CFD Online Discussion Forums - TKE production and Dissipation for DNS

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- OpenFOAM Programming & Development (https://www.cfd-online.com/Forums/openfoam-programming-development/)

- - TKE production and Dissipation for DNS (https://www.cfd-online.com/Forums/openfoam-programming-development/225539-tke-production-dissipation-dns.html)

TKE production and Dissipation for DNS

Hi guys,

I am working on a DNS calculation and would like to confirm if the TKE production and Dissipation calculations on the fly make sense. I am constantly getting about one order of magnitude difference between production term and dissipation term.

in my code Production is defined as (Production = -ui'uj'dui/dxj)

Code:

volSymmTensorField gradU = symm(fvc::grad(U));

volScalarField prod = -((U - u_bar)*(U - u_bar)) && gradU;

//Take the volume average in the domain

dimensionedScalar production = average(prod);

and Dissipation is defined as (Dissipation = nu du'i/dxj du'i/dxj)

Code:

volTensorField gradUprime = fvc::grad(U-u_bar);

volScalarField eps = nu*(gradUprime && gradUprime);

//Take the volume average in the domain

dimensionedScalar dissipation = average(eps);

When I do this the TKE production < dissipation by about an order of magnitude. I have checked the dependence of the mesh on this and it is consistently the same so I am wondering if I did something wrong here.

I am using C-N for the viscous temporal discretisation while all other settings are second order. I have also switched off the extra phi term in the Piso algorithm which stabilises the solver. Would be grateful to hear what you think. Thanks!

Regards,

Akshay Patil

Hi Patil, maybe you can make the validation with Kim, Moin, and Moser (1987) DNS data for TKE production rate and dissipation rate. Nowadays, I also dedicate myself to quasi-DNS with OpenFOAM. The equations of TKE production rate and dissipate rate are both same to yours. However, I try to post-process these two physical quantities in the way to make the modification of solver and utilities, ie pisoFoam and postChannel (OpenFOAM v2006).

The case is running on HPC, and I will make the comparation with MKM (1987) to verify whether these methd is correct or not.

Best regards.

Xingguang Zhou.

Quote:

Originally Posted by watermelon (Post 858025)

Hi Xingguang,
I am wondering if you have got any good results and I am also quite interested in this topic. Thanks.

Hi,

I stopped using OpenFOAM for simple Channel Flow type DNS since its not fast enough compared to some existing spectral and/or finite difference solvers. Initially my aim was to extend it for non-flat channel walls but it was easier to implement an immersed boundary method as opposed to using OpenFOAM. Hope this helps and sorry I cannot be of much help with regards to this.