[Seismic] Fiber Beam-Column Element #6116

zidanmahmoud · 2019-12-17T21:01:33Z

This element is to model the response of reinforced concrete beam-column elements due to seismic loading. The element was first developed as part of StructuralMechanicsApplication, but because it is very specific as per the discussion in #6107, it is put in its own application for now.

In details, the objective of this element is to model reinforced concrete under cyclic loading, typically beyond the yielding point. It models a 1D element under axial force and biaxial moments. The formulation of the element is a flexibility (inverse stiffness) based and solved with displacement control non-linear FEM developed in StructuralMechanicsApp (#5799 & #5968) so that it handles the material softening in the concrete. Essentially, the element is cut into sections, where the numerical integration points live. The sections are discretized into fibers, where each fiber can have its own constitutive law (Figure). The stiffness of the element is derived from the flexibility matrices of the sections, which in turn are derived from the tangent moduli of the fibers.

[Fabio Taucer, Enrico Spacone, and Filip Filippou. A fiber beam-column element for seismic response analysis of reinforced concrete structures. 01 1991.]

…onstit. law

loumalouomega

Several comments, please @AlejandroCornejo check the CL implementations, I think it can be simplified

loumalouomega · 2019-12-18T09:20:02Z

applications/SeismicApplication/custom_conditions/fiber_beam_column_section.cpp

+void FiberBeamColumnSection::CalculateBMatrix(Matrix& rBMatrix)
+{
+    KRATOS_TRY
+    rBMatrix = ZeroMatrix(3, 5);


it is possible to assign manually the zeros? (is cheaper), otherwise:

Suggested change

rBMatrix = ZeroMatrix(3, 5);

noalias(rBMatrix) = ZeroMatrix(3, 5);

This is just to ensure resizing the matrix and fill it to zeros. I pass a matrix of 0,0 size then resize it with zeros then fill the numbers. noalias would not work because it requires size conformity. Would it be a better solution to do resize then fill manually or simply pass the correct size of the matrix? This thing is also done in a couple other places in the code ..

OK, in that case you should check the size first, resize if required and use noalias then

Will do. Thanks

loumalouomega · 2019-12-18T09:41:37Z

applications/SeismicApplication/custom_conditions/fiber_beam_column_section.hpp

@@ -0,0 +1,304 @@
+//  KRATOS  ____       _               _


The header are .h for agreement

loumalouomega · 2019-12-18T09:42:14Z

applications/SeismicApplication/custom_conditions/fiber_beam_column_section.hpp

+    IndexType mId;       // id of the section
+    double mPosition;    // position of the integration point
+    double mWeight;      // weight of the integration point
+    double mTolerance;   // tolerance of the section equilibrium


This should be aproeprty, saving in each section is super expensive

This in the other hand should be defined in a shared object to reduce memory needs

It is used in one function. I could simply pass it to the function from the element without saving it.

For example

loumalouomega · 2019-12-18T09:42:38Z

applications/SeismicApplication/custom_conditions/fiber_beam_column_section.hpp

+
+    IndexType mId;       // id of the section
+    double mPosition;    // position of the integration point
+    double mWeight;      // weight of the integration point


Can this be evaluated on running time?

Ok, I see this is not a condition, nevermind my comment

loumalouomega · 2019-12-18T09:43:16Z

applications/SeismicApplication/custom_conditions/fiber_beam_column_section.hpp

+ * @author Mahmoud Zidan
+ */
+
+class KRATOS_API(SEISMIC_APPLICATION) FiberBeamColumnSection


If this is not a condition should be ain a different folder, not in custom_conditions

loumalouomega · 2019-12-18T09:48:50Z

applications/SeismicApplication/custom_constitutive/uniaxial_kent_park.cpp

+
+// Project includes
+#include "includes/properties.h"
+#include "custom_constitutive/uniaxial_kent_park.hpp"


Idem. headers are .h

loumalouomega · 2019-12-18T09:49:47Z

applications/SeismicApplication/custom_constitutive/uniaxial_kent_park.cpp

+    rSerializer.save("mStrainR", mStrainR);
+    rSerializer.save("mStrainP", mStrainP);
+    rSerializer.save("mUnloadSlope", mUnloadSlope);
+    rSerializer.save("mTangentModulus", mTangentModulus);


Do you really need to save everything?, particularly the tangent modulus

The tangent modulus, in particular, I am saving in memory because the computation of the material response involves many steps so that we do not have to do all the computations when we just need the tangent modulus. I am looking into reducing the number of saved variables ..

