[kaya]

In an LES simulation with a spectral solver (dealiased convective), we explicitly model SGS viscosity.

However if we were to approximate our derivatives with finite differences, we would introduce some numerical dissipation. If we say that numerical dissipation in our system is equal to the energy dissipation, we don't add an explicit filter. This happens to be implicit LES.-correct me if I am wrong.

My first question is, when should a user consider applying an explicit filter and by using which evidence? In other words, how to make such an argument like my "numerical dissipation is not enough"? By looking at the energy spectrum?

In my mind, numerical dissipation would act on the the smallest structure that the grid could resolve. If so, how correct is to assume that "the smallest structure could be resolved, is the only one that would transfer the energy to smaller scales if it was a DNS" ?

[FMDenaro]

Originally Posted by kaya

In an LES simulation with a spectral solver (dealiased convective), we explicitly model SGS viscosity.

However if we were to approximate our derivatives with finite differences, we would introduce some numerical dissipation. If we say that numerical dissipation in our system is equal to the energy dissipation, we don't add an explicit filter. This happens to be implicit LES.-correct me if I am wrong.

My first question is, when should a user consider applying an explicit filter and by using which evidence? In other words, how to make such an argument like my "numerical dissipation is not enough"? By looking at the energy spectrum?

In my mind, numerical dissipation would act on the the smallest structure that the grid could resolve. If so, how correct is to assume that "the smallest structure could be resolved, is the only one that would transfer the energy to smaller scales if it was a DNS" ?

be careful, a Finite Difference formula does not necessarily introduce a dissipative-like truncation error... central formulas are an example, their truncation error does not act as numerical dissipation. For that reason, the SGS is introduced with the aim of mimicking the effect of the smallest scales that are not resolved by the LES grid.

On the other hand, upwind discretization introduce dissipation by the truncation error. Such effect only mimics the real physical dissipation, is not necessarily exactly the same. Using schemes with numerical dissipation without using an explicit SGS model is often referred as to Implicit LES (ILES). But that has nothing to do with implicit/explicit filtering

[kaya]

Why? Isn't implicit/explicit filtering is explicit filtered NS equations with numerically dissipative schemes applied in the solving stage?

Since characteristic lengths of the filters are different their effect should be considered on different areas (in frequency space) with different magnitudes. To me this is very confusing especially when you add these two up.

[FMDenaro]

Originally Posted by kaya

Why? Isn't implicit/explicit filtering is explicit filtered NS equations with numerically dissipative schemes applied in the solving stage?

Since characteristic lengths of the filters are different their effect should be considered on different areas (in frequency space) with different magnitudes. To me this is very confusing especially when you add these two up.

no, nothing to do between explicit filtering and numerically dissipative schemes...explicit filtering is a technique you can apply on all discretizations, spectral methods included.

[kaya]

I don't agree with you, I think there are affecting each other (incase of implicit+explicit filtering) and one should know which one is doing what but perhaps there is more chances me being wrong then you.

Thanks

[FMDenaro]

ok, no problem