cemoore

I've been making 3d printed models for molds for glass casting, concrete casting, etc, for about seven years and now delving into plaster mold making for slipcasting. I've never had success making a plaster cast directly from a plastic model. Also, plastic models printed on low to mid-priced FDM printers are always going to need 'post-processing' to smooth out the surface (especially the tell-tale 3d printer striations) and seal it properly.  I use automotive primer/filler paints or Smooth-On's XTC-3d epoxy to fill and sand, then shellac, wax, etc. It's an iterative and time-consuming process but can't be skipped.  Once I print the model (I use Rhinoceros software to design - and check for undercuts within the software), I post-process significantly to get a 'perfect' surface as mentioned, cast a rubber mold (I use mostly Smooth-On Vytaflex 40 urethane rubber), then cast the plaster duplicates from the rubber mold. If I have a concave surface model that is small, say a teacup where it would be difficult to post-process the printed model and/or you don't want to make your knuckles bleed, I take one more step and make a convex positive instead, where I'm then able to post-process more easily an exposed outside/convex surface.  From that I then cast two rubber molds - the first rubber mold from the plastic model, and then the second and final rubber mold from the first rubber mold. When casting the rubber molds, I always use a release agent recommended by the rubber manufacturer. No release is necessary to cast the plaster from the rubber mold.  The process is much work, but it's worth the effort, if not for the learning when the failures happen (usually design-related)! 

I've been making 3d printed models for molds for glass casting, concrete casting, etc, for about seven years and now delving into plaster mold making for slipcasting. I've never had success making a plaster cast directly from a plastic model. Also, plastic models printed on low to mid-priced FDM printers are always going to need 'post-processing' to smooth out the surface (especially the tell-tale 3d printer striations) and seal it properly.  I use automotive primer/filler paints or Smooth-On's XTC-3d epoxy to fill and sand, then shellac, wax, etc. It's an iterative and time-consuming process but can't be skipped.  Once I print the model (I use Rhinoceros software to design - and check for undercuts within the software), I post-process significantly to get a 'perfect' surface as mentioned, cast a rubber mold (I use mostly Smooth-On Vytaflex 40 urethane rubber), then cast the plaster duplicates from the rubber mold. If I have a concave surface model that is small, say a teacup where it would be difficult to post-process the printed model and/or you don't want to make your knuckles bleed, I take one more step and make a convex positive instead, where I'm then able to post-process more easily an exposed outside/convex surface.  From that I then cast two rubber molds - the first rubber mold from the plastic model, and then the second and final rubber mold from the first rubber mold. When casting the rubber molds, I always use a release agent recommended by the rubber manufacturer. No release is necessary to cast the plaster from the rubber mold.  The process is much work, but it's worth the effort, if not for the learning when the failures happen (usually design-related)! 

I was definitely NOT suggesting that you used bone china, it warps/slumps too much. Just using it as the classical example of high-bisque low-glost approach. If you do read up on it I would concentrate on how they get the glaze to stick to a non-porous body: pre-heating, additives, ...? 
Will you need the puck to be glazed on all surfaces? If so using stilts is one option, another might be to  glaze the "bottom" in an additional lower-fired glazing.

Casting bodies don't need to be very plastic, and can have comparatively low shrinkage.

https://digitalfire.com/glossary/slip+casting
Slip cast ware can dry-shrink as little as 1.5% (compared to 6%-8% for plastic stoneware bodies).
https://www.potterycrafts.co.uk/Products/pottery-powdered-clays/P3101
Shrinkage (w-d) 2 - 4% & Shrinkage (d-f) 6 - 8%
So maybe 4% for bisque to fully fired?
However you could high-fire  to maturity, grind, and then either high or low-fire to glaze -- so zero shrinkage. Bone-china is traditionally fired high-bisque low-glost (it needs to be supported when fired to maturity).
Glazing a mature body presents problems, as it isn't porous, but is probably doable (AFAIK industrially various additives are used).
PS Solid casting is possible, unless the gradual build-up of clay obstructs the inflow of fresh slip. A hockey puck seems a pretty undemanding shape.

Thanks Tom/(Hulk)! I like your idea of unglazed and hollow but with enough extra thickness to account for removal. It seems there would be so many variations, cutting it down seems the most predictable and practical. Without a glaze, the body would have to be semi-vitreous or vitreous. As to what I make from concrete to the narrow specs - they are OEM parts that go in a bathroom/damp setting, that are attached to cast metal parts. The part shape looks about like an ice hockey puck - nothing truly fancy in shape.  At the end of the day, while my 3d printed models looked round to my eyes, they were slightly off using caliper measures (eg, not perfectly round). A partial solution to that might be spending thousands on machined models, but that's not a financial option. As well, the casting process is always introducing some variations. Milling them after casting made them beautifully round and within spec. As they were small I ended up using a drill press and attaching into a hole on their underside a blank drill bit with a flexible plastic gasket to get tight on the bit. I used a velcro-backed flexible diamond hand pad affixed to a 3d printed arc'd pad holder to get the curved sidewall surface to 90 degrees. A sort of poor-man's milling. Just thinking about that project in ceramic but also future projects where I might need to meet these kind of specs and ways I might be able to machine/grind my own designs.  
Your input is greatly appreciated! Always learning.
 

