[RobertW]

Dear All,

I hope you are well.
I start my work with Ansys Fluent (I did a few airflow analyses long time ago).

I need to model the airflow through a bunch of little balls (see attached draft and picture on it, these balls are for illustration only, the actual balls look somewhat different).
Balls are stored inside 1mx1mx1m container (see attached draft). Each ball has approx. diameter of 10mm.
These balls are inserted into the container after their coating is finished (balls are a product of the food industry, they have softer inside and harder/coated outside).
The balls are conditioned in the container, i.e. the air flows through the container/between balls, to allow some water evaporation from balls and to prevent lumping between balls.
I need to model this airflow process (between balls) and verify whether the airflow is approx. uniform across the container and to check whether there are some stagnation points.
The attached draft presents the location of air inlet and outlet in the container. The air is ‘pushed’ into the container (through inlet hole, see attached draft) and is allowed to flow through the container (between balls) toward the outlet hole. The temperature of air at inlet, inside the container and at outlet is a ‘room’ temperature.
I plan to conduct, at least for a start, two 2D analyses on the planes that correspond to sections A-A and B-B (see attached draft).

I have a few questions:
1. Should I consider the airflow inside the container (especially near inlet) as compressible? What do you think? There are tens of thousands of balls in the container. The actual conditioned product may also have rectangular or pillow like shape (instead of ball/round shape), and can cause (perhaps) considerable resistance to the airflow.
2. I consider using model: viscous laminar, k-epsilon/realizable, non-equilibrium wall functions, and solution method: pressure velocity coupling. Would you apply different models and solution methods?
3. Do you think that 2D simulation will do to effectively simulate the distribution of airflow in selected areas of the container? Is 3D simulation (of a portion of container) necessary? For now, I consider 2D simulation as a good start point (on a plane of section A-A). I want to run a few analyses and see what results I will get from 2D simulations.
4. I wonder what would be the best and simplest way to validate the simulation model. My idea is to locate a few little sensors inside the container (filled with balls). I am not sure what kind of sensors they could be (humidity, air speed, …?). The easiest could be using little thermocouples, but the air temperature is everywhere approx. the same. Any ideas?
5. I know the water content (in %) in the balls. Is it possible to simulate the drop rate of water content (in the ball) at a specific magnitude of airflow?

Please help guys.
Regards

[vinerm]

1. Whether the flow is compressible or not depends on Mach number. Since you can determiner average voidage, after filling in the balls, you can determine average flow velocity. This would give an idea about the Mach number. If it is close to or higher than 0.3, then you need to consider compressibility, otherwise no.

2. That depends on the objective. The flow would most likely be turbulent. k- might work but preferable will be k, if you are planning to simulate the balls by creating their geometric model.

3. 2D simulation cannot represent the scenario. There are two options in 2D, planar and axisymmetric. Planar cannot represent ball and axisymmetric can only represent balls in a line. A preferable option would be to consider a 3D but symmetric model.

4. Simplest way is to validate against the outlet temperature. As far as flow is concerned, you can validate using some balls marked with markers, such as drying chips. Those will dry up only if enough air flows over those.

5. Yes, you can simulate the drying process.

Usually, such processes are modeled and not simulated, i.e., the effect of balls is taken via porous media representation. That makes the life easier.

[RobertW]

@vinerm,
Thank you for your kind reply. Your response was very informative. I am grateful.

Please allow me to ask two more questions:
1. I think how to represent (in simulation) a large bunch of little balls. I consider a porous media formulation to model three dimensional models for balls mass configurations.
I read some papers presenting similar simulations (e.g. airflows through the grains stored in silos). But to be honest, I am still not sure how should I actually develop a porous media (in what way) in Fluent to physically represent a bunch of tens of thousands of little balls and 'voids' between these balls. Any tip would be much appreciated.

2. In terms of modeling of the drop rate of water content (in the ball) at a specific magnitude of airflow, I am looking in the literature, what material properties I need to know in order to effectively model it. Is the material heat capacity one of the most important? Would there be any other critical material properties to consider?

[vinerm]

Use of porous model to model flow through a packed bed requires data showing flow rate vs. pressure drop. This data can come from either experiment or detailed CFD model. If you have a lab where this testing can be done, then use experiment and get a few plots showing flow-rate or velocity vs. pressure drop across the inlet and outlet. If you do not have a lab where experiment can be done, then you need to take a small subset of the box containing balls and do a simulation. However, in this simulation, you should have real balls. You need to consider rather a small set and the top, bottom, left, and right boundaries should be considered translational periodic or symmetric depending upon where do you draw the line to create subset. Then, run flow simulations over this for multiple, at least 3, flow-rates and get a graph. Using this graph, you can calculate resistance coefficients required to use porous model. The procedure to determine these coefficients using flow-rate vs. pressure-drop graph is given in Fluent user manual.

[RobertW]

@Vinerm
Thank you for your kind clarification on how to deal with the porous media modeling. It helped me a lot and I am grateful.

Input data (to collect before modeling starts)
As I understand the input data to the airflow analysis are the geometry of balls (diameter) and the input velocity of the air penetrating these balls. Besides balls I also need to conduct the airflow analysis for rectangular tablets of convex and flat shape, so I will also need the precise CAD models of these. Would there be anything else that I may have overlooked?

