Timestep becomes very small

[Santanu Das]

Dear all,

I am currently simulating the case of rising bubbles, specifically using multiple bubbles while avoiding coalescence by incorporating the no-coalescence.h file. For this simulation, I have implemented a periodic boundary condition with a fixed grid size of 1024 x 1024 using a box size of 20. Additionally, I have set the maximum timestep (DTMAX) to 1e-3.

The code runs smoothly until t ~ 70 with a DT of 1e-3. However, beyond this point, I have noticed that the DT consistently decreases. I have attached the logfile that records the time versus DT for your reference. The bubbles reach a maximum velocity of 1.5. Given the CFL condition of 0.5, along with the defined cell size and velocity, I would expect the timestep to be around 5e-3.

Could anyone provide insight into why the timestep drops so significantly and suggest how I might address this issue?

Best,

Santanu Das

[Patrick Fischer]

Hi,
I observe the same behavior in my simulations (see my post from October 29). But the program eventually crashes after a certain time. I have studied the flow and found out that the values are blowing up at the exit of the channel due to the Neumann boundary condition. How do the velocity/pressure fields look like ?

Regards

[Santanu Das]

Hi Patrick,

In my case, I am using a periodic boundary condition. The velocity field appears to be correct and does not blow up. The simulation runs correctly for some time, but then there is a sudden decrease in the timestep. If I force the DT to be fix in timestep.h file, then the program continues to run with that fixed DT. I also checked the CFL condition, which remains unchanged.

Best regards,

Santanu

[Patrick Fischer]

Hi Santanu,
thank you for your reply.
Do you mean that if you fix the time step according to the theoretical CFL condition, the program does not decrease suddenly the time step  and the velocity does not blow up ?
In my case, I observe a blow up of the values despite the time step being automatically decreased. And the program eventually crashes. And I actually observe these huge values appearing at the exit of the channel.
Have a good day

[Santanu Das]

Hi Patrick,

Yes, that's what I meant. If I fix the time step based on the theoretical CFL condition, the program runs according to this time step and does not decrease suddenly, and the program continues to run. Also, I do not observe any velocity blow-up in my case.

Have a good day!
Santanu

[Wojciech (Voitek) Aniszewski]

You gentlemen should consider printing several quantities to stdout/err. Such as the max values of all velocity components, surface tension and/or interface curvature. There are several kinds of CFL conditions and conditioning, you could that know better the growth of which quantity limits your dt. (In 8.5 cases in 10 it's velocity, but still...). Then, methodically, find where and when it happens, and maybe that'll let you find out why. Especially since you have similar cases only one of which runs smoothly...

good luck

Voitek

--
You received this message because you are subscribed to the Google Groups "basilisk-fr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to basilisk-fr...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/basilisk-fr/54568485-5121-4911-9308-32e99bda7e67n%40googlegroups.com.

[Santanu Das]

Hi Wojciech,

Thank you for your suggestions. I will check for very large values of various quantities, particularly velocity. I hope to find out something that can help resolve the timestep issue.

Best regards,

Santanu