RP machine sensitivity

From: Jon Hunwick (Delcam  International PLC)
To: RP-ML
Date: Friday, September 16, 1994
Subject: RP machine sensitivity
Hello all,

Some while ago I broadcast a message regarding the European research
project known as CARP, (not to be confused with EARP), which is
currently investigating the whole RP issue, combining CAD CAE CFD and
RP. Because we are most interested in components with complex surface
forms we are using surface modelling software (DUCT) to generate the
models, and producing STL (binary) format files. To do this we have
had to produce a suite of programs which can take a damaged mesh (one
with holes in it) and fix it up. We do find, however, that there are
circumstances where the amount of work needed to completely fix the
mesh is excessive, and would therefore request some input from both
users and vendors (please) to answer the following questions.

1 - How sensitive is the system to boundary edges, i.e. faces which
    meet, but do not have properly coincident nodes.

2 - How sensitive is the system to 'sick' triangles. (We define sick
    triangles as those which share edges with more than three others,
    and cause 'stray' vectors).

3 - How sensitive is the system to triangles which overlap. (These are
    usually caused by poor trimmig of the base surfaces).

4 - How sensitive is the system to 'thin' triangles whose aspect ratio
    is enormous.

Having a list (machine by machine) of the relative sensitivities would
be a great help to users, as they would be able to know when to say
'enough is enough' in their fixing, and pass the mesh on to the
machine for building. 

Many thanks in advance.