[praveengambhir16]

Hello All!
I have recently simulated a multiphase water/air simulation. The scenario was water dripping from a hole into a container. I have considered a 5 degree wedge domain with symmetry faces which resulted in a considerable decrease in cell count. However, I have observed a very troubling issue with my run.
I have previously simulated the scenario of capillary rise phenomenon considering a 1e-5 time step. I've also tried adaptive timestep from another post in cfd-online, but the 1e-5 timestep worked very well, with a faster than considered computation time.
This current example, though, even with a 1e-5 timestep, took considerably longer to compute than expected (upwards of 15 hours with 128 cores distributed HPC, the number of mesh elements were less than 30,000). I've restricted the transient solution to 5 coefficient loops per timestep, with a convergence criteria RMS of 1e-4. What would be a more nominal transient time step value in order to reproduce the solution with reasonable accuracy, and reducing solve time?

Thanking You,
Praveen
TTI

[ghorrocks]

The recommendation remains the same. Use adaptive time stepping homing in on 3-5 coeff loops per iteration. This is especially important if you are doing surface tension modelling. CFX requires VERY small time steps when you are doing surface tension modelling.

This results in very long run time with CFX. Try fluent, it has much better options for surface tension modelling and can run much faster than CFX.

[praveengambhir16]

Thank you very much, Mr.Horrocks. We are in the process of acquiring a Fluent license soon. I will use Fluent to solve the same case whenever we acquire a license. Are there any other tips, tricks or workarounds to get around the problem of unusually large computation time for surface tension problems?

Regards,
Praveen
TTI

[ghorrocks]

With Fluent and surface tension - note the defaults are not the best for most simulations. You can make massive increases in speed and performance by optimising the settings. But the settings required for each simulation are different so you are going to have to try things out to find what works for you. But in particular look at NITA, and Geo-Reconstruct. But many other options are worth looking at as well.