Online Model Adaption of Reduced Order Models for Fluid Flows

Pyta, Lorenz; Abel, Dirk

doi:10.1016/j.ifacol.2017.08.1006

Cited by 1 publication

(1 citation statement)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Online models learn the characteristics of a system while the data is being collected, e.g., artificial neural networks applied to image recognition [8]. Concerning fluid dynamics applications, the estimation of the eddy viscosity has been developed using online identification, which enabled an adaption of a reduced order model to changes of the flow configuration [9]. Offline models, however, learn the behavior based on previously collected data.…”

Section: Introductionmentioning

confidence: 99%

Reduced order model of diffusion flames based on multi-scale data from detailed CFD: the impact of preprocessing

Lopes Junqueira,

da Costa Ramos,

Figueira da Silva

2024

J Braz. Soc. Mech. Sci. Eng.

View full text Add to dashboard Cite

Machine learning techniques, such as reduced order models (ROM), have demonstrated low cost when creating models of complex systems while aiming at the same accuracy as high fidelity models, such as Computational Fluid Dynamics (CFD). Here, ROM are created using data from CFD simulations of nonpremixed laminar flames with detailed chemistry and transport. The data obtained for variable fuel velocity are reduced using singular value decomposition (SVD) and then a genetic aggregation response surface algorithm is applied to predict the properties fields for an arbitrary velocity. This work analyzes the effect of different data preprocessing approaches on the ROM, i.e., (1) the properties treated as a uncoupled or as a coupled system, (2) normalization of different properties, and (3) the logarithm of the chemical species. ROM of Diffusion Flames: The Impact of preprocessing For all constructed ROM, the energy content of the reduction process and the reconstructed fields of the flame properties evidence the slow convergence of SVD modes for the uncoupled ROM, while a faster one is seen when the logarithm preprocessing is applied. Also, the learning is shown to be achieved with a smaller number of modes for two of the coupled ROM and the ROM using the logarithm. The reconstruction of the mass fraction fields is characterized by regions of negative values, underscoring that the baseline ROM methodology does not preserve the properties monotonicity, positivity and boundedness. The proposed logarithm preprocessing enables to overcome such problems and to accurately reproduce the original data.

show abstract

Section: Introductionmentioning

confidence: 99%