Integration point variables and section variables in
Abaqus/Standard
can be written as field output to the output database in four different
positions: the integration points, the centroid, averaged at nodes, or
extrapolated to the nodes. Integration point variables and section variables in
Abaqus/Explicit
can be written as field output to the output database in three different
positions: the integration points, the centroid, or the nodes. By default,
output is provided at the integration points.
In most cases
Abaqus/Explicit
writes only integration point data to the output database. Transferring of
results from the integration points to the user-specified position in
Abaqus/Explicit
is done by the postprocessing calculator. See
The Postprocessing Calculator
for details.
Element history output to the output database is always provided at the
integration points.
Obtaining Output at the Integration Points
By default, the variables are output at the integration points where they are calculated. In
Abaqus/Standard you can obtain the position of the integration points by using output variable
COORD.
Obtaining Output at the Centroid of Each Element
You can choose to output the variables at the centroid of each element
(the midpoint between the end nodes of a beam or a pipe element). Centroidal
values are obtained by interpolation of the integration point values if the
integration scheme for the element does not include a centroidal integration
point. Element output of the element centroidal values is not available for
recovering results within substructures; for more information, see
Using Substructures.
Obtaining Element Output Extrapolated to the Nodes
You can choose to extrapolate the element integration point variables to
the nodes of each element independently, without averaging the results from
adjoining elements. Element output at the element nodes is not available for
recovering results within substructures; for more information, see
Using Substructures.
Obtaining Element Output Averaged at the Nodes in Abaqus/Standard
You can choose to extrapolate the variables to the nodes and to then
average them over all of the elements in the set that contribute to each node.
For derived variables, such as stress invariants,
Abaqus/Standard
first averages the extrapolated tensor components over all of the elements
connected to the node to obtain unique components at each node and then
calculates the derived value based on the averaged components.
By default,
Abaqus/Standard
partitions the elements in the model into averaging regions. The partitioning
is based upon the structure of the elements: element type, number of section
points, type of material, single layer or composite, etc. Partitioning is not
based upon the values of element properties (such as thickness), material
orientations, or material constants. Averaging occurs only over elements that
contribute to a node and belong to the same averaging region.
In some situations you may want the averaging regions to take into account
the values of element properties. For example, since variables may be
discontinuous between elements with different material constants, you may not
want elements with different property definitions included in the same
averaging region. In such cases you can force
Abaqus/Standard
to take into account values of element properties by setting the
Abaqus
environment parameter average_by_section to
ON
. However, in problems with many section
and/or material definitions the default value of
OFF
will, in general, give much better
performance than the nondefault value of ON
.
Extrapolation and Interpolation of Element Output Variables
The shape functions of the element are used for purposes of extrapolation
and interpolation of output variables. Extrapolated values are generally not as
accurate as the values calculated at the integration points in the areas of
high stress gradients, particularly in the case of modified triangles and
tetrahedra. Therefore, adequately detailed meshing is necessary around nodes
where accurate nodal values of such element results are needed. If a
cylindrical or spherical coordinate system is defined for the element (see
Orientations),
the orientation at each integration point may be different. When the values at
the integration points are extrapolated to the nodes, the difference in the
orientation is not taken into account; therefore, if the orientation varies
significantly over the elements connected to a node, the extrapolated values
are not very accurate. If the material orientation undergoes significant
spatial variation in a region of the model where the material behavior is truly
anisotropic, a finer mesh is required to obtain accurate results even at the
integration points. In that situation once the overall solution has converged
with respect to the mesh density, the interpolation or extrapolation away from
the integration points can also be assumed to be reasonably accurate. You
should also be particularly careful when interpreting output variables
extrapolated to the nodes for second-order elements with midside nodes outside
the quarter-point region, such as when one edge is collapsed in two dimensions
or one face is collapsed in three dimensions.
For derived variables, such as Mises equivalent stress, the components are
first extrapolated or interpolated. The derived value is then calculated from
the extrapolated or interpolated components. However, in linear mode-based
dynamic analysis procedures where derived values are obtained as nonlinear
combinations of modal response magnitudes (Random Response Analysis
and
Response Spectrum Analysis),
the nonlinear combinations are first calculated at the integration points.
These derived values are then extrapolated to the nodes or interpolated to the
centroid.