Ken,
You have clearly outlined a procedure that will allow people to implement
adaptive slicing for SLA. If enough users start to practice such methods,
there would be additional pressure on RP manufacturers to implement
automated adaptive slicing in their pre-processing software.
The value of adaptive slicing is determined by the dynamics of layer
overhead (recoat for SLA), "draw" time, and desired quality, and as you
point out, we must consider the entire process (finishing time as well), and
not only the build time. This extends to the pre-process prep and slice time
as well. My following comments apply to SLA, as that is what you mention in
your post. The same consideration applies to FDM, 3DP, etc.
I was interested in these methods at one time, and investigated layer
thickness a couple of years ago. I started with an analysis of the benefits
with regard to part quality (which as you point out, also impacts finishing
time) and build time. The results showed that for us, with the types of
builds we were doing at the time on an SLA 250, that 0.004" to 0.0045" was
the optimal layer thickness for minimizing the build cycle time. We were
building at 0.006" at the time. Going below 0.005" caused major problems
with dewetting of the SL5170 resin. Zephyr recoating blades were just being
announced, so leading edge bulge was also a consideration.
BTW, the input for this analysis included height, surface area, and volume
data for several hundred parts (I think around 800), comprising about 75 or
so builds. I calculated a "typical" build using this data.
This analysis considered the entire build, and therefore, the analysis is
flawed due to the lumped approach I took, but still somewhat useful. By
lumped, I mean that each layer in the build was treated the same, as an
average. I am guessing that the maximum scanning time for any layer usually
occurred in the first inch of a build. I am willing to guess that 0.003"
would have been more optimal in that range. The final layers (ignoring stair
stepping for now) usually required little scan time, and therefore, thicker
layers would be more optimal.
Zephyr changed the equation somewhat, due to the fact that we could minimize
recoat time to some extent. I have done no such analysis on an SLA500 or
5000. We recently added a 5000 to the lab, and none of this analysis was
completed, as we are not strained for capacity.
The history of our lab is important for understanding why such a detailed
optimization was done. Demand was growing faster that we could add capacity
(which takes a while in a large company). We needed to make the best use of
our resources during the growth stage from one SLA250 to three, and now
going beyond that level of capacity. I recall working with a colleague to
estimate the typical scan length during the first layers of a build in which
only supports were being built, such that we would optimize the slice
thickness for a typical build.
BTW, we were slicing supports at 0.012" and parts at 0.006" for the entire
build. This had to do with minimizing what was happening in the first layers
before the parts started drawing, but ignored the rest of the build. I think
an average of an hour per build was saved by slicing all supports at 0.006"
Of course, all of this changed when we switched to 0.003" layers on the
SLA250's about 1.5 years ago. An internal service must remain competitive
with service bureaus, or risk being outsourced. Quality was increasing in
importance, and we were now constrained for finishing capacity, and didn't
want to add additional headcount. We increased the build time somewhat in
exchange for better quality and fewer post finishing requirements.
Well, Ken, thanks for prompting the discussion. Thinking back on where we've
been was a fun diversion. There is clearly opportunity for improved
efficiency of the SLA and other RP processes, and I believe these will come
in time. The technology will proliferate further as these improvements drive
down the price. I recall saying a few years ago that the competitive
pressures in the RP industry would turn SLA into a commodity, even with lot
sizes of one. Who would have imagined.
As an aside (but related), I recall a recent post by Al about RP stock
prices. These efficiency improvements do reduce the price of a part, which
in turn leads to more usage of the technology (aided by some other drivers),
which in turn leads to more machine sales for the manufacturers, which in
turn leads to more profits (hopefully). The shareholders, however, expect
something bigger than prototyping applications. We as industrial users are
chasing tooling as the next big area, but it is clear to me that this
application benefits the users, yet does little to raise the manufacturers
to the next level. Shareholders are waiting for the big jump, that 10X
change, a strategic inflection point as Andy Grove calls it. This is what
any sane CEO or President of an RP equipment producer must be focused upon
delivering, if they want to keep their jobs.
Which brings me back to adaptive slicing. I wouldn't look to the machine
manufacturers for this nearly as much as the 3rd party software market.
Companies such as Materialize will benefit from the maturity of the rapid
prototyping industry. The manufacturers can't ignore these efficiencies
either, however, or they will be behind in these areas when the "killer app"
is introduced. Strategic decisions are always the most difficult, and that
is what makes it all so interesting! I could go on for a while about the
strategic choices made by our RP equipment producers, but better bite my
tongue. (Not the right thread anyway).
If anyone is still with me at this point, thanks for listening and have a
good day.
