"Warnings" in Gravity Analysis

[soko]

Hello!
I am trying to perform a 3D Gravity Analysis of an 8-storey buidling model. This particular building consists mostly of bare frames and has 5 small walls surrounding the elevetor and the stairwell. I have allready modeled this building using both SAP2000 and Opensees with elastic frame elements and performed an eigenvalue analysis in order to compare the periods and frequencies of the modal analysis from both programs. The comparison results were satisfactory enough, so I came up with the conclusion that the 3D elastic model I created in Opensees was correct. I find it important to inform you about the way I designed the walls: I used linear finite elements and simulated their great stiffness with rigid frame elements designed at the same level as the beams with length equal to the wall's long side. The values of the rigid properties I used were A=1000000, J,Iy,Iz=1000000. Furthermore, I designed the perimeter infill masonry using frame elements with realeases at the element's both ends in SAP while in Opensees I used truss elements.

Subsequently, I created a nonlinear model of the building in Opensees using inelasticBeamColumn elements with fiber sections for all beams and columns, while for the rigid elements I maintained the elasticBeamColumn elements from the elastic model. When I assigned the Torsional Stiffness for 3D Model while creating the fiber sections I used Ubig=1.e4 because in my structure G*J is near to this value. When I performed the eigenvalue analysis for this model, the results were close enough to those of the elastic model in SAP so I persumed that my inelastic model was correct.

In order to perform a Gravity Analysis in Opensees for both elastic and inelastic models, I assigned them firstly with NodalLoads (computed according to the mass previously assigned) and secondly with eleLoads (which were the same as those I used in SAP model). I ran both models (elastic and inelastic) with and without rigid diaphragm at the floor levels and using both Linear and Newton Algorithms. The results of each analysis were the following:

1)The elastic model ran without any problems for all possible combinations of Loads assigned (eleLoads or nodalLoads) and chosen algorithm (Linear or Newton) but the values of the nodal reactions of the foundation nodes (which I designed to be completely fixed) differed from those I obtained from SAP, especially the reaction values corresponding to the wall's foundation.

2)The inelastic model with elelLoads and rigid floor diaphragm assigned ran while using Linear algorithm, printing the requested nodal reactions in the end but gave the following warning:
"ForceBeamColumn3d: update-failed to get compatible element forces&deformations for element 618..."
The reaction's values differed from those of SAP as previously stated for the elastic model.

3)The inelastic model with elelLoads and rigid floor diaphragm assigned did not run while using Newton algorithm and printed the following message:
"ForceBeamColumn3d: update-failed to get compatible element forces&deformations for element 618...
Domain: failed in update...
LoadControl: model failed to update for new dU..
NewtonRaphson: the Integrator failed in update..
StaticAnalysis: the Algorithm failed at iteration 8 with domain at load factor 0.9
OpenSees > analyze failed, returned: -3 error flag"

4)The inelastic model with nodalLoads and rigid floor diaphragm assigned ran while using either Linear or Newton algorithm but the reaction's values differed from those of SAP as previously stated for the elastic model.

5)The inelastic model with elelLoads without rigid floor diaphragm assigned did not run while using Linear algorithm and printed the following message:
"ForceBeamColumn3d: update-failed to get compatible element forces&deformations for elements 371, 404, 453, 486, 519, 552, 585, 618, 700
Domain: failed in update..
LoadControl: model failed to update for new dU..
Linear: the Integrator failed in update..
StaticAnalysis: the Algorithm failed at iteration 6
OpenSees > analyze failed, returned: -3 error flag"

6)The inelastic model with either elelLoads or Nodal Loads without rigid floor diaphragm assigned did not run while using Newton algorithm and printed the following message:
"testNormDispIncr: failed to converge after 6 iterations...
NewtonRaphson: the convergenceTest object failed...
StaticAnalysis: the Algorithm failed at iteration 0...
OpenSees > analyze failed, returned: -3 error flag"

7)The inelastic model with nodalLoads without rigid floor diaphragm assigned ran while using Linear algorithm but the reaction's values differed from those of SAP as previously stated for the elastic model.

Can you please help me understand the meaning of each warning and suggest a way to correct them?
I would also like to know if it is "normal" to get different results of the base nodal reactions between the two software.
Finally, how come the inelastic model with diaphragms assigned runs for most of the above combinations and the one without diaphragms doesn't?

Thanks a lot in advance!

[fmk]

1) Linear: the elastic models should give you the exact same results if both models and analysis are the same. check one or other of the models. an a seperate note, nothing is rigid and making values too big can cause problems numerically, just use values that reflect the wall properties.

