CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > ANSYS > FLUENT

DPM Iteration Fluent

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree6Likes
  • 1 Post By vinerm
  • 1 Post By vinerm
  • 1 Post By vinerm
  • 1 Post By vinerm
  • 1 Post By vinerm
  • 1 Post By vinerm

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   May 2, 2020, 16:54
Default DPM Iteration Fluent
  #1
New Member
 
CFDSim72
Join Date: Apr 2020
Posts: 17
Rep Power: 4
sirsammie72 is on a distinguished road
Could someone explain to me the way the DPM iteration works in fluent with steady tracking? it makes sense to me that the number tracked would remain constant, and all other particle fates would remain zero until particles began to reach their fates, and then the total number of fates should add to the number tracked.

However this is not what happens. Is this what is supposed to happen?

I am having issues with incomplete particles in cyclone separator and have tried increasing the max number of steps and also confirmed that the incompletes have a residence time of much shorter than the total flow solution.

Thanks
sirsammie72 is offline   Reply With Quote

Old   May 2, 2020, 17:13
Default Particle Tracking
  #2
Senior Member
 
vinerm's Avatar
 
Vinerm
Join Date: Jun 2009
Location: Nederland
Posts: 2,946
Blog Entries: 1
Rep Power: 34
vinerm will become famous soon enough
Lagrangian Particle Tracking or DPM is solution of Newton's second law for particles (actually parcels or representative particles). Incompleteness could be either numerical or parallel. Parallel incompleteness is represented as parallel incompleteness and shows up only for parallel simulation. Incomplete particles in serial or due to numerics are those whose fate could not be determined within given time-steps for particle tracking. Do note that particle tracking is always unsteady since time is the only independent variable. However, if the number of incomplete particles do not change even after increasing the number of steps for DPM, say, even for 200000 steps, then the particles are stuck in some recirculation zone, which is possible for cyclone separator. Another reason could be absence of gravity or incorrect specification of gravity and operating density. So, check for these two parameters.

For physical understanding of DPM, in case you are rather new to DPM, you can refer

https://www.cfd-online.com/Forums/blogs/vinerm/
8cold8hot likes this.
__________________
Regards,
Vinerm

PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority.
vinerm is offline   Reply With Quote

Old   May 4, 2020, 11:18
Default
  #3
New Member
 
CFDSim72
Join Date: Apr 2020
Posts: 17
Rep Power: 4
sirsammie72 is on a distinguished road
Thank you for your reply. I have checked my gravity and operating density settings and they are ok. Would you be able to explain more about the recirculation zones? I first thought this meant the vortex of the cyclone when reading the user guide, but realized it most likely means something within the cells, as the residence time for particles before becoming incomplete is ~0.3s and the flow is running for much longer, and particles would have time to circle in the vortex many times. I have tried with max number of steps at 1000000 and still find many incompletes. I also have checked that many of the incompletes do not have zero velocity at the point where they stop being tracked.

I read the DPM post you provided. I am new to DPM as well as CFD. Could you explain to me why the number of incompletes and escaped particles fluctuate throughout the calculation when using steady tracking? ( I am just using unsteady/steady as defined by fluent but understand all particles are transient as you mentioned.) I would think that the number of tracked would remain constant and the escaped and incomplete numbers would only increase as more particles reach their fate. How is DPM reporting fates for all particles each iteration, even after 1 time step (0.005s) in which particles have not reached their final destination?

Additionally, do you think the incompletes and ~0.3s residence prior to incompletes/getting stuck in the recirculation zones could be solved using unsteady DPM? I know the user guide states that only unsteady DPM is suitable for particles in suspension, though I do not understand why.
sirsammie72 is offline   Reply With Quote

Old   May 5, 2020, 08:30
Default Steady and Unsteady DPM
  #4
Senior Member
 
vinerm's Avatar
 
Vinerm
Join Date: Jun 2009
Location: Nederland
Posts: 2,946
Blog Entries: 1
Rep Power: 34
vinerm will become famous soon enough
DPM equations are always unsteady, so, steady and transient DPM have the meaning in terms of time of evolution; steady implies tracking if time is not the controlling factor, e.g., if a sand particle is left in water, eventually it will settle down, while transient tracking will track it over time until the time specified by the user for the simulation. That's why suspensions require transient particle tracking because steady scenarios for suspensions are either all particle floated up or all settled down but nothing suspended.