...eDerek
____________________________________
E. Derek Smith
3DP Program Manager
Technology Scout
Motorola Communications Enterprise
8000 West Sunrise Blvd., Room 2329
Ft. Lauderdale, Florida 33322
954-723-4790 (Phone)
954-723-5584 (Fax)
eds014@email.mot.com
____________________________________
> ----------
> From: Foamcaster@aol.com[SMTP:Foamcaster@aol.com]
> Sent: Wednesday, October 28, 1998 2:08 PM
> To: caporusso@picolit.diegm.uniud.it; rp-ml@bart.lpt.fi
> Subject: Re: Adaptive slicing
>
> n a message dated 10/28/1998 3:23:24 AM Mountain Standard Time,
> caporusso@picolit.diegm.uniud.it writes:
>
> << Is there a commercial RP system that wotks with model sliced with
> differen
> layer widths? >>
>
> Yes.
>
> SLA machines (3D Systems) will build multiple thicknesses during the same
> run.
> The process of setting up a build for multiple layer thicknesses is as
> follows:
>
> When setting up the part(s) the operator must change layer thicknesses at
> vertical positions that are EVENLY DIVISIBLE BY ALL LAYER THICKNESSES used
> in
> the build.
>
> Example: A part that is 2.5 inches tall is to be positioned with the
> bottom of
> the part at a Z height of .25 inches. It is desirable to run the first
> .25
> inches of the build (supports) at .006 inch layer thickness, then switch
> to
> .004 inch layer thickness for the bottom 1.5 inches of the part, then
> switch
> to .0025 inch layer thickness for the next .5 inch, and finish the last .5
> inches of the part at .006 inch layers.
>
> Test 1) (All Z height levels where the slice thickness is to change must
> be
> evenly divisible by ALL of the layer thicknesses used in the build.) .25
> inch
> / .006 = 41.666667 .25 inch / .004 = 62.5 .25 inch / .0025 = 100 . . .
> .006 and .004 inch layer thicknesses DO NOT evenly divide into .25 inches.
> Only the .0025 layer thickness divides evenly. What is the nearest Z
> height
> (below the part) at which all 3 proposed layer thicknesses will divide
> evenly?
> .24 inches will divide evenly by all three layer thicknesses. Set up the
> part
> to switch from .006 inch layers to .004 inch layers at the .24 inch Z ht.
>
> Test 2) Using the same process, determine the nearest Z ht at the 1.75
> inch Z
> ht level which is evenly divisible by all three layer thicknesses. Either
> 1.74 or 1.80 can be used since they are the two values closest to the
> desired
> 1.75 inch Z ht where we desired to change slice thicknesses. The needs of
> the
> particular part that is being built will determine whether it is best to
> switch at 1.74 or wait until 1.80 to switch from .004 to .0025 inch layer
> thickness.
>
> Test 3) Using the same process, determine the nearest Z ht at the 2.250
> inch
> Z ht level which is evenly divisible by all three layer thicknesses.
> Either
> 2.220 or 2.280 can be used since they are the two values closest to the
> desired 2.250 inch Z ht where we desired to change slice thicknesses. The
> needs of the particular part that is being built will determine whether it
> is
> best to switch at 2.220 or wait until 2.280 to switch from .0025 to .006
> inch
> layer thickness to finish the build.
>
> If the part is set up as described above, the 3D Systems, Maestro software
> will successfully slice the part(s) and create the build files needed for
> a
> successful build.
>
> Math simplification: A relatively easy way to determine where the
> division
> lines may be successfully made is as follows: 1) Determine what the
> minimum Z
> ht (above zero) which can be evenly divided by all of the layer
> thicknesses
> that are to be used in the build. In the above example, that value is
> .060
> inches. 2) Divide the desired Z ht value where you want to switch layer
> thicknesses by .060. (Example: 1.75 / .060 = 29.1666667). 3) Multiply the
> next even number (above or below) the result of step 2 by .060. (Example:
> 29 *
> .060 = 1.74 or 30 * .060 = 1.80) See Test 2, above, for actual
> application
> wherein these numbers are valid.
>
> It takes a little longer to set up the build this way, but a significant
> amount of build time can be saved, while producing the highest quality
> part if
> the SLA operator will take the time to learn this process. This process
> usually also saves a significant amount of post finishing time.
>
> If any of the above process is not clear, feel free to contact me for
> further
> explanation.
>
> Ken Miller
> Miller Technologies
> 395 S. 1100 W.
> Farmington, UT 84025
> Voice and Fax: (801) 451-7997
> Email: foamcaster@aol.com
>
> For more information about the rp-ml, see http://ltk.hut.fi/rp-ml/
>
For 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 - 22:47:04 EEST