Integration of parameterized physics-informed neural network for modeling gap- and pressure-dependent heat transfer coefficients in sequentially uncoupled casting simulations (T13#)

• Subject Area: Metallurgical, Thermal and Thermomechanical Treatment of Materials
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 236616214

Casting simulations are essential for efficient and high-quality processes but are often time-consuming. Uncoupled simulations save time but lose accuracy. This project leverages physics-informed deep learning based on operator learning, which learns governing equations in a parametric way. These models are data-efficient, requiring minimal simulations, while adhering to fundamental physical and thermodynamic laws. Advanced HTC models and multi-scale coupling significantly improve prediction accuracy. Integration into commercial software WinCast enables practical applications and testing. The method provides fast, precise simulations and enhances competitiveness, especially for SMEs.

• DFG Programme: Collaborative Research Centres (Transfer Project)
• Subproject of: SFB 1120: Precision manufacturing by controlling melt dynamics and solidification in production processes
• Applicant Institution: Rheinisch-Westfälische Technische Hochschule Aachen
• Business and Industry: RWP GmbH
• Project Heads: Dr. Markus Apel; Dr.-Ing. Björn Pustal; Dr.-Ing. Shahed Rezaei