In case of steady tracking, Fluent injects particle at a particular instant, say every 10 or 20 iterations, and then tracks all the particles all the way down to their fate or until the number of steps for DPM are exhausted. Every time this is done, the flow-field is different, hence, number of particles escaping or incomplete are different. Instead of using DPM, just plot Streamlines from inlet and observe where do they stop. Streamlines are also DPM particles but massless. You do not need to define any injection for this since surface injection is used automatically. Observe the location where streamlines stop. That's where your particles are getting stuck.
8cold8hot likes this.
__________________
Regards,
Vinerm

PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority.
vinerm is offline   Reply With Quote

Old   May 5, 2020, 12:23
Default
  #5
New Member
 
CFDSim72
Join Date: Apr 2020
Posts: 17
Rep Power: 4
sirsammie72 is on a distinguished road
Ok, thank you. this makes much sense to me and is very helpful.

One thing I am still unsure about -- what then is the physical representation of incomplete particles?

Are these not particles in suspension as well, or what is physically implied by "recirculation zones"?

Would not tracking the particles as "unsteady" as defined by fluent allow for the stuck particles to just be calculated as in suspension?
sirsammie72 is offline   Reply With Quote

Old   May 5, 2020, 13:14
Default Suspension
  #6
Senior Member
 
vinerm's Avatar
 
Vinerm
Join Date: Jun 2009
Location: Nederland
Posts: 2,946
Blog Entries: 1
Rep Power: 34
vinerm will become famous soon enough
In Steady scenario, if particles have not reach a boundary, then those are known as incomplete. So, in a way suspended but since suspension is not a steady-state, those are called incomplete instead of being called suspended. As far as transient is concerned, most of the times particles are suspended until they hit a boundary. That's why transient tracking is required for suspension modeling.
8cold8hot likes this.
__________________
Regards,
Vinerm

PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority.
vinerm is offline   Reply With Quote

Old   May 6, 2020, 17:18
Default
  #7
New Member
 
CFDSim72
Join Date: Apr 2020
Posts: 17
Rep Power: 4
sirsammie72 is on a distinguished road
Ok, I understand. And identifying the streamlines was very helpful. However, what I do not understand is what is physically implied by recirculation zones common in cyclone separators. Does this imply small eddies that trap the particle in infinite loops? Is this a function of the vortex? What is the physical phenomenon occurring, or is this just a modelling problem?

I have analyzed the particle data and found that the velocity becomes quite low and then flat lines at right above zero while path length continues to rise as particle becomes stuck and is marked as incomplete. However, there are particles that escape and also reach zero velocity (I assume while rounding a turn) but do not get stuck.

Also, I have found that the increasing the particle and flow velocity significantly decreases the number of incomplete particles, which also adds to my wanting to understand the physical phenomenon.

I have seen the posts that suggest incomplete particles can be considered collected if above critical diameter, but if this was the case my current results would be incredibly inaccurate.

With regard to transient tracking for suspension -- how is the amount of time that particles are tracked determined? is it the particle time step size*max number of steps. I know the injection time parameters can be altered, but I mean how long particles stay in system
sirsammie72 is offline   Reply With Quote

Old   May 7, 2020, 09:35
Default Dpm
  #8
Senior Member
 
vinerm's Avatar
 
Vinerm
Join Date: Jun 2009
Location: Nederland
Posts: 2,946
Blog Entries: 1
Rep Power: 34
vinerm will become famous soon enough
RANS based turbulence models cannot resolve sub-grid scale eddies, so, the particle can only get stuck in large scale eddies, the eddies much bigger than the particles. This is similar in size to the central vortex in cyclone separator.

Along with Newton's second law, the first law is also at work here. Second law is used to solve for the particle motion, however, if the forces acting on the particles become 0 or insignificant, it keeps on doing what it is doing. So, if it is moving in a circular path, it will keep on moving in a circular path. Do note that it is only net force that is zero not individual forces.

Incomplete particles cannot be considered as collected; their fate is unknown. Given a slight disturbance those can move towards gas outlet or collection bins. Best assumption for incomplete is to keep them with the particles of similar size.

Total amount of tracking time is based on flow-time. Particle time-step and max number of steps are used only for each iteration when particles are tracked. However, how long they are tracked depends on how long the flow runs. If total flow time for which you are simulating is 100 s, then particles will be tracked for 100 s. However, at each tracking step, say every 10th continuous flow iteration, particles are tracked for time given by particle time-step and maximum number of steps.
8cold8hot likes this.
__________________
Regards,
Vinerm

PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority.
vinerm is offline   Reply With Quote

Old   May 7, 2020, 12:56
Default
  #9
New Member
 
CFDSim72
Join Date: Apr 2020
Posts: 17
Rep Power: 4
sirsammie72 is on a distinguished road
Ok, thank you. So for transient tracking, I was able to find no incomplete particles by increasing the particle time step size significantly. While many particles do not exit many are still in suspension. I am thinking increasing length of flow time will help then based on your explanation.

