Exploring an Agent-Based Generative AI CAD Framework: Enhancing CAD Design Automation with AI-Driven Parametric Modelling and Adaptive Query Generation
2025 (English)Independent thesis Advanced level (degree of Master (Two Years)), 80 credits / 120 HE credits
Student thesis
Abstract [en]
This thesis presents the development of an AI-driven, agent-based framework designed to make Computer Aided Design (CAD) more accessible, intuitive, and efficient for the CAD users. Traditional CAD workflows require technical expertise and manual operations, creating barriers for those unfamiliar with complex software tools. Our goal was to create a system where users can simply describe their design intent using plain language, sketches, or both, and have those ideas automatically converted into accurate CAD models.To achieve this, we developed a modular framework that integrates advanced AI components capable of understanding user input, inferring missing details, and executing design operations inside a professional CAD environment (CATIA V5). The system uses language and vision models to interpret input, a structured database to manage commands, and a graph-based workflow engine to ensure correct execution order. Following an iterative development method (Rapid Application Development), we implemented and tested the framework in stages, gradually improving its capabilities. The result is a working prototype that successfully performs CAD functions based on natural input, demonstrating the feasibility of more intelligent and user-friendly CAD automation
Place, publisher, year, edition, pages
2025. , p. 68
Keywords [en]
CAD, AI, Large Language Models, Natural Language Processing, Agentic, CATIA, P
National Category
Other Mechanical Engineering
Identifiers
URN: urn:nbn:se:liu:diva-218259ISRN: LIU-IEI-TEK-A--25/05160—SEOAI: oai:DiVA.org:liu-218259DiVA, id: diva2:2002963
Presentation
2025-06-04, Linkoping University, Linkoping, 13:10 (English)
Supervisors
Examiners
2025-10-022025-10-022025-10-02Bibliographically approved