BlueprintAgent automates accurate frame models from scanned blueprints
BlueprintAgent: Constraint-Triggered Targeted Revisits for Simulation-Ready Generation from Scanned Structural Blueprints
Computation and LanguageArtificial IntelligenceComputer Vision and Pattern Recognition
Summary
Making detailed, computer-ready building models from scanned blueprints is usually slow and done by hand. The authors introduced BlueprintAgent, which uses a smart language model combined with vision tools and engineering rules to check and fix errors locally in the plans. This method helps get very precise frame details suitable for safety checks and simulations. Their tests show it works much better than previous methods, especially on complicated building drawings.
What this means in practice
- •For structural engineering teams: Automatically generate engineering-consistent simulation models from scanned blueprints to speed up building safety and retrofit analyses.
- •For construction project managers: Produce reliable digital building frame representations from legacy paper plans to facilitate modern construction planning tools.
Authors
Zhouyuan Xu, Chen Yang, Linhao Wang, Jiansheng Fan, Chen Wang
Abstract
Converting in-service reinforced-concrete (RC) building blueprints into simulation-ready models---structured frame representations that support deterministic FEM export and qualified-engineer review---underpins safety assessment and seismic retrofit, but the process remains manual. Direct prompting of a multimodal large language model (MLLM) over a scanned sheet is unreliable: outputs often violate engineering constraints on beam--column support, span count, or 3D continuity. We present BlueprintAgent (BPA), a constraint-triggered multimodal agent for simulation-ready frame extraction from scanned blueprints. BPA treats the MLLM as the primary reader and decision maker, with OCR and computer vision supplying localized evidence. Its central mechanism realizes engineering constraints as callable validators whose entity-level conflict reports trigger targeted MLLM revisits over the local region---an inference-time control distinct from fixed pipelines and free-form self-reflection. We evaluate BPA on 300 real scanned blueprint sheets from 20 anonymized RC frame projects, against five baselines and six ablations. BPA reaches a macro-averaged Beam F1 of 0.994, against 0.301 for single-MLLM zero-shot and 0.820 for a fixed pipeline; removing MLLM-led axis adjudication collapses Beam and Column F1 on complex multi-sheet projects. For dense technical drawings, engineering constraints are best deployed as triggers for entity-level targeted revisits rather than as post-hoc output filters.