Learning Forum

[javat33489]

Hello everyone. My previous post about erosion was deleted. I am writing a new, more updated post.
I am doing an erosion calculation.
I made a simple test model of 150K cells.
I am using the k-epsilon realizable turbulence model.
53 kg/s of water at the input, atmospheric pressure at the output.
I am doing the calculation without a discrete phase for 100 timesteps, everything is going well, there is convergence.
I am injecting through a proppant inlet with a diameter of 14 mm, 25 kg/s (half of the water is proppant beads).
I am tracking the particles, it shows 290 particles entering and exiting.
I am turning on erosion and the erosion grid.
I select the oka model and set the density of the material subject to erosion as for steel 8700.
I set up the calculation, select a fixed timestep (otherwise my program crashes, I don't know why). I select a timestep of 1 sec, 100 substeps, recording every 100 steps, set up a grid for broadcasting in real time during the calculation.

Here are the main screenshots of the settings that I described above:

 

 

Problems:
1. At first, everything goes well. Gradually, the graph shows erosion. But then I get a floating point exception. Why?
2. I do not see the deformation of the grid in real time, and I get an error in the console:

Please help. Please give me some advice.
P/S. Also, please give me some advice on which timestep is better to choose for such problems? If you select too small, for example 0.001 sec and the total calculation time is 3600 sec, it will take a very long time to calculate, considering that you still need to set substeps. If you set a very large step, a negative volume grid error occurs. I need your advice. Which timestep and substeps are best suited.

[Rob]

The Forum's had a few issues - if the other thread reappears I'll remove it. 

Looking at the above erosion is working as expected but the views don't seem to be behaving. If you stop the model and plot distortion/displacement on the wall has it got a non-zero value? Can you also check in Views (top ribbon) and see if the views are present. That looks to be triggering a graphic error rather than killing Fluent. 

I suspect the error is down to the deformation and remeshing. The erosion MDM does (or did - looks like I need to rerun my test model) just smooth the mesh to suit the new shape. With inflation that could cause problems, and given the shape is changing inflation meshing isn't necessarily a good idea anyway. 

There is no recommended erosion time step size, other than it's how long it takes to measure the lengh of a piece of string. It's very dependent on the erosion material pair: ie polystyene beads onto inconel may need a time step in the order of years but steel pellets onto chocolate may need microseconds. Given you're modelling a known real system how long is it expected to last? Then base the time scale on that. Eg if I expect an item to last a month I might run a one hour erosion time step and run in blocks of 24. 

 

[javat33489]

Thank you for your reply sir.

1. Yes, I realized that inflation is not the best solution and I removed it but it did not help.

I also tried to make the step smaller, for example, timestep 1 second, total calculation time 3600 substeps 100. I still get the point exception error no matter what I do.

I also tried to refine the grid to 10 million cells. Nothing helps.

 

2. Sir, I noticed some warnings during the calculation, which leads to the error:

 

3. Sir, I also tried to make the grid using FLUENT MESHING and its mesher. Now I get the negative volume error almost immediately when I start the calculation, and I also see artifacts in the form of a knocked out grid. I tried to make timesteps smaller, but it does not help.

 

Sir, I really hope for help.

[Rob]

That's a classic symptom of a very large parcel hitting a fairly small cell. How many parcels are in the domain?

[javat33489]

Sir, at a flow rate of 53 l/s of water, I have 25 kg/s of proppant. The size of the proppant is 1.4 mm, here is the setting:

170 particles pass through the interface:

[Rob]

Which means each parcel is around 25/170 kg/s which is a lot of mass hitting what I assume is a small facet. Try using stochastic tries in Turbulent Dispersion, 10-20 may be suitable to reduce the parcel mass. 

[javat33489]

thanks for the reply sir but it didn't help:

[javat33489]

Sir, can you give me some more advice?
I also tried increasing the number to 50 in stochastic, it didn't help.

[Rob]

OK, from the image you can see the problem. Given erosion is a function of parcel mass and facet size what do you think you can do?

[javat33489]

I don't know what to do sir so I turned to the forum.

[Rob]

Read what I wrote carefully. You don't need to look at the software to answer. 

[javat33489]

OK, from the image you can see the problem. Given erosion is a function of parcel mass and facet size what do you think you can do?

-Sir, do I need to reduce the parcel mass?

Sir, did I set the condition correctly?
At the input, I set 53 kg/s of water, I also needed to specify that half of this solution is proppant and I specified the number of particles 25 kg/s:

Will this mean that the total volume of this mixture will be 53 kg/s (28 water, 25 proppant particles)?

 

[Rob]

Reducing the injected mass changes the problem. So, you could increase the number of tries or decrease the mesh resolution. I'd also be wary of scaling injection flow by face area as you can get a fairly large variation in parcel mass. 

[javat33489]

Thanks for the reply sir.
But I don't fully understand what settings I should make now.

[Rob]

Think carefully about what I said. If a heavy parcel hits a small facet you have a problem. So, what are your options? 

[javat33489]

Sir, I understood that you are talking about reducing the package. I mean, what setting should I edit in fluent?

[Rob]

Please answer the question. No need to look at panels - look at the problem. We can worry about what panel to change later. 

[javat33489]

Yes sir I answer the question you need to change the mass of the parcels, but then I will worry that the calculation will be wrong, because in reality they upload exactly that many particles.

Sir I also tried to reduce the number of particles to 10 kg/s to 5 kg/s, I still get an error, I also tried to increase the number of attempts to 250, I also improved the mesh in this place, but nothing helped.

[Rob]

If you make the facets bigger it'll help. Reducing the mass injected will help. As will increasing the stochastic tries. Or reducing the erosion time step and duration.  

[javat33489]

Ok sir thank you I will try again

[javat33489]

sir, I tested different particle behavior settings, regardless of the stochastic number of attempts setting, I see the same picture in particle tracking, I see that the particles are bunched up in lines, these lines are destroying my grid. How can I scatter them over the entire area?

[Rob]

That implies the isn't a great deal of turbulent dispersion, so the parcels aren't separating. You can try adapting the inlet region as that'll increase the number of facets and so increase the parcel count. However it also suggests the particles may not be arriving at the unit like that in reality so some more thought into where/how many of particles need to be added into the model: that's a question for the experimentalists. 

[javat33489]

Sir. Please tell me. Is it possible in Fluent or in SFD postprocessor to paint the deformed mesh from erosion in a different color to highlight it and visually show the affected areas?

[Rob]

Have a look in the contours, in (I think Erosion) there's a distortion/displacement value. That should show the unchanged geometry as blue and then the usual colour map. No reason you can't play with the scale and/or isoclips to further highlight those zones. 

[javat33489]

Sir, when solving a problem in small time steps, can I solve the problem for 60 seconds and then perform further averaging linearly on the graph for 3600 seconds?

[Rob]

I can't answer that as it's entirely case dependent. If you want to take that approach why not just post process the erosion on a fixed mesh? 

[javat33489]

yes sir you are right!
and what is the correct way to do it with a fixed grid? just don't turn on the dynamic grid?

[Rob]

Correct - erosion can be done as a post processing exercise and was until the moving mesh module was written as a UDF. 

[javat33489]

sir, I didn't find among the standard options to display a deformed or eroded mesh in SFD prepost. Maybe it's some kind of expression? Can you tell me in detail how to display it?