product
L-bracket
Minimum mass under a stress limit. The finished part peaks at 1.007× the limit, at the re-entrant corner, and a tighter limit costs material as it should.
Optimisation
Let the loads shape the part.
Give Naksha a design space, the loads and the limits. It returns a lighter, stiffer part as a smooth, watertight surface, and checks it again as built before you see a number.
Topology and topography optimisation are included in Naksha Free.
Topology optimisation
Material only where the load goes.
Start from a block of design space. Naksha removes what doesn't carry load and keeps the paths that do, for the goal you set.
- Stiffness. The stiffest part for a mass budget, over several weighted load cases with forces and pressures.
- Mass. The lightest part that keeps a deflection or stiffness target.
- Stress. The lightest part that stays under a von Mises limit everywhere.
- Vibration. Raise the lowest natural frequencies, or keep them above a floor.
- Buckling. Hold a required buckling load factor while removing material.
Column
Half the material, buckling factor held at its bound. Material moves to the outside and the base: the core empties and the result is a tube.
±4%as-built frequency vs the optimiser
Natural frequency
Clamped beam
The two lowest frequencies are raised at a fixed mass. The part as built lands within 3–4% of the optimiser's prediction, so the gain is real, not grey material.
Manufacturing controls
Designs you can actually make.
An optimised part is only useful if it can be cast, machined or printed. Naksha builds the process into the design, not into a clean-up afterwards.
- Keep and cut regions. Mark bolt bosses, faces and clearances as solid or empty.
- Minimum member size. No struts thinner than your process can make.
- Symmetry. One, two or three planes, held exactly.
- Extrusion. A constant section along an axis.
- Draw direction. Single or split draw, with no undercuts and no internal voids.
- Tolerance check. The part rebuilt slightly thinner and thicker, to show how much stiffness it keeps.
Sheet metal, shells and shape
Thickness, beads and boundaries.
Not every part starts as a block. For sheet and shell structures Naksha sizes the thickness, stamps beads and moves boundaries.
Beads on a plate
Beads stamped into a flat plate cut its compliance by 96%, about 24× stiffer, with no added mass. Bead height, draw angle and symmetry are under your control: the steepest flank is 16° against a 30° limit.
Thickness where it works
Free-size gives every shell element its own thickness. Here it follows beam theory to within 0.4%. Gauge sizing then snaps the result to standard sheet gauges, and the chosen set matches an exhaustive search of all 7,776 combinations.
Free-shape
Boundaries that even out stress
Boundary nodes move to level the stress along an edge, with a mesh-quality guard on every step. On a plate with a hole the peak stress falls 8.9%. Under biaxial load, where a circle is already best, the shape is left alone.
Morphing
Shape changes you define
Describe the changes you're willing to make, such as a moved rib or a wider flange, and Naksha finds the best combination of them.
Crash optimisation
Lighter structures that still protect.
Crash is nonlinear, noisy and expensive. Naksha runs the OpenRadioss simulations in parallel and checks every run before it's trusted.
- Gauge studies. Find the lightest wall thicknesses that meet crush and peak-force limits.
- Crash topology. Redistribute material to cut intrusion at the same mass, or to cut mass while intrusion stays under a limit.
- Run quality gates. Energy balance, hourglass energy, added mass, time-step collapse, impulse and momentum, and SAE J211 channel filtering. A run that fails a gate never steers the design.
Verification
The numbers are for the part you'll build.
An optimiser's own figures describe a cloud of partial material. Naksha turns the result into real geometry and solves that geometry again, so the number you report is the finished part's.
- 1Set upDesign space, keep and cut regions, loads, supports and limits.
- 2OptimiseThe design evolves until it settles, with the best feasible design kept.
- 3Make it solidPartial material is resolved into a clean part, and floating pieces are removed.
- 4Smooth surfaceA closed, watertight STL, ready for CAD and printing.
- 5Solve it againThe part on its own new mesh, with the loads transferred: the as-built answer.
An example from the cantilever benchmark: the solid part on the optimiser's mesh agrees within 0.2%. On its own finer mesh, the part is 16% more flexible than the optimiser predicted, and that is the figure Naksha reports.
Benchmark results
Every benchmark can fail.
Each case has checks that can fail, and they print pass or fail. Here are the results, including the two that don't fully pass yet.
| Benchmark | What it checks | Result | Checks |
|---|---|---|---|
| CantileverStiffness at 30% volume, four variants | Exact symmetry, no undercut along the draw, as-built stiffness vs the optimiser | As built within 0.2%, symmetric to machine precision | 11 / 11 |
| L-bracketMinimum mass under a stress limit | As-built peak stress, no floating pieces, tighter limit costs mass | Peak 1.007× the limit | 5 / 5 |
| Clamped beamTwo lowest frequencies, 49–50% volume | As-built frequencies vs the optimiser, watertight surface | As built within 3–4% | 5 / 5 |
| ColumnTopology with a buckling bound | Bound met, material moves outward and to the base | Bound met at 49.6% volume | 5 / 5 |
| Bead plateTopography with draw angle and symmetry | Stiffness gain, slope limit, symmetry, bead discreteness | −96% compliance, 16° vs 30°; some beads stay part-height | 3 / 4 |
| Shell stripFree-size and gauge sizing | Thickness law vs beam theory, gauge set vs exhaustive search | Exponent 0.498 vs 0.5; the exhaustive optimum | Pass |
| Plate with a holeFree-shape | Peak stress, leaves an already optimal shape alone | −8.9% peak stress | 5 / 5 |
| 3-point bendCrash topology, OpenRadioss | Intrusion at equal mass; lightest design under an intrusion limit | 23.45 → 12.82 mm; 42% mass at 15.96 mm (limit 16) | 9 / 9 |
| Crush tubeGauge study, OpenRadioss | Lighter than the best sampled design, limits met, model accuracy | 271 g vs 298 g; predicted force 10.8% off at the optimum | 4 / 5 |
The optimisation engine also passes 20 of 20 internal tests, including every sensitivity checked against finite differences. The optimisation module is in active development and is being brought into the Naksha workbench. The results above come from the engine on our benchmark set. See the solver benchmarks →
Naksha by Effileap
Bring a bracket. We'll bring it back lighter.
Send us a part, its loads and its limits. We'll show you the optimised design and its as-built numbers.