[MBDyn-users] How to use the shell4easans element

Pierangelo Masarati masarati at aero.polimi.it
Wed Feb 1 09:18:35 CET 2012


On 01/31/2012 09:01 AM, Pierangelo Masarati wrote:
> On 01/30/2012 10:08 AM, Resch Reinhard wrote:
>> Dear all,
>>
>> If I run a simulation with a shell4eas element, I get the following
>> warning message: Shell4EAS(4001): warning, "shell4eas" is deprecated;
>> use "shell4easans" instead
>
> Exactly, the EAS/ANS variant is more appropriate.
>
>> But if I replace the keyword shell4eas by shell4easans the simulation
>> does not converge any longer. Maybe the node order is different
>> between shell4eas and shell4easans?
>
> No, everything from the user's perspective is the same.
>
>> Any suggestions?
>
> The reason could be that right now portions of the Jacobian matrix are
> not computed exactly, but either approximated or neglected. You could
> try generic strategies like reducing the time step or increasing the
> tolerance or, if the issue is related to not reducing tolerance below
> some value, to use the "at most" keyword in the "max iterations"
> statement to allow temporary bailout if the test could be reduced but
> not below the requested threshold.

About your specific model, I see that the test is not going below some 
1e-5; if you increase the tolerance to 1e-4 it converges as expected. 
Results slightly differ from the EAS only case, but this could well be 
because the two formulations are indeed different.  We have compared 
both models with results from the literature, and we did not find a 
significant difference, but the EAS/ANS form is expected to be more 
accurate and reliable.  In any case, we encourage a review of the 
implementation, especially of AssJac, which is not complete.

Cheers, p.


-- 
Pierangelo Masarati
Associate Professor
Dipartimento di Ingegneria Aerospaziale
Politecnico di Milano


More information about the MBDyn-users mailing list