Every ceramic AM part leaves the printer larger than its final drawing. The green body contains resin, wax and ceramic powder. When those organics exit and the particles fuse, the part shrinks. We turn that predictable shrinkage into a controlled dimension.
Where the volume goes
Shrinkage happens in two stages. During debind, the binder burns out and the part loses mass but changes shape very little. During sintering, the ceramic particles rearrange and bond, pulling the whole geometry inward. The linear shrinkage for our oxide ceramics is 15 to 20 percent. Volumetrically, that is roughly 38 to 49 percent, which explains why the green part looks nothing like the dense final part.
The good news is that the shrinkage is repeatable for a given material, layer thickness and furnace profile. Once we characterise the shrinkage factor for a grade, we can predict final dimensions before we print.
Scaling the build file
Compensation is straightforward on paper: divide the target CAD dimensions by the expected shrinkage factor. If a zirconia grade shrinks 18 percent linearly, we scale the model by 1/0.82, or about 1.22, before slicing. The printer then builds a part that is 22 percent larger in every direction. After sintering, the dimensions return to the nominal drawing.
In practice, we maintain a per-grade shrinkage table and update it from production measurements. Wall thickness, part size and cross-section can shift the effective shrinkage slightly, so the first article of a new geometry is always measured before we release production parameters.
Anisotropy and tolerance bands
Shrinkage is not perfectly isotropic. Gravity, support orientation, thermal gradients in the furnace, and local geometry can cause minor asymmetry. Tall thin features may shrink slightly more in the Z direction than in XY. Thick sections heat differently from thin walls. We manage this by orienting parts to minimise Z-height variation and by keeping the thermal profile slow enough that gradients even out.
For parts with tight positional tolerances, we add machining allowance to critical surfaces. After sintering, those surfaces are ground or lapped to final dimension. The ±40 µm tolerance we quote is for as-sintered dimensions on well-suited geometries; tighter tolerances are possible with post-processing.
Design assumptions that break compensation
Shrinkage compensation fails if the CAD file contains details that cannot survive the process. Holes below 0.5 mm, walls below 100 µm as-sintered, and unsupported overhangs will distort or collapse before the final dimensions are reached. We review models for these risks and recommend changes before printing. The earlier we see the geometry, the cheaper it is to adjust the design rather than scrap parts.