Re:[2] CAD software?

From: Christopher Hicks
Date: Monday, January 30, 1995

From: Christopher Hicks
To: RP-ML
Date: Monday, January 30, 1995
Subject: Re:[2]  CAD software?
     >I'd be real interested to know
     >if anyone else out there spends as much time staring at individual 
     >triangles as I do.
     >Chuck Kirschman                     ckirsch@eng.clemson.edu
     Chuck, 
     
     We design and manufacture orthopedic implants (primarily knee and hip 
     joints).  These parts contain numerous curved surfaces and blended 
     radii... very similar to turbine blade but with more features.  
     
     The curved surfaces look terrible if the meshing is too coarse.  In 
     fact the surfaces can look bad even with a LOT of triangles.  I did 
     some experimenting and came up with a simple explanation as to why the 
     parts looked so rough, even though when inspected by a CMM they were 
     within the allotted profile tolerance  I included an excerpt below 
     from an internal report I did a couple of years ago.  (sorry if this 
     is too long, or too basic)
     
     In conclusion, if I have to have smooth curved surfaces (for example, 
     an RP (HF) master to make a mold) I set the CordHeight to zero during 
     the tessellation in ProEngineer.  ProE will return with an error 
     saying that the specified CordHeight is too small.  ProE will tell you 
     the minimum ChordHeight permitted.  I typically use this or some small 
     multiple of it to give me the smoothest curve possible.
     
     Excerpted from "Analysis of RP for use in Manufacturing" 9/1/93>>"    
     3.1  Tessellation inaccuracies
     
     Rapid prototype part accuracy begins with the CAD model.  As discussed 
     in section 1.1.1, the solid model is converted into an STL file.  The 
     STL file format is accepted by all RP systems and is the common format 
     for data exchange at this time.  When a solid model is converted
     into an STL file, all surfaces are converted into a collection of 
     triangles.  This process is called tessellation  Flat surfaces can be 
     represented by triangular sections with almost no loss in information. 
      Curved surfaces however, can only be approximated by a series of 
     triangles.  The way curved surfaces are approximated by triangles is 
     controlled in ProEngineer through the use of two parameters: 
     ChordHeight and AngleControl.  
     
     AngleControl is important when a surface feature is bounded by a curve 
     with very small radii relative to its part size, such as a dimple 
     (small R) on a golf ball (large R).  If additional improvement in 
     AngleControl is not specified, those features will have little 
     definition in the tessellated output.  AngleControl is less important 
     than ChordHeight for most parts designed at Intermedics Orthopedics. 
     
     ChordHeight specifies the maximum distance between a facet edge and 
     the precise surface of the solid model, as shown in figure 9.  Smaller 
     values of ChordHeight produce more accurate models but increases the 
     size of the resulting STL file.  ProEngineer will impose a lower limit 
     upon how small cord height can be specified for a given model.  
     ChordHeight use is analogous to applying a profile tolerance to a 
     curved surface.
     
     For all the parts built in this accuracy study the lowest permissible 
     value of ChordHeight was used.  For the stem tibia test part discussed 
     in section 3.1, a ChordHeight value of .0002" was used.  An 
     interesting visual phenomenon occurs as a result of the tessellation
     process.  For even small values of ChordHeight, the model may appear 
     to have been coarsely approximated giving the false impression of an 
     inaccurate part.  For example, a typical curved profile used on knee
     articulating surfaces has a 1.3" radius.  Given the ChordHeight, we 
     want to calculate the resulting length of the facet edge since the eye 
     will see the intersections of the facet edges as sharp corners. This 
     can be solved using basic geometry.  {sorry, couldn't include the 
     figures.  Basically, intersect a circle with R=1.3" by a line offset 
     toward the center of the circle by the distance CordHeight.  The 
     length of the straight line between the two points where it intersects 
     the circle is the Facet Length.}
     
     ChordHeight       Facet length
     .010"              .321"
     .001"              .102"
     .005"              .228"
     .0005"             .072"
     .0001"             .032"
     .00005"            .023"
     
     Thus for even a small ChordHeight of .005" the eye will see facets 
     which are .228" long. This is in spite of the fact that the part 
     profile was built to a geometric profile tolerance of +/-.0025".  
     (This value contains only the variation due to the computer model, 
     tolerances due to process inaccuracies are not included in this 
     number.)  Facet lengths for other values of ChordHeight are shown in 
     the table below.  Compared to a machined part, the RP facets fool the 
     eye into believing the RP part is rough because on standard machined 
     part profiles the surface will gradually vary from the highest spot to 
     the lowest spot over a large distance.  
     
     With tessellated RP parts and tessellated CAD solid models, the part 
     profile tolerance is being completely used up as the actual surface 
     zig-zags between the upper and lower profile limits. Parts requiring 
     smooth visual surfaces will require secondary finishing to fill in the 
     corners where tessellated surfaces join or remove material until the 
     surface is level with the intersecting tessellated surfaces.  
     functionally the tessellation "corners" on a part do not affect
     form and fit testing.  "<<
     
     Chris
     
     Chris Hicks       Intermedics Orthopedics, Inc. 
     Advanced Manufacturing Development


Previous message | Next message
Back to 1995 index