CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > General Forums > Main CFD Forum

Why particle size should be smaller than mesh size to use DEM?

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

Like Tree6Likes
  • 2 Post By sbaffini
  • 2 Post By q__
  • 2 Post By sbaffini

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   January 6, 2021, 07:21
Default Why particle size should be smaller than mesh size to use DEM?
  #1
Senior Member
 
Mandeep Shetty
Join Date: Apr 2016
Posts: 185
Rep Power: 9
granzer is on a distinguished road
Hello,
I am trying to simulate the flow of discreet particles in a continuous medium using the CFD-DEM coupling.
Let the continuous phase be water and particle be high-density thermocol, which has considerable size (assume chunks of themocol) but its density is smaller than water so that the chunks float (if the density of water is taken as 1000 kg/m3 let the themocol's density be 800 kg/m3)

i) I read that to use DEM the size of the particles should be less than that size of the mesh. Why is this?

ii) For this simulation can I consider the interia of the thermocol chunks to be negligible (and so flowing freely with water) as its density is less?

iii) Are all Euler-Lagrangian solver CFD-DEM solvers, since I see that Euler_Lagrangian and CFD-DEM are used interchangeably?
granzer is offline   Reply With Quote

Old   January 6, 2021, 07:29
Default
  #2
Senior Member
 
sbaffini's Avatar
 
Paolo Lampitella
Join Date: Mar 2009
Location: Italy
Posts: 2,150
Blog Entries: 29
Rep Power: 39
sbaffini will become famous soon enoughsbaffini will become famous soon enough
Send a message via Skype™ to sbaffini
Not an expert in this specific field but, a general reasoning you can adopt here as well as in other cases is: if you have a grid but it doesn't capture your phenomena of interest, then a model for these phenomena requires them to happen at scales smaller than those described by the grid, ideally much smaller.

For particles, if they are the size of a mesh cell or bigger, the mesh should probably take them into account.

I'm pretty sure that DEM and Euler-Lagrange are not, in general, interchangeable. There are DEM approaches that are purely Lagrangian and don't have a grid at all (at least not in the common Eulerian sense)
Eifoehn4 and aero_head like this.
sbaffini is offline   Reply With Quote

Old   January 6, 2021, 14:39
Default
  #3
Senior Member
 
Mandeep Shetty
Join Date: Apr 2016
Posts: 185
Rep Power: 9
granzer is on a distinguished road
Quote:
Originally Posted by sbaffini View Post
Not an expert in this specific field but, a general reasoning you can adopt here as well as in other cases is: if you have a grid but it doesn't capture your phenomena of interest, then a model for these phenomena requires them to happen at scales smaller than those described by the grid, ideally much smaller.

For particles, if they are the size of a mesh cell or bigger, the mesh should probably take them into account.

I'm pretty sure that DEM and Euler-Lagrange are not, in general, interchangeable. There are DEM approaches that are purely Lagrangian and don't have a grid at all (at least not in the common Eulerian sense)
Hello Mr.Paolo Lampitella, thank you for the answer. Here is what I understand from it. Can you please correct me if I am wrong in my understanding:
Quote:
"...if you have a grid but it doesn't capture your phenomena of interest.."
You mean if the grid is not sufficient enough to 'resolve' the phenomena of interest ryt?.
Quote:
"then a model for these phenomena requires them to happen at scales smaller than those described by the grid, ideally much smaller."
If we are not able to "resolve" the phenomena we have to model for the said phenomena...but using a model for the phenomena requires the mesh size to be larger than the phenomena itself?

Quote:
"For particles, if they are the size of a mesh cell or bigger, the mesh should probably take them into account."
This sentence seems to answer my questions (i) and (ii). Are you suggesting that when considering particle flows, even though the particles may be bigger than the mesh, it shouldn't cause a problem for simulations?

Quote:
I'm pretty sure that DEM and Euler-Lagrange are not, in general, interchangeable. There are DEM approaches that are purely Lagrangian and don't have a grid at all (at least not in the common Eulerian sense)
(In the last sentence I am assuming you are talking about particle methods like SPH or MPM?) As you say since there are DEM that is purey Lagrangian are the terms 'DEM' and 'Lagrangian' synonymous or is it wrong to say so?
granzer is offline   Reply With Quote

Old   January 6, 2021, 17:56
Default
  #4
q__
New Member
 
q
Join Date: Sep 2020
Posts: 19
Rep Power: 5
q__ is on a distinguished road
Quote:
Originally Posted by granzer View Post
Hello,
I am trying to simulate the flow of discreet particles in a continuous medium using the CFD-DEM coupling.
Let the continuous phase be water and particle be high-density thermocol, which has considerable size (assume chunks of themocol) but its density is smaller than water so that the chunks float (if the density of water is taken as 1000 kg/m3 let the themocol's density be 800 kg/m3)
i) I read that to use DEM the size of the particles should be less than that size of the mesh. Why is this?