2-5) as for the nonlinear beam failing, is the failure at a beams or a columns? if beams it is because this element needs to have elongation or contraction from the section response. if you constrain these 2 be 0, large axial forces are induced to make these axial deformation go away. Either use a Penalty method to enforce the constraints (again nothing is rigid) or use dispBeamColumn elements (also nonlinear but numerically inacurate compared to forceBeamColumn when curvature is not varying linearly from one end to next of the element)

6) see 1 above. the models or analysis are different to start with, the nonlinear case will probably just make things worse.

[soko]

Hello again!
Thank you for your instant reply!

After some changes to both elastic models (sap and opensees) I managed to achieve satisfactory convergence between the reactions obtained from both software, especially those of F2 (axial reaction of the elements) but the rest of them (F1,F3,M1,M2,M3) appear to have still significant differences. The changes I made concern making sure the model built and loads assigned to both models are exactly the same. As for the rigid properties (A,J,Iy,Iz) in opensees, I assigned them with their real values as assigned in sap and multiplied them with the modifier's values (1000000) I used in sap so that their final values would be the same! After making these changes the outcome of the analysis changed as follows:

1)The elastic models, with and without diaphragms, assigned with either nodalLoads or eleLoads ran perfectly for either Linear or Newton algorithms and the obtained results were those I mentioned at the beginning of my reply.

2) The inelastic models with diaphragms, assigned with either nodalLoads or eleLoads ran without any warnings for either Linear or Newton algorithms but appart from the diffenses of F1,F3,M1,M2,M3 reactions between opensees and sap, the axial reactions of the walls differed significantly as well.

3)The inelastic models without diaphragms, assigned with eleLoads, did not run for either Linear or Newton algorithms and printed the following messages:
Linear algorithm:
"ForceBeamColumn3d: update-failed to get compatible element forces&deformations for elements 604
Domain: failed in update..
LoadControl: model failed to update for new dU..
Linear: the Integrator failed in update..
StaticAnalysis: the Algorithm failed at iteration 7
OpenSees > analyze failed, returned: -3 error flag"

Newton algorithm:
"testNormDispIncr: failed to converge after 6 iterations...
NewtonRaphson: the convergenceTest object failed...
StaticAnalysis: the Algorithm failed at iteration 0...
OpenSees > analyze failed, returned: -3 error flag"

I tried using DisplacementBeamColumn elements and the analysis ran only for Linear algorithm but the reactions were significantly different from those obtained from sap.

4)The inelastic models without diaphragms, assigned with nodalLoads, ran for Linear algorithm but had unsatisfactory results while it did not run for Newton algorithms and printed the following messages:
"testNormDispIncr: failed to converge after 6 iterations...
NewtonRaphson: the convergenceTest object failed...
StaticAnalysis: the Algorithm failed at iteration 0...
OpenSees > analyze failed, returned: -3 error flag"

Concerning the inelastic models without diaphragms, I also tried to perform the alanysis with smaller time step but all I got was achiving more succesfull iterations without being able to complete the Gravity analysis for Newton algorithm.

Is there a reason why the models without diaphragms cannot perform the Gravity analysis for Newton algorithm?
I also still cannot understand the differences of the reactions between the two software as I made sure the models are exactly the same!

[fmk]

you obviously have a problem with your model. As mentioned SAP and OpenSees for linear elastic case will give the EXACT same answer if the models are identical. The fat that you are getting 'satisfactor answers' for some but not all indicate that the models are different. I would not bother going to the nonlinear case until you sort out the linear case.

ps. if you are comparing the results for some model you have sketched on paper, don't assume that the OpenSees one is the one with all the errors!

[NikosCV]

