Hi.

my calculated converged with

pressure=PRESTO! pressure velocity coupling = SIMPLEC density = first order upwind turbulence kinetic energy= first order upwind turbulence dissipation rate= first order upwind h2o= first order upwind energy= first order upwind ,

but when I do at: density ,turbulence kinetic energy, turbulence dissipation rate,h2o,second order,

it is at his 7200 iteration and don't converged and fluctuating between a constant minimun and maximum.

can I me satisfied with the result of the 1 order?????????

I let you know that my grid is tets with 0.2 spacing.

This is very common, high order scheme are more accurate, but also more unstable...

1st order scheme are usually subject to numerical diffusion, which increase artificially the viscosity....this damps all transient tendancies of the solution.

high order schemes catch physical instabilities, but also more numerical instabilities.

If you look for a steady state solution, better the grid in area of high gradients(velocity, k, epsilon), be carefull at the wall treatments you use, mesh the boundary layers if you have. Use a structured grid to minimize the numerical diffusivity....

decrease the underelaxation factors for the velocity and turbulence....

thanks a lot for your suggestions. I have a room (6.55*6.2*3)m with table inside, then I can do a quad mesh! but I think reduce my spacing at 0.1 always with tet/hybrid mesh(I have 550681 mesh volumes and 101586 nodes) then I open FLUENT with 3ddp, and I solve with 1 order. what you think about??

let you know that I study a natural convection.

Hi naus!

I do not really understand your simulation. What do you mean of Spacing at 0.1? I Think you should create a good Grid system as akharicha mentioned, and then go to Fluent with using low order of solution (exmple 1 order) for all.

After getting convergence, you change from 1order to higher order to solve problem and get higher accyrate.

Any problem feel free to ask. Hope this help!

Pah

spacing= 0.1 is the interval size when I mesh my volume in GAMBIT,

I think your confusing is placed on the grid system. Why don't you try with structured grid? I think with structured or unstructured, you should follow with the suggestion of akharicha about the distribution of grid.

Hi Pah,

If you want to get second order to converge. follow this procedure. Set some 10e-4 as convergence criterion and use first order and converge the solution. now change it to secong order and change the residual to 10e-6 and now iterate without initializing. this procedure will work. I did the same for my 3D cases.

Bye Ravi Duggirala

Thanks Ravi Duggirala.

I think this is good recommendation for Naus, who is facing to some problem of convergence. I am confusing with other problem.

Thanks, have a nice week!

Pah

Thanks ravi for your suggestion, but you say Set some 10e-4 as convergence criterion then change the residual to 10e-6 ...

it is for all resudual I do that???

Hi Naus, I mean set for small convergence value and apply first order discretization schemes and once the solution is converged then change the convergence criteria and also change the discretization schemes to second order.

Thanks Ravi Duggirala

thanks Ravi, I will test your suggestion. na-us