The backbone of the particle tracking methods in Lagrangian coordinate system is that they usually treat particles like points of assumed properties (D, T, v etc.) and they are not “resolved”. Since the parameters need to be coupled by the interpolation particle sizes need to be smaller than the computational cell for the interpolation to be meaningful.
Eifoehn4 and granzer like this.
q__ is offline   Reply With Quote

Old   January 7, 2021, 05:54
Default
  #5
Senior Member
 
sbaffini's Avatar
 
Paolo Lampitella
Join Date: Mar 2009
Location: Italy
Posts: 2,150
Blog Entries: 29
Rep Power: 39
sbaffini will become famous soon enoughsbaffini will become famous soon enough
Send a message via Skype™ to sbaffini
Hello granzer, let me try to be more specific:

Quote:
Originally Posted by granzer View Post
You mean if the grid is not sufficient enough to 'resolve' the phenomena of interest ryt?.
yes

Quote:
Originally Posted by granzer View Post
If we are not able to "resolve" the phenomena we have to model for the said phenomena...but using a model for the phenomena requires the mesh size to be larger than the phenomena itself?
Yes, that was the general idea I was trying to promote but, as in most circumstances, there are exceptions. It really depends from the phenomena, the model and scope of the model. There are indeed models that
don't satisfy this but are known to work. Let me give you some examples:

1) Continuum model in place of molecular dynamics. We know that for the continuum hypothesis to be valid there must be enough molecules in a volume for the related averages to be meaningful. The model will still work for very small grids but the results might not be meaningful anymore.

2) Roughness models and wall functions. Wall functions are wall boundary conditions used when the grid can't resolve some features. It is only in this context that it makes sense to augment the wall function with roughness effects. If the grid can resolve the flow at the wall, by definition, it must resolve the roughness as well. So there is no roughness model outside wall functions (i.e., when the grid is not coarse enough to use them).

3) I used to work with models for micro vortex generators, which are devices with size of the order 0.1 the boundary layer thickness. The related models are just forcing terms in the equations and their intended use is for grids that, in theory, could actually describe them. Their use case is, typically, in the sense of an immersed boundary approach, you accept some error in order to more easily move them around and not having to make a mesh again. This is not a physical model, it is more a smart model that doesn't want to describe the physics but be smart and useful.

So, to be more specific, what I said is for models of physical phenomena that happen at smaller scales. Still, I know no physical model that works like, say, the micro vortex generator model I was mentioning above (that is, general scales that might or not be described by the grid).

Quote:
Originally Posted by granzer View Post
This sentence seems to answer my questions (i) and (ii). Are you suggesting that when considering particle flows, even though the particles may be bigger than the mesh, it shouldn't cause a problem for simulations?
Actually no, I meant quite the opposite. That is, if your particles are the size of, say, 10 grid cells, what is the meaning of the solution in those 10 cells which, instead, would ideally be occupied by a particle? To correctly take into account a particle of that size on your grid, you probably need the grid to describe the particle (i.e., go around it, not trough it), with boundary conditions etc.

This does not mean that there might not be models for this scenario that still work (think about my micro vortex generator example), but that is certainly not a physical model in the common sense. It is some smart model or some trick (say, immersed boundary treatment).

Quote:
Originally Posted by granzer View Post
(In the last sentence I am assuming you are talking about particle methods like SPH or MPM?) As you say since there are DEM that is purey Lagrangian are the terms 'DEM' and 'Lagrangian' synonymous or is it wrong to say so?
Again, I am not an expert but, what I know about DEM explicitly points to what I classify as Lagrangian treatment. For me, Lagrangian basically means any sort of N-body treatment, where position and velocity of N objects (particles, planets, etc.) are obtained by integrating in time their respective Newton 2nd laws.

Now, there are physical applications where these objects just interact between themselves and thus no additional field needs to be defined on an underlying Eulerian grid (yet, nearest neighbor searching would still, typically, require a sort of grid for algorithmic implementation).

To be more specific, Lagrangian is a more mathematical term that, despite what anyone can think, univocally identifies a certain treatment that we can simplify in "moving particles around without an underlying grid".

