Non-linear analysis of a masonry arch bridge

AndreaC91

Quick update on the issue of masses: I tried to run 3 non-linear analyses by putting the density of all the material except one equal to zero.
It came out that when only the density of the backfill is considered, the sum of external force Y on BASE-SX is about 1231 kN.

However, if I select Component BACKFILL > Tools > Mass, I get a mass 268900 kg, which should correspond to a total external force Y on BASE-SX of 1319 kN.

The difference 1319 kN - 1231 kN = 88 kN is the error I measured before on the total reaction. So, it looks like something goes wrong when dealing with the backfill.

Victor

@AndreaC91, sorry, I must have read 801kN at t=0 instead of t=0.5 s.

I agree that half the external force should be 1702 kN.

The difference between 1610 and 1702 is probably because external force is partly cancelled by part of the weight of the elements directly using the support nodes (weight is also counted as an external force!). It goes up to 1658 kN on mesh refinement so I think it's fine.

EDIT: Your comment about backfill is consistent with this reason since most of the support is on the backfill.

AndreaC91

Yes, you are probably right. As a matter of fact, the elastic analysis provides correct results because it gives reaction forces instead of external forces. So, there is no disturbance due to the weight of the boundary elements.

I guess that most of the little problems I've found so far might be solved with the mesh refinement.

Thanks for your help

AndreaC91

Hi,

I'm still working on the arch bridge we were talking about some time ago. Please find here attached the final model dealing with.

I'd like to understand if there's a way to compute the vertical capacity curve of the arch.
I have tried to graph out the displacement at midspan against a vertical tandem load (2 x 3000 kN), applied at about one third of the span, but the outcome is linear regardless of the compression-only material assumption.

As far as you know, through the "custom model definition" is there any other non-linear model (for instance to be used for concrete) that I can try to use to model the behaviour of the masonry?

Thanks,

disla

Hi AndreaC91,

Your vertical tandem load (2 x 3.000 kN) is actually 1 load of -600KN distributed into two nodes, meaning (2 x 300 KN) .
When considering 3.000KN each one the bridge seems unable to carry the load.


Regards

AndreaC91

Yes, in the shared file I left the 2 x 300 kN tandem load, which is the standard service load.
I've tried to run the analysis with the 2 x 3000 kN and it does stop before reaching the total tandem load, this probably because the tensile stress grows up significantly. However, if I try to graph out the vertical displacement at midspan against the applied tandem load, I get a linear curve, which I'm not completely sure to be correct.

In addition, I'd like to understand if there is any other non linear constitutive model to be used with concrete that I can try.

Thanks,

disla

There is nonlinear behavior, but you don’t see it because you are not looking at the nodes where the arch is failing.




There are other material behaviors but I’m not Civil engineer and I don’t know how to set up them. Check ccx manual calling shared libraries (triggered by putting the
@ character in front material name )

UnilateralMazars
SignoriniHyperviscoelasticity
Signorini
SaintVenantKirchhoffElasticity
RusinekKlepaczko_ssrt
RambergOsgoodNonLinearElasticity
Ogden
Norton
Mazars
Marigo
JohnsonCook_ssrt
JohnsonCook_ssr
JohnsonCook_s
Iwan
IsotropicViscoplasticityAmstrongFredericKinematicHardening
IsotropicLinearKinematicHardeningPlasticity
IsotropicLinearHardeningPlasticity
IncompressibleMooneyRivlinBehaviour
ImplicitNorton
HosfordPerfectPlasticity
Hayhurst
GreenPerfectPlasticity
GeneralizedMaxwell
FiniteStrainSingleCrystalBrickHuang
DruckerPragerEcroLin
DruckerPragerCap
DruckerPrager
DruckerPerfectPlasticity_Cazacu2001
CompressibleNeoHookeanBehaviour
ChabocheViscoplasticity
Chaboche2
Chaboche

It would be awesome if you manage to set up one of them and you can share it. You will probably have to go to the Calculix forum.

mmartin

A month ago, I set up CXX +MFRONT in order to using MAZARS material model for brittle materials. I also wanted to model a masonry arch.

Unfortunately the library failed after the calculation due to a bug while calculating the tangent matrix. I hope that it will work out soon. MAZARS model will let you reproduce real life single macro cracks instead of the damage portion shown using "compression_only_materials".

Last weekend, I set up a masonry arch on SALOME MECA_CODEASTER using MAZARS, it is not converging by now. I hope it will by changing internal non linear solver parameters. If I succeed, I will send you screenshots.

MANUEL

disla

Hi mmartin,

This is a very very very simple liml file where UnilateralMazars "works" (better said - runs) for me. I think this is the proper way to proceed and understand the new material behavior.

Maybe we could work all together to see how to set up a more complex problem later.

I don't have any remote idea about this material behavior but seems kind of weird.

-Dynamic Explicit change the Displacements results.
-With just a little pressure seems like the Jacobian of the cube easily will become negative.
-Gravity makes it to fail.

I'm almost sure material properties (units) are wrong in the material definition.

Well, just a starting point. Nonlinear Dynamics seems mandatory as was warned about density during the initial set up. (Also because nonlinear Static stops )

Let me know if I can help in something. There are many people around asking about this material behaviors for concrete and maybe we could help them.

