Learning Forum

[massimiliano.nori]

Dear Sir / Madam,

I would like to know how to speed up particle tracking simulation for the two phase eulerian model in Ansys Fluent, As i have  12 million elements it takes much longer time to proceed simulation and also continuous phase experiences divergence after injecting the particles. Kindly help on this matter. 

 

Regards

Nori Massimiliano 

[Rob]

Please can you confirm the model settings? With the Eulerian Multiphase model the particle phase would be added at a boundary; an injection is used for DPM.  Images will help. 

[massimiliano.nori]

I am using K-epsilon with realizable, Eulerian two phase (vapour and liquid), Iiquids are injected as particles at the inlet surface by enabling pressure gradient force and two way turbulence coupling. Continous phase runs without any divergenece but the divergence of residuals are detected and rises to high value when particles injected. DPM settings; number of time steps 500000, step length - 5, DPM iterations update every 1 iteration of continuous phase. Coupled solver choosen.

 

[Rob]

You may find the solution is transient as the Euler models tend to have results that vary with time.  However, with the mass input from the injection, what volume fraction are you seeing? You may find DPM, pure Eulerian Granular or Ansys Rocky to be better options. 

[massimiliano.nori]

I am using steady state continuous and DPM phase simulation. I am injecting the particles after the continuous phase has reached convergence, the moment after injected the particles in few iterations k and epsilon values becomes very high and getting floating point exception error. i cant proceed further to run the simulation. i am using mass flow inlet and pressure outlet BC. DPM - mass flow rate of particles and velocity calculated for the inlet boundary provided. I am using eulerian granular option. I dont see any volume fraction at outlet as the particle injection are given afterwards. Could you clarify what is ansys rocky?

[Rob]

If you do a hand calculation what volume fraction of particles are you injecting? DDPM links into the Eulerian Multiphase model, hence my comments: in many cases you're better off with DPM or Eulerian or Rocky. Ansys Rocky is a DEM solver, and can be coupled to Fluent to extend the multiphase capabilities. https://www.ansys.com/en-gb/products/fluids/ansys-rocky covers the basics. 

[massimiliano.nori]

Volume fraction of liquid is around 0.8 % and we face the divergence problem after injection of liquid particles. Moreover we run the simulation with two eulerian phases and  convergence achieved. After injection reverse flow warning appears almost in 1000 faces near the outlet because of which solution not able converge and ends up with floating point error.  

[Rob]

With 0.8% I'd not bother with DDPM. As I have no idea what you're seeing it's very difficult to make any suggestions. 

[Sudalai Manikandan]

Considering the above scenario, Is it possible to use Single phase + particle injection to represent the two phase flow ith injection? 

[Rob]

Quite probably. DPM will probably run steady, so will be quicker too. 

[Sudalai Manikandan]

How can i get the optimised value of number of time steps and step length so that i dont get incomplete particles? Could you suggest any desired settings to start with? 

[Rob]

Start with the default number of steps and see what happens: they're usually good enough. If you get a high number of incomplete tracks look to see why before increasing the maximum number. If tracks are stuck somewhere you may just use more cpu for no gain; if the tracks are still heading somewhere but (for example) swirling as in a cyclone try doubling the max number. Fluent uses up to that number, but if you get a rogue parcel or few and add a silly maximum value you could be waiting for a while for no good reason. 

 

[Sudalai Manikandan]

 

I uses more than 5 million time steps but still results 10% of injected are incomplete. my geometry is about 3 by 15 m horizantal cylindrical vessel. could you suggest any other settings shall improve my result. As i am investigating liquid particle diameter at the outlet section.

 

[Rob]

Not without seeing why those particles aren't reaching the outlet. You can't use single metrics in isolation, we'd typically use a mix of images and numercial data when investigating a model result.