Concrete Cylindral Crushing Test failure

[Ofsi]

Hello,

I'm currently modelign a squat shear wall using Flexure Beam Elements and comparing the results with ANSYS FEM model. So far the results haven't been good so I'm trying to work from "scratch" that is to begin with simpler things and test my way from there. So a basic test would be to make a simulation of Cylindral Crushing Test to verify that my material models for concrete are similar. However I run into a convergence problem at step 1 using OpenSees for 1.0MN, so what I do is I reduce the load to 10kN I get convergence. Then I increase the load by factor of 10 and similarily increase number of substeps which should result in the same load increase per step, however I run into convergence problem at step 1 again even with load steps increased (if I recall correctly it happened @ 100kN). Oh and I'm using Displacement Beam for the patch cylindral command, shouldn't matter since I'm only trying to verify material properties.

So another method tried was to just keep low load, step size pretty high and analyze over the step size, apparantly it kept on and on and on so the cylinder never crushed. Following is the code, input and comments would be greatly apreciated:

wipe; wipeAnalysis; model basic -ndm 2 -ndf 3


#@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

# SI units [m, N, Pa, kg]

#@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

#Geometry of the wall

set height 0.3
set diameter 0.112838

#nodes

node 1 0 0.0; node 2 0 $height;

#Concrete parameters, compressive strength and strain have negative values

set f_c [expr -25*pow(10,6)]; #Concrete compressive strength
set eps_0 -0.002; #Concrete strain at maximum strength
set E_c [expr 2*$f_c/$eps_0]; #Young's modulus
set dens_c 2400;
set f_t [expr -0.1*$f_c]; #Tensile capacity, the - is needed for tensile strength
set eps_20 -0.005; #Concrete straint at crushing strength
set f_u [expr -5*pow(10,6)]; #Concrete crushing strength
set lambda 0.1;
set E_t [expr $E_c*$lambda];


uniaxialMaterial Concrete02 1 $f_c $eps_0 $f_u $eps_20 $lambda $f_t $E_t


# Define cross-section

section Fiber 1 {
patch circ 1 20 5 0.0 0.0 0.0 [expr $diameter/2.0] 0.0 360.0
}


geomTransf Linear 1

element dispBeamColumn 1 1 2 3 1 1

fix 1 1 1 1

# Set axial load

pattern Plain 1 Constant {

load 2 0.0 1000000 0.0;

}


# Create a recorder to monitor nodal displacements
recorder Node -file "CylindralTest_node.out" -time -node 2 -dof 1 disp

# Create a recorder to monitor element forces in columns
recorder EnvelopeElement -file "CylindralTest_element.out" -time -ele 1 forces

recorder Element -file "CylindralTest_element_stressStrain.out" -time -ele 1 stressStrain




# Create the system of equation, a sparse solver with partial pivoting
system BandGeneral

# Create the constraint handler, the transformation method
constraints Transformation

# Create the DOF numberer, the reverse Cuthill-McKee algorithm
numberer RCM

# Create the convergence test, the norm of the residual with a tolerance of
# 1e-12 and a max number of iterations of 10
test NormDispIncr 1.0e-6 1000 3

# Create the solution algorithm, a Newton-Raphson algorithm
algorithm Newton

# Create the integration scheme, the LoadControl scheme using steps of 0.1
integrator LoadControl 0.001

# Create the analysis object
analysis Static

# Perform the analysis
analyze 1000

[Ofsi]

The load should have negative sign . I also need to swap out the convergence test to the unbalance one. Still strikes me as odd that if I put on 1.e4 load with load increament of 0.001 and get convergence that if I put 1.e5 load on with load increament of 0.0001 I get failure at the first step.

[Ofsi]

Ofsi wrote:
> The load should have negative sign . I also need to swap out the
> convergence test to the unbalance one. Still strikes me as odd that if I
> put on 1.e4 load with load increament of 0.001 and get convergence that if
> I put 1.e5 load on with load increament of 0.0001 I get failure at the
> first step.


This problem still exists even if I change the load direction as well as change the convergence criteria.... 1e5 N it solves 1e6 it doesn't, any internal limits going on here?

[vesna]

I would not use Load Control integrator but Displacement control. Also, for such boundary conditions Plain constraints fit better.

[Ofsi]

Thanks for the answer vesna, however what I'm to achieve is that if I put a -1000000 load on the section it should equal to 100MPa pressure since the area of the cylinder is 0.01m^2. So if I put this load on the column I'd like to get convergence failure @ timestep 0.25 showing that the failure occurs where the material is at its limit. I'm not sure how this would be done with the Load Control Integrator reading the description of it. Also what strikes me as odd is that I can prescribe -1e5 load with LoadControl 0.001 and return successful analysis after 1000 analyses, but I can not prescribe -1e6 load with a LoadControl 0.0001 cause interation failure happens at the first load step.

Also after some more testing I seem to be able to go up to -245000 in load and it analyzes fine, so I'm close to breaking point there. if I put -246000 it fails at loadstep 0.001 so it feels like the program has already decided a failure and won't carry out analysis?

[vesna]

The convergence problem is not indication of a failure. It is just inabillity of your solver to solve the problem given the model and analysis objects.

[Ofsi]

Most interesting, so if I want to find a yielding load to some given system what would be the best method?

It's just I find it strange that the solver discovers the inability of the model at the first loadstep or in some precheck. I've toyed with changing the material properties and I can change the loads accordingly to be very close to the peak crushing strength and I can analyse it, but soon as I adjust the load to be less than 0.5MPa range with the crushing strength it gives error at step 0.001.

This system is simple enough for me to know the maximum displacement of 6e-4 based on the peak compressive strength/strain so I can probably use Displacement control, what worries me is the shear wall section which I'm now trying to figure out if I'm dooing something wrong in since the "breaking point" (now I'm not sure if I can even call the failure of convergence breaking point) and where non-linear behavior starts to occur in FEM and the Flexure Beam macro element is quite large.

[vesna]

When convergence problem is encountered in the first step of analysis, most of the time it is indication of the error in the model.