Generative Design with Fusion 360 and Siemens NX

Generative design is the AI workflow most likely to change what you actually draw at your desk. Instead of you sketching a bracket and asking the simulation whether it survives, you tell the software the load case, the keep-out volumes, the manufacturing method, and the material list, and the AI produces dozens of valid shapes that you would never have drawn by hand. You pick the one that wins on weight, strength, or printability.

This lesson walks you through the workflow in Autodesk Fusion 360, with notes on how the same ideas apply in Siemens NX and other tools.

What You'll Learn

• What generative design is, what it is not, and where it pays off
• The five inputs every generative study needs
• A step-by-step Fusion 360 workflow you can practice with the student license
• How to evaluate the AI's output critically before sending it to a real part
• What it costs (Fusion 360 generative design extension pricing as of 2026)

Generative Design Is Not Topology Optimization (Quite)

Both topology optimization and generative design start from a design space and produce an organic-looking shape. The difference is in the inputs and outputs.

Topology optimization (the older technique) takes one starting design and finds the optimal material distribution within it. Output: one shape.

Generative design (the newer, AI-driven approach) takes a problem definition and produces many candidate designs across different manufacturing methods and materials. Output: a Pareto front of options you can sort, filter, and pick from.

In Fusion 360, the generative workflow can produce designs for additive manufacturing, 5-axis CNC milling, 3-axis CNC, 2-axis cutting, and casting — all in the same study. That breadth is what makes it useful for early concept work.

The Five Inputs of a Generative Study

Every generative study needs:

1. Preserve geometry — the surfaces that must exist (bolt holes, mounting faces, mating surfaces). These cannot change.
2. Obstacle geometry — keep-out volumes the design cannot intrude into (other parts, clearance envelopes, hands of an operator).
3. Loads and constraints — forces, moments, pressures, fixtures. Same as you would set up for a static FEA study.
4. Manufacturing method(s) — additive, milling, casting, 2-axis cutting. The AI generates different geometry for each, because what a printer can do is not what a 3-axis mill can do.
5. Materials — usually 2-5 candidates. The AI explores each and reports the tradeoffs.

If any of those are wrong, the output is wrong. Garbage in, garbage out — except now the garbage looks beautifully organic.

A Fusion 360 Generative Workflow

Here is the workflow you can practice on a student license. Fusion 360 is free for students and educators. The Generative Design extension is a separate paid add-on for commercial use (approximately $1,600 per year for an annual subscription as of 2026, with around 200 cloud solve credits per month included). Students get a free trial and academic access through Autodesk's education program — check the current terms on the Autodesk site before you rely on the pricing.

Step 1. Build the preserve geometry. Start a new design and model only the surfaces that must be preserved — the bolt-hole pattern, the bearing seat, the mounting face. This is what your part must contain.

Step 2. Build the obstacle geometry. Model the surrounding parts, fasteners, and any clearance volumes as separate bodies in the same file. These are what your part cannot intrude into.

Step 3. Enter the Generative Design workspace. From the workspace selector at the top of Fusion. Create a new study.

Step 4. Assign your preserve and obstacle bodies in the study setup.

Step 5. Apply loads and constraints under the Design Conditions section. This works just like the Simulation workspace — pin a face, apply a force, apply a moment.

Step 6. Choose manufacturing methods. Enable additive, milling, etc. For aerospace brackets you almost always want at least additive (titanium DMLS) and 3-axis CNC (aluminum).

Step 7. Choose 2-5 materials. Common starting set: 6061-T6 aluminum, Ti-6Al-4V (titanium), 17-4PH stainless steel, AlSi10Mg (printed aluminum), Inconel 718.

Step 8. Generate. The solve runs in the cloud. Depending on complexity, this takes 30 minutes to several hours. You get an email when it is done.

Step 9. Review outcomes. The output is a grid of candidate designs. You can sort by mass, factor of safety, max stress, or manufacturing cost. Filter by manufacturing method and material.

Step 10. Export the winner. Pick the design that wins on your priority axis, export it as a STEP or mesh file, and pull it into your real CAD/FEA workflow for verification.

Reading the Output Critically

Generative output is gorgeous and very easy to over-trust. Apply this checklist before you commit:

• Is it manufacturable for real? A "printable" design from the AI may still have unsupported overhangs, trapped powder cavities, or thin features below your printer's resolution. Talk to your manufacturing engineer or print bureau.
• Is the load case complete? AI optimizes for the loads you specified. If you forgot a vibration case, a thermal case, or a transport handling load, the design is optimized for the wrong problem.
• Do the safety factors match your shop standard? Generative tools often default to a factor of safety of 2.0. Aerospace primary structure typically wants 1.5 ultimate / 1.0 yield with margins documented separately.
• Verify in your trusted FEA tool. Run the geometry through Ansys, Abaqus, or NASTRAN before you commit. Fusion's internal solver is good for screening; it is not a substitute for verified analysis.
• Check the "AI smell". Watch out for parts that look like sea sponges with no clear load path, sharp interior corners that will crack in fatigue, and parts that ignore obvious assembly access requirements.

Siemens NX, PTC Creo, and SOLIDWORKS

The same workflow exists in:

• Siemens NX — Topology Optimization and the broader NX AI offerings, popular in aerospace primes.
• PTC Creo — Generative Topology Optimization extension.
• SOLIDWORKS — Topology Study within Simulation Professional.
• nTopology / nTop — used heavily for advanced lattice and lightweighting work in aerospace.

The inputs are the same five categories. The outputs look similar. If you learn the workflow in Fusion 360 because the student license is accessible, you can transfer the skill to whichever tool your future employer uses.

When Generative Design Wins

Generative design is most useful when:

• The part is weight-driven (aircraft brackets, drone frames, satellite structure).
• You are early in concept and need to explore alternatives quickly.
• The part will be additively manufactured, so the organic geometry is free.
• The load case is well understood and stable.

It is less useful when:

• The part is dominated by assembly or mating constraints (an injection-molded housing).
• The geometry is prescribed by a standard (e.g. AN/MS fasteners, flange dimensions from a spec).
• You have multiple competing load cases that change frequently in design iteration.

Key Takeaways

• Generative design takes a load case, keep-outs, materials, and manufacturing methods, and produces many candidate geometries.
• Fusion 360 makes it accessible to students; commercial users add the Generative Design extension on top of a Fusion subscription.
• The five inputs that determine output quality are preserve geometry, obstacles, loads, manufacturing methods, and materials.
• Always verify the winning candidate in a trusted FEA tool before committing.
• Generative wins on weight-driven, additively manufactured, early-concept parts; not on housings or prescribed geometries.