Hello,

when I model the behavious of bacteria in a flow using particle tracking, it seems that no any bateria particles enters the strong recirculation zones. Can someone suggest if this is physically correct--considering that the bacteria are so small particles?

Thanks,

Jack

The only mechanism that could enable them to enter recirculation zones would be diffusion.Do you take diffusion into account in your tracking algorithm?

correction: I only thought about recirculation behind an object. If you have a recirculation zone in front of an object (e.g. forward facing step), then particles could enter due to their inertia. But this would rather happen with bigger particles. The smaller they are the better they follow the streamlines.

Thanks for your reply.

When using species transport equation, the bacteria can always diffuse to the recirculation zones. Does that mean that species transport method over-estimates the effects of diffusion, and particle tracking is more sutable for such a situation?

Have you seen some publications discussing this?

Thanks

Jack

Is diffusion modelled in the tracking algorithm (random movement)?

What is the Peclet number for your problem? I guess if it is big and the numerical diffusion in the Eulerian approach is high, you might overpredict the effects of diffusion. To check this you could set the diffusion to zero and check if the bacteria still enters the recirculation regions.

What was the motivation to use particle tracking? I would assume it is computationally more expensive.

Thanks again.

Yes, the diffusion(or dispersion) was modelled in the particle tracking algorithm.

I do not know how to get Peclet number in commercial code. But using species transport method, even when I set the turbulence diffusion close to zero(very big Peclet number), the bacteria can still enter the recirculation regions(caused by molecular diffusion).

I understand that particle tracking is computationally expensive. But my current concern is to identify which approach is more suitable or physically correct for my case.

Theoritically, do you think that the two approaches can be equally suitable for my case(with strong recirculation) if modelled correctly?

Best regards,

Jack

Theoretically they should do exactly the same, I think.

Practically my opinion is that when diffusion plays a major role you'll need a lot of particles to get good statistics. Maybe you just did not compute enough particles. What is the concentration inside the recirculation zones compared to the free stream (for the Eulerian approach)? How many particles did you trace?

Maybe you can look at a very simple example to compare the two methods. Is it possible to switch off the diffusion completely (turbulent + molecular)? Anything that enters your recirculation zone in that case is a result of numerical errors (for recirculations on the downstream side).

Another point that could have an influence is the step size for your particles. If the steps are too big the particles will have problems to follow the streamlines and get deviated away from vortex cores, just as it would happen with large particles. Can you compute something with a forward facing recirculation? In that case the particles would be shot into the vortex, due to their large velocity.

What are the physics of your particles. Is their drag coefficient big enough for them to follow the streamlines perfectly?

I very much appreciate your excellent discussion and comments.

The number of particles I input is 5000,which should be suffcient. For the Eulerian approach, the residence time of flow in recirculation zone is much more than 1 fold higher than in the free stream(I did not monitor the concentration). The physics of your particles are the same as those of water with a diameter of 10 um. The drag force is big enough (Schiller Naumann model).

I will try to switch off the diffusion completely (turbulent + molecular) and see what happen. It's also a good point to change time step. I will reduce my time step for my next run. But do you expect that will make the particle tracking follow the streamline--that's actually not what i want, since they are very close in my current results.

Thanks again,

Best regards,

Jack

Hi - sorry, but it took me a bit longer this time ;-)

If you reduce the time step of the particle tracking, you'll increase the accuracy. Hence if the particles would follow the streamlines in nature, they will deviate less with a decreasing time step. Btw, what do you mean by they are very close?

5000 doesn't sound incredibly big to me. My experience is somewhat old now, but I remember using something like 50000 particles for a flow around a circular cylinder in 2D (with random movement). But I don't actually rememeber all the details and it might not have been the smartest approach.

What is your geometry? And how do you place the particles - completely randomly, or in regions of interest?

Cheers, Oliver

Dear Oliver,

Great to see your reply again.

I did try to reduce my time step. The result has very slight change. Probably the time is sill not small enough. I will try a smaller one later.

In the CFD code I use, I can draw the streamline and prticle trackingr results. They are very close. Streamline is just more smooth and it also does not pass recirculation zone.

I do not know if my computer can run 50000 particle cases (I will do a simple test). But in my sense, many literatures use around 3000 particles. If the particles can enter the dead zone, the thousands of particles should be sufficient to get a visible result when they enter the dead zone. So I am still suspect that something was wrong with my particle tracking model.

My geometry is very simple, a reactangular tank connecting with a circular cylinder in 2D. I input the particles within 0.1s from the inlet of the geometry. If you could send me your email address to hgdsj@yahoo.com, I can show you more details.

Thanks again for the comments.

Best regards,

Jack