I trust you, just in case

I reduced many of the mConverged* variables, and I believe the results are still the same. 🤞

I recommend to create a test before change anything and change the things later, so you know you haven't broke anything

Sorry for the delay. I fixed the test case and added it to CI, but the test case fails and the results are slightly different. So, I think keeping the extra variables is needed for now 😐

loumalouomega · 2019-12-18T09:51:35Z

applications/SeismicApplication/custom_constitutive/uniaxial_menegotto_pinto.cpp

+void UniaxialMenegottoPintoMaterialLaw::FinalizeMaterialResponsePK2(Parameters& rValues)
+{
+    KRATOS_TRY
+    mConvergedLoadingIndex  =                mLoadingIndex;


This is not properly done, please look the plasticity laws implemented by @AlejandroCornejo. You don't save everything twice, you revaluate and update the converged value

loumalouomega · 2019-12-18T09:52:26Z

applications/SeismicApplication/custom_elements/fiber_beam_column_element_3D2N.cpp

+void FiberBeamColumnElement3D2N::PrintData(std::ostream& rOStream) const {
+    pGetGeometry()->PrintData(rOStream);
+}
+


What the main difference with the beam element already existing?

Differences are:

Geometry is a line_gauss_lobatto_3d_2, similar but another integration rule

The integration points are sections, which are discretized into fibers

After every non-linear iteration, a loop of element equilibrium is performed

The stiffness matrix is the inverse of the integral of section flexibility matrices

Ok, can we look for an approach in order to reduce code duplication?

Technically, some methods are exactly the same, because the global dofs are the same. The local however are different. It is only one axial force and two biaxial moments in this element. So in principle maybe they can share a base class?

I suggest to wait with this for two reasons:

Many methods in the structural elements are duplicated, I was talking to @KlausBSautter that we should do a proper refactor soon

@KlausBSautter has implemented a new beam element, which still needs to be merged, it will replace completely the current one

loumalouomega · 2019-12-18T09:53:13Z

applications/SeismicApplication/seismic_application.cpp

+    // calling base class register to register Kratos components
+    KratosApplication::Register();
+
+    KRATOS_INFO("") << "    KRATOS  ____       _               _      \n";


We decided to avoid this, or at least doing it in only one line

zidanmahmoud added 14 commits October 21, 2019 12:32

classes for fiberBeamColumn element

177bab3

adding all functions of beam-column element + adding steel uniaxial c…

6bebb6a

…onstit. law

updates to fiberBeamColumn element

186414f

coverging fiber beam example

b169da6

Merge branch 'master' into StructMech/FiberBeamColumn-Element

13b8e18

adding uniaxial linear elastic material law

897ff5a

Merge branch 'master' into StructMech/FiberBeamColumn-Element

a63a31e

documentation + correction of the constitutive laws

06d6406

Merge branch 'master' into StructMech/FiberBeamColumn-Element

10253a3

Removing commented out lines

9302908

migration of fiberBeam element to SeismicApp

1c68304

removing unnecessary comment

18da17d

cleanup

6b1d9ef

Merge branch 'master' into StructMech/FiberBeamColumn-Element

83b1311

zidanmahmoud requested review from philbucher, longchentum and KlausBSautter December 17, 2019 21:01

minor changes

c73c86f

loumalouomega requested changes Dec 18, 2019

View reviewed changes

Mahmoud added 8 commits December 18, 2019 12:07

adding small fiberBeam element test

224fdd0

overall code optimization

a584ac0

renaming .hpp to .h

8c2d8b5

removing extra variables from CLaws

881b341

base class for section and fiber

47aa634

fix test case

75fb2b9

bug fix CLaws

46db15c

Merge branch 'master' into StructMech/FiberBeamColumn-Element

38bed0f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Seismic] Fiber Beam-Column Element #6116

[Seismic] Fiber Beam-Column Element #6116

zidanmahmoud commented Dec 17, 2019 •

edited

Loading

loumalouomega left a comment

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Jan 15, 2020

loumalouomega Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

loumalouomega Dec 18, 2019

philbucher Dec 18, 2019

loumalouomega Dec 18, 2019

zidanmahmoud Dec 18, 2019

	rBMatrix = ZeroMatrix(3, 5);
	noalias(rBMatrix) = ZeroMatrix(3, 5);

[Seismic] Fiber Beam-Column Element #6116

Are you sure you want to change the base?

[Seismic] Fiber Beam-Column Element #6116

Conversation

zidanmahmoud commented Dec 17, 2019 • edited Loading

loumalouomega left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

zidanmahmoud commented Dec 17, 2019 •

edited

Loading