A Three-Layer Runtime Constraint Verification Framework with Self-Correction for AI-Generated Parametric CAD Models
Source abstract
Abstract
Large language models (LLMs) frequently generate physically impossible parametric specifications in computer-aided design (CAD). This paper presents FMforME, a three-layer runtime constraint verification framework that integrates LLM-based generation with six engineering standard-grounded defect detection predicates (D1–D6) and execution in Autodesk Fusion 360. To address detected violations, a Self-Examine subsystem iteratively provides structured, standard-referenced feedback to the LLM, enabling verifier-guided error correction. We evaluated the framework across 564 test cases using four LLM backends from three model families. The system showed consistent buildability within the evaluated benchmark: across all models, every specification validated by the Monitor successfully built in Fusion 360 (548/548). Across the four evaluated backends, realized seeded defect detection ranged from 95.5% to 95.8%, and three backends showed an observed 100% natural pass rate in this benchmark. Crucially, this runtime safety barrier adds little computational overhead relative to LLM inference, averaging only 0.1 ms per in-memory verification check. These results show that standards-referenced runtime verification with structured self-correction can serve as a practical safeguard for LLM-generated CAD specifications within the tested single-part workflows, while broader validation on complex assemblies remains future work.
Source: crossref · Quality: baseline
Affiliations
Zhejiang University—University of Illinois at Urbana-Champaign Institute, Zhejiang University, Haining 314400, China
College of Energy Engineering, Zhejiang University, Hangzhou 310027, China
Funding
National Natural Science Foundation of China · National Key Laboratory for Bio-based Transportation Fuels
References (48)
Show source-provided references
1. Cook, R.D., Malkus, D.S., Plesha, M.E., and Witt, R.J. (2001). Concepts and Applications of Finite Element Analysis, John Wiley & Sons. [4th ed.].
2. Shafiq, O., Rahmat, A., Alexiadis, A., and Ghiassi, B. (2025). Evaluating the performance of large language models for geometry and simulation file generation in physics-based simulations. Appl. Sci., 15.10.3390/app152212114 ↗
3. From text to tech: Shaping the future of physics-based simulations with AI-driven generative models10.1016/j.rineng.2023.101721 ↗
4. Fusion 360 Gallery: A dataset and environment for programmatic CAD construction from human design sequences10.1145/3450626.3459818 ↗
5. Seff, A., Ovadia, Y., Zhou, W., and Adams, R.P. (2020). SketchGraphs: A large-scale dataset for modeling relational geometry in computer-aided design. arXiv.
6. Koch, S., Matveev, A., Jiang, Z., Williams, F., Artemov, A., Burnaev, E., Alexa, M., Zorin, D., and Panozzo, D. (2019, January 16–20). ABC: A big CAD model dataset for geometric deep learning. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, CA, USA.10.1109/CVPR.2019.00983 ↗
7. Sharma, G., Goyal, R., Liu, D., Kalogerakis, E., and Maji, S. (2018, January 18–22). CSGNet: Neural shape parser for constructive solid geometry. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA.10.1109/CVPR.2018.00578 ↗
8. Wu, R., Xiao, C., and Zheng, C. (2021, January 11–17). DeepCAD: A deep generative network for computer-aided design models. Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, QC, Canada.10.1109/ICCV48922.2021.00670 ↗
9. FeatureNet: Machining feature recognition based on 3D convolution neural network10.1016/j.cad.2018.03.006 ↗
10. Chien, C.T., and Chien, C.H. (2025). Hybrid AI-driven computer-aided engineering optimization: Large language models versus regression-based models validated through finite-element analysis. Appl. Sci., 15.10.3390/app151810123 ↗
11. Baker, C., Rafferty, K., and Price, M. (2025). Large language models in mechanical engineering: A scoping review of applications, challenges, and future directions. Big Data Cogn. Comput., 9.10.20944/preprints202508.1938.v1 ↗
12. Large language model-empowered next-generation computer-aided engineering10.1016/j.cma.2025.118591 ↗
13. Jones, B.T., Hähnlein, F., Zhang, Z., Ahmad, M., Kim, V.G., and Schulz, A. (2025). A solver-aided hierarchical language for LLM-driven CAD design. arXiv.10.1111/cgf.70250 ↗
14. Mudur, N., Cui, H., Venugopalan, S., Raccuglia, P., Brenner, M.P., and Norgaard, P. (2025). FEABench: Evaluating language models on multiphysics reasoning ability. arXiv.
15. Wang, L., Meng, H., Xiang, Z., Liu, J., Zhou, P., Chen, L., and Tang, Y. (2026). Text2CAD-Bench: A benchmark for LLM-based text-to-parametric CAD generation. arXiv.
16. Madaan, A., Tandon, N., Gupta, P., Hallinan, S., Gao, L., Wiegreffe, S., Camburu, O.-M., and Clark, P. (2023, January 10–16). Self-Refine: Iterative refinement with self-feedback. Proceedings of the Advances in Neural Information Processing Systems 36 (NeurIPS), New Orleans, LA, USA.10.52202/075280-2019 ↗
17. Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., and Cao, Y. (2023, January 1–5). ReAct: Synergizing reasoning and acting in language models. Proceedings of the Eleventh International Conference on Learning Representations (ICLR), Kigali, Rwanda.
