[danadanao]

Hi everybody,

I have a question for you concerning Fluent. I tried to run a transcient simulation by setting :
- time step size = 1s (just an example)
- number of time steps = 0
I got a converged result, but it was written Time=0.00000e+00 into the scaled residuals window.

Do you know if it is a bug, a steady solution, or the first step solution?

[srjp]

Fluent did not perform any iterations in this case, and it had only the initial solution (and did not have the next time step data), so it shows as converged. If you look at the contours and other variables, you would see that the solution is the initial condition itself.

[danadanao]

Thank you for you answer srjp, but Fluent did iterate and shows another result than initialisation.
I can do this with all the cases I work on, but I still don't understand what is computed...

[sandisk]

Hi,
I had a query, I am also trying Unsteady Simulation. I have not yet fully understood the criteria to decide the number of time steps. Is it related to the physical time. I mean if I want simulate flow in a pipe (unsteady state condition) , what is the criteria for Number of time steps.
Thanks.

[srjp]

You would run a simulation in unsteady mode in two cases:

1. If we are interested in the final steady state result, but it is not possible to run the steady state solver. In this case, the simulation is run in unsteady mode till there is no change in critical variables (velocity, shear stress, etc) with time.
For example, in a pipe flow with uniform velocity inlet boundary condition, it takes some time for the boundary layer and the velocity profile to develop. So, once the velocity at different spatial locations do not vary with time, we can stop the iterations. We care only about the end result, and not about the intermediate results.

2. If the flow itself is unsteady, and we are interested in studying the transient nature of the flow. In this case, the number of time steps depend on how long we want to study the flow.
For example, rise of bubbles through a stagnant liquid. Here we would run the simulations till the bubbles leave the liquid, or after studying enough number of bubbles.

Finally it is the physical time (flow time) that matters, since flow time = number of timesteps x timestep size

[sandisk]

oh okay, i see.. thanks.. i am simulating slurry flow, where the flow itself id transient. I have given the flow time such that the fluid traverses the entire length of the pipe..i.e my velocity is 3.5 m/s and length of the pipe is around 6 m,with diam 0.052 m. I have given flow time around 1.2 sec. Is that fine ? Also, it takes days to simulate transient simulations. I read in some post that, transient simulation running into days is common. If you have any experience with transient simulation, Could you let me kn ow if transient simulation take such a long time.??

[srjp]

If the flow is transient in nature, then you have to run the unsteady simulation. Transient simulations running for days is very common. The simulation time depends on many factors including model setup, processor speed, memory and so on. You can review Fluent manual on how to speed up the process.
You can also consider the following methods:
1. Assigning very accurate boundary and initial conditions
2. Running a steady state simulation first, and then starting the transient simulation from the results.
3. Using first order upwind scheme for faster convergence
4. Using a simpler grid
5. Using a faster computer with bigger memory
6. Using parallel computing, HPC (High performance computing) etc.

Don't forget to 'autosave' while running the simulation for days !

Also in your case, the residence time of the liquid is 6/3.5 = 1.7sec. So you may have to run at least for this time.

[sandisk]

Thanks a lot !!