[CORE] Adding new Lobatto quadratures to linear geometries

[AlejandroCornejo]

📝 Description
Adding a new entry to the enum of thequadraturesfor 1st order Lobatto quadratures (integrate exactly polynomials up to order 1).

Added option to triangles, quads, test and hexas.

[loumalouomega]

FYI @KratosMultiphysics/technical-committee

[AlejandroCornejo]

Now passes! @rubenzorrilla

[rubenzorrilla]

I think there are still some issues.

[AlejandroCornejo]

I think there are still some issues.

I modified them accordingly

[rubenzorrilla]

Suggested change

Quadrature < HexahedronGaussLobattoIntegrationPoints1, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(),

Why repeated?

[AlejandroCornejo]

It is not "repeated" actually. Previously, in the EXTENDED_GI_GAUSS thy were using the Lobatto so I just maintained the previous code.

[rubenzorrilla]

Still I don't get it. But aren't we calling thrice the same method? In other words, shouldn't Quadrature < HexahedronGaussLobattoIntegrationPoints1, 3, IntegrationPoint<3> >::GenerateIntegrationPoints() appear once? (right now it appears three times).

[AlejandroCornejo]

I may be totally wrong but, by analogy with other geometries I was implementing, on the one hand:

    static const ShapeFunctionsValuesContainerType AllShapeFunctionsValues()
    {
        ShapeFunctionsValuesContainerType shape_functions_values =
        {
            {
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_GAUSS_1 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_GAUSS_2 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_GAUSS_3 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_GAUSS_4 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_GAUSS_5 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_EXTENDED_GAUSS_1 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_EXTENDED_GAUSS_2 ),
                Hexahedra3D8<TPointType>::CalculateShapeFunctionsIntegrationPointsValues( GeometryData::IntegrationMethod::GI_LOBATTO_1 )
            }
        };
        return shape_functions_values;
    }

and then we need to assing an individual quadrature to each os the previous entries:

    static const IntegrationPointsContainerType AllIntegrationPoints()
    {
        IntegrationPointsContainerType integration_points =
        {
            {
                Quadrature < HexahedronGaussLegendreIntegrationPoints1, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(),
                Quadrature < HexahedronGaussLegendreIntegrationPoints2, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(),
                Quadrature < HexahedronGaussLegendreIntegrationPoints3, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(),
                Quadrature < HexahedronGaussLegendreIntegrationPoints4, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(),
                Quadrature < HexahedronGaussLegendreIntegrationPoints5, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(),
                Quadrature < HexahedronGaussLobattoIntegrationPoints1, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(), // was there already
                Quadrature < HexahedronGaussLobattoIntegrationPoints1, 3, IntegrationPoint<3> >::GenerateIntegrationPoints(), // was there already
                Quadrature < HexahedronGaussLobattoIntegrationPoints1, 3, IntegrationPoint<3> >::GenerateIntegrationPoints() // My contribution! :)
            }
        };
        return integration_points;
    }

Are we now on the same page?

[rubenzorrilla]

OK. Now I understand. Your addition is correct. What is weird to me is that GI_EXTENDED_GAUSS_1 and GI_EXTENDED_GAUSS_2 feature the HexahedronGaussLobattoIntegrationPoints1 integration points (something that was already there).

[rubenzorrilla]

@loumalouomega together with @AlejandroCornejo we've discussing about this. We think that it might be something weird with the hexa GI_EXTENDED_GAUSS_* integration as it was using the Lobatto integration points (rather than the collocation ones). I think that you can shred some light on this.

If it is actually wrong, or we put is as a boilerplate for a future implementation, my suggestion would be to remove it.

[AlejandroCornejo]

For now, I've removed the extended ones amd maintained the standard gauss and the new Lobatto one. FYI @loumalouomega

[ipouplana]

THX @markelov208 for your tips. I kept doing some research and the problem seems to be this code (for the 2d quadrilateral elements), GeoMechanicsApplication\custom_elements\U_Pw_small_strain_interface_element.hpp:

    virtual void CalculateOnLobattoIntegrationPoints(const Variable<array_1d<double, 3>>& rVariable,
                                                     std::vector<array_1d<double, 3>>&    rOutput,
                                                     const ProcessInfo& rCurrentProcessInfo);

    virtual void CalculateOnLobattoIntegrationPoints(const Variable<Matrix>& rVariable,
                                                     std::vector<Matrix>&    rOutput,
                                                     const ProcessInfo&      rCurrentProcessInfo);

    virtual void CalculateOnLobattoIntegrationPoints(const Variable<Vector>& rVariable,
                                                     std::vector<Vector>&    rOutput,
                                                     const ProcessInfo&      rCurrentProcessInfo);

@ipouplana this code seems to be yours, could you give me a hint of what's being done here? I see that you are computing something using Lobatto quadratures (it is using the mThisIntegrationMethod which by default is GeometryData::IntegrationMethod::GI_GAUSS_2 so it is not Lobatto) and then interpolating. Now the Lobatto quadratures can be used via GeometryData::IntegrationMethod::GI_LOBATTO_1

We already discussed this internally, but I will mention it here too for the record.
The interface elements are defined with a special geometry which uses a Lobatto quadrature file. And thus it actually uses a Lobatto quadrature regardless of the label used to call it, e.g. GI_GAUSS_2.

[AlejandroCornejo]

The problem is SOLVED! :)

[rubenzorrilla]

#12918 (comment) is not clear to me yet. Aside of this, looks good to me.

[AlejandroCornejo]

#12918 (comment) is not clear to me yet. Aside of this, looks good to me.

For now I've removed the EXTENDED_GAUSS_GI in the hexa since it is not consistent

[rubenzorrilla]

All my concerns have been addressed.