Overview#
Setup (design space, loads, constraints, manufacturing methods) is covered on the Generative Design Workspace page. This page picks up at the point where you are ready to generate: how to run studies without wasting tokens, how to explore the outcomes, how to bring one back into Design, and how to prove it works. Treat the generated body as a draft that must be validated and remodeled.
Licensing and cost (Autodesk help):
- Generating outcomes needs tokens or the Fusion Simulation Extension. The charge for outcome generation and use of outcome-based designs is 11 tokens for a generative study. Pressing Generate charges the tokens; the system consumes them when at least one iteration of any outcome is generated.
- The features previously offered by the Generative Design Extension are now part of the Simulation Extension, which covers generative studies without limits (or a 14-day extension trial).
- The Generative Design workspace requires a commercial subscription, an active trial, a start-up license or an educational license; a Personal-use licence is not listed.
- Cancelling before one iteration of any outcome has been generated credits your tokens back; once the Explore environment shows at least one outcome, the generation is not cancelled and tokens are consumed. A study with status Failed (no outcome iteration) is refunded.
- Autodesk may limit job timeouts, queued jobs, concurrent solves and model size (Cloud computing services in Fusion), without publishing numbers.
Where to find it#
- Define tab > Generate panel: Pre-check, Previewer, Generate. In the Generate dialog tick the studies; choose Token costs & balances from the Show list to see Required, Available and Will Remain. You can also right-click a study in the browser and choose Generate, and select several studies with Shift or Ctrl.
- Job Status (Application bar > Job Status icon > View Job Status), Generative Designs tab: processing status and Cancel (shown in the Action column only while cancellation is possible). Clicking Failed opens Generate Details.
- Explore: Define tab > Explore > Explore Generative Results opens the Explore contextual environment, with Thumbnail, Properties, Scatter Plot and Table views, plus Outcome View, Comparison View, Outcome filters and Recommended outcomes.
- Explore tab > Create > Design from Outcome, or Mesh Design from Outcome.
Step by step: running studies economically#
- Run Pre-check and Previewer first (both free). Previewer shows where material goes using the Unrestricted method and one material for active load cases.
- Prototype with low Outcomes Resolution (Study Settings), one material and one or two manufacturing methods, and read the result as a diagnostic: is material connected to every load and constraint, are obstacles respected?
- Widen the study deliberately. The number of outcomes depends on the variability of the setup, such as the number of materials and manufacturing methods (and directions). Remove combinations you would never build.
- Run the final study at higher resolution if needed (High gives more detail but may take longer).
- Explore, choose two or three candidates and create designs.
- Validate each candidate in a Static Stress study created from the generative setup.
A budgeting example using Autodesk's 11 tokens per study (illustrative project plan): one prototype study plus one final study is 22 tokens; repeating the final study twice because of mistakes brings it to 44. With the Simulation Extension the cost is the subscription, but run time still matters. Autodesk does not publish run-time figures; keep a log of your own.
Study and outcome status#
| Status | Meaning (Autodesk) |
|---|---|
| Study: Converged | All outcomes Converged |
| Study: Completed | Outcomes are Converged or Completed |
| Study: Failed | No outcome iteration generated; tokens refunded |
| Outcome: Converged | All objectives achieved; minimum factor of safety is equal to or greater than the Safety Factor limit |
| Outcome: Completed | Generated, but the final iteration does not meet all criteria (for example iteration limit reached or regions separated); may still be valid |
| Outcome: Failed | No iteration generated |
If an outcome has errors, Autodesk suggests a starting shape (a Completed outcome can itself be used as a starting shape), reviewing obstacle and preserve geometry, changing Outcomes Resolution, applying realistic loads, and reviewing milling or 2-axis cutting parameters.
Exploring outcomes#
- Thumbnail View: explore visually. Properties View: compare properties side by side. Scatter Plot View: relationships between properties. Table View: spreadsheet with sortable columns such as name, recommendation, processing status, material, manufacturing method, volume, mass, maximum von Mises stress, factor of safety limit, minimum factor of safety and maximum displacement (global). From each view you can open the Outcome View and the Comparison View, and export a table to CSV.
- Outcome filters: filter by processing status, study, design file, manufacturing method, materials and objective ranges (sliders). Labels and favorites help you organize iterations, and outcomes are grouped by visual similarity after all generations finish.
- Recommended outcomes: a recommendation engine shows the top four outcomes. In Recommendation settings you rate Low mass, Factor of safety achievement, Low displacements, Low cost, Low manufacturing complexity and Solver maturity on a scale from Undesired to Essential.