I am currently using LES but have tried with RSM as well. More incompletes are found in RSM but I would imagine due to mesh.

Also, I don't understand what you been by best assumption is to keep them with particle of similar size?

Thanks again for your help.
sirsammie72 is offline   Reply With Quote

Old   May 7, 2020, 13:01
Default LES and Particle Fate
  #10
Senior Member
 
vinerm's Avatar
 
Vinerm
Join Date: Jun 2009
Location: Nederland
Posts: 2,946
Blog Entries: 1
Rep Power: 34
vinerm will become famous soon enough
If you are using LES, then you are resolving very small eddies as well but still larger than particles so the previous comments about RANS are still valid.

By assumption of keeping with same particle size I mean this - if you find some incomplete particles, check their diameter and then determine what happened to rest of the particles of same size. You can safely assume same fate for incomplete particles then.
sirsammie72 likes this.
__________________
Regards,
Vinerm

PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority.
vinerm is offline   Reply With Quote

Old   May 15, 2020, 11:47
Default
  #11
New Member
 
CFDSim72
Join Date: Apr 2020
Posts: 17
Rep Power: 4
sirsammie72 is on a distinguished road
One addition question. Could you help me understand why the path length is so large for incomplete particles? I am struggling to understand how increasing the number of steps removes the particles from the recirculation zones. For example, with 1e5 max. number of steps, there are 60/90 incomplete particles. Those particles that are incomplete have incredibly long path lengths with constant low-velocity when compared to those that escape.

When increasing max. number of tracks to 1e6, there are now only 30/90 incomplete particles. Those particles that were previously stuck with very long path length now no longer get stuck and have very short/normal path length. What I don't understand is, how does increasing max. number of steps. prevent them from getting stuck and decrease path length to the normal length?

With my current understanding, I would think that increasing max. number of steps would mean that they still get stuck and have the same long path length but have enough steps to be able to emerge from the recirculation zone after the long path length and then escape. Help with understanding would be appreciated. I am sure I am missing something as to how the tracking/calculation occurs
sirsammie72 is offline   Reply With Quote

Old   May 15, 2020, 13:52
Default Tracking
  #12
Senior Member
 
vinerm's Avatar
 
Vinerm
Join Date: Jun 2009
Location: Nederland
Posts: 2,946
Blog Entries: 1
Rep Power: 34
vinerm will become famous soon enough
For steady tracking, there are three time related parameters; two used at a time. One is maximum number of steps. This is essentially the number of time-steps for which integration is done. If this limit is reached before particles reach some prescribed fate, such as, escape or bounce or evaporate, depending upon applicable laws, then particles are reported as incomplete. Second parameter is Step Length Factor. This is the number of steps that a particle takes to cross one mesh cell. Smaller the factor, longer it takes a particle to cross the cell in terms of time-steps, not in time. Time to cross depends upon particle's velocity but smaller value improves resolution and accuracy, however, also increases time required for integration. Third aspect is Length Scale. This is used if enabled and in that case, Step Length Factor is not used. One of the, Step Length Factor and Length Scale, is used to determine time-step for integration of Lagrangian equation. Each time integration is done, the velocity field, and hence, values of net force is different. The results are reproducible since the equations are deterministic, however, increasing the number gives particles more time to penetrate deeper. It is possible that the default value of 50000 steps is not enough. The limit on this value is 1e9. And a good estimate of maximum number is steps required to cross all the cells 1.5 times. So, if particles are traveling primarily in x-direction, then maximum number of time-steps should be number of cells in x-direction multiplied by Step Length Factor, which is 5 by default.

As far as path length is concerned, how are you determining it?
8cold8hot likes this.
__________________
Regards,
Vinerm

PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority.
vinerm is offline   Reply With Quote

Reply

Tags
dpm, fluent

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Fluent - DPM - Rotation - Mesh methods huangdf FLUENT 0 September 10, 2018 04:07
Iteration Number in Fluent 5 Alfonso Ferrandez FLUENT 8 October 4, 2017 12:27
Fluent Cd*A and local time for each iteration print out benjaminhogan FLUENT 1 November 4, 2014 13:32
Fluent DPM - incomplete particles richard ben FLUENT 4 September 26, 2014 12:37
Reading a file from 32-bit Fluent 15.0 through a UDF for DPM in parallel anthony05 Fluent UDF and Scheme Programming 0 July 11, 2014 02:52


All times are GMT -4. The time now is 03:42.