Skip to main content
QUICK REVIEW

[论文解读] DeepM&Mnet for hypersonics: Predicting the coupled flow and finite-rate chemistry behind a normal shock using neural-network approximation of operators

Zhiping Mao, Lu Lu|arXiv (Cornell University)|Nov 1, 2020
Model Reduction and Neural Networks参考文献 31被引用 24
一句话总结

本文提出 DeepM&Mnet,一种深度学习框架,利用预训练的 DeepONet 预测法向激波后的耦合高超声速流与有限速率化学反应,精度高且相比 CFD 速度提升超过 100,000 倍。通过将稀疏测量值集成到即插即用架构中,该方法实现了对多尺度、多物理场系统的高效、精确数据同化,适用于实时应用。

ABSTRACT

In high-speed flow past a normal shock, the fluid temperature rises rapidly triggering downstream chemical dissociation reactions. The chemical changes lead to appreciable changes in fluid properties, and these coupled multiphysics and the resulting multiscale dynamics are challenging to resolve numerically. Using conventional computational fluid dynamics (CFD) requires excessive computing cost. Here, we propose a totally new efficient approach, assuming that some sparse measurements of the state variables are available that can be seamlessly integrated in the simulation algorithm. We employ a special neural network for approximating nonlinear operators, the DeepONet, which is used to predict separately each individual field, given inputs from the rest of the fields of the coupled multiphysics system. We demonstrate the effectiveness of DeepONet by predicting five species in the non-equilibrium chemistry downstream of a normal shock at high Mach numbers as well as the velocity and temperature fields. We show that upon training, DeepONets can be over five orders of magnitude faster than the CFD solver employed to generate the training data and yield good accuracy for unseen Mach numbers within the range of training. Outside this range, DeepONet can still predict accurately and fast if a few sparse measurements are available. We then propose a composite supervised neural network, DeepM&Mnet, that uses multiple pre-trained DeepONets as building blocks and scattered measurements to infer the set of all seven fields in the entire domain of interest. Two DeepM&Mnet architectures are tested, and we demonstrate the accuracy and capacity for efficient data assimilation. DeepM&Mnet is simple and general: it can be employed to construct complex multiphysics and multiscale models and assimilate sparse measurements using pre-trained DeepONets in a "plug-and-play" mode.

研究动机与目标

  • 解决使用传统 CFD 模拟耦合高超声速流与有限速率化学反应时计算成本过高的问题。
  • 开发一种数据高效、快速的代理模型,利用稀疏测量实现实时预测。
  • 实现在未见马赫数下对七个耦合场(五种物种密度、速度和温度)的精确预测。
  • 展示预训练 DeepONet 即插即用集成在复杂多物理场系统中的可行性。
  • 通过正则化和全局守恒约束确保数据同化过程的稳定性和准确性。

提出的方法

  • 使用 DeepONet 近似非线性算子,将输入场(如物种密度)映射到输出场(如速度、温度和其他物种密度)。
  • 在高保真 CFD 数据上独立训练各个 DeepONet,以从系统其余部分独立预测每个场。
  • 构建 DeepM&Mnet 为复合神经网络,将多个预训练 DeepONet 与一个可训练神经网络结合,用于数据同化。
  • 使用损失函数,结合数据保真度(稀疏测量的均方误差)和算子一致性(匹配 DeepONet 预测)。
  • 应用 L2 正则化和全局质量守恒约束,以稳定训练并提升泛化能力。
  • 测试两种 DeepM&Mnet 架构:一种由流场预测物种,另一种由物种密度预测流场,两者均使用稀疏数据输入。

实验结果

研究问题

  • RQ1DeepONet 是否能以极低计算成本准确预测法向激波后的耦合流场与有限速率化学反应场?
  • RQ2当在有限 CFD 数据范围内训练时,DeepM&Mnet 在未见马赫数下的泛化能力如何?
  • RQ3稀疏测量是否能显著提升 DeepM&Mnet 在训练分布外泛化时的预测精度与稳定性?
  • RQ4正则化和全局守恒约束在稳定 DeepM&Mnet 数据同化过程中的作用是什么?
  • RQ5预训练 DeepONet 的即插即用集成是否能实现复杂多物理场系统高效、模块化的建模?

主要发现

  • 与 CFD 参考解相比,DeepONet 的相对均方误差约为 10−5,证明其具有高精度。
  • DeepONet 的推理速度比用于生成训练数据的 CFD 求解器快超过 100,000 倍。
  • 当仅提供少量稀疏测量时,DeepM&Mnet 在未见马赫数下仍保持高精度,实现稳健泛化。
  • 在 6 个稀疏数据点下,配合适当的正则化(ωR = 10−5)和全局守恒(ωG = 1),DeepM&Mnet 实现最低均方误差(如 ρN2 的误差为 2.73×10−4),且训练稳定。
  • 表现最佳的配置(ωG = 1,ωR = 10−5)使物种密度和温度的预测误差低于 1.2×10−5,速度误差为 7.25×10−6。
  • 若无正则化,即使采用全局守恒,训练仍会变得不稳定,凸显正则化在可靠数据同化中的必要性。

更好的研究,从现在开始

从论文设计到论文写作,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。