Attention to Detail: Fine-Scale Feature Preservation-Oriented Geometric Pre-training for AI-Driven Surrogate Modeling

2025-04-27

AI Summary

AI surrogate models for 3D physics simulation often fail to preserve CAD-level geometric details—such as thin walls and fillets—due to inadequate implicit encoding of fine-grained geometry. Method: We propose a self-supervised pretraining framework that decouples geometric representation learning from physical task learning. It introduces a near-zero-level mesh sampling strategy and a batch-adaptive attention-weighted reconstruction loss to enhance the model’s capacity to encode subtle geometric structures implicitly. The pipeline follows a three-stage paradigm: geometric pretraining, latent-space reconstruction refinement, and few-shot physics fine-tuning. Contribution/Results: Our method achieves high-fidelity geometric reconstruction and high-accuracy physical prediction on structural mechanics tasks. Experiments demonstrate substantial accuracy gains over conventional parametric surrogates under data-scarce conditions, effectively bridging the semantic gap between geometric representation and physical modeling.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

AI-driven surrogate modeling has become an increasingly effective alternative to physics-based simulations for 3D design, analysis, and manufacturing. These models leverage data-driven methods to predict physical quantities traditionally requiring computationally expensive simulations. However, the scarcity of labeled CAD-to-simulation datasets has driven recent advancements in self-supervised and foundation models, where geometric representation learning is performed offline and later fine-tuned for specific downstream tasks. While these approaches have shown promise, their effectiveness is limited in applications requiring fine-scale geometric detail preservation. This work introduces a self-supervised geometric representation learning method designed to capture fine-scale geometric features from non-parametric 3D models. Unlike traditional end-to-end surrogate models, this approach decouples geometric feature extraction from downstream physics tasks, learning a latent space embedding guided by geometric reconstruction losses. Key elements include the essential use of near-zero level sampling and the innovative batch-adaptive attention-weighted loss function, which enhance the encoding of intricate design features. The proposed method is validated through case studies in structural mechanics, demonstrating strong performance in capturing design features and enabling accurate few-shot physics predictions. Comparisons with traditional parametric surrogate modeling highlight its potential to bridge the gap between geometric and physics-based representations, providing an effective solution for surrogate modeling in data-scarce scenarios.

Problem

• Preserving fine-scale geometric features in AI-driven surrogate modeling
• Addressing scarcity of labeled CAD-to-simulation datasets via self-supervised learning
• Bridging geometric and physics-based representations for accurate few-shot predictions

Innovation

• Self-supervised geometric representation learning for fine-scale features
• Decouples geometric feature extraction from physics tasks
• Uses batch-adaptive attention-weighted loss for detail preservation