[KiruBiru]

Hi guys,

Another challenge: need to simulate an air particle filter like e.g. HEPA. HVAC application, the medium is air.

Not sure how to do that.
As far as I understand, I need to draw the filter with some specific thickness and amount of folds, and then make a porous domain out of it? (the settings below are not mine).

Are there any good tutorials on how to do that and what these settings actually mean and where do they come from? Thanks in advance!

[ghorrocks]

Before you touch the software you need to consider what role this filter plays in your system and how detailed a model of the filter you require. If the filter is a small part of a larger system then a simple plane with a pressure loss as a function of flow rate might be enough. If you are looking at the details of flow around the filter you might need to model the shape of the filter. If you are looking at how the filter traps particles you might need to model the fibres which make the filter. It all depends on what you are trying to do.

[KiruBiru]

Hi ghorrocks,

You made it clear. It is apart of a larger system, so the first option should suffice. As for creating a plane with a pressure loss as a function of flow rate, could you please suggest where can I find info / tutorials on how to set this up?

[ghorrocks]

In that case the easiest way forward is to make your filter an interface and apply a pressure drop to it. You could also use a source term to do it as well.

[Qkarl]

Originally Posted by ghorrocks

Before you touch the software you need to consider what role this filter plays in your system and how detailed a model of the filter you require. If the filter is a small part of a larger system then a simple plane with a pressure loss as a function of flow rate might be enough. If you are looking at the details of flow around the filter you might need to model the shape of the filter. If you are looking at how the filter traps particles you might need to model the fibres which make the filter. It all depends on what you are trying to do.

Hi, what if 2nd and 3rd option? any idea can enlighten me? thanks

[ghorrocks]

They are just increasing levels of fidelity in the filter model. Along with the increased fidelity will come a massive increase in computational costs, meaning that what you can practically simulate will get smaller and smaller. But, as I said first time, it is all about what you are trying to achieve.

[AntonioDesah]

Hi Ghorrocks. I’m interesting on studying how the filter traps the particle. How shall I model the fibers? The only information I know at the moment is for an assigned material a ratio expressed in grams. Any help ?

[ghorrocks]

What particle trapping physics are you intending to model? Many things are happening in a filter at the micro-scale. If you don't know what physics is occurring then you cannot model it.

[AntonioDesah]

Originally Posted by ghorrocks

What particle trapping physics are you intending to model? Many things are happening in a filter at the micro-scale. If you don't know what physics is occurring then you cannot model it.

Thanks for your super prompt reply ghorrocks! I would like to study the power of filtering of different materials for a facial mask. The physics of the problem would be the following :

Steady state flow air domain.
At a certain point of the domain as source will spray aerosol because a virus travels inside droplet of water contained in the air.( the source will simulate a person who sneezes)

In the middle of the domain I can insert a membrane and assign to it a porous wall B.C.

At the first level of micro-scale simulation I can study how the droplets of water are trapped through the fiber of the filter. If at this stage I can already identify which material among the ones I have to simulate, can trap more droplet my simulation is already enough detailed.

If a deeper level of simulation is required I can add some particles inside the droplets which will simulate a pathogen agent ( diameter and mass known)

My concern is how to model the fibers of material. Can I consider, at first attempt, a plane transverse to the direction ?( for the sake of simplicity just a vertical line in the rectangular fluid domain ) or I need to simulate it as a grid very very fine ?

Thanks in advance for your reply.

[ghorrocks]

You comment about using a porous material for the filter means (in my opinion, I am no expert on filter modelling) that you could model the amount of air which goes through the filter compared to any leakage paths at the edges of the filter. But I suspect CFD won't tell you much here because a simple back of the envelope calculation making some reasonable assumptions about pressure loss is likely to be just as accurate.

But I cannot see how you can model particle trapping with a porous model. There is no particle trapping model in the porous material model.

Your other comment appears to suggest that you would directly model the filter as fibres and directly model the trapping process. Most filters these days are of the HEPA type which do not have a single filter element, but rely on lots of fibres closely packed together. The paths through the HEPA filter are often larger than the particles you are trying to filter out - they work by making the path convoluted such that the particle is bound to strike the wall some where, and when it does it sticks. This convoluted path is made up of hundreds of fibres layered on each other is random arrangements.

So I cannot see how you can practically model a HEPA filter directly. They are just too complex. Even getting a reasonable measurement of the filter structure sounds like a massive task to me.

Are you intending modelling a HEPA type filter? If so how do you propose to overcome these problems? Or are you proposing a different type of filter?

[AntonioDesah]

Originally Posted by ghorrocks

You comment about using a porous material for the filter means (in my opinion, I am no expert on filter modelling) that you could model the amount of air which goes through the filter compared to any leakage paths at the edges of the filter. But I suspect CFD won't tell you much here because a simple back of the envelope calculation making some reasonable assumptions about pressure loss is likely to be just as accurate.

But I cannot see how you can model particle trapping with a porous model. There is no particle trapping model in the porous material model.

