NativeCAD · topology optimisation

The GE jet engine bracket, 84.3% lighter

The bracket 629 engineers raced to lighten in the 2013 GE challenge — 2.04 kg of titanium — optimised by NativeCAD for all four challenge loads at once and handed back as parts you can build: printed, cast, machined from plate or welded from flat stock, each checked in CalculiX.

321 glightest version that carries every load (C), from 2.04 kg

1.76×stiffness per kilogram of the original (A)

5 of 5buildable versions that carry all four challenge loads, pin loads to 42.3 kN

321 gC is lighter than the 327 g published challenge winner

Load paths

Switch the viewer to Load paths to see the bracket reduced to its skeleton: 17 members and 14 joints, each drawn at the thickness the optimiser gave it and coloured by the stress it carries across all four cases. The hardest-working members are the two arches under the clevis lugs, at 55% of yield under the 42° load. Versions D, E and F are laid out on that skeleton, and every version has its own: pick one with the chips to see where its load actually goes.

Every version, one scoreboard

Version	Mass	Vertical stiffness	Safety factor, worst case	Build
The 2013 design envelope, rebuilt in NativeCAD	2.04 kg	400 kN/mm	1.55 CalculiX, 175k-node tet mesh	Machined billet
Topology optimisation, 20 % of the envelope	419 g 79.5% lighter	224 kN/mm 1.76× stiffness per kg	1.28 NativeCAD FEA, 1 mm grid	Reference shape
A with exact bores, seats and lug eyes	419 g 79.5% lighter	235 kN/mm 1.76× stiffness per kg	1.29 CalculiX, 148k-node tet mesh	Metal AM (DMLS) 24,000-face B-rep
Optimised at 15 %, same interfaces	321 g 84.3% lighter	184 kN/mm 1.59× stiffness per kg	1.36 CalculiX, 154k-node tet mesh	Metal AM (DMLS) 24,000-face B-rep
Four straight legs on the load paths, editable feature tree	391 g 80.9% lighter	225 kN/mm 1.05× stiffness per kg	1.60 CalculiX, 150k-node tet mesh	Investment casting or metal AM 24,000-face B-rep 14 features
Plates and ribs on the load paths, nothing overhangs from the top	763 g 62.7% lighter	122 kN/mm 0.8× stiffness per kg	1.76 CalculiX, 149k-node tet mesh	3-axis CNC from plate, 2 setups 24,000-face B-rep 18 features
Seven flat plates on the load paths	790 g 61.3% lighter	129 kN/mm 0.83× stiffness per kg	2.10 CalculiX, 149k-node tet mesh	Waterjet-cut plate, TIG welded 24,000-face B-rep 18 features
Published winner, 2013 challenge	327 g	Published result, Designs 5(4):77

Safety factor against the Ti-6Al-4V yield of 903 MPa (131 ksi, the challenge figure), worst of the four load cases. The four bolt bores are held rigid, as the challenge specifies; the idealised rigid edge where a bore meets the underside is singular, so the CalculiX peak is taken 3 mm clear of the bolt bores. Stiffness is the pin load over the pin's travel under the 35.6 kN vertical case.

B STEP (gz) 6.8 MB C STEP (gz) 6.8 MB D STEP (gz) 6.8 MB D feature tree 3 KB E STEP (gz) 6.5 MB E feature tree 2 KB F STEP (gz) 6.6 MB F feature tree 5 KB

One bracket, every way to make it

Metal AM (DMLS)B · Print-ready · 419 g · SF 1.29Printed in one piece; the bores and bolt seats are finish-machined.

Metal AM (DMLS)C · Lightest · 321 g · SF 1.36The lightest shape, printed in one piece with the same finish-machined interfaces.

Investment casting or metal AMD · Strut frame · 391 g · SF 1.60One casting from a printed or moulded wax pattern. Four straight legs fill cleanly and inspect easily.

3-axis CNC from plate, 2 setupsE · Machined · 763 g · SF 1.76One top setup mills everything from a 78 mm plate, a second drills the pin bore. Straight walls and cutter-radius corners: no 5-axis work.

Waterjet-cut plate, TIG weldedF · Welded · 790 g · SF 2.10Seven waterjet-cut plates, fillet-welded, then both bores drilled in one fixture. Near-net: very little titanium ends up as chips.

The four challenge loads

35.6 kN vertical8,000 lbf straight up through the pin

37.8 kN horizontal8,500 lbf out, across the pin

42.3 kN at 42°9,500 lbf, 42° from vertical

565 N·m torsion5,000 lbf·in about the clevis centre

Bolts rigid, pin rigid, Ti-6Al-4V — the challenge's own rules. All four cases are optimised together, not one at a time.

Run it on your part

The same optimisation is one call in the NativeCAD MCP server — cad_topopt: pick the faces that must stay, the faces that are bolted, the faces that are loaded, and a weight budget. It returns the optimised shape and its numbers, ready for cad_mesh_to_brep and a CalculiX check with cad_fea_case.