Machine learning-based statistical closure models for turbulent dynamical systems

Machine learning-based statistical closure models for turbulent dynamical systems
复制标题

基于机器学习的湍流动力系统统计闭合模型

DOI:
10.1098/rsta.2021.0205
复制
发表时间:
2022
期刊:
Physical and Engineering Sciences
影响因子:
--
通讯作者:
Harlim, John
Harlim, John
中科院分区:
--
文献类型:
--
作者:
Qi, Di;Harlim, John

文献摘要

参考文献

被引文献

相似文献

我们提出了一种机器学习(ML)非马尔可夫闭包建模框架,用于准确预测受外力作用的湍流动力系统的统计响应。这个统计闭包问题的困难之一是缺乏训练数据,这种配置在神经网络模型的监督学习中是不可取的。在这项使用 40 维 Lorenz-96 模型的研究中,数据短缺是由于去相关时间之外的统计平稳性所致。因此,训练数据中唯一的信息内容来自短时瞬态统计。我们对各种截断机制采用统一的闭包框架,包括和排除方差的详细动态方程。封闭框架采用长短期记忆架构来表示高阶未解决的统计反馈,并选择 ansatz 来解释内在的不稳定性,但产生稳定的长期预测。我们发现这种统一的不可知机器学习方法在各种截断场景下都表现良好。数值上表明,ML 闭合模型可以准确预测各种与时间相关的外力的长期统计响应,这些外力具有较大的最大强迫幅度且不在训练数据集中。本文是主题问题“动态系统中的数据驱动预测”的一部分。
We propose a machine learning (ML) non-Markovian closure modelling framework for accurate predictions of statistical responses of turbulent dynamical systems subjected to external forcings. One of the difficulties in this statistical closure problem is the lack of training data, which is a configuration that is not desirable in supervised learning with neural network models. In this study with the 40-dimensional Lorenz-96 model, the shortage of data is due to the stationarity of the statistics beyond the decorrelation time. Thus, the only informative content in the training data is from the short-time transient statistics. We adopt a unified closure framework on various truncation regimes, including and excluding the detailed dynamical equations for the variances. The closure framework employs a Long-Short-Term-Memory architecture to represent the higher-order unresolved statistical feedbacks with a choice of ansatz that accounts for the intrinsic instability yet produces stable long-time predictions. We found that this unified agnostic ML approach performs well under various truncation scenarios. Numerically, it is shown that the ML closure model can accurately predict the long-time statistical responses subjected to various time-dependent external forces that have larger maximum forcing amplitudes and are not in the training dataset.This article is part of the theme issue ‘Data-driven prediction in dynamical systems’.
DOI: 10.1073/pnas.1313065110
发表时间: 2013-08-20
影响因子: 11.1
作者:
Sapsis, Themistoklis P.;Majda, Andrew J.
通讯作者: Majda, Andrew J.
用于过滤具有模型误差的非线性多尺度系统的线性理论
DOI: --
发表时间: 2013
期刊: Proceedings of the Royal Society A
影响因子: --
作者:
Tyrus Berry;J. Harlim
通讯作者: J. Harlim
DOI: 10.1016/s0167-2789(02)00446-3
发表时间: 2002-06-15
影响因子: 4
作者:
Chorin, AJ;Hald, OH;Kupferman, R
通讯作者: Kupferman, R
DOI: --
发表时间: 2016
期刊:
影响因子: --
作者:
F. Lu;Kevin K. Lin;A. Chorin
通讯作者: A. Chorin
DOI: 10.1098/rspa.2017.0385
发表时间: 2017-09-01
影响因子: 3.5
作者:
Gouasmi, Ayoub;Parish, Eric J.;Duraisamy, Karthik
通讯作者: Duraisamy, Karthik