[ajjadhav]

Hello,
I am doing solid-liquid flow in a stirred vessel.
Case 1: liquid phase water (90 vol%) and solid phase: Glass beads 1 mm (10vol%)

Case 2: liquid phase water (80 vol%) and solid phase: Glass beads 1 mm (20 vol%)

I use two approaches Eulerian multiphase model or Lagrangian Particle Tracking.

In case of Eulerian multiphase model I simply use water as continuous phase and glass bead a as dispersed phase and there respective volume fractions for both cases ( as mentioned above). My results are matching with experiment very well.


I am struggling to perform Lagrangian Particle Tracking.
I don’t know how do start.
Lagrangian Particle Tracking uses the representatives particles to assess the behaviour of solid phase.
So should I take water as a continuous phase and (1000) glass beads as a transport solid? What about the volume fractions of both phases? How does cfx knows that i have 10 vol% or 20vol% of solid?
Or should i use eulerian multiphase model results as a initial for Lagrangian particle tracking?
Please guid me.
Thank you in advance.

[ghorrocks]

In Lagrangian Particle Tracking, each tracked particle represents a number of real particles. So you choose the number of modelled particles to be high enough that it accurately captures the distribution of tracks (in fact, you should do a sensitivity study to confirm this, just like you do with mesh size and convergence criteria). The number of real particles is determined by the volume fraction and the particle size distribution you define.

Yes, you can use the Eularian model as an initial condition for the Lagrangian - I suspect it will only initialise the continuous phase, the particle phase will need to be calculated from scratch but that is better than nothing.

As you have seen, getting accurate results with Eularian Particle Tracking is much easier than with Lagrangian Particle tracking. People really should use the Eularian approach more often rather than just defaulting to the Lagrangian approach.

[ajjadhav]

Thank you Glenn for your reply,
As you mentioned real number of particles can be calculated using volume fraction, size distribution and density.
Can you please let me know how many particles do i have in the first case where volume fractions is 10 vol%, particle size is 1 mm and density 2485 kg/m3 and stirred tank is 288 mm?
Thank you

[ghorrocks]

You need to be able to do calculations like this yourself if you expect to be able to do multiphase CFD.

[ajjadhav]

That’s my point, how is this possible, using only volume fraction and size we can calculate number of particles? We also need total volume of the system.

[ghorrocks]

"We also need total volume of the system." - Yes, that is correct.

Are you commenting on my post #2? Yes, you would need the total volume of the system as well, if you are calculating the total number of particles.

[ajjadhav]

Yes Glenn I am commenting on your post.



My basic problem is that what should be the mass flow rate in my case (for 10 vol% solid loading)? I read that cfx needs mass flow rate for calculation of number of real particles in the system.
Hows does cfx calculate number of real particles?

[ghorrocks]

If you have an inlet or opening then you can define the conditions of the particles at the inlet/opening, including the particle mass flow rate. So will have to do some work to calculate the particle mass flow rate required to give the volume fraction you want.

[ajjadhav]

Thank you Glenn for your reply. That's a good suggestion.

The issue is I don't have an inlet or outlet.
I have sittered vessel with impeller and baffles.

[ghorrocks]

I have not found an easy way to define initial conditions for Lagrangian particles. You might need to do user fortran to do this.

[JuPa]

Hi Ajjadhav - I had a similar issue where I was modelling a stirred tank reactor with no inlets and outlets.


I was not doing Lagrangian particle tracking but I had an issue with initializing the simulation before running. I solved it in the following way:

• In real life, physically there MUST be an inlet and an outlet pipe to pump material into and out of the tank, right?
• In my physical domain I included the inlet and outlet pipes.
• I first ran a "tank filling" simulation where the tank filled via the inlet.
• I then used that as the initial conditions for an Eulerian multiphase simulation (setting the inlet and outlets as walls).

Now I suspect you need to do the same thing:

1. Run a simulation filling the tank from the inlet, but include your particles!
2. Use this as the initial condition for your "closed" lagrangian tank mixing simulation.

It's going to be very difficult for you to do this as 1 simulation - you need to break it down into a few mini simulations and build a story.

[ajjadhav]

Hello JuPa,
Thank you for the suggestion, I will definitely work on it.

What do you think about, using the eulerian multiphase model simulation results as an initial for Lagrangian particle tracking eventhough it doesn't make sense?

Thank you
Regards
Ananda

[JuPa]

In theory it should make sense. As long as your material names are the same then you'll have volume fractions, velocity fields etc for your water phase and the same variables for your glass beads.


Now weather or not CFX is smart enough to pick that up is another question!

I suggest do a simple 2D test. Initialise a simple straight forward two-phase Eulerian simulation. Run it for 10 iterations or so. Then see what results you get when you try to use the Eulerian results to initialize particle tracking results.

If Eulerian results can't be read by particle tracking results there's potentially another way...manually export Eulerian results from CFD post and manually enter them into a particle track file which contains your initial conditions.

It'll be a pain in the arse, and it might take you a full day of boring copy and pasting - BUT then it'll be done. You might be able to script it in Python or something.


Good luck!

[ajjadhav]

Hello JuPa,
I will work on your suggestion.

I worked on using Eulerian simulation as an initial condition for Lagrangian Simulation and the results (velocity profile) exactly match with the experiments. It means, the idea is right.

Do you know how to extract the individual particle track?

I am using 1000 Lagrangian particles, when I export it I get a .csv file containing data but is it mixed, I want data to be separate with particle number ot ID.


Thank you
Regards
Ananda

[JuPa]

I don't know I'm afraid - you might have better luck asking on the Ansys forums.


There might be a way to cheat and use Eulerian...


Let's say your glass beads are 5 mm diameter. You could potentially set up a MUSIG model simulation, with diameters of 5 mm plus a statistically insignificant distribution.


E.g.
Group 1: 5 mm + 1 micron
Group 2: 5 mm + 2 microns
...
Group n: 5 mm + 10 microns

That way you could use Eulerian to track individual particle groups.


If you do go down this route do calculate the particle Stokes number number first to confirm the added diameters are statistically insignificant.



As Glenn would say - do a sensitivity check!