Michael Hancock Debashis Basu Ashish Das Nilanjan Mukherjee

Michael Hancock, Debashis Basu, Ashish Das, Nilanjan Mukherjee ( Michael. Hancock@sdrc. com )

Need for Abstraction • Interoperability through STEP, IGES and Vendor to Vendor translator compatibility -- a must !! - CAD Data quality becomes a major issue. • • Sliver surfaces • Unstitched geometry Industry Approach for FE Analysis - FIX the CAD and mesh • Fix the CAD ( geometry ) with another CAD tool • Make additional changes considering Analysis Intent - Create and mesh an Abstraction of the geometry using virtual topology that only refers to the geometry • • Geometry is never modified Analysis Intent built-in to some degree Good mesh quality achievable Define Boundary Conditions on the abstraction SDRC Confidential – Do Not Copy

Abstraction Process Legacy FE data STL data Create abstraction Group to criteria De-feature Re-feature Mesh to meet requirement Analysis cycle(s) Geometry abstraction Group to criteria De-feature Re-mesh to requirement Synthesized geometry Native geometry Imported Geometry SDRC Confidential – Do Not Copy

Analysis Intent • What is Analysis Intent ? - Feature Removal, Simplification • Suppression • • - Features (thru hole, blind hole, etc) - Edges - Vertex Small feature removal (auto-merge) Bead Abstraction Boundary Smoothing Isthmus removal - Design Alternatives (need New features) • Bead Creation • Replace existing curve to a new boundary SDRC Confidential – Do Not Copy

Virtual Topology from Surfaces • Virtual Topology - The area entity of the virtual topology is called a Section - Section can reference one or more surfaces - Section ->Loops ->Curves->connectors - Section-surface relation is tracked - If curve is on edge(s), curve-edge relation is tracked • Characteristics - Loops are always closed - There could be multiple loops, but at least one - A loop cannot self-intersect - Section must be meshable - Adjacent sections share a common boundary SDRC Confidential – Do Not Copy

Shell of Sections forms a Water-Tight Virtual Volume • Sections can be made water-tight even if there are unstitched surfaces with gaps and overlaps Free edges of unstitched surfaces Volume is not water-tight Sections define a water-tight volume SDRC Confidential – Do Not Copy

Section Creation Options • Basic option - Hole suppression Curve Merging Surface Grouping Tolerance • Advanced option - Total Curvature Pre-process fillets Combine cylinders Surface merge based on target element size SDRC Confidential – Do Not Copy

Section Creation • Abstraction on forward creation of sections Auto create -- std. Mesh on counter-bore SDRC Confidential – Do Not Copy Auto create -- adv.

Manual Modification Tools • Manipulating the virtual topology entities - changes section definition meshing reacts to the change boundary conditions react to the change Any removal operation can be undone by an add operation • Some capabilities - Add, Remove, Replace connectors • Split, Merge, Stretch curves - Add, Remove, Replace curves • Split, Merge, Stitch sections ( sometimes loops ) - Add, Remove loops • Un-suppress, suppress features attached to loop(s) SDRC Confidential – Do Not Copy

De-featuring: Remove Loop • De-featuring --- entity suppression ( remove hole ) At times it is helpful to be able to create hard points for mesh to snap to. SDRC Confidential – Do Not Copy

De-featuring: Isthmus Removal • De-featuring --- Isthmus removal ( replace curve ) 3 sections created over 72 surfaces The 2 isthmus sections contain 2 outer loops each. An isthmus section Still 3 sections !! Replace curve operation removes the isthmus. Still one section with two outer loops. 30 mm mesh ( free mapped, allow tri ) SDRC Confidential – Do Not Copy 30 mm mesh ( free mapped, allow tri )

De-featuring: Auto-merge Section • De-featuring --- Auto-merge (small feature removal) 518 surfaces 410 sections Using auto-merge with 13 mm - Higher order operator built to suit analysis intent SDRC Confidential – Do Not Copy

De-featuring: Bead Abstraction • De-featuring --- Bead Abstraction Two sections on surfaces defining beads Extract median line defining bead Mesh may go across the entire section Ensure mesh to capture stiffness May remove the rail curves of the beads Final representation for bead and fillet Create median line on fillets SDRC Confidential – Do Not Copy

De-featuring: Boundary Smoothing • De-featuring --- Boundary smoothing Laplacian smoothing for nodes that has no projection space SDRC Confidential – Do Not Copy

Re-featuring: Bead Creation • Re-featuring --- Bead Creation SDRC Confidential – Do Not Copy

Re-featuring: Replace Curve • Re-featuring --- Replace curve to a new boundary - Mesh will get projected to underlying surfaces - Additional options • Ignore surface for projection • Add surface for projection (surface is not a part of the solid ) SDRC Confidential – Do Not Copy

Example: A Shell Part • Completely unstitched geometry 1. Automated Abstraction: <5 minutes 2. Manual Editing of Abstraction <15 minutes 3. Mesh Generation < 5 minutes 393 surfaces ->112 sections -> 5481 elements created SDRC Confidential – Do Not Copy

Example: A Solid Part • 1481 surfaces, 1062 sections, 32985 par. tri, 56578 par. tet SDRC Confidential – Do Not Copy

CAD neutral ? • The Abstraction works on a CAD NEUTRAL level while providing flexibility for analysis and design modifications. • Future direction – Abstraction on Mesh • Industry reaction - Ford Power-train says “time to mesh large power-train model was reduced by 75% in last one year” - ZF Friedrichshafen AG says “section meshing is a key in the process chain” SDRC Confidential – Do Not Copy

Abstract • AN ABSTRACTION AND MESHING TECHNIQUE FOR INDUSTRY PROBLEMS. - Michael Hancock, Debashis Basu, Ashish Das and Nilanjan Mukherjee • The most commonly used data exchange methods between the CAD and CAE application are (a) Direct translators, which are vendor-to-vendor data exchange, (b) Indirect translators, like IGES, STEP, STL and the like and (c) Consistent kernel, like ACIS, Parasolid and the like. Although there has been many thoughts on creating a CAD neutral framework, in reality the design data containing surface information is handed over to the analysis land through one of the above data exchange channels. Any surface data from an industry model, coming through one of the above data exchange channels, brings imperfect geometry with gaps, overlaps and surface degeneracy. Meshing such surfaces is highly unlikely to produce quality mesh with a desired density. This paper proposes an abstraction technique that creates an auxiliary simplified topology referencing the underlying imperfect (sometimes perfect) geometry. A meshing strategy is presented that works directly on the abstraction layer. The synergy between the abstraction and meshing techniques provides one way of achieving “CAD neutrality”, trying to embed the concept in the process. The idea here is to accept any surface data from any data exchange channel, be it stitched or unstitched, and create a watertight topology layer (the abstraction) spanning multiple surfaces which is then used for meshing. An application based on this abstraction technique provides the flexibility to remove topological details irrelevant for finite element analysis while providing the facility to perform design modification (for example elimination of holes) without altering the geometry. The technique eliminates the need to repair any underlying imperfect geometry. The meshing technique smoothes out effects from geometry deficiencies and works on any unspecified (void) region within the abstraction to produce a quality mesh of the desired density. SDRC Confidential – Do Not Copy

Solutions to. . . Get There Faster™