DEM is a more specific approach, but still broad enough, and what it means is basically defined by its community. Still, there are examples of applications that are DEM and don't use any underlying grid. Indeed, the presence of the underlying grid, intended in the Eulerian sense, is necessary when an interaction (backward, forward or both) between the two has to be taken into account in order to correctly describe the physics at hand.

So, clearly, in CFD, it typically happens that your approach is indeed Euler-Lagrange, because you are actually interested in the Euler part or what it does on the Lagrange part. But there are cases where the Lagrangian part is self consistent (within the approximation of the modeling assumptions, of course) and doesn't require the Euler part at all. SPH is one such example, but I think that this also applies to several applications in the realm of solid dynamics, where I don't think they use any global description for the fluid between the solid particles.
granzer and aero_head like this.
sbaffini is offline   Reply With Quote

Old   January 7, 2021, 12:32
Default
  #6
Senior Member
 
Mandeep Shetty
Join Date: Apr 2016
Posts: 185
Rep Power: 9
granzer is on a distinguished road
Quote:
Originally Posted by sbaffini View Post
Hello granzer, let me try to be more specific:



yes



Yes, that was the general idea I was trying to promote but, as in most circumstances, there are exceptions. It really depends from the phenomena, the model and scope of the model. There are indeed models that
don't satisfy this but are known to work. Let me give you some examples:

1) Continuum model in place of molecular dynamics. We know that for the continuum hypothesis to be valid there must be enough molecules in a volume for the related averages to be meaningful. The model will still work for very small grids but the results might not be meaningful anymore.

2) Roughness models and wall functions. Wall functions are wall boundary conditions used when the grid can't resolve some features. It is only in this context that it makes sense to augment the wall function with roughness effects. If the grid can resolve the flow at the wall, by definition, it must resolve the roughness as well. So there is no roughness model outside wall functions (i.e., when the grid is not coarse enough to use them).

3) I used to work with models for micro vortex generators, which are devices with size of the order 0.1 the boundary layer thickness. The related models are just forcing terms in the equations and their intended use is for grids that, in theory, could actually describe them. Their use case is, typically, in the sense of an immersed boundary approach, you accept some error in order to more easily move them around and not having to make a mesh again. This is not a physical model, it is more a smart model that doesn't want to describe the physics but be smart and useful.

So, to be more specific, what I said is for models of physical phenomena that happen at smaller scales. Still, I know no physical model that works like, say, the micro vortex generator model I was mentioning above (that is, general scales that might or not be described by the grid).



Actually no, I meant quite the opposite. That is, if your particles are the size of, say, 10 grid cells, what is the meaning of the solution in those 10 cells which, instead, would ideally be occupied by a particle? To correctly take into account a particle of that size on your grid, you probably need the grid to describe the particle (i.e., go around it, not trough it), with boundary conditions etc.

This does not mean that there might not be models for this scenario that still work (think about my micro vortex generator example), but that is certainly not a physical model in the common sense. It is some smart model or some trick (say, immersed boundary treatment).



Again, I am not an expert but, what I know about DEM explicitly points to what I classify as Lagrangian treatment. For me, Lagrangian basically means any sort of N-body treatment, where position and velocity of N objects (particles, planets, etc.) are obtained by integrating in time their respective Newton 2nd laws.

Now, there are physical applications where these objects just interact between themselves and thus no additional field needs to be defined on an underlying Eulerian grid (yet, nearest neighbor searching would still, typically, require a sort of grid for algorithmic implementation).

To be more specific, Lagrangian is a more mathematical term that, despite what anyone can think, univocally identifies a certain treatment that we can simplify in "moving particles around without an underlying grid".

DEM is a more specific approach, but still broad enough, and what it means is basically defined by its community. Still, there are examples of applications that are DEM and don't use any underlying grid. Indeed, the presence of the underlying grid, intended in the Eulerian sense, is necessary when an interaction (backward, forward or both) between the two has to be taken into account in order to correctly describe the physics at hand.

So, clearly, in CFD, it typically happens that your approach is indeed Euler-Lagrange, because you are actually interested in the Euler part or what it does on the Lagrange part. But there are cases where the Lagrangian part is self consistent (within the approximation of the modeling assumptions, of course) and doesn't require the Euler part at all. SPH is one such example, but I think that this also applies to several applications in the realm of solid dynamics, where I don't think they use any global description for the fluid between the solid particles.
Wow! Thank you sir, for taking the time to write such a descriptive answer. This clears a lot of questions in my head. I am still a beginner at this so I still need to learn a lot here to completely understand it but kinda gave me the path for further study to solve my problem.


