Katz, Paul wrote:
> Recently I have been
> assigned the task of confirming/maintaining scale factors on our DTM
> sinterstations and I am suddenly realizing what a black art this is. Am
> I missing something? The Nyscale file and spread sheet that my company
> currently uses for this task would be great if materials actually grew
> or shrunk at a consistent rate, but we all know that this is not the
> case. Growth/shrink rates are all geometry dependent (both size and
> mass) and it hasn't been uncommon to see radical rate differences for a
> part that is 2" wide as opposed to a part that is 5" wide.
May I offer a (introductory) explanation:
- the correction factors applied attempt to compensate for observed part
errors, due generally to attributes of EACH particular process type
thus:
- injection moulding has a shrinkage and distortion
- machining has tool error, machine motion errors, tool/part deflection
errors
thermal distortion, stress relief errors
- RP has error sources personal to each process type
The easy things to adjust are the process inputs:
- model geometry (scale, surface offsets, position/orientation in the
machine)
- attributes of the process (temperature, pressure, speed, etc.)
Unfortunately most error sources are not tied to a single input, but
rather
dependent on the local geometry (as you indicated) or the local sequence
in the manufacturing process.
Thus, ideally one would like to vary the input parameters continuously
during
the build so as to get the best quality (shape) output. Obviously, few
manufucturing processes offer easy control to all parameters during the
process
(rather than just setting them once at the start), and its even harder
to
link observed errors to controllable process parameters.
The classic way to solve this problem is model the process and correct
for
(predicted) errors before you make the part. This is generally good for
detecting and correcting gross errors (e.g., cracks, porosity) but not
yet
to the level of dealing with fine tolerances.
<plug for the work I was involved in>
I worked for a few years at the National Research Council of Canada on
an
empirical approach to this problem:
- make a part (with errors)
- observe the (shape) errors (using a laser scanner or other measurement
device)
- change the shape (locally) to reflect the (local) errors (using
software that
compares the observed shape with the desired/designed shape)
This lead to a patent application (successful) for a method to do this
based
on a numerical model of the part/tool (e.g., Finite Element, Finite
Difference, Boundary element). If you're interested in this talk to Dr.
Jeff Xi (jeff.xi@nrc.ca). The general approach has been applied
successful for
CNC machining and Electro Chemical Machining. Basically this approach
automates what toolmakers do manually, and offers a chance to get better
quantitative info on errors.
<end of plug>
In general its still an art, with relatively few "levers" to play with.
-- Bert van den Berg E-mail: bert@hymarc.com Hymarc Ltd. http://www.hymarc.com 38 Auriga Dr., Unit 5 Tel: (613) 727-1584 Ottawa, Ontario K2E 8A5 Fax: (613) 727-0441For more information about the rp-ml, see http://ltk.hut.fi/rp-ml/
This archive was generated by hypermail 2.1.2 : Tue Jun 05 2001 - 22:44:56 EEST