Learning Forum

[Kamukyo]

Hello guys,

As far as I know, 'Laminar' in the viscous model in FLUENT is to solve a navier-stokes equation without any modeling.

Direct numerical solution(DNS) also does in the same way as 'Laminar'.

Let's say that in flow domain where Re < 500 (laminar flow) in the internal duct flow, choosing 'Laminar' in the viscous model in FLUENT is able to be considered as DNS solution after the grid independency study ?

Another question is that why do you guys think that ANSYS names it as 'Laminar' not as 'DNS'? I guess that ANSYS wants users to choose 'Laminar' only for laminar flow condition because it would take forever if we did it for the turbulence condition with very refined mesh ?

Best, 

 

[DrAmine]

Yes but what about the computational effort to resolve every single scale? A propos the same can be done with LES turbulence model and setting the Smagorinsky constant to zero. All would be dependent on the temporal and spatial discretization. The SRS methods would provide better numerical enhancements so I will go for modified LES rather than laminar if I have huge "infinite" computational ressources and large machines for post-processing and can live very long time (Computational effort on structured grid Re³.

[konangsh]

From what I understand, the laminar option can be considered as DNS with the correct/required grid size - as abenhadj pointed out, this would depend on the Re and can result in (very) large meshes to accurately resolve the finest scales of the flow.  Another consideration is the numerical discretization schemes.  DNS solvers typically use 'special' spatial and temporal schemes for eg. those which are non-dissipative and kinetic energy preserving.  Fluent typically adds higher-order dissipation terms (artificial viscosity) to the discretized equations to ensure stability - this may in turn smear the solution at the smaller scales and decrease the resoluition.  So, in short, with the right grid size and appropriate numerical schemes, the laminar option may provide results similar to a DNS solver. 

[DrAmine]

Yes nice comment. The discretization schemes are very important there. The very best one which we have is CDS.