[MBDyn-users] question about eigenanalysis

Pierangelo Masarati pierangelo.masarati at polimi.it
Sat Nov 18 12:13:17 CET 2017


Without matrix scaling, the analysis works as intended.  By default, 
only permutation is enabled, to avoid scaling issues.  I noticed that 
after changing the reference system, the Jacobian matrices are slightly 
different.  The differences are very small (of the order of 1e-7), so it 
shouldn't alter the result, but apparently this affects the behavior of 
Lapack's scaling process.  I'm tracking the origin of that difference.

Sincerely, p.


On 11/18/2017 09:21 AM, lengjun wrote:
> Dear Louis,
> Thank you for your advice and I think it is a good way to deal 
> with the problem. I also found that
> the problem can be well solved if I change the final_time and time step.
> Maybe the key to the problem is choosing suitable initial sets and 
> methods to solve matrix equations.
> Best wishes,
> -Iris
>
>
>
>
>
> At 2017-11-17 23:10:29, "Louis Gagnon" <louis.gagnon at polimi.it> wrote:
> >Dear Iris,
> >I am not familiar with eigenanalysis, but this is indeed a peculiar issue.
> >I however noticed that if you use a position with z not equal to 1, the 
> >issue is resolved.
> >Try
> > > reference: arm,
> > >   reference, global, 0., 0., 1.01,
> >Alternatively, you can change the way lapack handles the problem and 
> >change to the following line:
> >use lapack,balance,permute;
> >which also solves your issue
> >Best Regards,
> >-Louis
> >
> >
> >On 14-Nov-2017 08:12 AM, lengjun wrote:
> >> Dear all,
> >> I have a problem about the eigenanalysis of a cantilever. I set a 
> >> reference called arm:
> >> reference: arm,
> >>   reference, global, 0., 0., 1.,
> >>   reference, global, eye,
> >>   reference, global, null,
> >>   reference, global, 0., 0., omega;
> >> and all nodes and elements are in the 'arm' reference.
> >> the result of frequency are incorrect:
> >> ##############################################################################
> >> # Key for lines starting with "Step":
> >> # Step Time TStep NIter ResErr SolErr SolConv Out
> >> Step 0 0 0 0 0 0 0 1
> >> Step 1 0.001 0.001 0 0 0 0 1
> >> # Derivatives solution step at time 0 performed in 0 iterations with 0 error
> >> Info: 0, success BALANC="B" ILO=7 IHI=226 ABNRM=4.62438e+16 
> >> BBNRM=4.62438e+16
> >> Mode n.       Real           Imag             Damp %     Freq Hz
> >> # Key for lines starting with "Step":
> >> # Step Time TStep NIter ResErr SolErr SolConv Out
> >> Step 0 0 0 0 0 0 0 1
> >> Step 1 0.001 0.001 0 0 0 0 1
> >> ##############################################################################
> >> However ,when I changed the reference 'arm' into:
> >> reference: arm,
> >>   reference, global, 0., 0., 0.,
> >>   reference, global, eye,
> >>   reference, global, null,
> >>   reference, global, 0., 0., omega;
> >> 
> >> it goes well. the results are:
> >> # Derivatives solution step at time 0 performed in 0 iterations with 0 error
> >> Info: 0, success BALANC="B" ILO=7 IHI=226 ABNRM=1.24038e+10 
> >> BBNRM=1.24038e+10
> >> Mode n.       Real           Imag             Damp %     Freq Hz
> >>         7: -3.07437e-08 +       38.281 j     8.03108e-08          6.0926
> >>         9:  1.11572e-09 +       38.281 j    -2.91457e-09          6.0926
> >>        11:   1.4287e-08 +      234.684 j    -6.08777e-09         37.3511
> >>        13:  1.41642e-08 +      234.684 j    -6.03545e-09         37.3511
> >>        15:  7.43393e-08 +      635.257 j    -1.17022e-08         101.104
> >>        17:  1.41226e-07 +      635.257 j    -2.22314e-08         101.104
> >>        19:  1.84635e-09 +      697.431 j    -2.64735e-10             111
> >>        21:  1.25765e-07 +       1189.2 j    -1.05756e-08         189.267
> >>        23: -1.27215e-07 +       1189.2 j     1.06975e-08         189.267
> >>        25:  1.05445e-07 +      1863.02 j     -5.6599e-09         296.508
> >>        27: -1.74265e-09 +      1863.02 j     9.35393e-11         296.508
> >>        29:  6.88544e-08 +      2622.68 j    -2.62534e-09         417.413
> >>        31: -4.36833e-08 +      2089.71 j      2.0904e-09         332.588
> >>        33: -7.66471e-08 +      3438.49 j     2.22909e-09         547.253
> >>        35:  9.22549e-08 +      2622.68 j    -3.51757e-09         417.413
> >>        37: -1.77564e-08 +      4283.13 j     4.14566e-10         681.682
> >>        39: -6.07606e-08 +      3438.49 j     1.76707e-09         547.253
> >>        41:  7.44935e-09 +      3474.15 j    -2.14422e-10         552.928
> >>        43:  -6.6307e-08 +      4283.13 j      1.5481e-09         681.682
> >>        45: -3.87825e-08 +      4845.22 j     8.00427e-10         771.141
> >>        47:  1.29202e-07 +      5124.28 j    -2.52136e-09         815.554
> >>        49:  1.03002e-06 +      5881.29 j    -1.75136e-08         936.036
> >>        51:  4.21351e-07 +      5124.28 j    -8.22264e-09         815.554
> >>        53:  2.84468e-07 +      5881.29 j    -4.83682e-09         936.036
> >>        55:  2.56009e-07 +      6196.96 j    -4.13121e-09         986.277
> >> I am wodering if the results are related to the position of the 
> >> reference. The attachments are the input files.
> >> Best wishes.
> >>   Iris
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> _______________________________________________
> >> MBDyn-users mailing list
> >> MBDyn-users at mbdyn.org
> >> https://mail.mbdyn.org/cgi-bin/mailman/listinfo/mbdyn-users
> >> 
> >
> >-- 
> >Louis Gagnon, Ph.D.
> >Postdoctoral fellow
> >Dipartimento di Scienze e Tecnologie Aerospaziali
> >Politecnico di Milano
> >+39 02 2399 8642
> >http://louisgagnon.com/research/
>
>
>
>
> _______________________________________________
> MBDyn-users mailing list
> MBDyn-users at mbdyn.org
> https://mail.mbdyn.org/cgi-bin/mailman/listinfo/mbdyn-users


-- 
Pierangelo Masarati
Professor
Dipartimento di Scienze e Tecnologie Aerospaziali
Politecnico di Milano

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mbdyn.org/pipermail/mbdyn-users/attachments/20171118/0361ea85/attachment.html>


More information about the MBDyn-users mailing list