[tomcatbobby]

Hi

I could use some help with an error I am getting in CFX.

| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| Floating point exception: Overflow

Details of error:-
----------------
Error detected by routine POPDIR
CRESLT = ILEG

I have no idea why I am having this error but it is driving me insane! I have seen many forum posts for this type of error but none of the solutions solve my problem.

I am trying to model a simple problem of natural convection in a room. The room is a box with walls set as adiabatic. Inside the room is a heater raised off the floor with a static temperature. The fluid is air at 25 degrees C with buoyancy enabled and no turbulence (laminar).

Any ideas.

Thanks

[ghorrocks]

It means your simulation is diverging. You need to improve numerical stability - better mesh quality, smaller timesteps, better initial conditions, start off with first order discretisation, check the physics is correct.

In your case is your simulation steady state? If so what is the heat source (you said it is a heater), and what is the heat sink?

[tomcatbobby]

Originally Posted by ghorrocks

It means your simulation is diverging. You need to improve numerical stability - better mesh quality, smaller timesteps, better initial conditions, start off with first order discretisation, check the physics is correct.

In your case is your simulation steady state? If so what is the heat source (you said it is a heater), and what is the heat sink?

Ok, I can appreciate those considerations. In terms of yours questions I have been trying to keep the problem as simple as possible. Mesh default size is set to 100, no timesteps, no initial conditions, not sure about where to set first order discretisation within CFX?

Heat source is effectively a box at a constant temperature of 145 degrees C. So as far as know, it is unsteady and a transient solution is required...but....my limited knowledge of CFD was telling me run a steady state analysis first so that the transient solution had some results to start with? There is no heat sink at the moment (but I want to add this later), all I want to do at the moment is moniter the room as it gets hotter and hotter.

Am welcome to any suggestions and fully accept what im doing could be completely wrong, I have only started learning CFD this year.

Thanks

[ghorrocks]

You are never going to get convergence using a steady state model on a model which is not steady state. If you put a heater in a room with adiabatic walls, it is going to slowly heat up as time progress. This means there is no steady state so of course it diverges.

You could say that the steadys state result is the entire room at the heater temperature and no flow, but the steady state solver will have a hard tiem getting there as the initial condition is so far from the final result.

So I would forget the steady state bit and just start off with a transient run.

[tomcatbobby]

Thanks for the reply. This is very helpful. I have tried to run transient analysis in the past and have become confused by which timesteps to use?

I was referred to here for choosing them

http://my.fit.edu/itresources/manual...ug/node572.htm

in particular, equations 13.2-21 and 13.2-22.

How do I know my length and velocity scales?

Thanks

[ghorrocks]

The best way to set time step size when you are doign new work is to use adaptive timestepping, targetting 3-5 coeff loops per iteration (assuming a single phase simulation). Then let the solver find the timestep size itself.

[tomcatbobby]

ok I changed to transient solution and used adaptive settings but am still getting the overflow error? I think its because of the initial conditions are not enough but I dont know why? I have set an (t0) initialisation criteria to 0velocity in all directions, pressure the same as the reference pressure and a temperature of 10 degrees c?

What am I missing?

Thanks

[ghorrocks]

Are you sure your initial conditions are sensible? If so then try starting with a smaller timestep.

[tomcatbobby]

Tried smaller timestep and a smaller mesh today and yet its still a no go. Either an overflow error or and invalid number error.

Inititial conditions are as follows (by initial, I clicked the global initialisation button)

Velocity= 0 in all directions.

Pressure=same as reference pressure=atm

Temperature=15 degrees centigrade.

Thanks

[ghorrocks]

What is your timestep? I bet your timestep is still too large.

Also you are running a laminar model on a flow which is almost certainly turbulent. That will also cause convergence problems.

Please post your CCL.

[tomcatbobby]

Yes I have been selecting laminar for two reasons. Thought it would keep the problem simple and also as flow initially has 0 velocity in all directions. I figured the flow would be turbulent anyway.

I've exported my CCL but how do i get it onto the post?

Thanks

[tomcatbobby]

Here it is.

Just delete the .txt at the end. wouldnt let me load a .ccl.

[tomcatbobby]

Ok i changed the reference pressure for the domain and the initialisation to 1 atm. This meant that the run continued without any errors. Havent checked results yet though. Im sure i have made mistakes in my setup anyway so would you mind checking the CCL anyway.

Thanks

[ghorrocks]

Check your adaptive time stepping settings. You are currently doing the first update after one minute, should be after the first timestep. Also your initial timestep and minimum timestep are invalid.

[tomcatbobby]

Can you elobrate, how are they invalid? How do i set the first update to be after the first timestep?

Other than that is it ok?

Thanks for your help on this.