System Solver Problem!

[Ameri]

Hi there,

I have modeled a conventional 3D 6-story, 4-bay RC moment frame building in OS and when I want to assign gravity loads to the model by using a simple static analysis, I encounter a 'LValue too small' error! It's really strange for me as the model is not a huge one and I had never faced such a problem in OpenSees. I changed the "LValue" from 10 to 100, but it doesn't work yet. What is so frustrating is that none of the system solvers work for me (BandGeneral, BandSPD, etc ...) and I really don't know what to do! Besides, as my thesis defense session is approaching, I have zero time to make myself familiar with SP and try "Mumps" solver (Although I don't think if it is compulsory for my rather small model!). Another odd thing is that I have a 3-bay model with the same characteristics and story numbers, in which no problem occurs when analyzing. When it become 4-bay the error happens. I wonder if you could provide any clue.

Best

[fmk]

if none of the solvers are working i suggest you look at the model .. the LVALUE needs to get larger as the matrix becomes more ill-conditioned .. if all the other solvers are failing in the solve (and not due to memory limitations) it means the system is probably singular, a.k.a. your model is incorrect.

[Ameri]

Thanks you Dr. Mckenna. As always, your advice was a troubleshooting guide. However, I'm still a bit confused about the way my problem was solved! The problem was that I had distributively assigned all the mass and loads only to beams, and not to the columns! I used to think that in Finite Element Analysis this is not that important whether you assign the loads and the mass to elements or nodes, as they will be finally put on nodes and stiffness/mass matrices will be made accordingly! I also thought that ill-conditioning usually occurs owing to instability of the structure or sparsity of the matrix (large condition number indeed). Now I really don't understand why assigning loads and the mass only to beams makes the stiffness matrix some kind of singular, and how ill-conditioning is related to this. I would be most thankful if you could guide me in this regard.

[fmk]

it shoudln't! unless you were doing an eigenvalue analysis as beam elements would not contribute anything to mass.

[Ameri]

Maybe the forceBeamColumn element is involved herein! I think I should be checking out the other elements so as to better understand what's happening. By the way thanks a lot for your permanent support sir.