I am running a case of film boiling in fluent. When I increase time step size I get smooth curves for Nusselt number and the volume fraction but the bubble seems elongated, if I decrease the step size, it makes the calculations too slow although I get a better shaped bubble as well as I could capture all the variations precisely in Nusselt number and volume fraction.
For explicit simulations: Satisfy Courant–Friedrichs–Lewy condition.
For implicit simulations: Simulations are A-stable so based on accuracy.
Trial and error sometimes is the best way if you are dealing with a specific problem repeatedly. Plus you will learn about the limits of your model/simulation.
Generally, it's based on the ratio of smallest cell size to free stream velocity. However, for variable step size several cases courant no. requirement have to fulfill.
For VOF simulation with mass and energy source terms, perhaps a check on Courant number below the default global Courant no. of 0.25 will always be good for increasing stability of the solution by reducing the time step size to 1 micro sec or less.
For explicit simulations: Satisfy Courant–Friedrichs–Lewy condition.
For implicit simulations: Simulations are A-stable so based on accuracy.
.
I will only strengthen the response Priyank Maheshwari that is entirely accurate, but it is followed strictly can bring surprises.
.
Implicit numerical schemes due to its strong stability often filter oscillations with higher frequency giving the impression that the results are correct, then at least as a simulation with time intervals and length satisfying the Courant-Friedrichs-Lewy condition is good practice.
.
In simulations of hydraulic transients in networks of storm water runoff Installations that operate either as free surface, either as closed conduit surface, the wave velocity perturbation can pass values 2m/s to 1000m/s, many hydrologists to take advantage of the surface simulation schemes free "fit" with its implicit call "slot Preissmann" penstock in the simulation schemes. As in this case the mesh simulation should be divided by at least 500 points, making it extremely heavy calculation, there are adjustments in the speed of propagation of the wave, wave simulations of 50m/s or less, what is a huge misconception.!
.
The same authors check the quality of the numerical scheme by the same stability, and how this scheme becomes stable it is considered correct, even with a strong problem in accuracy!
.
I will not name here international publications that fall into this error, because this should be the task of the reviewers and not those of the comments, however I note the following assumption:
.
Stability in simulations with explicit systems clearly shows the errors of the simulation, however to simulate through an implicit scheme and is not tested accuracy simulation shows no sign of apparent errors, giving a false impression that the calculation is correct!
.
Hello
yes you should try the above procedure. But in case if it didn't work out for you then you go for a small time step size and then choose adaptive instead of fixed. This way fluent will automatically calculate your time step size and you will get an idea about it.
- Badges
- Science topic