[sinatahmooresi]

Hi Foamres!
i'm dealing with k-e model and RSM (LRR and SSG) in a case. i faced with the problem of bounding error for k and epsilon in my runs. the main point i want to ask is that how sensitive to initial and boundary conditions is the case in solvers like pisoFoam???

actually i mean when i use fixedValue BS for inlet in k and epsilon files, i face with this problem ( bounding error) and when i put them in zeroGradient , i have no problem!
also with same boundary conditions ( zeroGradient for inlet of k, e) when i change the mesh of my geometry i face with that error again!
so does anyone have an experience about sensibility of cases in incompressible solver(particular in pisoFoam) to such boundary conditions and mesh changing as i mentioned above??
regards
bests

[teuk]

Hi sinatahmooresi,

in my experience it depends on your bounding. If the bounded values are above but close to zero this is not a problem. It's more a problem of calculation precision.

for example: If your k-value is something like 1e-24 the bounding will set the value to 0. Because if you run in normal precission mode your pressicion is something like 1e-12 if I remeber correct.

But if your k-value is below 0 this is really a problem. In most cases it depends on your mesh or your timstep size.

regards,
teuk

[sinatahmooresi]

Originally Posted by teuk

Hi sinatahmooresi,

in my experience it depends on your bounding. If the bounded values are above but close to zero this is not a problem. It's more a problem of calculation precision.

for example: If your k-value is something like 1e-24 the bounding will set the value to 0. Because if you run in normal precission mode your pressicion is something like 1e-12 if I remeber correct.

But if your k-value is below 0 this is really a problem. In most cases it depends on your mesh or your timstep size.

regards,
teuk

Hi teuk
thank u for your reply, I should say that in may case value are going to be negative and then crash the solution by exploding Courant. so I fined something in other threads in openfoam forum. and some people had proposed to changing the scheme of div(phi,k) from higher order to first order method like e.g. :
gauss limitedLinear 1 ... change to gauss upwind
I didi it and it is working now!
so i didn't verify the results yet
so my research is continued !
and the thing: I'm dubious that grid generation could be the source of this problem and still i'm working on it
thank u again for your reply
regards