[atul1018]

Hello All


I am simulating a particle-laden flow (backward facing step), which is turbulent flow case. Particles are injected from inlet after certain time and fluid and particle velocity profiles are compared with the experimental data. The case is simulated using DPMFoam.



firstly, I used RANS (with dispersion model to get effect of turbulent fluctuations on particles) to resolve the fluid field, the resolved fluid fields are the used to predict the particle trajectories in the domain. The simulated fluid velocity profiles give excellent agreement with the data but particles dispersion is under-estimated, thus giving very different particle velocity profiles than that of observed in experiment.



As the RANS gives the mean fields and effect turbulent fluctuations are modeled by some dispersion models, which seems to under predict the dispersion. I decided to use LES in order to get fluid fields which 80-90 % resolved, thus no need to use any dispersion model.


When LES-DEM is performed, the fluid velocity profiles gives good agreement and particle velocity profile has improved than that of RANS-DEM. But still I get very less particles below the step (still particle dispersion is under-predicted. I am wondering what could be reason for that?


Any Expert comments?


Best Regards
Atul

[FMDenaro]

First, since you are simulating a backward facing step you should check if your flow velocity is correct. There are many database for such flow to compare the statistics.
If your solution is correct, you have to consider that a filtered velocity is not exactly what appears in an experiment. The particle can have some inertia and the real velocity is a DNS-like solution. Check better the condition of the experiment.

[atul1018]

Hello FMDenaro


Thanks for your answer.

First, since you are simulating a backward facing step you should check if your flow velocity is correct. There are many database for such flow to compare the statistics.

-the flow velocity profiles for fluid phase give really good agreement with the experimental data. I compared Umean and Urms, both seems to be giving good match with experiment. isn't it sufficient to say that I got correct solution for fluid phase?

If your solution is correct, you have to consider that a filtered velocity is not exactly what appears in an experiment. The particle can have some inertia and the real velocity is a DNS-like solution. Check better the condition of the experiment.

-As said, I think my fluid solution is correct and I also understand that LES doesn't resolve/represent actual solution. But as I am resolving 80%-90% kinetic energy, I should get at least some particle below the step. The problem I face is that I get almost no particles below the step at measurement locations (till 1/3rd of domain), particles start to get disperse towards the outlet of the domain (in remaining 2/3rd of domain), which is opposite to observations in physical experiment. In the physical experiment the particles are fully dispersed till the reattachment location.


Am I doing anything wrong or missing something?
What do you suggest in order to get better dispersion of particles?


Best Regards
Atul

[FMDenaro]

Many reasons could be possible I cannot say what happens.

From your description seems that the recirculation region is steady and isolated from the main flow. And that is wrong. You should provide the plot of the velocity profile at several stations.
Could you show also the experimental data?

[atul1018]

From your description seems that the recirculation region is steady and isolated from the main flow. And that is wrong. You should provide the plot of the velocity profile at several stations.
Could you show also the experimental data?

-no, the re-circulation region is not steady, I performed averaging in order to get mean values (UMean). I plot the velocity profiles (Umean) at several stations and you can see the flow gets reattached ~x/H=7 as observed in the experimental data.
-I am also attaching plot of the particle velocity profiles at several stations, you can see that particles below y/H <1 are almost zero before reattachment point (x/H=7), they start to get dispersed after this reattachment location.



Any idea, what could be the reason of it?


Best Regards
Atul

[FMDenaro]

Originally Posted by atul1018

-no, the re-circulation region is not steady, I performed averaging in order to get mean values (UMean). I plot the velocity profiles (Umean) at several stations and you can see the flow gets reattached ~x/H=7 as observed in the experimental data.
-I am also attaching plot of the particle velocity profiles at several stations, you can see that particles below y/H <1 are almost zero before reattachment point (x/H=7), they start to get dispersed after this reattachment location.



Any idea, what could be the reason of it?


Best Regards
Atul

That fact that the particles follow apparently only the mean velocity is strange.
How do you distribuite the particles at the inflow?

[atul1018]

How do you distribuite the particles at the inflow?