- Outcome View modes include Model, Transparent, Stress and Design Preview, and you can show preserve geometry, obstacle geometry and starting shape on an outcome.
A good candidate is not simply the lightest: compare mass against displacement and safety factor, check the manufacturing method and consider how easy it is to remodel (general guidance).
Creating a design from an outcome#
| Command | Result | Notes |
|---|---|---|
| Create > Design from Outcome | An editable design in a separate tab | Runs on the cloud; track in Job Status; up to four designs at once; can be edited with T-Spline tools |
| Create > Mesh Design from Outcome | A mesh design | Runs on the cloud; up to four at once |
After the operation completes, click Open Design in Job Status (or the green button in the Outcome or Comparison View). For further analysis you can create a Static Stress study from a generative outcome (Simulation help topic "Set up a Static Stress analysis of a generative design outcome"), and you can send the study to Ansys products.
General remodeling advice (not Autodesk data): rebuild interface features with exact dimensions, add fillets at junctions, and re-run the load cases on the final geometry.
Validation in Simulation#
Autodesk says you need to validate outcomes by calculating stress, strain, displacement and safety factor from the applied loads, by creating and solving a Static Stress study from a generative setup. Suggested checks on the created design (general practice):
| Check | Pass criterion |
|---|---|
| Static stress under every load case | Safety factor at or above your target (default thresholds are 2.0 and 4.0) |
| Mesh convergence at the highest stress | Stress stops changing on refinement |
| Modal frequencies | First mode away from excitation, if vibration matters |
| Buckling (slender members in compression) | Load factor above 1 with margin |
| Fatigue or cyclic load | Outside Fusion's simulation scope; estimate separately |
The study's safety factor can differ from the remodeled part, because geometry changes; leave margin in the study target (general guidance).
Comparing generative design and shape optimization#
| Aspect | Generative design | Shape optimization |
|---|---|---|
| Where | Own workspace | Simulation study |
| Outputs | Many outcomes across methods and materials | One mesh per solve |
| Manufacturing constraints | Unrestricted, additive, milling, 2-axis cutting, casting | Min Member Size and symmetry planes |
| Token cost (Autodesk) | 11 per study | 3 per study (basic) |
| Best for | New parts, process comparison | Lightening a modelled part |
Failure modes and fixes#
| Symptom | Cause | Fix |
|---|---|---|
| Study fails or returns no outcomes | Over-demanding limits, geometry errors, Buckling with Remove rigid body modes | Loosen limits; check Pre-check; read Generate Details |
| Study cannot be generated | Already in progress, already generated, or setup errors | See Autodesk FAQs: wait or cancel, or fix errors reported by Pre-check |
| Outcomes look similar | Narrow study matrix | Add methods or materials; shrink preserve bodies |
| Chosen outcome fails validation | Remodeling or setup assumptions differ | Add fillets, margin, and re-validate |
| Cost higher than expected | Several studies selected, or no extension detected | Read Required in the Generate dialog; check account |
| Connection problems | Network blocking | Allow HTTPS (port 443) and the Autodesk domains in Autodesk's network requirements |
Practice exercise#
Run an economical study end to end on the bracket from the setup page.
- Prototype: one aluminium material, Unrestricted and 2.5-axis milling, Outcomes Resolution Low. Record the outcome count and run time.
- Diagnose: confirm material reaches every preserve body and none enters the obstacle. Fix and rerun if needed.
- Final: add a second material and keep only the milling method you can build.
- Explore: filter by status and objective ranges, sort by mass, check Recommended outcomes and compare the top candidates in the Comparison View.
- Create a design from the best outcome, remodel the interfaces, and run a Static Stress study with each load case. Write down the final safety factor and its difference from the value reported in the study.
- Count tokens used (11 per study without the extension) and decide whether the saved mass justified the cost.
Resolution, run time and result size#
Outcomes Resolution has two positions, Low (faster, less detail) and High (more detail, may take longer). Autodesk notes that the default setting does not work best for every model. Keep a log of each study (setup change, resolution, number of outcomes, elapsed time, tokens): after a few studies this predicts the next one better than any rule of thumb. Delete outcomes you will not use and keep only a few favorites so the document stays responsive (general advice).
Checking the study against a manual design#
Before trusting a generative result, build a simple manual design for the same loads (for example a plain pocketed plate) and run a static stress study. Compare mass, displacement and safety factor. If the best generative outcome is not clearly better, the study may not be worth its cost, or its constraints may have been too conservative (general guidance).
Related#
- Generative Design Workspace
- Shape Optimization
- Cloud Solve and Credits
- Results Interpretation