Learning Forum

[fatih_g]

I simulate a coughing scenario with inert particles using DPM and also turbulent airflows in Fluent. At the outlet, I count the particles and try to observe the effect of randomness due to the turbulence and also interaction of air and particles. So, I use a transient scenario for continuous phase and unsteady tracking for discrete phase. For stochastic tracking, I use discrete random walk model and random eddy lifetime. Hence, I expect to see a random fluctuation in the counted particle number at the outlet, whenever I initialize and run the simulation. However, I get the same results for each run in the same Fluent session. But, if I restart Fluent and run the simulation for the same scenario, I can observe the randomness in the counted particle numbers. My final aim is to see the effect of randomness for several runs and create a statistical data to process further.

So, is there a way other than restarting Fluent to observe the effect of the randomness? What can be done to see the effect of randomness by running the simulations several times automatically? Can I write a script to control Fluent by another program or just use a property of Fluent for that?

[Rob]

If you're running transient particles the number of particles leaving every step should vary if turbulence is sufficient to alter their trajectories and the source is continuous. Restarting will probably give a slightly different result due to the stochastic randomness. However, if the variation is noticeable it also suggests you don't have enough parcels in the domain.

[fatih_g]

Thank you for the answer. But is there a way to see this slight randomness without restarting Fluent? Or how can I make this restarting process automatically, e.g. by using a user defined function?

[Rob]

You'd need to trigger a new cough to see the effect, so potentially you need to re-initialise the flow and particles. Please can you post some images as it's a little unclear what you're seeing.

[fatih_g]

You can see the picture of my scenario below. When I re-initialize the solution, I see the same counted number of particles at the outlet on the left side.

[Rob]

Please replot with residence time. Given the lack of dispersion I suspect you'll get much the same result every time.

[fatih_g]

The figure with residence time is below. I don't understand why I get the same result after re-initialization.

[Rob]

You have a fairly coarse mesh in places (for other's reading this there are a couple of other threads looking at other parts of this model) and those parcels look to enter the system, drop (gravity or initial velocity) and then head in the +z direction, possibly due to the bulk flow. Given the colour overlaps most parcels aren't deviating much from the mean trajectory.
As you get more and more parcels in the domain the effects of stochastic tracking diminish as the parcel "cloud" gives the average result. Depending on what you're monitoring you may also be averaging out some of the turbulent effects.

[fatih_g]

Ok, I will try the same scenario with a finer mesh and see if the results change. But, how can I do these runs automatically?

[Rob]

If you remesh and then read that mesh into the current case it'll do all the set up in the background assuming the boundary labels don't change.