PDA

View Full Version : Initialization failure ICM



Will_Rust
February 7, 2014, 05:37 AM
Hi,

I'm having some trouble initialising a 1D-2D ICM model where I receive the following error in the time-set log:


Halving: H: SZ95997852.1-, Q: SZ95997852.1-, N: , Reason: Cond Sing
Message 261: Too many time step halvings - cannot continue.
Message 317: SIM will now attempt a new initialisation phase with a phase in 10**4 times the value in simulation parameters.

The initialisation keeps trying to increase the Sim param until it can't go any further (presumably).
I have tried increasing 'Max number of timestep halvings - initialisation' in the Sim params to 30 but I still get this error.

I am also not sure what 'Cond Sing' refers to, and I can't see anything wrong with the long-section near SZ95997852.1.

Anyone's thoughts would be much appreciated!

Cheers

Robert Dickinson
February 7, 2014, 09:24 AM
Hi Will,
Often when you have this issue increasing the number of iterations does not help but lowering the time step and/or using the 2D option "Link 1D and 2D calculations at minor timestep" will help.
I hope this helps.
Regards,
Bob Dickinson
Innovyze

Will_Rust
February 10, 2014, 04:25 AM
Hi Bob,

Thanks for the reply.
I have reduce the timestep down to 10, 5 and 1, with 1D-2D minor time step linking, and I am still getting this same error. Are there any other fixes I could try or aspects of the model I should look into the find the problem?

Cheers,
Will

Duncan Kitts
February 10, 2014, 05:32 AM
Hi Will,

Poor initialisation usually occurs for 3 reasons:-
*Initial Flooding at time zero.
*A large amount of storage which requires filling. This is usually related to large rivers with tidal boundaries.
*Model instability and poor convergence.

Rather than forcing the model through the initialisation phase by changing the simulation parameters it is always worth inspecting the geometry and boundary data to ensure it is appropriate. Feel free to send your model to the support@innovyze.com email address and we'll take a look.

Will_Rust
February 10, 2014, 06:29 AM
Hi Duncan,

I have had a look at the conduit geometry but can't find anything out of the ordinary. I'll send over a transportable - I will probably have to use Thru Drop Box due to the size of the model.

Much appreciated.
Thanks,
Will