[mcc007]

Hello everyone,

I have some questions about my simulation. I use Fluent 6.3.26 to simulate plasma interaction with liquid (water). When I tried time step size 1 s with different number of time steps (e.g., 100, 300, 600, etc.), I could get stable results. But the result (electric potential) seems too high than I expected. Because the average electric potential shown in the figure is around -5.e+10 V. But I only applied -2.5e+3 V. See the attachment in the first figure.

So I reduced my time step size to 1e-4 s and used a small number of time steps (less than 100), the result seems to make sense. However, ff I increased the number of time steps to 150 or even larger, then I have errors for my simulation: under convergence history of a UDS: 1.#QNBe+00.

My questions are:

1. why at high time step, the result is higher than expected?

2. why I choose a smaller time step size, the result is reasonable but the increase of number of time step leads to simulation errors: under convergence history of a UDS: 1.#QNBe+00.

I am wondering if you could help me with this.

Thanks a lot,
mcc007

[rupak504]

Originally Posted by mcc007

Hello everyone,

I have some questions about my simulation. I use Fluent 6.3.26 to simulate plasma interaction with liquid (water). When I tried time step size 1 s with different number of time steps (e.g., 100, 300, 600, etc.), I could get stable results. But the result (electric potential) seems too high than I expected. Because the average electric potential shown in the figure is around -5.e+10 V. But I only applied -2.5e+3 V. See the attachment in the first figure.

So I reduced my time step size to 1e-4 s and used a small number of time steps (less than 100), the result seems to make sense. However, ff I increased the number of time steps to 150 or even larger, then I have errors for my simulation: under convergence history of a UDS: 1.#QNBe+00.

I guess it depends on the Courant number. try this https://www.simscale.com/blog/2017/08/cfl-condition/

My questions are:

1. why at high time step, the result is higher than expected?

2. why I choose a smaller time step size, the result is reasonable but the increase of number of time step leads to simulation errors: under convergence history of a UDS: 1.#QNBe+00.

I am wondering if you could help me with this.

Thanks a lot,
mcc007

I guess it depends on the courant number. try this: https://www.simscale.com/blog/2017/08/cfl-condition/

[mcc007]

Originally Posted by rupak504

I guess it depends on the courant number. try this: https://www.simscale.com/blog/2017/08/cfl-condition/

Thanks for your reply. I checked that page. It seems doesn't work. Any suggestions? Thank you.

[vinerm]

Just like space discretization, temporal discretization has its own rules. You cannot use long, straight lines to draw a curve or a circle. Smaller are better. But that doesn't mean way too small.

You need to look at the time-scales of the phenomena taking place in the case you have. If those are of the order of milli-second, then 1e-4 is good but if those have large time-scales, then you can use a larger time-step. E.g., if you want to study melting of ice kept at room temperature, you don't expect any significant changes even over 5 seconds. So, you use a time-step of 15-20 second.

As far as iterations are concerned, it should never be more than 40. If the case requires more than 40 iterations to converge in each time-step, then the time-step is too larger and should be reduced, but the number of iterations should not be increased.

[mcc007]

Originally Posted by vinerm

Just like space discretization, temporal discretization has its own rules. You cannot use long, straight lines to draw a curve or a circle. Smaller are better. But that doesn't mean way too small.

You need to look at the time-scales of the phenomena taking place in the case you have. If those are of the order of milli-second, then 1e-4 is good but if those have large time-scales, then you can use a larger time-step. E.g., if you want to study melting of ice kept at room temperature, you don't expect any significant changes even over 5 seconds. So, you use a time-step of 15-20 second.

As far as iterations are concerned, it should never be more than 40. If the case requires more than 40 iterations to converge in each time-step, then the time-step is too larger and should be reduced, but the number of iterations should not be increased.

Make sense. It's very helpful. Thank you.