[SMApp] Adding a mixed Displacement/Strain element #12897
Conversation
| // Compute the symmetric gradient of the displacements | ||
| noalias(rKinVariables.SymmGradientDispl) = prod(rKinVariables.B, rKinVariables.NodalDisplacements); | ||
|
|
||
| noalias(rKinVariables.NodalStrain) = prod(rKinVariables.N_epsilon, rKinVariables.NodalStrains); |
There was a problem hiding this comment.
Are we sure of this? Maybe I'm wrong, but it seems that we are using the same container for the nodal values and the integration point interpolation.
There was a problem hiding this comment.
noalias(rKinVariables.NodalStrain) = prod(rKinVariables.N_epsilon, rKinVariables.NodalStrains); The left term is a interpolated strain Vector and the right one is the shape function matrix and a element size strain vector
There was a problem hiding this comment.
Ufff... now I see the s at the end... This is very misleading IMO.
Taking into account that NodalStrain is only used to calculate the equivalent strain, I'd avoid saving this in the kinematic variables container and directly do the interpolation it in the CalculateEquivalentStrain.
...tion/custom_elements/solid_elements/small_displacement_mixed_strain_displacement_element.cpp
Outdated
Show resolved
Hide resolved
...tion/custom_elements/solid_elements/small_displacement_mixed_strain_displacement_element.cpp
Outdated
Show resolved
Hide resolved
...tion/custom_elements/solid_elements/small_displacement_mixed_strain_displacement_element.cpp
Outdated
Show resolved
Hide resolved
| protected: | ||
|
|
||
| /** | ||
| * Internal variables used in the kinematic calculations | ||
| */ | ||
| struct KinematicVariables | ||
| { | ||
| Vector N; | ||
| Matrix B; | ||
| Matrix N_epsilon; // Used to interpolate nodal strains | ||
| double detJ0; | ||
| Matrix J0; | ||
| Matrix InvJ0; | ||
| Matrix DN_DX; | ||
| Vector NodalDisplacements; // Displacement DoFs -> U | ||
| Vector NodalStrains; // Strains stored at the nodes (strain DoFs) -> E | ||
| Vector EquivalentStrain; // Stabilized strain field E = (1-tau) N_e · E + tau Bu · U | ||
| Vector SymmGradientDispl; // Symmetric gradient of the nodal displacements: Bu·U | ||
| Vector NodalStrain; // N_e · E | ||
|
|
||
| /** | ||
| * The default constructor | ||
| * @param StrainSize The size of the strain vector in Voigt notation | ||
| * @param Dimension The problem dimension: 2D or 3D | ||
| * @param NumberOfNodes The number of nodes of the element | ||
| */ | ||
| KinematicVariables( | ||
| const SizeType StrainSize, | ||
| const SizeType Dimension, | ||
| const SizeType NumberOfNodes | ||
| ) | ||
| { | ||
| detJ0 = 1.0; | ||
| N = ZeroVector(NumberOfNodes); | ||
| B = ZeroMatrix(StrainSize, Dimension * NumberOfNodes); | ||
| DN_DX = ZeroMatrix(NumberOfNodes, Dimension); | ||
| J0 = ZeroMatrix(Dimension, Dimension); | ||
| InvJ0 = ZeroMatrix(Dimension, Dimension); | ||
| NodalDisplacements = ZeroVector(Dimension * NumberOfNodes); | ||
| NodalStrains = ZeroVector(NumberOfNodes * StrainSize); | ||
| EquivalentStrain = ZeroVector(StrainSize); | ||
| SymmGradientDispl = ZeroVector(StrainSize); | ||
| NodalStrain = ZeroVector(StrainSize); | ||
| N_epsilon = ZeroMatrix(StrainSize, StrainSize * NumberOfNodes); | ||
| } | ||
| }; | ||
|
|
||
| /** | ||
| * Internal variables used in the kinematic calculations | ||
| */ | ||
| struct ConstitutiveVariables | ||
| { | ||
| Vector StrainVector; | ||
| Vector StressVector; | ||
| Matrix D; | ||
|
|
||
| /** | ||
| * The default constructor | ||
| * @param StrainSize The size of the strain vector in Voigt notation | ||
| */ | ||
| ConstitutiveVariables(const SizeType StrainSize) | ||
| { | ||
| StrainVector = ZeroVector(StrainSize); | ||
| StressVector = ZeroVector(StrainSize); | ||
| D = ZeroMatrix(StrainSize, StrainSize); | ||
| } | ||
| }; | ||
|
|
There was a problem hiding this comment.
I think that at the moment we can make all these private.
Let's see if I've addressed them all :)
|
...tion/custom_elements/solid_elements/small_displacement_mixed_strain_displacement_element.cpp
Outdated
Show resolved
Hide resolved
| switch (integration_order) | ||
| { | ||
| case 1: | ||
| mThisIntegrationMethod = GeometryData::IntegrationMethod::GI_GAUSS_1; | ||
| break; | ||
| case 2: | ||
| mThisIntegrationMethod = GeometryData::IntegrationMethod::GI_GAUSS_2; | ||
| break; | ||
| case 3: | ||
| mThisIntegrationMethod = GeometryData::IntegrationMethod::GI_GAUSS_3; | ||
| break; | ||
| case 4: | ||
| mThisIntegrationMethod = GeometryData::IntegrationMethod::GI_GAUSS_4; | ||
| break; | ||
| case 5: | ||
| mThisIntegrationMethod = GeometryData::IntegrationMethod::GI_GAUSS_5; | ||
| break; | ||
| default: | ||
| KRATOS_WARNING("SmallDisplacementMixedStrainDisplacementElement") << "Integration order " << integration_order << " is not available, using GI_LOBATTO_1" << std::endl; | ||
| mThisIntegrationMethod = GeometryData::IntegrationMethod::GI_LOBATTO_1; |
There was a problem hiding this comment.
I think it is possible to directly specify the quadrature. I think that it is more convenient in this case to do so rather than the order.
There was a problem hiding this comment.
How can we do this?
|
There are still things to be addressed, specially the issue with the variable. |
…lid_elements/small_displacement_mixed_strain_displacement_element.cpp Co-authored-by: Rubén Zorrilla <[email protected]>
📝 Description
Adding a mixed u/E high accuracy element in small strains. This element is based on:
with some improvements such as the usageof tangent constitutive matrix in the balance of linear mom equation and the elastic constitutive matrix in the strain compatibility equation.