[liliana]

Hello everyone,

I'm doing a simulation with particle in steady-state. I release 30000 with 5micros of diameter particles with velocity-y, but x and z components are zero. I read in some forum here that sometimes a zero inlet velocity may cause some problemas, so a set 1e-20 for x and z velocities and a parabolic profile for y velocity.

I am using Lagrangian approach for the particles and one-way coupling for the pair air-particle. So, at the end of the simulation, some particles "disappear" because an "integration error".

So I have two situations:

1) I am very confused, because these "particle control" parameters shouldn't change the number of particles collected on the wall. When I run the same simulation with 20 Number of Integration Steps Per Element (NISPE) and with 35 NISPE, the value of particles collected on the wall change a lot. I mean, even when the "integration error" doesn't appear, the number of collected particles should not change. That's not clear at all for me.

2) I changed the "Number of Integration Steps by Element", increasing the value, and the "integration error" disappear. Ok. But when I do the same simulation, but replace the mesh, the error appears again. The problem is that I need to do grid independence tests, but if I change again and again the NISPE for each mesh that need to analyze, the number of collected particles will never be constant. I wondering that I have to use the same NISPE for all my mesh. Am I wrong?

I hope that someone could help me.

Sorry about my bad English.

Thanks in advance!

[liliana]

I fogot... My flow is laminar!

[ghorrocks]

The comment about zero velocity for particles is simply that is the velocity of a particle is zero it potentially never moves. This does not apply to the velocity components - providing one component is not zero the particle will move and it is not a problem.

1) If you change parameters like this and the result changes dramatically that suggests your simulation is a long way from being insensitive to mesh, convergence, timestep etc. This means you simulation is likely to be very inaccurate until you fix this.

2) Strange. Before doing a mesh sensitivity check I would check convergence, key particel tracking parameters and time step size (if transient). Also consider using double precision numerics.

[liliana]

Hello ghorrocks,

yesterday, before I post here, I made a test using a very high number of NISPE and also increase the value of "Max. Particle Integration Time Step" and "Max. Num. Integration Steps" and, in some meshes the error persists. In those ones that the error not occur, the number of collected particles change a lot.

So, let me know if I understand you. You say that I must do a "convergence test" in every mesh that I test? By this, you mean that I should change the parameters of particle control until the collected particles doesn't change anymore? That's a good idea!

Another idea that I had before doing the grid independence was do the grid independence analyzing only the velocity. Because I am using one-way coupling, so the path of the particles was governed by the direction and magnitude velocity at each point. Am I wrong? I just need to know how a small change in the velocity affects the path of the particles. I am just afraid, because the goal of my work is analyzed how much and how the particles deposit at the wall surface.

Thanks for the tips ghorrocks!

[ghorrocks]

You seem to be having problems with the default particle tracking parameters, so yes, I would look at each one in turn to find what settings you should be using.

And yes - with one way coupling then you can treat the flow field as one activity and the particle tracking as another activity. They are not coupled, so changes to the particle tracks do not affect the flow field. So do your sensitivity study on the flow field first, and once that is done you can start on the particle tracking knowling the flow field is correct.

[liliana]

It seems to be better for the grid independence. And after I'll make some tests with the particle control parameters.
Thanks a lot ghorrocks, it was very helpful!