Tillämpning av artificiell intelligens inom mekanisk konstruktion

Abstract

The increasing availability of large language models and AI-based development tools has opened new possibilities for integrating artificial intelligence into engineering workflows. This thesis investigates how AI can be applied as a support tool in a CAD-based design process, with focus on geometry generation, automated assembly operations, and component dimensioning. The study was conducted through a practical implementation in Autodesk Fusion, where an AI agent in the development environment Cursor was connected to the CAD system via a Model Context Protocol server, enabling direct interaction between the AI agent and the CAD environment. The results demonstrate that AI-assisted CAD workflows are practically feasible and can contribute to increased efficiency in well-defined and recurring design tasks. However, the reliability of AIgenerated results was found to be highly dependent on the clarity of user instructions, and manual verification remained necessary throughout the process. A key finding is that the time required to verify AI-generated outputs in many cases exceeds the time savings gained through automation, particularly for more complex tasks. For AI-based CAD workflows to become genuinely competitive with traditional methods, model reliability must improve to a level where users can trust generated results without reviewing each step individually. The study concludes that while AI holds significant promise for the future of CAD-based engineering, its full potential is contingent on continued improvements in both model quality and cost-effectiveness