Your other comment appears to suggest that you would directly model the filter as fibres and directly model the trapping process. Most filters these days are of the HEPA type which do not have a single filter element, but rely on lots of fibres closely packed together. The paths through the HEPA filter are often larger than the particles you are trying to filter out - they work by making the path convoluted such that the particle is bound to strike the wall some where, and when it does it sticks. This convoluted path is made up of hundreds of fibres layered on each other is random arrangements.

So I cannot see how you can practically model a HEPA filter directly. They are just too complex. Even getting a reasonable measurement of the filter structure sounds like a massive task to me.

Are you intending modelling a HEPA type filter? If so how do you propose to overcome these problems? Or are you proposing a different type of filter?

Thanks a lot once again! Considering that the scope of the analysis is not to Study then flow but to understand among 5 materials already built,which one has the highest filtering power and considering that due to the current circumstances the analysis has to be run fast enough I was thinking the following in a very simple way:

The fluid will be air plus water droplets.
The mask, due to indeed the very complexity of the fiber, will be model as pourous surface b.c.

Analysis of the results: loss of mass of water due to the presence of the mask.
I am expecting that it is a multiphase flow air plus water therefore I can control both phases mass flow conservation. The air will still go through the material whereas the water won’t or even if it does it will do it in a very long time potentially infinite( again due to the complexity of material would be very difficult to exact simulate what happens to the water but it is beyond the scope of the simulation I think)

I believe that I’m capturing enough for the scope of the simulation ? What do you think?
Is my approach completely wrong ?

[ghorrocks]

What data or measurements do you have of the 5 materials?

Why are you doing this by CFD rather than by experimentation?

[AntonioDesah]

Because it’s at a preliminary stage of design.

[Gert-Jan]

If you want to use a porous volume, how do you distinguish between various filters? For this, you need to feed your model with experimental data. So aren't you finshed by then?

Alternatively you need to model the physics on microscale with fibers and droplets. But that is more or less undoable. Too many details required. And a lot of unknowns as well.

In other words, I think a CFD-model is quite useless.

[AntonioDesah]

Originally Posted by Gert-Jan

If you want to use a porous volume, how do you distinguish between various filters? For this, you need to feed your model with experimental data. So aren't you finshed by then?

I could run different simulations with different material and identify the most filtering material with the drop of water mass flow. I can’t run experiments at the moment. I need just to chose among different material justifying the choice. That’s the reason why I was going for a CFD simulation.

[Gert-Jan]

That is not CFX works. It is not that you just tell the software: "These materials are cotton, nylon, asbest, glassfiber. Now calculate the waterflow through it for pressure drop x".

To perform that calculation, CFX requires the resistance per m of the material. So it demands the information that you want to know.

[ghorrocks]

Gert-Jan's comments are correct. Unless you know the precise details of what you are modelling (diameter ranges, packing factor, surface chemistry, how the fibres intersect each other etc) then you cannot model this. And even if you got this data the simulation would be massive and complex - think super computer and a year's development work - so completely impractical for a preliminary choice of material in your case. And a porous material model won't work either as that just affects the flow resistance, it has no model on how it affects particle trapping.

I cannot see how CFD is going to help you in this analysis. It should be done by experimentation.

[AntonioDesah]

Originally Posted by ghorrocks

Gert-Jan's comments are correct. Unless you know the precise details of what you are modelling (diameter ranges, packing factor, surface chemistry, how the fibres intersect each other etc) then you cannot model this. And even if you got this data the simulation would be massive and complex - think super computer and a year's development work - so completely impractical for a preliminary choice of material in your case. And a porous material model won't work either as that just affects the flow resistance, it has no model on how it affects particle trapping.

I cannot see how CFD is going to help you in this analysis. It should be done by experimentation.

https://link.springer.com/article/10...09-019-00218-7
In this research paper, there is the case study I would like to reproduce. Assuming that I could know the diameter of the fiber and assuming the same random distribution of the fibers for all the materials, do you think that it is still undoable ?

How would you guys do it if you don’t have the option at the moment, to run lab experiments?

[Gert-Jan]

So they do it on microscale. With fibers. That is something different than a porous volume.........

On microscale it is possible. But they use a certain volume with fibers. You need to draw something similar. You need to draw cottom fibers, or asbest, glass fibers, at least something...... How do you know how they look like and are orientated? You need literature or a microscope.

Then the fibers need to capture the particles. How do you include their interaction? You can do it like in the article. But is it representative for your case? Maybe. Maybe not. I don't have the time to read it in detail. If not, then you have to feed your model with experimental data or other literature data.

Then in the end, you might be able to caclulate how your volume with fibers works, and what pressure drop it has. Then you can scale it up to a larger porous volume. But this is a lot of work. At least a year with a lot of uncertainty.

[ghorrocks]

Assuming that I could know the diameter of the fiber and assuming the same random distribution of the fibers for all the materials, do you think that it is still undoable ?

If you know the diameters, orientations and packing, then simply the most packed filter with the smallest holes and most convoluted path is going to be the best filter. You don't need CFD to tell you that.