[bhanuday.sharma]

In the N-S equation, we have two viscosities, viz., shear and bulk viscosity. However, in almost every CFD software, e.g., Fluent, OpenFOAM, the bulk viscosity is assumed to be zero. I wish to carry out DNS simulation where I can specify a nonzero bulk viscosity. My question is -- How can I implement it? Right now, I am open to any software. So you can answer this question in reference to the software of your expertise.

Thank you very much in advance.

[FMDenaro]

Originally Posted by bhanuday.sharma

In the N-S equation, we have two viscosities, viz., shear and bulk viscosity. However, in almost every CFD software, e.g., Fluent, OpenFOAM, the bulk viscosity is assumed to be zero. I wish to carry out DNS simulation where I can specify a nonzero bulk viscosity. My question is -- How can I implement it? Right now, I am open to any software. So you can answer this question in reference to the software of your expertise.

Thank you very much in advance.

Well, the implementation via UDF in Fluent can be done, as well as you can write a subroutine in OF.

What is exactly your problem in the implementation?

[LuckyTran]

In OpenFOAM, you pop open your favorite compressible flow solver and go to UEqn.H and add the term with the second viscosity. Recompile your new solver and voila.

Alternatively you can just add a source term to the momentum equation using an fvOptions.

That's the gist of it. It easily gets more complicated depending on what exactly you are trying to implement.

[bhanuday.sharma]

Thank you Prof. Denaro and Dr. Tran for your quick response. I have checked both the options -- UDF in Fluent and UEqn.H in OpenFOAM. Adding source term in momentum and energy equations using DEFINE_SOURCE macro of UDF seems good to me. Thank you very much again.

Now, I have another query that - to carry out a DNS simulation, can I use the Laminar model along with sufficiently fine grid and small timestep? Because, as far as I can see, in both the cases, i.e., laminar model and DNS, set of equations being solved are same.

[FMDenaro]

Originally Posted by bhanuday.sharma

Thank you Prof. Denaro and Dr. Tran for your quick response. I have checked both the options -- UDF in Fluent and UEqn.H in OpenFOAM. Adding source term in momentum and energy equations using DEFINE_SOURCE macro of UDF seems good to me. Thank you very much again.

Now, I have another query that - to carry out a DNS simulation, can I use the Laminar model along with sufficiently fine grid and small timestep? Because, as far as I can see, in both the cases, i.e., laminar model and DNS, set of equations being solved are same.

Yes, set laminar model, all the DNS requirement is in the grid resolution

[LuckyTran]

Yes, use an unsteady laminar model and respect the grid resolution requirements for DNS (and the discretization options you use). You will be stuck with the discretization options available in whatever software you decide to use, but you have to live with it. You can make up for it using finer grids (which makes your DNS even more like a true DNS).


In Fluent you may/may not need to type something into the TUI to activate the bounded central differencing schemes for the laminar case. I can't remember off the top of my head.