About Submodeling

The model whose solution is interpolated onto the relevant parts of the boundary of the submodel is referred to as the “global” model (even though it may itself be a submodel of a larger “global” model). Driven variables are defined as those variables in the submodel that are constrained to match results from the global model. Driven variables can be degrees of freedom at nodes in the node-based technique, or they can be components of stress tensor at the integration points of element faces in the surface-based technique.

Submodeling can be applied quite generally in Abaqus. The material response defined for the submodel may be different from that defined for the global model. Both the global model and the submodel can have nonlinear response. See Shell-to-solid submodeling and shell-to-solid coupling of a pipe joint for an example application of the submodeling technique.

Vehicle-occupant/pedestrian interaction simulations are an example where the submodeling technique can be used efficiently. Crash safety simulation generally includes interaction between a vehicle and its occupant or a vehicle and a pedestrian. In some cases the influence of the human on the structural response of the vehicle is so small as to be negligible. In these cases, the global analysis of the vehicle is performed without the human or with a simple representation of the human, and the part of the vehicle surrounding the human is then used via the submodeling technique to study the detailed interaction with a number of human models.

Submodeling is classified first according to which of two basic techniques is used. The most common, and more general technique, is node-based submodeling, which uses a nodal results field (including displacement, temperature, or pressure degrees of freedom) to interpolate global model results onto the submodel nodes. The alternative surface-based technique uses the stress field to interpolate global model results onto the submodel integration points on the driven element-based surface facets.

You can choose either the node-based or surface-based technique or a combination of the two in your submodel. The following factors should be considered in deciding on the technique to be used:

• Whether you are performing solid-to-solid submodeling in a general static analysis in Abaqus/Standard:

  • Surface-based submodeling is available only for solid models and static analyses.

  • For all other procedures use the node-based technique.

• Whether the global model and submodel differ significantly in their average stiffness in the region of the submodel:

  • When the stiffness of the models is comparable, node-based submodeling of displacements will provide comparable results to the surface-based technique with a lesser likelihood of numerical issues associated with rigid-body modes.

  • When the stiffness of the models differs and the global model is exposed primarily to a load-controlled environment, the surface-based technique will generally provide more accurate stress results. Stiffness differences may arise due to additional detail in the submodel, such as explicit modeling of a fillet or a hole. In other cases stiffness changes may result from minor geometry changes for which a reanalysis of the global model is not warranted.

• Whether your model is subjected to large deformations or rotations:

  • Node-based submodeling of displacements will result in more accurate transmission of large deformation and rotation to the submodel.

• Whether the displacement response of the global model corresponds to the displacement response of the submodel:

  • When the displacements in the global model correspond closely with the expected displacements in the submodel, node-based submodeling is generally preferable.

  • Surface-based submodeling should be used when the submodel displacement response is expected to differ from the global model response. This situation can occur when thermal strains are modeled and the temperature history of the submodel differs from that of the global model; for example, when heat transfer submodeling is performed as part of a sequential thermal-structural analysis.

• The stiffness of the structure:

  • Surface-based submodeling may provide more accurate results for very stiff structures. When the structure is so stiff that only a small component of the global model displacement field contributes to the stress response, numerical roundoff in the displacement results can become significant; for example, when the global model displacement is dominated by a rigid-body motion component.

• The type of output you are interested in from the submodel:

  • Node-based submodeling of displacements will result in more accurate transmission of a displacement field.

  • Surface-based submodeling will result in more accurate transmission of a stress field, and determination of reaction forces in the submodel.

You can use both node-based submodeling and stress-based submodeling in the same model.

Two techniques are available for node-based submodeling: one technique uses the submodel interface and the other technique uses field import.

Surface-based submodeling is provided as a complement to the node-based technique, enabling you to drive the submodel with stresses from the global model.

A submodeling analysis consists of:

• running a global analysis and saving the results in the vicinity of the submodel boundary;

• defining the total set of driven nodes or driven surfaces in the submodel;

• defining the time variation of the driven variables in the submodel analysis by specifying the actual nodes and degrees of freedom or element-based surfaces to be driven in each step; and

• running the submodel analysis using the “driven variables” to drive the solution.

By default, the global model in the vicinity of the submodel is searched for elements that encompass the locations of driven nodes or driven surfaces' faces; the submodel is then driven by the response of these elements. In some cases more than one element can encompass the location of a driven node. For example, adjacent bodies in the global model may have temporarily coincident nodes or surfaces, as depicted in Figure 1.

A global model with coincident surfaces in the area of the local model's driven nodes.

In this case the location of the driven node in the corresponding global model is touching both element A and element C; however, only the results from element A should drive the node in the submodel.

To preclude certain elements from driving the submodel, you have the option of specifying a global element set to limit the search to an appropriate subset of the global model.

The size of the results file or the output database can be minimized for a submodeling analysis by requesting output for only those global nodes and global elements that are used to drive the submodel. To determine which global nodes and/or elements are used to drive the submodel, do the following:

1. Run a data check analysis on the global model with any combination of results file or output database file output requests. A data check analysis is performed by using the datacheck parameter in the command for running Abaqus (Abaqus/Standard and Abaqus/Explicit Execution).

2. Run a data check analysis on the submodel.

A listing of the global nodes and/or elements that will be used to drive the submodel is output to the data file during the submodeling data check analysis.

Pay special attention to the frequency at which you request output in the global model. It is possible to define the results file output or nodal and element output to the output database file such that the information is written at different frequencies for different nodes and elements, although that should not be done for nodes and elements involved in the interpolation to define values at driven variables since Abaqus will take values at the coarsest frequency only. To avoid this problem, write the nodal and elemental output to the output database or the results file using the same frequency for all nodes and elements involved in the interpolation and choose a frequency that will allow the history in the submodel to be reproduced accurately.