Very different particle tracks with Discrete Random Walk Model activated between v14.5 and v18.2

  • 150 Views
  • Last Post 06 August 2018
  • Topic Is Solved
bassk posted this 26 July 2018

In v14.5 we see random motion of the particle as one would expect with the eddy interaction model. However, running the exact same case/data files and injections in v18.2, the particle track is very different. Notably, the particle track with/without eddy interaction activated is the same in v18.2, and the same path is seen in v14.5 if the model is deactivated. Therefore, it appears as though the model is having little to no effect on the particle path in v18.2, but did influence the particle path with random motion in v14.5.

What changes have been make to the particle tracking routines between these versions of Fluent and how can we get the desired behaviour in v18.2? Note that my colleague has v19.0, and the same issue is apparent in that version also. I have tried adjusting the Accuracy Control parameters and tracking scheme options, but with no improvement.

Order By: Standard | Newest | Votes
abenhadj posted this 27 July 2018

For post-processing, the tracking was made reproducible a couple of years ago, i.e. it now always uses the same pseudo-random numbers. DRW is still affecting the trajectories. So the behavior is by design.

 

Best regards,

Amine

bassk posted this 27 July 2018

Thanks for your prompt response!

Our problem is that in v14.5 the particles follow a random, erratic path when DRW is activated, as one would expect. In v18.2, the particle path is almost identical when DRW is either on or off, and random turbulence fluctuations do not appear to be affecting the particle motion. As I mentioned, we know that Accuracy Control can negate some of the effects of DRW, but adjustment of these parameters has no positive influence on results.  We want the same paths we see in v14.5 with DRW activated in v18.2.

I have attached some plots of the particle paths that hopefully illustrate my point. Paths are coloured by particle residence time. You can see that only when DRW is on in v14.5 do we get the kind of particle path that we desire and expect for our models. The path is identical in v18.2 with DRW on/off and matches the path in v14.5 with DRW off. This is what suggests to us that either DRW isn't affecting the particle trajectory in v18.2, or something else is negating it. Also note that particles are incomplete for the cases in v18.2 with DRW on/off and in v14.5 with DRW off. These plots are generated with the exact same case/data files (generated in v14.5) and injections, so we are positive that this is a version issue and not a set-up issue. I would be happy to provide you with the case so that you can verify and reproduce the issue.

Some background... particles are initialised from rest at the bottom of the domain, picked up by the flow, and carried to the outlet. They are 1µm aerodynamic diameter (rho=997 kg/m3) and we use them to sample turbulence parameters within the domain for correlation with experimental results. Our existing correlations are based on the v14.5 results, which have higher residence time and more accurate sampling, with more complete particles. Hence why we require the DRW model to behave same (or similar) in v18.2 as it did in v14.5.

  • v18.2 with DRW on

Paths when DRW is activated in v18.2

  • v18.2 with DRW off

Paths when DRW is deactivated in v18.2

  • v14.5 with DRW on

  • v14.5 with DRW off

abenhadj posted this 30 July 2018

For consistency start from scratch and do not rely on results obtained from old release by reading them into a newer version. You can use the most actual Student release for that.  As ANSYS Employee I cannot work on user's cases from Student Community.

Best regards,

Amine

bassk posted this 31 July 2018

My first approach in setting up this case and trying to reproduce our previous results from v14.5 was to set up the case from scratch as you suggest here. That is when we first noticed the difference in particle trajectories. I resumed the old case/data files to ensure that there wasn't a set-up issue with the new case. I have also checked the new case thoroughly and am sure there a no mistakes in my set-up (including using /file/write-settings in the GUI and doing a diff on the output). Therefore, setting up a case from scratch does not resolve this issue.

I have also increased/decreased the particle diameter to get more/less inertia to try and get the turbulence field to influence the trajectory, but there is still no effect. Next, I plan to artificially increase the inlet flow rate to get more turbulence in the domain. This will be different to the experimental set-up, but may help diagnose the problem.

Are you able to work on this case if my PhD advisor submits a support ticket? This is what we did originally, but were directed to this Student Community. What if he submits the same problem here? At the moment, it seems our only solution is to continue using v14.5 for these models, which obviously isn't ideal given the improvements between that and the latest versions.

abenhadj posted this 31 July 2018

Can you check the the turbulent viscosity ratio along the trajectory

Best regards,

Amine

rwoolhou posted this 31 July 2018

 What happens if you increase the number of particle steps? The last part of the R14.5 trajectory with DRW looks a bit odd compared to the rest of the plots. 

As an aside, if you plot the mesh with feature edges along with the particle tracks it may be a bit easier to figure out what's going on.  This may assist the wider community as we're unable to work on the files from here. 

 

bassk posted this 01 August 2018

We use 100,000 particle steps. In the v14.5 case with DRW off and v18.2 with DRW on/off, the particle gets stuck at that location after ~1200 iterations. That is, when I write x-, y-, and z-coordinates using the Step by Step report in the GUI, the coordinates do not change. It is not that the particle is stuck in some recirculating flow; it is stationary. This is a secondary issue, but we want to resolve the DRW problem first, as we hope the random fluctuating velocity components may prevent the particle from getting stuck. Note that walls are set to reflect with coefficient of restitution = 1. So the particle shouldn't be trapped on the wall. This is intentional.

Updated figures attached:

  • TVR vs Residence Time for all four cases:

  • TVR vs Residence Time for v18.2 case and zoomed in:

  • v14.5 track with DRW on:

  • v14.5 track with DRW off:

  • v18.2 with DRW on:

  • v18.2 with DRW off:

 

bassk posted this 01 August 2018

Comparing the TVR plots between the two v18.2 cases, it is hard to see from the attached image, but there are very subtle differences between the two. Though the two plots almost lie on each other identically. So the DRW model is having a small influence on the track, but it is negligible.

abenhadj posted this 06 August 2018

Hi,

The questions has been resolved and was due to to the TVR limiter which has been introduced to avoid prohibitive particle tracking at low turbulence. 

Best regards,

Amine

bassk posted this 06 August 2018

Thank you all for your help on this. The issue is resolved.

Close