soko wrote:
> Hello!
> I am trying to perform a 3D Gravity Analysis of an 8-storey buidling model. This
> particular building consists mostly of bare frames and has 5 small walls surrounding
> the elevetor and the stairwell. I have allready modeled this building using both
> SAP2000 and Opensees with elastic frame elements and performed an eigenvalue analysis
> in order to compare the periods and frequencies of the modal analysis from both
> programs. The comparison results were satisfactory enough, so I came up with the
> conclusion that the 3D elastic model I created in Opensees was correct. I find it
> important to inform you about the way I designed the walls: I used linear finite
> elements and simulated their great stiffness with rigid frame elements designed at
> the same level as the beams with length equal to the wall's long side. The values of
> the rigid properties I used were A=1000000, J,Iy,Iz=1000000. Furthermore, I designed
> the perimeter infill masonry using frame elements with realeases at the element's
> both ends in SAP while in Opensees I used truss elements.
>
> Subsequently, I created a nonlinear model of the building in Opensees using
> inelasticBeamColumn elements with fiber sections for all beams and columns, while for
> the rigid elements I maintained the elasticBeamColumn elements from the elastic
> model. When I assigned the Torsional Stiffness for 3D Model while creating the fiber
> sections I used Ubig=1.e4 because in my structure G*J is near to this value. When I
> performed the eigenvalue analysis for this model, the results were close enough to
> those of the elastic model in SAP so I persumed that my inelastic model was correct.
>
> In order to perform a Gravity Analysis in Opensees for both elastic and inelastic
> models, I assigned them firstly with NodalLoads (computed according to the mass
> previously assigned) and secondly with eleLoads (which were the same as those I used
> in SAP model). I ran both models (elastic and inelastic) with and without rigid
> diaphragm at the floor levels and using both Linear and Newton Algorithms. The
> results of each analysis were the following:
>
> 1)The elastic model ran without any problems for all possible combinations of Loads
> assigned (eleLoads or nodalLoads) and chosen algorithm (Linear or Newton) but the
> values of the nodal reactions of the foundation nodes (which I designed to be
> completely fixed) differed from those I obtained from SAP, especially the reaction
> values corresponding to the wall's foundation.
>
> 2)The inelastic model with elelLoads and rigid floor diaphragm assigned ran while
> using Linear algorithm, printing the requested nodal reactions in the end but gave
> the following warning:
> "ForceBeamColumn3d: update-failed to get compatible element
> forces&deformations for element 618..."
> The reaction's values differed from those of SAP as previously stated for the
> elastic model.
>
> 3)The inelastic model with elelLoads and rigid floor diaphragm assigned did not run
> while using Newton algorithm and printed the following message:
> "ForceBeamColumn3d: update-failed to get compatible element
> forces&deformations for element 618...
> Domain: failed in update...
> LoadControl: model failed to update for new dU..
> NewtonRaphson: the Integrator failed in update..
> StaticAnalysis: the Algorithm failed at iteration 8 with domain at load factor
> 0.9
> OpenSees > analyze failed, returned: -3 error flag"
>
> 4)The inelastic model with nodalLoads and rigid floor diaphragm assigned ran while
> using either Linear or Newton algorithm but the reaction's values differed from those
> of SAP as previously stated for the elastic model.
>
> 5)The inelastic model with elelLoads without rigid floor diaphragm assigned did not
> run while using Linear algorithm and printed the following message:
> "ForceBeamColumn3d: update-failed to get compatible element
> forces&deformations for elements 371, 404, 453, 486, 519, 552, 585, 618, 700
> Domain: failed in update..
> LoadControl: model failed to update for new dU..
> Linear: the Integrator failed in update..
> StaticAnalysis: the Algorithm failed at iteration 6
> OpenSees > analyze failed, returned: -3 error flag"
>
> 6)The inelastic model with either elelLoads or Nodal Loads without rigid floor
> diaphragm assigned did not run while using Newton algorithm and printed the following
> message:
> "testNormDispIncr: failed to converge after 6 iterations...
> NewtonRaphson: the convergenceTest object failed...
> StaticAnalysis: the Algorithm failed at iteration 0...
> OpenSees > analyze failed, returned: -3 error flag"
>
> 7)The inelastic model with nodalLoads without rigid floor diaphragm assigned ran
> while using Linear algorithm but the reaction's values differed from those of SAP as
> previously stated for the elastic model.
>
> Can you please help me understand the meaning of each warning and suggest a way to
> correct them?
> I would also like to know if it is "normal" to get different results of the
> base nodal reactions between the two software.
> Finally, how come the inelastic model with diaphragms assigned runs for most of the
> above combinations and the one without diaphragms doesn't?
>
> Thanks a lot in advance!

Hello there soko! I am a fellow undergraduate civil engineering student in AUTh and interested in modelling concrete core-walls with linear elastic elements (rigid beams etc.) for my thesis, just as you did.
I have tried inserting the exact same input for frame rigid properties, but after a modal analysis on OpenSees & SAP2000 I get different eigen-periods (>10% divergence). I think I've tested every possible alternative values, but in vain.
(Just to mention, when I model individual only walls, I get a better convergence for the first natural period of the structure)

Do you have any clue for this?
Do you get identical results from your analyses?

[kadirtopcu]

hi,
when you find a solution for these issues, could you please write here? I have been struggling against the problem that you mentioned for a long time. But I coulndt find any solution yet..