Runs with CalculiX v19 that Victor has include in last Mecway version.

AndreaC91

I'm trying to set up the analysis with the Mazars model but the analysis stops at first step with the messages below:


Using up to 1 cpu(s) for the stress calculation.
Using up to 1 cpu(s) for the symmetric stiffness/mass contributions.
Factoring the system of equations using the symmetric spooles solver

I don't understand how I could work it out. Please find attached my file if you want to have a look.

disla

As I comment , I think the proper way is to solve the simple cube and check it behaves as espected.

mmartin

I enclose(a rar file including steps and liml files) you the masonry arch I have been working on. But unfortunately, no displacement are produced after calculation. Could you try it out? Maybe is because of my MFRONT version??? I do not know. thank you.

disla

Hi Manuel,

I don't think it is practical to experience a new material behavior on a 1 Gb model.

¿Are we talking about Mazars or UnilateralMazars?. It is not the same

UnilateralMazars

*User Material, constants=9
, , , , , , , ,


Mazars
*User Material, constants=8
, , , , , , ,


I have put side by side to cubes. One with UnilatteralMazar and another with conventional Isotropic. Both have the same young modulus and are under the same Pressure.
Displacements has nothing in common. I think we first need to have clear the material properties that are introduced in the definition and which unit system is needed.

Could you provide the units you are using in your model and that useful Mazar material definition excel spreadsheet from Code_aster forum you talk in the Calculix forum?

AndreaC91

I've tried to run some analyses on the concrete cube case changing the parameters but results are someway meaningless.
By the way, I don't understand why a dynamic analysis is required.

It looks like there is no much material on how to set up an analysis with CalculiX and the Mazards constitutive model.

MikeMcMullen

I am quite interested in progress on this.
What I ran into is a lack of material models that model both compression and tension behavior of the arch. Also the actual behavior is not easily captured once there is some tension failure as the stiffness can locally become negative which may or may not allow the solution to converge. Victor indicated to me once that the best way to solve this is with incrementing forced displacements rather than with loads. In any case final local failure occurs when displacement is about 0.26 of the thickness, and is by material crushing. This is why a material model that captures the compression non-linearity is important. Also the tension non-linearity is not well modeled with the compression only materials modeled in Callculix unless there is some reinforcing to take the tensile forces that the compression only model implies at higher strains. Note the problems I ran were flat, never the less genereate substantial internal arching forces, especially for localized loads.

The reference to using dynamic may be referring to the use of an explicit analysis to model the behavior as there are usually not convergence issues if time steps are short enough and stress/strain behavior with areas of negative stiffnesss can be used. There seems to be a lack of examples and documentation for use of calculix or Mecway for this type of analysis.

mmartin

Find attached the excel file modelling MAZARS material that I downloaded from the CA forum.

MikeMcMullen

Oh, I am quite familiar with these models and sometimes use them by hand or spreadsheet. They are not available in calculix, and I am not familiar with MFront enough to compile a material model for calculix and integrate it with Calculix. I do appreciate the nice summary though and think it will inform the understanding of the origionator of this thread.
I had lost my references to Eurocode simplifications on this issue and they are generally not available here in the US ($$$). US codes don't use these advanced material models much as simpler models work well enough for most work. I have been working on bridge rail crash modeling and ultimate failure behavior prediction which is one of the exceptions as there are no service states governing to worry about.

disla

HI,

I have set up three cubes , elastic, Mazar, UnilateralMazar with values I consider reasonable according to the Excel and the attached document.
All three have the same young modulus to have comparable deformations, at least in a first order approximation.
All the other values are the same when possible.

Mazar and UnilateralMazar models accumulates damage but only in a particular direction (the direction of the load applied first)
What doesn’t have any sense to me is that Density needs to be huge to be able to do the models comparison.
There must be some material parameter or internal conversion that is not right.

By other hand, Mazar and UnilateralMazar seems to be exactly the same. ?¿?.
Maybe someone else can advance and provide some enlightenment on this example cubes.

AndreaC91

I have tried to play with the cubes comparison.
The physical behaviour of cubes with Mazars' models might be meaningful, since the tensile stress produced by the dynamic load is higher than the strength. The cubes tend to bear the compression stress only.
However, I've noted that:

1. there is no recover of displacement even if the pressure is much smaller than the tensile strength;
   
2. no stress is computed;
   
3. no difference in displacements is obtained if the Young Modulus and the tensile / compression strength are specified in MPa insted of Pa.
   

As you were saying, the effect of the density is really unclear. Since no gravity load is specified, in principle density should not affect the analysis.

mmartin

At least in my case, It seems, according to the odd results, that MFRONT is not working properly. MANUEL

AndreaC91

I've also tried to look for support on the CCX forum but with no outcome.

Did someone manage to use any other model for concrete on Mecway (e.g. Drucker-Prager, Mohr-Coulomb, etc.)?

mmartin

Good morning ANDREAC91:

I also asked for help on the CCX forum with no answer.
And there is no Mfront update. It seem that the mazars model routine does not deliver material matrixes properly.

In my case, I have not tried another damage model for concrete.
Manuel