Porous media modeling
Modeling in 3D certainly takes much more time than in 2D. I do not have much time allowed for this modeling.
For this reason, in order to shorten the overall modeling time, I wondered whether it would make sense to apply the following procedure:
Step 1 - I would take a small subset of the box containing balls and do a simulation (with real balls), as you suggested. So this part would be done in 3D. (I do not have an access to the lab and I cannot get experimental data showing flow rate v's pressure drop)

Step 2 - Then, I would use porous media modeling with use of 2D model. If something like that makes sense. I would conduct such 2D analysis for the entire cross-section of the container that is filled with the balls. The customer is predominantly interested in general information on how good their process is in an effective penetration of balls and in identification of areas where there might be a risk of stagnation point.

Thank you.
Regards

[vinerm]

Your understanding is correct. If you simulate the balls instead of modeling via porous domain, then you have to use 3D because in 2D, balls can only be represented along only one line, i.e., x-axis. However, once you have determined resistance coefficients that represent the box as a porous zone, then you can certainly use 2D. However, 2D would be based on the assumption that the resistance to flow is only in two-directions and there is essentially no flow in the z-direction.

Furthermore, while using porous model, since you won't have to simulate the balls, the geometry becomes quite simple; just a box. So, you can model even the entire box or half of it (symmetric model) and still the simulation would run quite fast because you can use rather uniform mesh.

[RobertW]

@Vinerm
Thank you for your confirmation.
I think that I will also need to validate the small 3D model, before 2D modeling. As you have suggested, I could use marked several balls inside the real container filled with balls (in a company), measure the water content in these, and compare to the simulated water drop rate.

May I ask you, do you think that I would need to specify the surface porosity of the ball in 2D and 3D model?

Thank you.
Regards

[vinerm]

If the balls are not highly porous, I'd recommend modeling those as solid balls. Surface porosity does not exist in itself, this is outcome of porosity of the body. So, either you have to consider all the balls as porous or solid. You may consider the roughness of the balls' surfaces though.

[RobertW]

@vinerm
Sorry, my mistake. I meant balls' surface roughness. I guess that once I get data of the surface roughness I can introduce it into Fluent.

[vinerm]

Yes, you can do that. However, surface roughness can only be used with wall functions.

[RobertW]

@vinerm
Thank you for your kind reply.

[RobertW]

@vinerm,
Thank you again for your kind answers to my earlier questions. I started my modeling in CAD software and Ansys. I wondered whether you could give me a few more hints.

Input data (from the 'real' world)
I developed a list of input data that I need to collect in order to develop a porous media model 'representing' a real scenario. The list includes as follows:
1. Geometry of a container that contains balls
2. Geometry of balls (besides balls, studied element are also rectangular tablets/pills - flat and pillow shape). Container can be filled with one element type at the time (e.g. balls only)
3. Application of random arrangement of pills (in preliminary 3D simulation inside a small box)
4. Ball's surface roughness
5. Material of the container (in my opinion not needed unless I would like to conduct a thermal analysis)
6. Material of ball's coating (again in my opinion not needed unless the analysis will involve something more than the airflow study).
7. The inlet - Air velocity/pressure or volumetric air flow rate, room temperature
8. The outlet - room temperature, atmospheric pressure
9. The 'inside' of container - room temperature, atmospheric pressure
Is it the complete list?

Roughness of a ball
Ball's (and rectangular pill's) surface roughness is rather a 'regular' one for tablet-like products of food/pharmaceutical industry. I cannot measure the roughness of studied elements. I wondered whether definition in Ansys of such surface makes a big difference (in particular for rectangular pills) if the surface is not 'extremally' rough or deformed.

Also, is there any set of ready to use 'standard' (approximated) values of roughness? Quick check in Google did not deliver anything interesting.

Definition of contact between balls
I develop 3D model of the container and balls in CAD software which I prefer over in-built CAD in Ansys. I wondered how to define the contact between balls. How important contact definition is for non-thermal analysis?

If I need to define contact between balls, would this be a good idea to overlap the balls a little bit and develop one solid from bunch of balls. If so, will that create a problem for meshing (small meshing elements in 'contact' areas between balls v's large meshing elements elsewhere). Will little elements of meshing in areas of balls 'contact' cause 'strange' results.
If I do not overlap balls little bit, what would be the best approach to define contact between balls (I will probably have a few hundreds of balls in a small 3D box - in reality there are a few hundred thousands of them in a container).

Porous media model - alternative
While reading papers on airflow study through a bunch of little elements (such as grains), I can read about porous media modeling only. Is it by far the most popular model for such analyses? Did you came across a reasonable alternative.

Thank you.
Regards

[vinerm]

Yes, this would suffice. Since there is not flow through the balls, you don't really need to be very precise about their shape. You may assume that balls are hard-pressed against each other. Essentially, you should be able to generate a good mesh with a domain representing balls.

The standard option is to use porous media since this is accurate enough and rather inexpensive. Other options are to use Packed Bed model but this is available only under Multiphase flow and you cannot apply any surface roughness. The benefit is that you don't have to create model of balls rather just the box with size of balls specified. Results might be comparable to porous model. Another, more accurate, option is to use Macro Particle Model in Fluent. This is not really meant for such scenarios, rather where particles move, but can be used for you case.

[RobertW]

@vinerm
Thank you for your quick answers. I am grateful.
I started my modeling in CAD software. I think that I will stay with porous media model.
But I will take a closer look at the Packed Bed model and the Macro Particle Model.

Thank you again.

Regards

[RobertW]

@vinerm
I have just started another thread (titled as '3D simulation of airflow through small objects', 'Fluent' section of this forum), related to the topic discussed in this thread. I present there a little bit different scenario for modeling. I thought that perhaps you may also be interested in that new thread. Thanks again vinerm for your help in this thread.