[论文解读] Compressed Convolutional LSTM: An Efficient Deep Learning framework to Model High Fidelity 3D Turbulence

Arvind Mohan, Don Daniel+2 more|arXiv (Cornell University)|Feb 28, 2019

Model Reduction and Neural Networks参考文献 44被引用 59

一句话总结

作者提出了 Compressed ConvLSTM (CC-LSTM)，将卷积自编码器与 ConvLSTM 结合，用以学习3D湍流的低维吸引子并高效生成时空流场实现，通过基于物理的诊断进行验证。

ABSTRACT

High-fidelity modeling of turbulent flows is one of the major challenges in computational physics, with diverse applications in engineering, earth sciences and astrophysics, among many others. The rising popularity of high-fidelity computational fluid dynamics (CFD) techniques like direct numerical simulation (DNS) and large eddy simulation (LES) have made significant inroads into the problem. However, they remain out of reach for many practical three-dimensional flows characterized by extremely large domains and transient phenomena. Therefore designing efficient and accurate data-driven generative approaches to model turbulence is a necessity. We propose a novel training approach for dimensionality reduction and spatio-temporal modeling of the three-dimensional dynamics of turbulence using a combination of Convolutional autoencoder and the Convolutional LSTM neural networks. The quality of the emulated turbulent fields is assessed with rigorous physics-based statistical tests, instead of visual assessments. The results show significant promise in the training methodology to generate physically consistent turbulent flows at a small fraction of the computing resources required for DNS.

研究动机与目标

• 在工程与地球科学应用中，说明数据驱动、低成本高保真湍流建模的需求。
• 开发一个可扩展的框架，在3D湍流流动中学习一个低维吸引子。
• Demonstrate a training strategy that jointly compresses data and models spatio-temporal dynamics.
• Assess physical fidelity of generated turbulence using physics-based diagnostics rather than visual inspection.

提出的方法

• 使用卷积自编码器 (CAE) 将3D湍流快照压缩为潜在空间。
• 将 ConvLSTM 扩展到3D，以建模潜在空间中的时空演化。
• 以两步方式训练 CC-LSTM：基于 CAE 的维度降低，随后在潜在空间进行 ConvLSTM 的时序建模。
• 通过 CAE 解码器动态地将潜在预测解压回原始3D场。
• 采用循环播种方法，从短期预测跳跃生成长时间预测。

实验结果

研究问题

• RQ1 CC-LSTM 在减小潜在空间维度的同时，是否能准确捕捉3D湍流动力学并保留关键的空间相关性？
• RQ2 物理基诊断（如能量谱、速度梯度统计、Q-R 面）是否指示重建场具备物理一致的湍流特性？
• RQ3 基于 CAE 的压缩对重建质量和随时间的预测稳定性有何影响？

主要发现

• CC-LSTM 通过在压缩潜在空间中工作实现了巨大的参数量化减少，同时不牺牲本质的空间相关性。
• CAE+ConvLSTM 流程能够在远低于 DNS 的计算成本下生成物理上一致的湍流场。
• 基于物理的诊断表明，对于 HIT 和 ScalarHIT 数据集，压缩表示能够 reasonably well 重现能量谱和速度梯度统计。
• 对 HIT 的压缩比达到 z=125，对 ScalarHIT 达到 z=20，显示在可接受保真度下实现了有效的降维。
• 该方法通过播种和循环预测实现了长时间预测，同时在单个 GPU 上保持训练可行性。