- I inject the particles from the inlet after some time has passed (e.g. total simulation is 1 sec, particles are injected @0.5sec till the end of simulation). The numbers of particles per second are calculated as 10% of fluid mass flow rate. The particles are defined by some diameter and density. the particles injected have uniform distribution @ inlet.

[FMDenaro]

Originally Posted by atul1018

- I inject the particles from the inlet after some time has passed (e.g. total simulation is 1 sec, particles are injected @0.5sec till the end of simulation). The numbers of particles per second are calculated as 10% of fluid mass flow rate. The particles are defined by some diameter and density. the particles injected have uniform distribution @ inlet.

First, I suppose that you inject only after the flow is fully developed, that is after the numerical transient is ended.
Then, your particles have a mass and therefore some inertia, they do not follow the lagrangian trajectories.

Maybe the diameter and the density is larger than the experimental ones? That could justify that fact that they are not so subject to the fluctuations.
Of course, some bug in the code could be also a simple reason.

[atul1018]

First, I suppose that you inject only after the flow is fully developed, that is after the numerical transient is ended.
Then, your particles have a mass and therefore some inertia, they do not follow the lagrangian trajectories.

-I have also performed the RANS-DEM simulations where I have injected the particles from the beginning, but it doesn't really make any difference. I will perform the LES-DEM following your suggestion where particles are injected from the beginning of simulation.

Maybe the diameter and the density is larger than the experimental ones? That could justify that fact that they are not so subject to the fluctuations.
Of course, some bug in the code could be also a simple reason.

-The description of particles are same as mentioned in the paper. These are copper particles (density 8800 kg/m3) having diameter 70 micrometer.
-I have tried almost everything in my knowledge but it seems the dispersion are under predicted in both LES-DEM and RANS-DEM (with dispersion model too). May be its the bug nut I am not sure.

[FMDenaro]

Originally Posted by atul1018

-I have also performed the RANS-DEM simulations where I have injected the particles from the beginning, but it doesn't really make any difference. I will perform the LES-DEM following your suggestion where particles are injected from the beginning of simulation.



-The description of particles are same as mentioned in the paper. These are copper particles (density 8800 kg/m3) having diameter 70 micrometer.
-I have tried almost everything in my knowledge but it seems the dispersion are under predicted in both LES-DEM and RANS-DEM (with dispersion model too). May be its the bug nut I am not sure.

I suggest to do a check by performing first a simulation of a passive lagrangian tracer. If it works you can try increasing the density and diameter step-by-step

[atul1018]

I suggest to do a check by performing first a simulation of a passive lagrangian tracer. If it works you can try increasing the density and diameter step-by-step

-Since my aim is to assess the particle tracking solvers in OpenFOAM and verify the findings of physical experiment, I can't change density or diameter randomly. So changing any parameter randomly wont be representing the experiment which i intent to simulate.

-I wanted to know what could be the reason of particle not dispersing properly below the step as found in the physical experiment.


Best Regards
Atul

[Msure]

Hi, have you solved the problem? I got the same problem a couple years ago and never figured it out. I doubt the lagrangian model in Openfoam is not correct.

[atul1018]

Originally Posted by Msure

Hi, have you solved the problem? I got the same problem a couple years ago and never figured it out. I doubt the lagrangian model in Openfoam is not correct.

Hi Msure

I have compiled the results as a journal paper and discussed the probable reason for particles not dispersing behind the step. This is mainly due to the simple dispersions model used for turbulent fluctuations. More sophisticated model dispersion models are required to recover lost turbulent fluctuations in both RANS-DEM and LES-DEM methods.

The detailed study can be found in my paper:
https://www.mdpi.com/2311-5521/7/10/337

Best Regards
Atul

[Msure]

Originally Posted by atul1018

Hi Msure

I have compiled the results as a journal paper and discussed the probable reason for particles not dispersing behind the step. This is mainly due to the simple dispersions model used for turbulent fluctuations. More sophisticated model dispersion models are required to recover lost turbulent fluctuations in both RANS-DEM and LES-DEM methods.

The detailed study can be found in my paper:
https://www.mdpi.com/2311-5521/7/10/337

Best Regards
Atul

Hi Atul, thanks for your nice work! It answered the question perfectly. Thanks.