This project leverages advanced Generative AI techniques to transform 2D engineering drawings into featurized 3D CAD models. By fine-tuning a vision large language model, the featurized 3D CAD enables accurate and scalable model creation for simulation, sustainment, and digital twin applications.
Engineering sustainment and simulation often require high-fidelity 3D models, but legacy designs exist only in 2D engineering drawings. Manual 3D conversion is time-consuming, costly, and prone to inconsistencies. Our solution leverages Generative AI to automate the conversion from 2D engineering drawings into accurate, scalable 3D models. The approach enables rapid model creation for digital twins, predictive maintenance, and simulation environments.
The system uses a fine-tuned vision LLM combined with a denoising network to remove non-structural information, then outputs CadQuery Python code for featured CAD models. This enables up to 80% reduction in model creation time with AI-driven geometric consistency and minimal manual intervention.
