'SQR_MLTPLT' simulation project by oaraujom


#1

I created a new simulation project called 'SQR_MLTPLT':

Square nozzle multiplier


More of my public projects can be found here.


#2

@oaraujom - are you looking into running a flow simulation through this nozzle? What’s the nozzle meant for? If you expect a turbulent flow, you could also looking into adding layers close to the physical wall to resolve the turbulent boundary layer better.


#3

Hello, and thank you for looking out.

Ideally yes, I’d like to refine my mesh boundary, but have been unsuccesful. The flow is meant to be laminar. This is supposed to be a layer multiplying mechanism for polymer flow. You have two different polymers going in through the inlets and you would get 4 layers out through the nozzle.

I know this might not be feasible, so I was thinking trying only with one fluid. Maybe even newtonian.

I’d appreciate your insight on this.


#4

@oaraujom - super interesting application! Will the polymers be physically very different? So would you need to account for surface tension effects and the likes? Otherwise the “passive scalar” approach might be worth a try.

Still not sure if I understood the application correctly - you would get 4 layers at the output? Why?


#5

The polymers should be relatively similar, say both ABS with a slight difference in viscosity and/or density.

Yes, after one pass through that layer multiplier you should be able to go from 2 layers to 4. Say you have polymer A and B (AB), then you separate them so you would have two AB arangements parallel, and then you would stack them on top of each other. That would give you an ABAB arrangement, if you were to drive that through another multiplier you would get 8 layers, and so on.

Not sure if I’m being too confusing.


#6

@oaraujom - ah, got you. When looking at the flow channel initially I didn’t get how 4 layers should be created, but if one really thinks it through, it makes sense :smile:

I quickly copied it and threw a bunch of layers near the walls into the mesh (the rest is your settings). The sim setup uses now a very simple model: The two fluids are considered physically equivalent, so no interaction whatsoever is computed for. For starters, I’m using a Newtonian model, but this could be changed easily. I ran it with 0.2 m/s is this too fast? Let’s see what happens.


#7

@oaraujom - interesting! It actually starts working with such a simple approach:

https://www.simscale.com/projects/dheiny/sqr_mltplt



You would need to tweak physical quantities such as viscosity, material model and flow speed, but the effect is clearly visible.


#8

That looks great.

I tried running non-newtonian, but it kept crashing on me. I ended up adjusting for different viscosities, and now I will try different speeds.

How did you post process for the layers snapshots and the tracing like yo did above?

BTW, thank you for the help. I found some of the simulations I did before did actually work.


#9

@oaraujom - thanks! Yeah, non-newtonian models can be tricky. I’m also not the expert when it comes to this, but @gholami has a lot of experience here, maybe he has a good hint?

Regarding the post-processing: I download the results and use Paraview for local post-processing. The first viz is simply showing the passive scalar quantity (“T1”) on the outside of the flow domain. As the BC on the walls says “Gradient to zero” you actually see the value from outside (other than velocity where it’s everywhere zero).

The second one uses the slice filter applied on a range across the z coordinate


The third one is a Contour plot set to 0.5 for the passive scalar “T1”. This basically is a little trick saying: The passive scalar 1 means fluid 1, the passive scalar 0 means fluid 0 so we consider 0.5 being the interface between them:


By the way: Using the passive scalar approach might also be a wrong assumption here: It’s only a passive species transported in a homogeneous flow field, so it actually moves also into the the other flow. It might be better to use here a real multiphase flow solver.


#10

Yeah, non-Newtonian can be tricky.

I ended up getting similar images using Simscale’s post processor yesterday. I’ll definitely keep on tweaking them here and there. Also, I don’ have much experience in Paraview yet.

Anyways, thank you for the explanation and for all the help! I really, really liked Simscale.


#11

@oaraujom - any luck running the non-newtonian flow through the nozzle multiplier? Would be very interesting to see the results.