Surface-Based Submodeling

You must define the source of the global solution results and provide the name of the output database file (in ODB or SIM format); the file extension is optional. If the file extension is omitted, Abaqus will use in order, the ODB output database file or the SIM database file.

A geometric tolerance is used to define how far driven element-based surface nodes in the submodel can lie outside the exterior surface of the global model, as that surface is interpolated in the global, undeformed finite element model. By default, surface nodes in the submodel must lie within a distance calculated by multiplying the average element size in the global model by 0.05. You can change the tolerance, which is useful in cases where submodel driven surfaces lie to a greater extent outside the global model exterior surface. Tolerances larger than this default value, however, can result in significantly greater computation times and lower accuracy in the driven solution for driven surface regions significantly outside the global model exterior surface.

You can define the geometric tolerance as a fraction of the size of the average element in the global model or as an absolute distance in the length units chosen for the model. If both tolerances are defined, Abaqus uses the tighter tolerance.

SUBMODEL, TYPE=SURFACE, ABSOLUTE EXTERIOR TOLERANCE=tolerance

SUBMODEL, TYPE=SURFACE, EXTERIOR TOLERANCE=tolerance

The Exterior Tolerance in Solid-to-Solid Submodeling

The exterior tolerance for a solid-to-solid submodel analysis is indicated by the shaded region in Figure 2. If the distance between the driven surface nodes and the free surface of the global model falls within the specified tolerance, the solution variables from the global model are extrapolated to the submodel.

Figure 2. The exterior tolerance in surface-based submodeling.

You specify the step of the global model history that is to be used for the driven variables in the current submodel analysis step.

DSLOAD, SUBMODEL, STEP=step

You can modify the submodel distributed surface load definitions from step to step to change the global step reference, you can remove surface load definitions, and you can reintroduce them later (see About Loads). Submodel distributed surface loads do not propagate between steps. At each new step all submodel distributed surface loads defined in previous steps will be removed unless they are modified or redefined. New surfaces cannot be added to the total set of driven surface defined for the submodel; this set of driven surfaces is a fixed part of the model definition.

DSLOAD, SUBMODEL, OP=MOD
DSLOAD, SUBMODEL, OP=NEW

You can improve accuracy and achieve reasonable submodel solutions by observing the following guidelines:

• Design your models so that your submodel surface intersects the global model in regions of relatively low stress gradients.

• Design your models so that your submodel surface intersects the global model in regions of uniform element size. A warning message is provided in the data (.dat) file in cases where significant nonuniform element size distributions are seen.

To understand whether your modeling approach results in a reasonably accurate solution, the following guidelines are recommended:

• Compare the stress distributions on the submodel-driven surfaces with the stress distributions in the global model. The degree to which the global model's stress distributions agree with those in the submodel-driven surface is generally an indication of the level of accuracy of your submodel solution.

• When using inertia relief in the submodel for cases where submodeling does not remove all rigid body modes, compare the inertia relief forces to the prevailing force level in your submodel. If the inertia relief force is large compared to the prevailing force level, your submodel results may be inaccurate.

When you use surface-based submodeling exclusively to drive your submodel response, your displacement solution will not be unique; you will generally encounter rigid-body modes and accompanying numerical issues. You can address these rigid-body modes by

• providing sufficient node-based submodel displacement boundary condition definitions in the submodel analysis,

• providing sufficient boundary condition definitions in the submodel analysis, or

• providing an inertia relief load definition in the submodel analysis (see Inertia Relief).

You can combine these definitions, as necessary and appropriate to your model, to address all rigid body modes.

Global model stress results are stored in the output database in the global coordinate system. Submodel tractions are calculated from these stresses and the current configuration surface normal in the submodel. Hence, when your global model result involves significant finite rotation, your submodel results will generally be inaccurate unless you provide sufficient node-based submodel displacement boundary condition definitions to impart similar rigid-body rotations to the submodel; exclusive use of surface-based submodeling definitions is not adequate to provide these rigid-body motions. You may also experience convergence difficulties in the submodel when it is not properly rotated.

When surface-based submodeling is used to drive a submodel region with an inelastic material definition, you may encounter rigid-body modes and accompanying numerical issues. For example, numerical issues will prevent convergence if the submodel material definition includes plasticity and the submodel loading results in a shear band formation beyond the material hardening definition, such that unconstrained motion can occur (i.e., if the submodel loads exceed the limit load capacity). In these cases node-based submodeling should be used.