Ensemble variational method with adaptive covariance inflation for learning neural network-based turbulence models

doi:10.1063/5.0199175

CORC > 力学研究所 > 中国科学院力学研究所 > 非线性力学国家重点实验室

	Ensemble variational method with adaptive covariance inflation for learning neural network-based turbulence models
	Luo, Qingyong 1,2; Zhang, Xin-Lei 1,2; He, Guowei 1,2
刊名	PHYSICS OF FLUIDS
	2024-03-01
卷号	36 期号:3 页码:21
ISSN号	1070-6631
DOI	10.1063/5.0199175
通讯作者	Zhang, Xin-Lei(zhangxinlei@imech.ac.cn) ; He, Guowei(hgw@lnm.imech.ac.cn)
英文摘要	This work introduces an ensemble variational method with adaptive covariance inflation for learning nonlinear eddy viscosity turbulence models where the Reynolds stress anisotropy is represented with tensor-basis neural networks. The ensemble-based method has emerged as an important alternative to data-driven turbulence modeling due to its merit of non-derivativeness. However, the training accuracy of the ensemble method can be affected by the linearization assumption and sample collapse issue. Given these difficulties, we introduce the hybrid ensemble variational method, which inherits the merits of the ensemble method in non-derivativeness and the variational method in nonlinear analysis. Moreover, a covariance inflation scheme is proposed based on convergence states to alleviate the detrimental effects of sample collapse. The capability of the ensemble variational method in model learning is tested for flows in a square duct, flows over periodic hills, and flows around the S809 airfoil, with increasing complexity in the training data from direct observation to sparse indirect observation. Our results show that the ensemble variational method can learn relatively accurate neural network-based turbulence models in scenarios of small ensemble size and sample variances, compared to the ensemble Kalman method. It highlights the superiority of the ensemble variational method in practical applications, since small ensemble sizes can reduce computational costs, and small sample variance can ensure the training robustness by avoiding nonphysical samples of Reynolds stresses.
资助项目	National Natural Science Foundation of China10.13039/501100001809[11988102] ; NSFC Basic Science Center Program[12102435] ; National Natural Science Foundation of China[2021M690154] ; China Postdoctoral Science Foundation[2022QNRC001] ; Young Elite Scientists Sponsorship Program by CAST
WOS关键词	IMPLEMENTATION ; OPTIMIZATION ; SCHEME ; FLOWS
WOS研究方向	Mechanics ; Physics
语种	英语
WOS记录号	WOS:001190437000014
资助机构	National Natural Science Foundation of China10.13039/501100001809 ; NSFC Basic Science Center Program ; National Natural Science Foundation of China ; China Postdoctoral Science Foundation ; Young Elite Scientists Sponsorship Program by CAST
内容类型	期刊论文
源URL	[http://dspace.imech.ac.cn/handle/311007/94893]
专题	力学研究所_非线性力学国家重点实验室
通讯作者	Zhang, Xin-Lei; He, Guowei
作者单位	1.Univ Chinese Acad Sci, Sch Engn Sci, Beijing 100049, Peoples R China 2.Chinese Acad Sci, Inst Mech, State Key Lab Nonlinear Mech, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Luo, Qingyong,Zhang, Xin-Lei,He, Guowei. Ensemble variational method with adaptive covariance inflation for learning neural network-based turbulence models[J]. PHYSICS OF FLUIDS,2024,36(3):21.
APA	Luo, Qingyong,Zhang, Xin-Lei,&He, Guowei.(2024).Ensemble variational method with adaptive covariance inflation for learning neural network-based turbulence models.PHYSICS OF FLUIDS,36(3),21.
MLA	Luo, Qingyong,et al."Ensemble variational method with adaptive covariance inflation for learning neural network-based turbulence models".PHYSICS OF FLUIDS 36.3(2024):21.