I am completing a transient simulation using Ansys's MPM (macroscopic particle model), but want to conduct a grid independence check on the geometry to make sure my mesh is refined enough for solutions to be accurate. Can I conduct these tests using the initial solutions - i.e. no additional injected particles, just a steady simulation with a mono-phase fluid domain? Or would I need to conduct the study whilst having the secondary phase/particle active? The former would obviously be preferable, as simulations take far less time to complete and so results can be obtained more quickly. Additionally, are there any other similar such tests which people would recommend I conduct in order to achieve accurate results?
Thank you in advance for your help.
Is this one-way or two-way coupling?
If one-way coupling, you can check the converged steady-state Eulerian solution for grid independence. Then after injecting particles, you would check the independence of the particles from the "step factor", which is less costly than refining the mesh and re-running.
if it's two-way coupling, you would first check the grid independence of the Eulerian solution. Then run the simulation with particles, look for high source term locations in the grid, and refine there. Then re-run to see if you are grid-independent. Finally, you would perform a check on the "step factor" for the particles to ensure they are independent for tracking purposes. This can be very tedious but is necessary to ensure the two-way coupling (source terms) are correctly applied and the solution is truly grid-independent. Note that in two-way coupling, you may also have to vary the URF for the source terms to get a converged solution, and reference the User Manual to see how URF affects the convergence of the DPM solution.
Hi Jonathan Komperda, is the 'step factor' related to altering the number of time steps, max interactions per time step and courant number, or is this something different? I am aware that for DPM simulations, the 'step length factor' is something which can be altered, however, this is not an option in the MPM user interface. The only other thing which I can see which is related in the MPM user interface is the 'particle sub time step with each flow time step'.
Furthermore, for convergence in a transient simulation, surely this is only achieved for each time step, and not across the entire simualtion? For my simulations, after a new time step begins, the solver appears to start again - residuals spike up, then converge down. This means that on the 'macro'/'global' scale, the residual graph does not appear to 'converge' (as residuals spike every new time step), however on the 'micro'/'local' scale, they do converge. Is this the result I should expect or not? Or do I need to re-evaluate my simulations?
Ethan McGrath, sorry, I missed the fact your simulation was transient. Yes, in your case it would be the sub-step for the particles to ensure they are independent. Essentially, you want to ensure that a particle does not cross many cells per step (which can be a possibility with high velocities), and using sub-steps for particles would ensure they have sufficient information to get a correct trajectory.
Regarding the residuals, are you seeing convergence at each step (like you are satisfying your convergence criteria) or do they spike, lower, then hit your maximum number of specified steps and stop? For each timestep, the residuals should drop and converge prior to moving on to the next timestep. This spiking/converging behavior is normal as long as each step converges after sufficiently many iterations. Typically, as the simulation goes on and approaches steady-state (or quasi steady motion), the solution for each step begins to converge much more quickly (fewer iterations and a steeper decline from the spike).
For reference, this information is available in the ANSYS Fluent User Guide.
Specifically see:
26.4.1 Specifying particle tracking parameters - notes how to select the particle time step
Hi Jonathan Komperda, that's good to know - I've selected an appropriate sub-time step for the particles assuring they don't cross multiple cells per step. For the second point, they do converge at each time step before moving to the next. Thank you again for the insight - always good to confirm you've done things correctly and I won't have to redo any results.
- Badges
- Science topic
- Similar topics
- Filing