Autodesk recently added a futuristic function to its Fusion 360 design software. The new Generative Design workspace uses cloud-based artificial intelligence (AI) software that designs mesh and T-spline models with parameters that you provide. It’s called generative because the software generates dozens or hundreds of solutions from those parameters. If you don’t like any of the results, you can input different parameters. The software will offer different, improved outcomes via machine learning, the ability to learn from and analyze data without being specifically programmed to do so.
Generative design is not exclusive to Fusion 360. Autodesk Revit, PTC’s Creo, and Siemens NX are other programs that have it. This software
has produced buildings, products, and construction materials.
WHY USE GENERATIVE DESIGN?
Generative design can produce lighter, stronger, more cost-effective, and aesthetically pleasing outcomes than you could think up on your own.
It frees you from your personal limits on imagination, time, history, engineering biases and experience. Outcomes often look different from traditional solutions, such as the wheel rims above and the airplane partition below.
This is because the AI uses biomimicry and evolutionary algorithms to produce designs
that are organic looking. You’re unlikely to get a perfect, finished design from the generative process, but you’ll get a base design to edit and develop.
ACCESS THE GENERATIVE DESIGN WORKSPACE
Find it in the workspace switcher of a paid, educational, or 30-day trial version of Fusion 360. Depending on your version, you may need cloud credits to run it. The trial version is limited to 300 cloud credits. Buy them through the Preferences panel: General → Cloud Credits (below). (Note: This system appears to be changing — check Autodesk’s website for the latest details). Generative design is not available in Fusion’s free version.
ACCESS IN-HOUSE LESSONS
Open the data panel. If it shows your design files, click on the house icon in the upper left to access the project list and scroll down to Generative Design Samples (above left). Click it open for projects to experiment with (above right). I clicked on the GE Bracket (this particular sample may not be available in all versions). All these files are locked, so right-click on the Browser root and choose Save Copy As (below).
Scroll back up the project list to Your Corner and find the copy there. Click it open, and then click on the Generative Design workspace.
The Generative Design workspace interface looks similar to all the other workspaces, except that it shows a milling (cutting) tool at the origin point.
The Browser has fields for inputting the information required to generate a study. There are also listings for any studies already done. You can right-click and rename past studies from the Browser, or just delete them. Each study listing contains all its inputs.
Click on the Ribbon’s icons from left to right to input your information in the correct order (above). The first icon, Guide, opens a learning panel on the right side of the screen (below).
Now let’s model something more relatable than the GE Bracket to show how generative design works.
GENERATIVELY DESIGN A TREE
Model a part in the Solid workspace to give the AI something to work with. It can be a single body, multiple bodies, or a component assembly. It can be very basic, maybe something that just has mounting holes, flanges, or parts that will carry loads. Remove any non-essential parts like fasteners, pins, fillets, and chamfers. Turn off the visibility of all parts that you don’t want considered in the study.
The image above shows a tree modelled in the Solid workspace. It consists of five simple bodies: three cones, a straight cylinder, and a tapered cylinder. Bring it into the Generative Design workspace and we’ll go through the icons from left to right.
In the Generative workspace, click on the New Generative Study icon to put a “Study1” entry in the browser (above). Everything you input will go here. Then right-click on the study. This brings up a window that lets you choose the result’s resolution: Coarse will generate faster than Fine (below).
You might want to generate several coarse studies to save your cloud credits before generating a fine one.
This lets you create additional bodies to serve as obstacle, preserve, and starting geometries (left). Those geometries don’t have to be created in the original design.
Click to access Solid and Surface submenus and run the mouse over the icons for tooltips. In the Modify menu, the red X removes features, and the blue dot replaces selected objects with primitives. Know that the changes you make in the Generative Design workspace are permanent, not temporary. The Edit Model function is there so that you don’t have to leave this workspace to go into other ones.
Let’s remove the lower two cones by selecting them in the Browser and deleting (above). Then click Finish Edit Model.
Choose what to preserve and what to serve as obstacles (above). Preserved geometry are items to remain untouched. Obstacle geometry are areas in the design where we don’t want the AI to place material. You can also choose a starting shape (this is optional), which is the shape the AI starts from.
