Learning Forum

[Henry]

I am facing a problem with the time it is taking me to get results from my model. My experiment is 5 years long. The temperature boundary condition is time-based temperature. With fluent runs in serial mode, it is taking weeks to get the results. The biggest time step size that my model is accepting without convergence problems is 1 sec. Is it true that natural convection and heat transfer problems can only be solved correctly with small time steps? Does a time step size say 1000 sec or 60*60*24=84600 sec is just not acceptable? Is it a problem with fluent or any other CFD code? Would I have been able to solve the problem with a bigger time step size?

[Karthik Remella]

Hello,

There is a physical time step, which is governed by the physics involved in the problem, and then, there is a time step based on numerics (that ensures the numerical stability of your model). The physical time scale of your problem is dependent on the physics. In your case, there might be diffusion and convective time-scales. These are to be computed based on your length scales of your domain (and other physical parameters such as velocity, diffusivity etc.). If you take a time-step larger than this physical time-step, your model will not be capturing the physics involved.

The numerical time-step commonly referred to as the CFL condition is there to ensure the numerical stability of your model. This is dependent on the smallest grid size and the velocity scale of your problem. The smaller the grid size is the smaller will be the time-step required to ensure numerical stability.

Generally in CFD models, your time-step will be based on the minimum of the two values - physical time-scale and time-step based on the CFL condition. Unfortunately, this is extremely important to ensure this to obtain good physical results from your model.

Thank you.

Best Regards,

Karthik

 

[Henry]

Thank you for your timely reply! Thank you very much for the detailed answer!