Hi Parth,

to be honest I don't have much experience with DPM modelling, but perhaps the following ideas might be helpful though:

First, the pressure based solver should be fine as long as the Mach number of your system isn't too high. Second, it sounds reasonable that you get better convergence prior to the injection, because then it's essentially a single-phase flow of your continuous phase without any interactions between the Eulerian and Lagrangian phases etc.. Consequently, your system should achieve some sort of (quasi-) steady state, which explains the low number of iterations required per time step. In particular during the start of injection you should obtain the largest deviation from this initial state, I assume...

Which convergence criteria do you use (Solution > Monitor > Residuals)? Which residual levels are you trying to achieve (default is typically 0.001)?  Did you already tried to use an adjustable time step size? The defined fixed time step size of 1e-5 might still be too large (I don't know how fast your process is). Since the Courant number is related to both, the time step size and the grid density, you should check if your settings fit this relationship. Also, how many particles are in your system and did you define any interaction with the continuous phase? If so, it might be advantageous to switch those interactions off for some time steps and see if this improves the solution (depending on what you want to model...).

Hope this helps!

Best regards,

Stephan

Hello Sir,

Thank you so much for your response.

• I am using the convergence criteria as 0.001 (Default in Fluent). 
• I have not tried to use the adaptive time steps, as I read somewhere that Adaptive time step technique is not useful in Discrete phase and VOF model. I am not sure about that. I think I should try adaptive technique first.
• I am defining the interaction between discrete phase and the continuous fluid.

I am not sure that my time steps are so large, that might be a good suggestion for me. I will try 1e-6. 

In my case, There is also evaporation of droplet is included and collision model is also included. I am using courant number as 50, then it works fine till the start of injection. After few time steps, even after 5000 iterations per time steps doesnt work for convergence. Then I have to increase Courant number after few time steps repeatedly and it again converges for some time. Then same problem arises. I dont know what should I do apart from this. 

I really appreciate your help Sir.

Thank you.

A Courant number of 50 seems to be rather high. What level for the residuals do you achieve (after the injection starts), and are all residuals in the same order of magnitude? Usually, I also start with simple discretization schemes (e.g. First-Order-Upwind), which typically show better convergence, and compare the outcomes with higher discretization schemes if required. Of course, in unsteady cases it also depends on which time scale you're interested in: do you need an accurate resolution of the transient behavior (i.e. the starting phase), or are you more interested in a final (quasi-)steady state, if there is any in your case?

Last but not least: I'm a little confused that you mentioned the VOF model in your answer. Do you have two Eulerian phases (plus the Lagrange phase) in your system?

Best regards,

Stephan

Hello Sir,

After starting the injection generally my residuals are in good manner towards the convergence except continuity. Continuity residual stays way far from convergence. After injection, the continuity fluctuates within the range of 1e-1 to 1e0 or even 1e+1. Sometimes I try to increase my fluent number in the range of few hundreds and again for few times it reaches near convergence and may be one or two time steps gets converged also, but then same problem repeats again and again. I am out of my mind now, what is the problem. I also tried with the smaller time step size (1e-6 s) and grid is also refined to 200k to 456k. ( But only 10% solution is resolved). 

I am more interested in the visualization of the development of the spray not the final steady state. 

About VOF, I am not using VOF. I have only one eularian phase with one langragian phase. This two phases have heat and mass transfer within interaction. 

I have attached my residual snap. Please have a look at it.

Thanks a lot for your response and hope to hear from you ahead.

Thanks again,

Parth.

Hi there,

I am also using unsteady particle tracking to track liquid jet breakup in a crossflow. I solve the continuous phase first, then I inject particles using dpm including the interaction with continuous phase but I am getting zero particles tracked. Is there a solution to this? My flow time after continuous phase convergence is 0.002s so my injection start time is also 0.002s is this wrong?

As far as I know, the injection start time is the absolute flow time at which the injection should start. Let's say you simulate the continuous phase for 5 s - until it reaches a steady state - and then you want to inject your particles. In this case, you should define the injection start time to be 5s and, of course, the injection stop time has to be defined too (with a value greater than the start time). Also, make sure that the location of your injection is set correctly. This should work! If not, please provide details about your injection and particle tracking settings to get further help.

Best regards,

Stephan

Hi stephan,

thank you for your answer I have managed to track the particles now however when I use the plain orifice atomiser model the nozzle is constantly in single phase and breakup is not occurring. I am trying to validate the spray model using the experimental data from h.k.suh and c.s.lee but the result I get are way off. I am now going to try using other injection models (group/single) if you had any suggestions or advice I would be very grateful.

koyal

Hello Koyal,

I'm sorry - haven't used the atomiser DPM model yet and don't know the work of H.K. Suh and C.S. Lee.  This is beyond my experience and field of researd, and that's why it's hard to give any suggestions or advice. Good luck anyway.

Stephan