- Preserve Geometry: Select the body or bodies you want to remain untouched (above left]. You don’t need to hold the Shift key down for multiple selections. I selected the base and the cone. The preserved shapes turn green after you click OK.
- Obstacle Geometry: Select the bodies you want the material to go around; that is, to not place any new material in that area. I selected the cylinder (above right). It will turn red after you click OK.
- Starting Shape: This is optional. Click on the body you want to serve as a jump-off point for the AI design. Nothing in this case was selectable.
Apply structural constraints and loads to your selected geometry (above).
- Constraints: These can be fixed, pin, or frictionless. You can apply constraints to multiple objects, and depending on the constraint type, to faces, edges, or vertices. I applied a fixed constraint to the cone (above left). Note the colors. Yellow is for starting shape (existing model), green is for preserve geometry, and red is for obstacle geometry.
- Loads: These simulate pushing, pulling, and twisting forces that your design should withstand. You can apply loads to multiple objects, and depending on the load type, to faces, edges, or vertices. I applied a load to the base and entered 1N (newton; a unit of force) in the dialog box (above right).
- Load case attributes: This is a dialog box that shows what you’ve applied (below).
Specify data that will help achieve the goals you want for the outcomes (above left).
- Objectives: What do you want optimized? I chose to minimize mass (above right).
- Manufacturing: Choose from Unrestricted, Additive, and Milling manufacturing methods (below). You can select all options and compare outcomes generated for each option.
Click on Manage Physical Materials (above) and then choose at least one material to use in the design process. An outcome is generated for each material you choose.
- Study Materials: Here you can pick a manufacturing method and a material from the built-in library (above left).
- Manage Physical Materials: This opens the Material Browser (above right), where you can manage and edit material libraries, choose favorites, modify properties, and create materials. I dragged the Aluminum icon up to the In This Study window to apply it to the design.
Pre-check, preview, generate designs, and find the generate status and details here (above).
- Pre-check: This checks the active generative study to ensure the setup meets the requirements to generate outcomes. You might get a dialog box listing warnings and errors to fix before continuing. You may be able to continue with warnings (they’re coded yellow), but not with errors (they’re coded red). But if everything is fine, you’ll get a message saying the study is ready to generate.
- Previewer: This generates an outcome preview after Pre-check verifies the study is ready to go, and before you run the job. You can use this to see how your setup affects outcomes, and adjust accordingly, if needed. Previewer ignores manufacturing and material. Previews can take hours to appear, and Fusion is unusable during that time. It doesn’t require cloud credits.
- Generate: This designs outcomes that meet your design requirements (above). It’s cloud-based and requires cloud credits. You can select multiple studies. Track the processing status in the Job Status dialog box. The amount of time it takes depends on how many loads you’ve applied and the overall complexity, but 3–4 hours or longer is normal. You can exit Fusion and the project will still generate.
- Generate Status : This displays a list of completed jobs and jobs in progress. You can cancel a job before one iteration of any outcome is generated.
- Generate Details: This displays information about outcome generation that you can save to a log file.
EXPLORE AND CREATE
Click the Explore icon to view your outcomes (above left). You can filter and compare multiple outcomes (above right).
Click on a thumbnail to see a larger version (above). The Create menu lets you choose to output a Design (solid) model and a Mesh model (below). After closing Fusion, you can return to the Explore icon to review the outcomes again.
EDIT THE GENERATIVE OUTPUT
THE DESIGN MODEL
This consists of a T-spline body and the preserved geometry. A boundary fill is done automatically. However, the model will generally need more editing.
You may be able to edit it in the Form space by right-clicking on a form icon in the Timeline and choosing Edit. But if that doesn’t work — an edit option doesn’t appear, or the model is ghosted in that workspace — bring it into the Solid workspace. Then click on each Timeline icon from left to right. They’ll highlight specific geometry and temporarily put that geometry in the Form workspace to edit (above).
THE MESH MODEL
Bringing this mesh model into the Mesh workspace is a bit counter-intuitive because typically the Timeline must be turned off. Right-click on the Browser root and choose Capture Design History. Then right-click on the mesh and choose Edit (above ). This will take you into the Mesh workspace (below).
Tweak your mesh model, then export it for 3D printing and see what your generative design looks like in real life!