Method · metal additive manufacturing · all 993 designs
A screen, not a print release
Every metal design,
sliced before it is printed.
Each part design with a metal route goes through the same geometric print screen. It is sliced over its full height at the layer thickness published for its alloy, fitted into a metal printer's build volume and oriented, then checked for supports, thin walls and trapped powder. A part that fails stays failed, with the reason recorded.
Register as of · solver check reviewed through
01 / The result
21 designs, three outcomes
02 / The register
Every design and its outcome
| Design | Alloy | Layer | Envelope, mm | Outcome | Run time |
|---|---|---|---|---|---|
| 993 engine cooling fan — archived Carrera F0 concept | AlSi10Mg aluminium | 30 µm | 280 × 280 × 30 | Rejected by the screen no orientation fits inside the machine | 6 s |
| 993 three-runner intake manifold, AlSi10Mg F0 concept | AlSi10Mg aluminium | 30 µm | 180 × 70 × 100 | Stopped at the memory limit the run exceeded its 10 GB container limit | 3 min |
| 993/993 Turbo connecting rod, Ti64 F0 topology concept | Ti-6Al-4V titanium | 30 µm | 186 × 84 × 19.6 | Rejected by the screen the surface mesh is not one closed body | 6 s |
| Stationary engine fan housing, AlSi10Mg F0 concept | AlSi10Mg aluminium | 30 µm | 300 × 300 × 170 | Rejected by the screen no orientation fits inside the machine | 7 s |
| 993 center cap, AlSi10Mg F0 concept without crest | AlSi10Mg aluminium | 30 µm | 76 × 76 × 46 | Screened | 39 s |
| 993 front impact support, AlSi10Mg graded-core concept F0 | AlSi10Mg aluminium | 30 µm | 139 × 100 × 53 | Screened | 45 s |
| 993 hollow Ti64 intake valve, F0 LPBF concept | Ti-6Al-4V titanium | 30 µm | 49 × 49 × 109 | Screened | 39 s |
| 993 K16 compressor wheel, F0 AlSi10Mg LPBF concept | AlSi10Mg aluminium | 30 µm | 60.5 × 60.5 × 18 | Screened | 19 s |
| 993 left turbo heat shield cover, F0 IN625 concept | Inconel 625 nickel alloy | 40 µm | 160 × 110 × 105 | Screened | 34 s |
| 993 oval exhaust tip, IN625 double-wall F0 concept | Inconel 625 nickel alloy | 40 µm | 120 × 85 × 120 | Screened | 1 min |
| 993 Turbo intercooler end tank, AlSi10Mg concept F0 | AlSi10Mg aluminium | 30 µm | 134 × 274 × 114 | Screened | 1 min |
| 993 Turbo three-into-one exhaust manifold, IN625 F0 concept | Inconel 625 nickel alloy | 40 µm | 146.4 × 66.4 × 215 | Screened | 2 min |
| 993 Turbo/GT2 intercooler bracket, titanium F0 concept | Ti-6Al-4V titanium | 30 µm | 255 × 80 × 23 | Screened | 37 s |
| Engine cooling fan guide-vane ring (stator), AlSi10Mg concept F1 | AlSi10Mg aluminium | 30 µm | 248 × 248 × 40 | Screened | 14.1 h |
| Engine cooling fan impeller, WE43 magnesium concept F1 redesigned from the Turbo rotor rebuild | WE43 magnesium | 30 µm | 243 × 244.8 × 56 | Screened | 2 min |
| Engine oil filter console with integrated galleries, AlSi10Mg concept F0 | AlSi10Mg aluminium | 30 µm | 144 × 90 × 50 | Screened | 1 min |
| Headlamp spring hook repair, aluminium F0 concept | AlSi10Mg aluminium | 30 µm | 16 × 8 × 15 | Screened | 12 s |
| K16 compressor wheel, Al2139 AM F1 iteration | Al2139 aluminium | 60 µm | 60.5 × 60.5 × 18 | Screened | 17 s |
| K16 turbine wheel, IN718 F0 concept | Inconel 718 nickel alloy | 40 µm | 55 × 55 × 20 | Screened | 14 s |
| Turbo oil return line, IN625 F0 concept | Inconel 625 nickel alloy | 40 µm | 140 × 60 × 94 | Screened | 45 s |
| Upper valve cover with COP towers, F0 AlSi10Mg concept | AlSi10Mg aluminium | 30 µm | 220 × 95 × 45 | Screened | 2 min |
Layer thicknesses follow the published process for each alloy on an EOS M 290 class machine: 30 µm for AlSi10Mg and titanium, 40 µm for the Inconel alloys, 60 µm for Al2139. WE43 magnesium is not a catalogue material for that machine; its route follows a published research process (Hyer et al., 2020).
03 / The limits
What a screen can and cannot say
What it catches
- Parts too large for the machine in any orientation, like the archived cooling fan and the fan housing.
- Geometry that is not one closed body, like the first connecting rod mesh.
- Walls too thin to print: the cooling fan's knife-edge blades were found at 0.07 mm and thickened.
- Powder trapped in closed cavities, and how much support a part would need.
What it does not do
No laser path, no supplier support design, no distortion or residual stress, no recoater check and no material property. A screened part is not printable, safe or authorised: it has only passed the first geometric gate. Printing stays prohibited for every design here.
Update / 15 October
Checking a stress solver before trusting it
Distortion and residual stress are outside a geometric screen. They need a thermo-mechanical solver, and a solver has to reproduce answers that are already known before it is pointed at a part. The checks recorded on 15 October exercised open-source codes on synthetic tests, not on any design in the register.
A defect found and corrected
Adamantine, our candidate open-source solver for build thermo-mechanics, counted one shear term twice and left another out when it exported its 3D von Mises stress. For a pure 1 Pa shear it reported 0 Pa or 2.449 Pa, depending on the plane, instead of 1.732 Pa (√3). A second defect ignored the requested output folder in coupled runs. A two-part correction to the exporter passed 16 of 16 direct cases, and another 16 in a separate automated test run. Earlier stress fields from the uncorrected exporter stay inadmissible.
One block, two solvers, one exact answer
A hypothetical 12 × 4 × 6 mm block, modelled as a single finite element, is clamped at its base and heated evenly from 300 to 301 K. For that element the answer is exact: a mean von Mises stress of 1,199,966.34 Pa. CalculiX 2.21 matched all six stress components within 0.432 Pa. The corrected Adamantine, in a restricted research build, came within 0.034 Pa of the reference and 0.050 Pa of CalculiX, against limits of 2 Pa and 4 Pa set in advance. The uncorrected formula would have reported 1,050,529 Pa on the same block, 12.5 % low.
04 / The evidence
Where this comes from
The register and video are imported from our private engineering repository at revision da4ff10. Each part runs alone in a container with a fixed memory limit; a crash or a limit is recorded as such, never turned into a pass. The solver check is summarised from a report dated 15 October 2026, reviewed at revision e697398; its inputs are synthetic and contain no part geometry.