I've been making 3d printed models for molds for glass casting, concrete casting, etc, for about seven years and now delving into plaster mold making for slipcasting. I've never had success making a plaster cast directly from a plastic model. Also, plastic models printed on low to mid-priced FDM printers are always going to need 'post-processing' to smooth out the surface (especially the tell-tale 3d printer striations) and seal it properly.  I use automotive primer/filler paints or Smooth-On's XTC-3d epoxy to fill and sand, then shellac, wax, etc. It's an iterative and time-consuming process but can't be skipped.  Once I print the model (I use Rhinoceros software to design - and check for undercuts within the software), I post-process significantly to get a 'perfect' surface as mentioned, cast a rubber mold (I use mostly Smooth-On Vytaflex 40 urethane rubber), then cast the plaster duplicates from the rubber mold. If I have a concave surface model that is small, say a teacup where it would be difficult to post-process the printed model and/or you don't want to make your knuckles bleed, I take one more step and make a convex positive instead, where I'm then able to post-process more easily an exposed outside/convex surface.  From that I then cast two rubber molds - the first rubber mold from the plastic model, and then the second and final rubber mold from the first rubber mold. When casting the rubber molds, I always use a release agent recommended by the rubber manufacturer. No release is necessary to cast the plaster from the rubber mold.  The process is much work, but it's worth the effort, if not for the learning when the failures happen (usually design-related)! 

I've been making 3d printed models for molds for glass casting, concrete casting, etc, for about seven years and now delving into plaster mold making for slipcasting. I've never had success making a plaster cast directly from a plastic model. Also, plastic models printed on low to mid-priced FDM printers are always going to need 'post-processing' to smooth out the surface (especially the tell-tale 3d printer striations) and seal it properly.  I use automotive primer/filler paints or Smooth-On's XTC-3d epoxy to fill and sand, then shellac, wax, etc. It's an iterative and time-consuming process but can't be skipped.  Once I print the model (I use Rhinoceros software to design - and check for undercuts within the software), I post-process significantly to get a 'perfect' surface as mentioned, cast a rubber mold (I use mostly Smooth-On Vytaflex 40 urethane rubber), then cast the plaster duplicates from the rubber mold. If I have a concave surface model that is small, say a teacup where it would be difficult to post-process the printed model and/or you don't want to make your knuckles bleed, I take one more step and make a convex positive instead, where I'm then able to post-process more easily an exposed outside/convex surface.  From that I then cast two rubber molds - the first rubber mold from the plastic model, and then the second and final rubber mold from the first rubber mold. When casting the rubber molds, I always use a release agent recommended by the rubber manufacturer. No release is necessary to cast the plaster from the rubber mold.  The process is much work, but it's worth the effort, if not for the learning when the failures happen (usually design-related)! 

I've been making 3d printed models for molds for glass casting, concrete casting, etc, for about seven years and now delving into plaster mold making for slipcasting. I've never had success making a plaster cast directly from a plastic model. Also, plastic models printed on low to mid-priced FDM printers are always going to need 'post-processing' to smooth out the surface (especially the tell-tale 3d printer striations) and seal it properly.  I use automotive primer/filler paints or Smooth-On's XTC-3d epoxy to fill and sand, then shellac, wax, etc. It's an iterative and time-consuming process but can't be skipped.  Once I print the model (I use Rhinoceros software to design - and check for undercuts within the software), I post-process significantly to get a 'perfect' surface as mentioned, cast a rubber mold (I use mostly Smooth-On Vytaflex 40 urethane rubber), then cast the plaster duplicates from the rubber mold. If I have a concave surface model that is small, say a teacup where it would be difficult to post-process the printed model and/or you don't want to make your knuckles bleed, I take one more step and make a convex positive instead, where I'm then able to post-process more easily an exposed outside/convex surface.  From that I then cast two rubber molds - the first rubber mold from the plastic model, and then the second and final rubber mold from the first rubber mold. When casting the rubber molds, I always use a release agent recommended by the rubber manufacturer. No release is necessary to cast the plaster from the rubber mold.  The process is much work, but it's worth the effort, if not for the learning when the failures happen (usually design-related)!