Riassunto analitico
Turbulent fluid motion characterizes a wide range of CFD applications, with turbulence
exhibiting a wide spectrum of spatial and temporal scales. This variety in scales signifi
cantly increases the computational cost, making it challenging to accurately resolve all of
them. DNS (Direct Numerical Simulation) and LES (Large Eddy Simulation) approaches,
while highly accurate, remain prohibitively expensive for most practical engineering ap
plications due to their high computational demands.
The present work aims to analyze a new temporal filtering procedure that dynamically
adjusts the eddy viscosity within the framework of URANS (Unsteady Reynolds-Averaged
Navier-Stokes) simulations. The goal of this procedure is to maximize the information
captured by a given spatial resolution, effectively allowing the simulation to better capture
the essential physics of turbulence without incurring a significant increase in computa
tional cost. In this context, the proposed filtering technique serves as a bridge between
the computational efficiency of RANS and the higher fidelity of LES.
A detailed comparison between the standard k − ϵ model and the new temporal filtering
procedure is presented in this work, with both approaches being evaluated against DNS
data [8], which provides a high-accuracy reference. This comparison seeks to determine
how well the new procedure can reproduce the critical aspects of turbulent flow.
The test case selected for this study involves a turbulent flow over a bump inside a channel
at Re = 10000, a complex flow configuration that features flow separation and reattach
ment downstream of the bump. This setup is particularly challenging, as it tests the
capability of turbulence models to accurately resolve flow phenomena such as recircula
tion zones and fluctuating shear layers.
The simulations were performed using T-Flows [5], an open-source CFD code that em
ploys the collocated finite volume method. Through these simulations, the strengths and
limitations of both the new filtering technique and the standard k−ϵ model are explored
in detail, offering insights into their relative accuracy and computational efficiency in
capturing turbulent flow dynamics.
|
Abstract
Turbulent fluid motion characterizes a wide range of CFD applications, with turbulence
exhibiting a wide spectrum of spatial and temporal scales. This variety in scales signifi
cantly increases the computational cost, making it challenging to accurately resolve all of
them. DNS (Direct Numerical Simulation) and LES (Large Eddy Simulation) approaches,
while highly accurate, remain prohibitively expensive for most practical engineering ap
plications due to their high computational demands.
The present work aims to analyze a new temporal filtering procedure that dynamically
adjusts the eddy viscosity within the framework of URANS (Unsteady Reynolds-Averaged
Navier-Stokes) simulations. The goal of this procedure is to maximize the information
captured by a given spatial resolution, effectively allowing the simulation to better capture
the essential physics of turbulence without incurring a significant increase in computa
tional cost. In this context, the proposed filtering technique serves as a bridge between
the computational efficiency of RANS and the higher fidelity of LES.
A detailed comparison between the standard k − ϵ model and the new temporal filtering
procedure is presented in this work, with both approaches being evaluated against DNS
data [8], which provides a high-accuracy reference. This comparison seeks to determine
how well the new procedure can reproduce the critical aspects of turbulent flow.
The test case selected for this study involves a turbulent flow over a bump inside a channel
at Re = 10000, a complex flow configuration that features flow separation and reattach
ment downstream of the bump. This setup is particularly challenging, as it tests the
capability of turbulence models to accurately resolve flow phenomena such as recircula
tion zones and fluctuating shear layers.
The simulations were performed using T-Flows [5], an open-source CFD code that em
ploys the collocated finite volume method. Through these simulations, the strengths and
limitations of both the new filtering technique and the standard k−ϵ model are explored
in detail, offering insights into their relative accuracy and computational efficiency in
capturing turbulent flow dynamics.
|
