Bubbleformer: Forecasting Boiling with Transformers

2025-07-28

AI Summary

Existing boiling surrogate models rely on future inputs (e.g., bubble positions), hindering autonomous prediction of nucleation dynamics and flow boiling velocity fields—especially under strong interface-momentum coupling, where generalization degrades significantly. To address this, we propose the first physics-generalizable Transformer model tailored for boiling: it incorporates factorized axial attention to capture anisotropic long-range dependencies, integrates frequency-aware scaling and thermophysical property conditioning for cross-fluid, cross-geometry, and cross-operating-condition generalization, and adopts a spatiotemporal Transformer architecture with frequency-domain-aware normalization. We further introduce physically interpretable evaluation metrics—heat flux consistency, interface geometric fidelity, and mass conservation. Our model achieves state-of-the-art performance in long-horizon prediction and extrapolation for both pool and flow boiling, enabling, for the first time, end-to-end autonomous prediction without simulation data. We publicly release the high-fidelity dataset BubbleML 2.0.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

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

Abstract

Modeling boiling (an inherently chaotic, multiphase process central to energy and thermal systems) remains a significant challenge for neural PDE surrogates. Existing models require future input (e.g., bubble positions) during inference because they fail to learn nucleation from past states, limiting their ability to autonomously forecast boiling dynamics. They also fail to model flow boiling velocity fields, where sharp interface-momentum coupling demands long-range and directional inductive biases. We introduce Bubbleformer, a transformer-based spatiotemporal model that forecasts stable and long-range boiling dynamics including nucleation, interface evolution, and heat transfer without dependence on simulation data during inference. Bubbleformer integrates factorized axial attention, frequency-aware scaling, and conditions on thermophysical parameters to generalize across fluids, geometries, and operating conditions. To evaluate physical fidelity in chaotic systems, we propose interpretable physics-based metrics that evaluate heat-flux consistency, interface geometry, and mass conservation. We also release BubbleML 2.0, a high-fidelity dataset that spans diverse working fluids (cryogens, refrigerants, dielectrics), boiling configurations (pool and flow boiling), flow regimes (bubbly, slug, annular), and boundary conditions. Bubbleformer sets new benchmark results in both prediction and forecasting of two-phase boiling flows.

Problem

• Forecasting chaotic boiling dynamics autonomously without future input
• Modeling flow boiling velocity fields with sharp interface-momentum coupling
• Generalizing boiling predictions across fluids, geometries, and operating conditions

Innovation

• Transformer-based spatiotemporal model for boiling dynamics
• Factorized axial attention and frequency-aware scaling
• Physics-based metrics for chaotic system evaluation