[Danhehe]

Hello, I am doing right now a transient simulation of an underground fire. The fire starts inside a train and the species and heat has to move arround the whole station.


I need to simulate like 1200(s) and my mesh has about 400k elements, which are polyhedra with maximum skewness 0.54 and minimum orthogonal quality 0.2.


I have done an steady case without the fire, just with natural flow inside the station to start the transient case from this solution.


For my transient simulation, I have set up a CFL=10 and an initial time step of 5e-4 s. This time step grows continualy by a factor of 2 every time step. It reaches something like 0.15 (meaning a CFL=7) and then, suddenly, at second 5 or 6 of simulation, viscosity ratio exceeds 10^5 and time step starts to fall.


My boundary conditions are ok as I checked them with my partners and the values are right.



Which could be my problem? Mesh? Schemes?...

Thanks

[vinerm]

Fire spread rate is usually quite high. Hence, even with a very coarse mesh, 0.15 s is quite large value for time-step. You have to use a smaller time-step. You can certainly increase the time-step by a factor of 2 but doing that every time-step is not a good idea. Use an increase factor of 1.1 or maximum 1.2 if you want it to increase with each time-step. A reduction factor of 0.9 would do good. Fix the maximum value such that the CFL does not go beyond 10.

[Danhehe]

Originally Posted by vinerm

Fire spread rate is usually quite high. Hence, even with a very coarse mesh, 0.15 s is quite large value for time-step. You have to use a smaller time-step. You can certainly increase the time-step by a factor of 2 but doing that every time-step is not a good idea. Use an increase factor of 1.1 or maximum 1.2 if you want it to increase with each time-step. A reduction factor of 0.9 would do good. Fix the maximum value such that the CFL does not go beyond 10.

So you are basically saying that my maximum time steps has to be in the order of 10^-2? Or are you saying that a growth rate of 2 for time step is too high?


I do not think the growth rate is the problem,as it goes to 0.15 and stay there for more than 6 or 7steps, I think it is more related to 0.15 being too high, which is a real problem for me.


Thanks for your answer

[vinerm]

If the objective is not to observe the initial spreading rate but the spreading extent over 1200 s, then the growth rate is not that important. However, if that is of concern, then a growth rate of 2 is rather high. Final time-step is certainly high and that's what causes the trouble with numerics.