不同脉冲激光沉积系统沉积的五组样品系列，成功预测了给定衬底温度下临界电流密度的磁场依赖性。我们的集成工艺和性能建模策略实现了用于涂层导体应用的YBaCuO薄膜制造的数据/模型驱动工艺设计。
