Standard nonlinear mesh adaptivity (NLAD) creates meshes at required substeps which are delinked from the boundary representation (bRep) of the geometry. Fidelity of the initial mesh to the model boundary (bRep) determines the subsequent fidelity of the successive remeshes during the NLAD process. A very coarse piecewise-linear initial representation of a curved model boundary does not improve even with mesh refinement, as the initial mesh replaces the true bRep by a faceted mesh-based geometry as a reference for subsequent remeshing.

To adequately represent the boundary through remeshing history in conventional NLAD, a carefully crafted fine mesh is necessary to represent the geometry. In terms of an adaptive analysis, this approach is inefficient, as starting with a fine initial mesh is equivalent to an early mesh refinement (which often leads to convergence problems).

The ideal solution is to maintain the evolving bRep during the entire deformation history and the corresponding evolving mesh/geometry relationships during a nonlinear adaptive analysis. NLAD-GPAD does so by using the bRep rather than the preceding deforming mesh object as reference geometry for the new mesh. The advantage is that the initial mesh does not need to be fine-grained to represent the boundary; instead, the adaptive process can refine the mesh in the regions of interest on the boundary and the interior of the body. The bRep representation ensures that the remeshed domains represent the actual model boundary accurately.

NLAD-GPAD supports small-deformation linear analyses (NLGEOM,OFF) only. The program enables the nonlinear solution procedure to solve the geometrically linear problem. Material nonlinearity and contact can be included in a small strain/deformation analysis.