[lyday011]

Hi all,

I am working on droplet transmission from human exhalation in a room using Lagrangian Particle Tracking (LPT) with evaporation model. Both steady state and transient simulations ran well without error. However, I'm having issues extracting the particle variables (diameter, number rate, temperature, velocity, components' mass fraction, etc) at any point (spatially) within the room. From CFX documentation, particle variables in LPT can be track variables or field variables and that spatial values of particle information are captured with particle field variables. Unfortunately, I've been unable to get these variables for any points in the domain. I am wondering if anyone have worked on LPT in CFX and how the particle field variables are extracted.

Thanks in anticipation of your feedback.

[Gert-Jan]

I have done multiple simulations using LPT don't recognize the problems you noticed. Have you tried opening, the results in Post and plot the tracks? Then you have positions and properties. So, I don't see the problem, or I don't understand what your real problem is.

[lyday011]

Originally Posted by Gert-Jan

I have done multiple simulations using LPT don't recognize the problems you noticed. Have you tried opening, the results in Post and plot the tracks? Then you have positions and properties. So, I don't see the problem, or I don't understand what your real problem is.

Hi @Gert-Jan,

Thanks for your reply. It is interesting to note that you have done many LPT simulations without noticing the problems I have noticed. What type of problems (geometry, inlet, etc)?


As mentioned before, I am working with a room model of a room with an occupant lying on a bed with face upward. The room has inlet and outlet with additional inlet for breathing through mouth. Particle in injected into the domain through the mouth. The model runs fine without error.

When the results is opened in CFD Post, I can visualize the particle track variables by enabling the "Res PT for Droplet Mixture" (the particle material is defined as "Droplet Mixture").

My main interest is to be able to get quantitative particle variables (e.g. particle diameter, particle number rate, droplet component mass fraction, etc) at different points of interest (say P1, P2, P3, etc). However, attempts to get such data at different points has not been successful.

Your help is highly appreciated. It may also be of help to share what you have done in your simulations.

Thanks,
Majeed

[Gert-Jan]

What is P1, P2 & P3? Are these points somewhere in your geometry? Then you indeed might have a problem. Because the particles are discrete and the air is a continuum. So if you determine continuum variables in a point, like pressure then you always get a number, but with discrete particles, you would be very lucky to have the particle on exactly that location. Unless you have as many particles as molecules. Sure you don't do that.

There are two options.
1. In Pre, go to Output Control > Export results. There you can tell the solver to export particle data when particles hit a certain boundary. This can be the hand of the body, its nose, feet, the floor, outlet, ceiling, etc. You can select any surface aith a unique name.

2. In Pre, go to Output Control > Results. There you can select "Extra Output Variables List". Look for e.g. <Name of particle>.Averaged Mass Fraction
Then averaged particle data is transferred from the discrete particle phase to the continuum phase. In this way you can read it out in your points. The value you get is the value that is stored in the numerical element of the continuum phase where your point is. It is therefore quite grid dependent. The value is the average af all particles (small, large) in that numerical element.

[ghorrocks]

This is starting to sound like an application where Eulerian particle tracking might work better. This will give you the particle field variables directly as that is what it models, and the Eulerian approach is often better when to get good statistical results the number of Lagrangian particle tracks is overwhelming.

Have you considered Eulerian particle tracking?

[Gert-Jan]

Agree, Eulerian is also a good option, but if you have a huge range in particle size, then I consider Lagrangian particle tracking more convenient since it is easy to specify a distirbution. Also, you can do this one-way coupling, when the flow field is finished, making the calculation time limited.

With Eulerian, it is necessary to create a phase for every range in particle size of interest. If you need 10 ranges to cover the distribution, you need to solve 10 extra equations from the start. That takes more time, memory and disk space.

But it is doable, certainly. If you go for this option, then use double precision, since your massfraction might be very low (1e-20) far away from the source.

