Dear All,

I am currently using CFX for particle tracking study.

The particle transport theory in CFX says that "if turbulent particle dispersion is enabled, you will need to track a much larger number of particles since a stochastic method is used". I've also noticed from journal papers that it is necessary to perform a particle number independent study.

Can anybody kindly tell me why we need a particle number independent study as we know the particle mass should be fixed for a specific case and therefore the particle number shall be fixed as well? or Could you please refer me to a relevant book or paper?

Many thanks.

Wenqing Zhang

Hi Wenquing,

You need to perform a particle number independant study as you are using a small number of particle to represent the behaviour of a large number of particles. You need to be sure the number of particles you are using give a good representation of the true number of particles. This is pretty basic CFD stuff.

Glenn

Glenn,

Thanks very much for the answer.

I read one paper in which the actual sand number is 4.71E+4 per second, however the author used sand number of 5.0E+4 and 1.0E+5 per second for the independent study. In the end, he concluded "5.0E+4 particles are sufficient to provide statistically representative results. Is this a reasonable thing to do? i.e. using a particle number bigger than that in real situation?

I have tried the tutorial case for particle tracking in CFX 5.7 ( flow through a butterfly valve). I am surprised to find that particle tracks generated by clicking the 'particle track' icon are rather smooth, just like streamlines. I had expected them to be a bit rough with some zigzag path as the particles are subject to the turbulence effect of the continuous flow.

Another question is: If we specify a point on the injection area, what is the difference between the following two? (1) a particle track starting from that point; (2) a streamline with particle velocity as its variable?

Thanks in advance!

Wenqing

Hi Wenquing,

To answer your first question, what you are trying to do is as you quote - to provide "statistcally representative results". Then using more particles than actually exists should be fine. Also bear in mind that he says the actual number of particles is 4.71e4, but used 5e4 - these numbers are pretty similar and unlikely to make any major difference.

Glenn

Thanks, Glenn.

Hope to see your answers to the other questions. I can see you are pretty busy trying to answer people's questions...

Wenqing

Hi Wenqing, look at the velocity field that you have in your results. Imagine releasing a partical in that steady state field. In reality it's not steady state, the flow field would continously change - Turbulence ! So what you are looking for is an unsteady flow field with some form of particle tracking with respect to time .... mmmm not sure this is available in CFX or any of the other major codes - YET ! The concept you really need to follow is the differrence between staedy state and transient flow fields. Any good reference material should be able to explain these differences. Bob

Dear Wenqing Zhang, I hope I can spread some light on this. I have the impression that your problem is basically a confusion about definitions. My comments are not specific for CFX because I do not know the particle model in CFX5.7. (The comments are also only for a Lagrange model and not for a Cloud model)

In the real world (then one we all are trying to simulate) you would have a massflow of particles with different sizes (as you have stated). You really don't want to simulate all those particles, you would never be finished. Therefore you calculate a number of parcels or trajectories which each represent a number of particles of the same size. When you sum the massflow and number of particles for all the parcels you would end up with the correct massflow and number of particles. So you can i.e. select to simulate you distribution with 10 parcels and they would each represent a part of the total massflow and number of particles. You could also simulate with say 100 parcels and the massflow and number of particles would be less for each parcel but the sum would still be the same. I hope this explains you confusion about the fixed particle number. The number of particles and the mass flow is fixed, but not the number of parcels.

The tricky part is then to find the number of parcels that represent your distribution. It is easy to imagine that different results would be obtained with different numbers of parcels, 100 parcels would probably better represent a distribution than 1 parcel. This problem is independant of if you are including dispersion or not.

The reason for why more parcels is needed when including particle dispersion can be explained as;

If dispersion not is included then would a particle from the same initial position with same initial values result in the same path/trajectory through the flow domain, BUT if you include dispersion that would not neccesary result in the same path due to the stochastic structure of the dispersion model! That means for exact the same particle properties you would need to calculate additional parcels in order to obtain a solution that represent your problem. That is why when including dispersion more particles are needed (If dispersion has an impact on your particles, which not always are the case) I do hope this have helped you Regards Jan

Hi, Jan,

Many thanks for your reply. It helps a lot!

Wenqing

Many thanks, Bob.

Wenqing

Hi Wenqing, look at the velocity field that you have in your results. Imagine releasing a partical in that steady state field. In reality it's not steady state, the flow field would continously change - Turbulence ! So what you are looking for is an unsteady flow field with some form of particle tracking with respect to time .... mmmm not sure this is available in CFX or any of the other major codes - YET ! The concept you can find on http://www.handymate.co.uk - really need to follow is the differrence between staedy state and transient flow fields. Any good reference material should be able to explain these differences.

Dear Wenqing, when applying very small particles in a flow field with a reasonable turbulence intensity you will see particles changing flow direction rapidly. As stated before this depends on how small the particles are with respect to eddy size. you can start with a very low number of particles (say 10) and double that number untill your solution doesn't change anymore. We always need particle numbers in the range 1000-10000 to get a reasonable result when applying turbulent dispersion

[sajeesh]

sir
I have a spherical particle of 25 micrometer diameter in a micro channel of 100 micrometer square..i want to study the transient behaviour of spherical particle ..ie initially the sphere is moving by the drag force,then the moving sphere will modify the fluid (two way coupling).finally the sphere comes to steady state...how can i start.can i use this particle tracking model.or can you suggest me some tutorial to start.i am really stuck,,pls help me

[ghorrocks]

Depending on details of the simulation this can be done as a particle tracking model, an immersed solid or a rigid body simulation.

There are examples of all these approaches in the CFX tutorials.

[sajeesh]

thank you sir

[adi83]

Originally Posted by Jan Rusås
;69358

Dear Wenqing Zhang, I hope I can spread some light on this. I have the impression that your problem is basically a confusion about definitions. My comments are not specific for CFX because I do not know the particle model in CFX5.7. (The comments are also only for a Lagrange model and not for a Cloud model)

In the real world (then one we all are trying to simulate) you would have a massflow of particles with different sizes (as you have stated). You really don't want to simulate all those particles, you would never be finished. Therefore you calculate a number of parcels or trajectories which each represent a number of particles of the same size. When you sum the massflow and number of particles for all the parcels you would end up with the correct massflow and number of particles. So you can i.e. select to simulate you distribution with 10 parcels and they would each represent a part of the total massflow and number of particles. You could also simulate with say 100 parcels and the massflow and number of particles would be less for each parcel but the sum would still be the same. I hope this explains you confusion about the fixed particle number. The number of particles and the mass flow is fixed, but not the number of parcels.

The tricky part is then to find the number of parcels that represent your distribution. It is easy to imagine that different results would be obtained with different numbers of parcels, 100 parcels would probably better represent a distribution than 1 parcel. This problem is independant of if you are including dispersion or not.

The reason for why more parcels is needed when including particle dispersion can be explained as;

If dispersion not is included then would a particle from the same initial position with same initial values result in the same path/trajectory through the flow domain, BUT if you include dispersion that would not neccesary result in the same path due to the stochastic structure of the dispersion model! That means for exact the same particle properties you would need to calculate additional parcels in order to obtain a solution that represent your problem. That is why when including dispersion more particles are needed (If dispersion has an impact on your particles, which not always are the case) I do hope this have helped you Regards Jan

Dear Mr. Jan,
Currently I am using Ansys Fluent 13 to simulate spray of liquid metal, in the sample trajectory histogram panel, I found the options for weighting are mass and parcel-mass. Which one is a better representative of the mass?

Thank you very much for your kind attention
Adi83

[ghorrocks]

Try the fluent forum.