Could not re-orthogonalise all Lanczos vectors

04-Unexpected behaviour/errors
Post Reply
Nico1978
Posts: 3
Joined: 27 Feb 2017, 21:18

Could not re-orthogonalise all Lanczos vectors

Post by Nico1978 »

Hello

I have a problem with my seismostruct file.

I have to calculate eigenvalues and I have done the exemple given in the manual. It worked perfectly. I have extended it to a 16 columns / 10 storeys building and I have the error shown in the subject title.
My file can be found here : http://ovh.to/nCcm3DW

For the periods that seismocstruct has given, they are very different from what I have been able to find by hand calculations.

Is it possible to check the stiffness / mass matrices given by the software? I have assumed a stiffness k=12EI/L^3 per column so multiplied by 16 for 1 storey. It seem to be less stiff in the software calculation.

Thanks

Nicolas
User avatar
z.gronti
Posts: 824
Joined: 16 Oct 2013, 08:14

Re: Could not re-orthogonalise all Lanczos vectors

Post by z.gronti »

Dear Nico1978,

Could you please check your file? In the uploaded file it seems that you have not assigned reinforcement to the column sections.
Zoi Gronti
Seismosoft Srl.
huffte
Posts: 978
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Could not re-orthogonalise all Lanczos vectors

Post by huffte »

Regarding the lower than theoretical stiffness - this is to be expected. The equation for column stiffness - 12EI/L^3 - is valid only for infinitely rigid beams. If you desired to do so, you could emulate such a condition by placing rotational restraints at all nodes which constitute a beam-column connection. This would effectively be equivalent to infinitely rigid beams. Alternatively, you could assign outrageously high E and I values to the beams, but this could create some numerical instabilities.
Tim Huff
Nico1978
Posts: 3
Joined: 27 Feb 2017, 21:18

Re: Could not re-orthogonalise all Lanczos vectors

Post by Nico1978 »

Thank you huffte for your answer. I will try to place a rotaional restraint at all nodes.

Dear z.gronti.
This was intentional to not assign any reinforcement into column section. I have to do an assignement where no reinforcement are given and I wanted to be as close as possible to the problem. Does the fact that not placing any reinforcment bar can create an instability for the result?

Thanks

Nicolas
User avatar
z.gronti
Posts: 824
Joined: 16 Oct 2013, 08:14

Re: Could not re-orthogonalise all Lanczos vectors

Post by z.gronti »

Dear Nico1978,

I don't think that if you don't assign reinforcement can cause instability to the eigenvalue analysis results. Please select the Jacobi with Ritz transformations eigen-solver in the Eigenvalue tab of the Tools > Project Settings to complete the eigenvalue analysis.
Zoi Gronti
Seismosoft Srl.
Nico1978
Posts: 3
Joined: 27 Feb 2017, 21:18

Re: Could not re-orthogonalise all Lanczos vectors

Post by Nico1978 »

Dear z.gronti

Thanks a lot. It's working well.

Huffte

Thanks to you also. I have obtained the same results than the ones obtained by hand calculation when I restrained all nodes in rotation.

Regards

Nicolas
huffte
Posts: 978
Joined: 22 Jul 2011, 10:19
Location: Cookeville, Tennessee, USA
Contact:

Re: Could not re-orthogonalise all Lanczos vectors

Post by huffte »

Good to hear Nico1978. It's a good way to verify software since the calculation is amenable by had with the restraints. Realize, however, that for modeling actual structures, you should likely delete the rotational restraints before proceeding with the analyses.
Tim Huff
Post Reply

Return to “04-Unexpected behaviour/errors”