18. Gou, Z., Shao, Z., Gong, Y., Shen, Y., Yang, Y., Duan, N., and Chen, W. (2024, January 7–11). CRITIC: Large language models can self-correct with tool-interactive critiquing. Proceedings of the Twelfth International Conference on Learning Representations (ICLR), Vienna, Austria.
19. Shinn, N., Cassano, F., Gopinath, A., Narasimhan, K., and Yao, S. (2023, January 10–16). Reflexion: Language agents with verbal reinforcement learning. Proceedings of the Advances in Neural Information Processing Systems 36 (NeurIPS), New Orleans, LA, USA.10.52202/075280-0377 ↗
20. Automatically correcting large language models: Surveying the landscape of diverse automated correction strategies10.1162/tacl_a_00660 ↗
21. Bubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., Lee, P., Lee, Y.T., Li, Y., and Lundberg, S. (2023). Sparks of artificial general intelligence: Early experiments with GPT-4. arXiv.
22. MyCrunchGPT: A ChatGPT assisted framework for scientific machine learning10.1615/JMachLearnModelComput.2023049518 ↗
23. A survey on 3D CAD model quality assurance and testing tools10.1016/j.cad.2016.10.003 ↗
24. Graph-based heuristics for recognition of machined features from a 3D solid model10.1016/0010-4485(88)90050-4 ↗
25. Spatial reasoning for the automatic recognition of machinable features in solid models10.1109/34.250845 ↗
26. Manufacturing feature recognition from solid models: A status report10.1109/70.897789 ↗
27. A review of automated feature recognition with rule-based pattern recognition10.1016/j.compind.2007.09.001 ↗
28. Shah, J.J., and Mäntylä, M. (1995). Parametric and Feature-Based CAD/CAM: Concepts, Techniques, and Applications, John Wiley & Sons.
29. A road map to solid modeling10.1109/2945.489381 ↗
30. Semantic feature modelling10.1016/S0010-4485(99)00090-1 ↗
31. A discourse on geometric feature recognition from CAD models10.1115/1.1345522 ↗
32. Knowledge based engineering: Between AI and CAD. Review of a language based technology to support engineering design10.1016/j.aei.2012.02.002 ↗
33. The status, evolution, and future challenges of multimodal large language models (LLMs) in parametric CAD10.1016/j.eswa.2025.127520 ↗
34. Lin, Z., and Nie, Z. (2025). MC Classifier: A classifier for 3D mechanical components based on geometric prior using graph neural network and attention. Appl. Sci., 15.10.3390/app15084399 ↗
35. Li, J., Ma, W., Li, X., Lou, Y., Zhou, G., and Zhou, X. (2025, January 11–15). CAD-Llama: Leveraging large language models for computer-aided design parametric 3D model generation. Proceedings of the IEEE/CVF CVPR, Nashville, TN, USA.10.1109/CVPR52734.2025.01730 ↗
36. CADialogue: A multimodal LLM-powered conversational assistant for intuitive parametric CAD modeling10.1016/j.cad.2025.104006 ↗
37. Bathe, K.J. (2014). Finite Element Procedures, Prentice Hall. [2nd ed.].
38. ASME (2019). Standard for Verification and Validation in Computational Solid Mechanics, American Society of Mechanical Engineers. (ASME V&V 10-2019, Reaffirmed 2025).
39. Zienkiewicz, O.C., Taylor, R.L., and Fox, D.D. (2014). The Finite Element Method for Solid and Structural Mechanics, Butterworth-Heinemann. [7th ed.].
40. Stimpson, C., Ernst, C., Knupp, P., Pébay, P., and Thompson, D. (2007). The Verdict Geometric Quality Library, Sandia National Laboratories. SAND2007-1751.
41. DeepSeek (2026, June 01). DeepSeek API Documentation. Available online: https://api-docs.deepseek.com/.
42. Autodesk (2026, June 01). Fusion 360 API Reference. Available online: https://help.autodesk.com/view/fusion360/ENU/.
43. (2017). Rolling Bearings—Radial Bearings—Boundary Dimensions, General Plan (Standard No. ISO 15:2017).
44. Budynas, R.G., and Nisbett, J.K. (2014). Shigley’s Mechanical Engineering Design, McGraw-Hill. [10th ed.].
45. Algebraic mesh quality metrics10.1137/S1064827500371499 ↗
46. (2004). Hexagon Socket Head Cap Screws (Standard No. ISO 4762:2004).
47. CYCLE: Learning to self-refine the code generation10.1145/3649825 ↗
48. Casey, E., Zhang, T., Ishida, S., McCarthy, W.P., Thompson, J.R., Khasahmadi, A., Lambourne, J.G., Jayaraman, P.K., and Willis, K.D.D. (2025, January 19–23). Aligning constraint generation with design intent in parametric CAD. Proceedings of the ICCV, Honolulu, HI, USA.10.1109/ICCV51701.2025.00806 ↗