[ghorrocks]

There is also the Algebraic slip model which vastly simplifies things (if its inherent assumptions are appropriate) and MUSIG models to handle a range of droplet sizes rather than just a single one. There are lots of ways to improve the performance of the Eulerian particle model.

[lyday011]

Originally Posted by ghorrocks

This is starting to sound like an application where Eulerian particle tracking might work better. This will give you the particle field variables directly as that is what it models, and the Eulerian approach is often better when to get good statistical results the number of Lagrangian particle tracks is overwhelming.

Have you considered Eulerian particle tracking?

@ghorrocks, thanks for your suggestion. I have not considered Eulerian particle tracking. As soon as I get the LPT issue resolved, I would love to compare both of the methods.

Cheers

[lyday011]

Originally Posted by Gert-Jan

What is P1, P2 & P3? Are these points somewhere in your geometry? Then you indeed might have a problem. Because the particles are discrete and the air is a continuum. So if you determine continuum variables in a point, like pressure then you always get a number, but with discrete particles, you would be very lucky to have the particle on exactly that location. Unless you have as many particles as molecules. Sure you don't do that.

There are two options.
1. In Pre, go to Output Control > Export results. There you can tell the solver to export particle data when particles hit a certain boundary. This can be the hand of the body, its nose, feet, the floor, outlet, ceiling, etc. You can select any surface aith a unique name.

2. In Pre, go to Output Control > Results. There you can select "Extra Output Variables List". Look for e.g. <Name of particle>.Averaged Mass Fraction
Then averaged particle data is transferred from the discrete particle phase to the continuum phase. In this way you can read it out in your points. The value you get is the value that is stored in the numerical element of the continuum phase where your point is. It is therefore quite grid dependent. The value is the average af all particles (small, large) in that numerical element.

Hi @Gert-Jan,

Thanks for your reply and useful insights.

Yes, Points P1, P2, & P3 are points within the geometry. They are defined as monito points (Output Control > Monitor > Add new item > [Set Name] > OK > Option > Catersian Coordinates).

On your first suggested option (i.e., Output Control > Export results), in my model setup, I didn't consider the Export results option as, similar to Particle Histograms, it based on surfaces and not monitor points within the domain. From your experience, is it possible to create a small surface rather than monitor points for the export results and/or particle histogram?

On the second option (i.e., Output Control > Results > Enable Extra Output Variable > Select Output Var. List). First I need a clarification on this. A similar setting is available under the Trn Results Tab of Output Control (Transient Results 1 > Option > Selected Variables > Output Variable List). Should both settings be made (in case of transient simulation) under Results and Trn Results Tab by selecting the same type of additional variables? In the model setup, I selected the same additional variables under both Tabs (Results and Trn Results). Then, while defining the monitor points (P1, P2, & P3), I added the ouput varaibles defined under the Results/Trn Results Tab to the "Output Variable List" field of the monitor points. This is done with the understanding that the selected variables under "Results Tab" are particle field variables that can be monitored (according to CFX Pre documentation). Are these correct? If YES, why are the monitor points returning zero for all the selected particle variables?

Once again, thanks for your help.

[Gert-Jan]

You cannot monitor LPT data during the run on points P1, P2, etc. The LPT data is discrete, the monitoring points look for continuum variables.

Monitoring LPT averaged data in a transient run ...... I don't know. I have no experience. Better save data to trn files and do the transient postprocessing in Post. Or do the Eulerian approach in double precision.

[lyday011]

Originally Posted by ghorrocks

There is also the Algebraic slip model which vastly simplifies things (if its inherent assumptions are appropriate) and MUSIG models to handle a range of droplet sizes rather than just a single one. There are lots of ways to improve the performance of the Eulerian particle model.

@ghorrocks, Yes, I'm aware of ASM and MUSIG options. The interest in LPT is because we are interested in comparing the LPT from an in-house CFD code with that of CFX. Both of the CFD results are to be validated with experimental results from the modelled room.