Towards Valid B-Rep Generation: Training-Free Wireframe Anomaly Detection and Repair

2026-08-05

AI Summary

This work addresses the challenge that geometric and topological anomalies—such as self-intersections and edge collapses—in intermediate wireframes often render generated B-Rep models invalid, while retraining large generative models remains prohibitively expensive. To circumvent this, the authors propose the Wireframe Debugging and Repair (WDR) framework, which intervenes at the intermediate wireframe stage without requiring model retraining. WDR first identifies anomalies by combining a vision-language model for coarse filtering with a Geometry-Topology Anomaly Detector (GTAD), then performs test-time optimization via an Energy-Guided Geometry-Topology Repair (EGGTR) module coupled with diffusion-based resampling. This approach establishes the first training-free pipeline for wireframe anomaly detection and repair, significantly improving B-Rep kernel validation success rates while preserving the diversity and distributional fidelity of generated CAD models.

Abstract

Multi-stage boundary representation (B-Rep) generation leverages intermediate wireframes to synthesize CAD models. However, geometric and topological risks in these wireframes -- such as self-intersections, edge collapses, and disconnected vertices -- can propagate to invalid final B-Reps. Mitigating such failures by retraining large generative models is computationally prohibitive. We propose Wireframe Detection and Repair (WDR), a training-free framework that intervenes at the intermediate wireframe stage to improve downstream B-Rep validity. WDR features a Geometric-Topology Anomaly Detector (GTAD) that combines parallel VLM-based coarse screening with geometric and topological detectors to predict downstream invalidity risk and route generation to dedicated branches. An Energy-Guided Geometric-Topology Repair (EGGTR) module then performs detector-triggered guided regeneration through geometry and topology branches. By scaling test-time computation via Energy-Guided Resampling and training-free guidance for diffusion models, WDR can be integrated into autoregressive and diffusion pipelines without retraining. Extensive experiments demonstrate consistent improvements in kernel-checked validity while largely retaining the measured diversity and distributional quality of synthesized CAD models. The code will be made publicly available upon acceptance.

Problem

• B-Rep generation
• wireframe anomalies
• geometric-topological validity
• CAD modeling
• invalid geometry

Innovation

• training-free
• wireframe repair
• B-Rep validity
• energy-guided resampling
• geometric-topological anomaly detection