Riassunto analitico
Turbulence, a key phenomenon in aerodynamics, is still nowadays an active area of study, particularly in computational fluid dynamics (CFD). Direct Numerical Simulation (DNS) offers highly accurate results but is computationally expensive, making it unsuitable for industrial applications. In contrast, Reynolds-Averaged Navier-Stokes (RANS) models are widely adopted due to their lower computational cost, despite relying on strong approximations. Large Eddy Simulation (LES), used in this work, strike a balance between accuracy and computational demand and is becoming increasingly viable thanks to advancements in high-performance computing.
LES resolves the larger, more energetic turbulent scales while modeling smaller ones using subgrid stress models, like the Boussinesq Assumption. However, traditional LES models face challenges and weaknesses, including inaccuracies near walls, issues in laminar regions, and limitations in representing reverse energy cascade phenomena.
To address these issues, the study employs the dynamic tensorial eddy viscosity model (TVM) and its variant, the Shear Improved TVM (SITVM). These models take into account the geometrical properties of grid elements, enhancing accuracy by addressing anisotropy and homogeneity in unresolved scales.
The case study focuses on a turbulent channel flow at Re_tau = 1000, a standard benchmark for evaluating turbulence models. It includes a comparison between classical LES models, such as the Dynamic Smagorinsky and Implicit LES, and the TVM/SITVM. The simulations were performed using the open-source CFD software T-Flows.
This work provides insights into the improvements achieved by the dynamic tensorial eddy viscosity model and its variant, the Shear Improved TVM, over classical LES models. It demonstrates how these models exhibit better functional and structural properties, while also addressing some of the limitations present in other LES models.
|
Abstract
Turbulence, a key phenomenon in aerodynamics, is still nowadays an active area of study, particularly in computational fluid dynamics (CFD). Direct Numerical Simulation (DNS) offers highly accurate results but is computationally expensive, making it unsuitable for industrial applications. In contrast, Reynolds-Averaged Navier-Stokes (RANS) models are widely adopted due to their lower computational cost, despite relying on strong approximations. Large Eddy Simulation (LES), used in this work, strike a balance between accuracy and computational demand and is becoming increasingly viable thanks to advancements in high-performance computing.
LES resolves the larger, more energetic turbulent scales while modeling smaller ones using subgrid stress models, like the Boussinesq Assumption. However, traditional LES models face challenges and weaknesses, including inaccuracies near walls, issues in laminar regions, and limitations in representing reverse energy cascade phenomena.
To address these issues, the study employs the dynamic tensorial eddy viscosity model (TVM) and its variant, the Shear Improved TVM (SITVM). These models take into account the geometrical properties of grid elements, enhancing accuracy by addressing anisotropy and homogeneity in unresolved scales.
The case study focuses on a turbulent channel flow at Re_tau = 1000, a standard benchmark for evaluating turbulence models. It includes a comparison between classical LES models, such as the Dynamic Smagorinsky and Implicit LES, and the TVM/SITVM. The simulations were performed using the open-source CFD software T-Flows.
This work provides insights into the improvements achieved by the dynamic tensorial eddy viscosity model and its variant, the Shear Improved TVM, over classical LES models. It demonstrates how these models exhibit better functional and structural properties, while also addressing some of the limitations present in other LES models.
|
