Skip to main content
QUICK REVIEW

[论文解读] Physics-aware deep neural networks for surrogate modeling of turbulent natural convection

Didier Lucor, Atul Agrawal|arXiv (Cornell University)|Mar 5, 2021
Model Reduction and Neural Networks参考文献 35被引用 29
一句话总结

本论文研究 PINNs 作为由 NS(以 Boussinesq 近似描述的三维湍流自然对流)代理的可行性。引入 padding 和放松不可压缩性以改进训练;即使在非常高的 Rayleigh 数值下,使用有限的 DNS 数据也报告了高精度。

ABSTRACT

Recent works have explored the potential of machine learning as data-driven turbulence closures for RANS and LES techniques. Beyond these advances, the high expressivity and agility of physics-informed neural networks (PINNs) make them promising candidates for full fluid flow PDE modeling. An important question is whether this new paradigm, exempt from the traditional notion of discretization of the underlying operators very much connected to the flow scales resolution, is capable of sustaining high levels of turbulence characterized by multi-scale features? We investigate the use of PINNs surrogate modeling for turbulent Rayleigh-B{é}nard (RB) convection flows in rough and smooth rectangular cavities, mainly relying on DNS temperature data from the fluid bulk. We carefully quantify the computational requirements under which the formulation is capable of accurately recovering the flow hidden quantities. We then propose a new padding technique to distribute some of the scattered coordinates-at which PDE residuals are minimized-around the region of labeled data acquisition. We show how it comes to play as a regularization close to the training boundaries which are zones of poor accuracy for standard PINNs and results in a noticeable global accuracy improvement at iso-budget. Finally, we propose for the first time to relax the incompressibility condition in such a way that it drastically benefits the optimization search and results in a much improved convergence of the composite loss function. The RB results obtained at high Rayleigh number Ra = 2 $\bullet$ 10 9 are particularly impressive: the predictive accuracy of the surrogate over the entire half a billion DNS coordinates yields errors for all flow variables ranging between [0.3% -- 4%] in the relative L 2 norm, with a training relying only on 1.6% of the DNS data points.

研究动机与目标

  • 评估 PINNs 作为由 NS 与 Boussinesq 近似所建模的三维湍流自然对流的代理的可行性。
  • 量化从部分 DNS 数据中准确恢复流场所必需的数据要求和残差采样策略。
  • 引入一种填充技术,将残差分布在带标签数据的区域周围,以规范化训练。
  • 提出放松不可压缩性约束,以提高优化收敛性和代理准确性。

提出的方法

  • 在 PINN 框架内,利用 Boussinesq 近似表述的非定常 Navier–Stokes 方程。
  • 使用多层感知机近似解向量 u = (v, p, T) 并通过自动微分计算偏微分方程残差。
  • 通过将带标签数据与 PDE 残差结合的损失进行训练,使用带分阶段学习率循环的 Adam。
  • 引入一种填充策略,在带标签数据的区域周围分布残差评估点,以提高边界精度。
  • 添加一个辅助的平移温度变量 T̄ = 1 − T 以提供额外的训练约束。
  • 放宽不可压缩性条件,以改善优化搜索和复合损失的收敛。

实验结果

研究问题

  • RQ1在高 Rayleigh 数下,使用部分 DNS 数据的情况下,PINNs 是否能准确代理三维湍流自然对流?
  • RQ2在训练 PINNs 处理湍 RB 对流时,哪些数据和残差采样策略能够在带标签数据与 PDE 残差之间达到最佳平衡?
  • RQ3在带标签数据周围的填充方案是否提高了 PINN 的准确性,特别是在训练边界附近?
  • RQ4放松不可压缩性约束是否能改善优化收敛性和代理性能?

主要发现

  • 在仅使用 1.6% 的 DNS 点进行训练时,针对高 Ra 的 RB 对流,在流变量上的相对 L2 误差范围为 0.3% 到 4%。
  • 表明 PDE 残差在训练初期占主导地位,而数据-标签误差随着循环更规律地收敛。
  • 引入一种填充技术,将残差点分布在带标签数据区域周围,在等预算下实现显著的全球精度提升。
  • 放宽不可压缩性约束可显著改善优化搜索和 PINN 损失的收敛性。
  • 在 Ra = 2×10^9 下,使用部分 DNS 数据训练的代理能在半十亿 DNS 坐标上高保真地预测整个场。

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。