Quote:
This does not mean that there might not be models for this scenario that still work (think about my micro vortex generator example), but that is certainly not a physical model in the common sense. It is some smart model or some trick (say, immersed boundary treatment).
If anyone else is looking into a similar kind of problem where you need to deal with particles significantly bigger than the mesh size I suggest looking into the "resolved DEM approach". This approach may be one of those 'smart model' mentioned by Mr.Paolo (not sure..I am still looking into it). But from my preliminary study, it looks like the model can be used to simulate particles much larger than mesh cell (Warning: So if the particle size is almost the same as mesh size this model may fail, although there seem to be some alterations of the model which seems to work.)

Quote:
But there are cases where the Lagrangian part is self consistent (within the approximation of the modeling assumptions, of course) and doesn't require the Euler part at all. SPH is one such example, but I think that this also applies to several applications in the realm of solid dynamics, where I don't think they use any global description for the fluid between the solid particles.
If anyone is interested in what is mentioned here Ref:
HTML Code:
https://blogs.3ds.com/simulia/particle-methods-in-abaqus-dem-vs-sph/
granzer is offline   Reply With Quote

Old   January 7, 2021, 13:05
Default
  #7
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,760
Rep Power: 71
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
I never worked with DEM but I wonder what the grid size is in this discussion... I assume that DEM has no requirement of a grid at all, thus the computational grid should be required by the coupling with some Eulerian approach. Is the coupling that should be the reason for discussing the proper size, not the DEM by itself, isn't it?
FMDenaro is offline   Reply With Quote

Old   January 7, 2021, 16:35
Default
  #8
Senior Member
 
Mandeep Shetty
Join Date: Apr 2016
Posts: 185
Rep Power: 9
granzer is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
I never worked with DEM but I wonder what the grid size is in this discussion... I assume that DEM has no requirement of a grid at all, thus the computational grid should be required by the coupling with some Eulerian approach. Is the coupling that should be the reason for discussing the proper size, not the DEM by itself, isn't it?
This is what I have understood so far(which is very limited) on how [CFD/FVM]-DEM seems to work...The application is to run simulation of particles (Lagrangian/granular particle) that are flowing in a fluid (which is taken as the continuum/Eulerian). Since the particles are effected by the flow(and flow by particle depending on the type of coupling is one way or two way etc, again depending on density of particle concentration etc), the particles(which are run by DEM) needs to know the flow environment locally around it. This is done by knowing in which mesh cell the perticular particle is present and taking the flow condition in that cell ( I am guessing this is what you mean by "coupling with some Eulerian approach" ?). So I guess it the particle is bigger then it becomes difficult to say which cell the particle belongs to.

So I yes it's the coupling that causes the size restriction.

If we use Lagrangian-Lagrangian model like say SPH-DEM then I don't think there is a matter of size begin restricted by mesh size(not sure about this though).
granzer is offline   Reply With Quote

Old   January 7, 2021, 16:48
Default
  #9
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,760
Rep Power: 71
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
I have no experience to say more... I found this recent paper


https://onlinelibrary.wiley.com/doi/...002/cjce.23773


and it seems here that DEM is mainly the solution of the lagrangian equation for a passive tracer, that is one-way mode.
FMDenaro is offline   Reply With Quote

Old   July 13, 2023, 16:41
Default
  #10
New Member
 
Join Date: May 2023
Posts: 1
Rep Power: 0
Halfpoetry is on a distinguished road
Thanks for the discussion above, which unfortunately seems very difficult for me to understand in my current stage. ..
What I just want to know is: Does StarCCM support what is called immerse boundary or fictitious domain? Because I must make the particle size larger than mesh size.
Halfpoetry is offline   Reply With Quote

Reply

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
decomposePar problem: Cell 0contains face labels out of range vaina74 OpenFOAM Pre-Processing 37 July 20, 2020 06:38
[mesh manipulation] How to write cellSet for different regions in constant/polyMesh/sets Struggle_Achieve OpenFOAM Meshing & Mesh Conversion 3 June 17, 2019 10:29
[mesh manipulation] Importing Multiple Meshes thomasnwalshiii OpenFOAM Meshing & Mesh Conversion 18 December 19, 2015 19:57
critical error during installation of openfoam Fabio88 OpenFOAM Installation 21 June 2, 2010 04:01
fluent add additional zones for the mesh file SSL FLUENT 2 January 26, 2008 12:55


All times are GMT -4. The time now is 02:53.