Repairing CAD Models
Barequet Gill and Subodh Kumar
Proceeding IEEE Visualization, Phoenix, AZ, IEEE, pp.363-370, October 1997
Johns Hopkins University
Abstract: We describe an algorithm for repairing polyhedral CAD models that have errors in their B-REP. Errors like cracks, degeneracies, duplication, holes and overlaps are usually introduced in solid models due to imprecise arithmetic, model transformations , designer's fault, programming bugs, etc. Such errors often hamper further processing like finite element analysis, radiosity computation and rapid prototyping. Our fault-repair algorithm converts an unordered collection of polygons to a shared-verte x representation to help eliminate errors. This is done by choosing, for each polygon edge, the most appropriate edge to unify it with. The two edges are then geometricall y merged into one, by moving vertices. At the end of this process, each polygon edge is either coincident with another or is a boundary edge for a polygonal hole or a dangli ng wall and may be appropriately repaired. Finally, in order to allow user- inspection o f the automatic corrections, we produce a visualization of the repair and let the user mark the corrections that conflict with the original design intent. A second iteratio n of the correction algorithm then produces a repair that is commensurate with the intent. Thus, by involving the users in a feedback loop, we are able to refine the correction to their satisfaction.
CAD Data Repair
Butlin, Geoffrey and Clive Stops
5th International Meshing Roundtable, Sandia National Laboratories, pp.7-12, October 1996
Abstract: Using CAD data for analysis is becoming more common, but is still fraught with dirty geometry problems such as slivers, crossovers, minute edge lengths, stray points 'on the moon', wonderful clean-as-a-whistle models that are useless for meshing, patchworks of faces that attract unnecessary elements, birds nests of draughting geometry stuck in a corner, etc. Even if binary, native CAD files are received, the engineering analyst is at the mercy of the idiosyncrasies of the CAD operator and the CAD software.
Data exchange through standards or native formats is only part of the solution. Indee d if IGES is used, the problems are often exacerbated. Data transfer only provides the exchange of geometric entities, 'as-is'. A CAD model usually requires significant modification, or transformation to get it into a form suitable for engineering analys is.
While the transfer function can be fully automatic with IGES, STEP or native formats, the transformation process is typically subjective, application specific and requires engineering judgement, and is therefore not readily automated.
FEGS CADfix software provides data transfer through IGES, STEP and native formats interactive repair tools for cleaning, and healing interactive de-featuring t ools for merging edges and faces and removing slivers and tangencies interactive editing tools for splitting up solid models into hex-meshable and CFD block-meshable regions.
Following the repair, smoothing, simplification, splitting up etc, the model can then be exported directly into the geometric modellers in analysis code pre-processors, such as ANSYS and Patran.
The typical target for CAD data transfer is auto tet-meshing. Such automatic algorithms are unfortunately at the mercy of idiosyncrasies of an apparently valid Br ep model. Minute slivers, fine tangencies, narrow ,necks' and face/edge sloppiness etc, can challenge even the most robust tet-mesher. A solution to this problem is proposed in the form of a batch macro of CADfix procedure calls that put the solid model 'through the laundry', washing out all minute features and ironing out all edge and f ace coincidences, discontinuities and discrepancies, according to a specified resolution. Such a model 'laundry' or super-cleaner, when invoked as a batch pre-processor promises to significantly raise the robustness of auto tet-meshing.
Give me a good mesh
5th International Meshing Roundtable, Sandia National Laboratories, pp.3, October 1996
Abstract :After much thought the title of my talk should be "Give me a good mesh". The following abstract should clarify the title.
This presentation addresses often asked questions that surround the mesh generation process. Answers to questions like the one in the title can involve numerical quantit ies that measure the shape of elements, descriptions of processes used to improve initial meshes and constraints on the density of elements. This talk aims to provide substantive material to politely but authoritatively respond to more provocative questions like "Is this mesh any good?"
About the Keynote Speaker: David Field earned an A.B., an M.S. and a Ph.D in mathematics from Bowdoin College, Oakland University and the University of Colorado, respectively. After teaching at the College of the Holy Cross, he joined the General Motors Research Laboratories where he holds the position of Staff Research Scientist. His research interests include numerical analysis, approximation theory, finite element analysis, and mathematics from computer aided design. In addition to cofounding and currently serving as president of the Great Lakes Section of SIAM he coedited for SIAM two volumes on geometrical and theoretical aspects of industrial design. He is a national visiting lecturer for the Society for Industrial and Applied Mathematics and for the Mathematical Association of America.
Constructive Solid Geometry, The CAD-FEM Connection
Marcal, Pedro V.
5th International Meshing Roundtable, Sandia National Laboratories, pp.157-168, October 1996
Abstract :A constructive solid geometry approach is adopted for FEM models that mirrors the approach in a CAD package. The topology is first divided into sub-regions using Boolean Geometry. The sub-regions are then mapped into their final shapes and a structured mesh is built. This enables the automatic meshing of all components built by the CAD package.
Midsurface Abstraction from 3D Solid Models for Engineering Analysis
As stated before, the most efficient analysis model for complex parts may consist of a union of 1D, 2D, and 3D geometries. Automatic meshing of such non-manifold geometries is not a trivial task. Recently, Shimada and Gossard  presented a method in which the mesh is obtained by closely packing spheres in the domain and placing nodes at the centers of the packed spheres. The distance between spheres is controlled by defining internode forces and finding a force-balancing configuration. Such methods are promising and may eventually lead to robust non-manifold meshing algorithms. Commercial automatic meshers do much better when the geometry is either 1D, 2D, or 3D. Therefore, the parts considered here will have on ly thin walls.