An LLM-assisted Framework of Gating System Design for Giga-casting

Giga-casting is a key enabling technology for automotive lightweighting, where die design plays a vital role in ensuring its casting quality. Among the key components of a giga-casting die, gating system is a core that determines the filling behaviour of molten metal. However, the larger dimensions and more complex geometries of giga-casting make its gating system design both more workload-intensive and experience-dependent. As a result, designers must conduct empirical and analytical study, with frequently switch among multiple specialized design tools, fragmenting the design workflow and reducing overall design efficiency. Therefore, an interactive design tool becomes crucial, as it could better understand designer’s intention and adaptively invoke the tools to assist giga-casting gating system design. To achieve this, this paper proposes an LLM-assisted framework for giga-casting gating system design. It consists of three parts: an interaction engine, a calculation module, and a knowledge base. The interaction engine, based on Large Language Models (LLMs), receives designers’ natural-language input and interactively provides design recommendation under the constraints of a knowledge base. Moreover, it also adaptively invokes corresponding calculation procedures in the module, based on its contextual understanding of the current design workflow. Finally, it outputs complete design parameters together with the underlying design rational and generates a 3D gating system model accordingly. The gating system design for a rear floor component is used as case study to demonstrate the proposed framework.