abaqus node/element numbering 2

[jisb]

Hello,

Could someone tell me if abaqus node/element numbering could affect abaqus simulation efficiency, and why? For example, one part has numbers from 1 to 100, and a second part has numbers from 1 million to 1 million + 1000. Would this affect abaqus efficiency?

Also, in Patran, there's "optimization" procedure basically to reorder the node/element number sequencies. What's the underlying reason for that? Thanks a lot!

Sean

 

[corus]

Abaqus numbers nodes as per the instances, so each instance annoyingly starts with node number 1. From that local numbering system it makes a global node numbering system which you can't see unless you use a preprint of the numbering. I would suspect that Abaqus would autmatically renumber your patran input mesh and then optimise that numbering to reduce the bandwidth as there are no commands within Abaqus to optimise the node/element numbering.
Patran probably renumbers the nodes too because it does.

corus

 

[johnhors]

Sean

The node and element numbering means nothing more than assigning an identification label and will have no effect on solution time or memory resource etcetera when using up to date sparse matrix solvers as found in Abaqus.

I suspect that the "optimization" procedure in Patran is a legacy routine once used to minimise the solution bandwidth with historical solvers like the wavefront method or any other method derived from Gaussian elmination where it was imperative to find solution orders that minimised the bandwidth.

 

[jisb]

Thank both of your answers!