Iteration displacement limit

[andy]

Hello,
I'm performing a moving load analysis along a cable (length 100 m), i.e. with a lot of load cases.
Results are OK, but the convergence speed is very slow: it seems that dlubal limits the displacement for iteration to 200 mm (see image attached), even if there is NO particular convergence problem.
Is it possible to increase this iteration increment limit (in order to reduce the number of iterations)?
Regards
Andy

[Frank Faulstich]

Hello Andy,

have you tried if the precision settings have any influence?



A tip would be to try the solver "Dynamic Relaxation". For cable structures it is sometimes very fast.



Frank Faulstich

[andy]

Hello Frank,
the convergence plot for a combination requires 27 iterations, moslty limited to 200 mm for each step.

If I relax tolerances

the benefit is just the reduction of one iteration, being convergence limits less stringent. But I'm not asking for less stringent limits: the problem I observe is that during convergence most iterations are limited to a displacement limit / per iteration (200 mm) that slows the solution. Is this limit "hardcoded" in the solver?


Dynamic relaxation is much slower (and takes 15000 iterations)


Thank you
Andy

[Frank Faulstich]

Hello Andy,

I am not aware of any settings to change the displacement per iteration. But maybe the value depends on other parameters. I would like to investigate this in more detail. Can you please attach the file for this purpose.

Frank Faulstich

[andy]

Hello Frank,
here ( cable0.zip ) is a minimal input file. Please change the filename to "cable0.rfx" in order to open directly with RFEM.
Best,
Andy

[Frank Faulstich]

Hello Andy,

I have taken a closer look at the problem. The deformation limit is hard coded and there is no way to influence it. It also does not depend on the size of the model.

The Dynamic Releaxation in attached example takes a time slightly twice as long as the Newton-Raphson solver.

Frank Faulstich

[andy]

Thank you Frank,
may I suggest to improve this "hardcoded" limit to be at least model-size dependent (200 mm for a small frame is a "huge displacement", whereas on a big cable structure is a little displacement) or better, to expose it, for instance in the menu "Change standard settings"?

Yours sincerely
Andy