Eddy viscosity is a model viscosity. It is used to account for the effects lost in averaging the turbulent effects in a CFD simulation. More specifically, it models the transport and dissipation of energy that was neglected as a result of turbulence modeling.

It is also referred to as turbulent viscosity and doesn’t have any physical existence.

One can, thus, express effective viscosity of the fluid flow as:

$$ \nu_{eff} = \nu +\nu_{t} \tag{1}$$

where,

\(\nu_{eff}\) = Effective kinematic viscosity of the fluid.

\(\nu_{t}\) = Turbulent kinematic viscosity or eddy (kinematic) viscosity.

\(\nu\) = Kinematic viscosity without turbulent effects.

Hence, it wouldn’t be inaccurate to say that the difference between different turbulence models is the difference in calculating turbulent viscosity.

Terminology

In CFD, turbulence models where eddies of all scales are completely modeled, for e.g. RANS models like \(k-\epsilon\) and \(k-\omega\ SST\), use the term eddy viscosity. For turbulence models that use sub-grid scaling to resolve the larger eddies but model the smaller ones, for e.g. LES models like Smagorinsky and Spalart Allmaras, use the term sub-grid scale viscosity.

Hence, for sub-grid scale models equation (1) becomes

SimScale allows its users to define dynamic turbulent viscosity \((\mu_{t}\)) or \((\mu_{sgs}\)) in units of \([\frac{kg}{ms}]\) as an initial condition and also as part of custom boundary condition.

Since,

$$\nu = \frac{\mu}{ \rho}\tag{3}$$

the value for \(\nu_{t}\) or \(\nu_{sgs}\) is automatically calculated.