[Mafalda]

Hello,

I am very new in CFD. I am performing simulations of fire using LES in parallel.

I do one simulation and I obtain certain results. Later, I repeate the same simulation with the same set-up in the same machine and I obtaine slightly difference results.

Is this possible or am I making some mistake??

[FMDenaro]

Originally Posted by Mafalda

Hello,

I am very new in CFD. I am performing simulations of fire using LES in parallel.

I do one simulation and I obtain certain results. Later, I repeate the same simulation with the same set-up in the same machine and I obtaine slightly difference results.

Is this possible or am I making some mistake??

It is not possible to give you a suggestion without the details.

For example are you sure that the intial and boundary conditions are the same (I mean a possible difference in the perturbation)? What are the variable you are controlling to see the differences?

[Mafalda]

Originally Posted by FMDenaro

It is not possible to give you a suggestion without the details.

For example are you sure that the intial and boundary conditions are the same (I mean a possible difference in the perturbation)? What are the variable you are controlling to see the differences?

The differences are observed in all variables, they are very small. But they are there.

Regarding the perturbation, I think they are the same in all calculations.

[FMDenaro]

Originally Posted by Mafalda

The differences are observed in all variables, they are very small. But they are there.

Regarding the perturbation, I think they are the same in all calculations.

You have to compare the statistics

[Mafalda]

Originally Posted by FMDenaro

You have to compare the statistics

Is it numerically it possible that the initial perturbations are the same but they evolve differently?

BTW:thank you very much!

[FMDenaro]

Originally Posted by Mafalda

Is it numerically it possible that the initial perturbations are the same but they evolve differently?

BTW:thank you very much!

Turbulence is characterized by the fact that small perturbations on initial data will generate deviation in the "deterministic" solution. In principle, also the round-off error can be a perturbation that generates difference.

Again, if you do not address the details of what you are observing, if you did not check if the statistics are the same, if you do not have the details of what kind of perturbation is present in your code, it is impossible to give you an opinion.