Generative Design Algorithms Prepare For Space

NASA is famously risk-averse, taking cautious approaches because billions of taxpayer dollars are at stake and each failure receives far more political attention than their many successes. So while moving the final frontier outward requires adopting new ideas, those ideas must first prove themselves through a lengthy process of risk-reduction. Autodesk’s research into generative design algorithms has just taken a significant step on this long journey with a planetary lander concept.

It was built jointly with a research division of NASA’s Jet Propulsion Laboratory, the birthplace of many successful interplanetary space probes. This project got a foot in the door by promising 30% weight savings over conventional design techniques. Large reduction in launch mass is always a good way to get a space engineer’s attention! Mimicking mother nature’s evolutionary process, these algorithms output very organic looking shapes. This is a relatively new approach to design optimization under exploration by multiple engineering software vendors. Not just Autodesk’s “Generative Design” but also “Topology Optimization” in SolidWorks, plus others. Though these shapes appear ideally suited to 3D printing, Autodesk also had to prove their algorithm could work with more traditional fabrication techniques like 5-axis CNC mills.

This is leading-edge research technology though some less specialized, customer-ready versions are starting to trickle out of research labs. Starting with an exclusive circle: People with right tiers of SolidWorks license, the paid (not free) tier of Autodesk Fusion 360, etc. We’ve looked at another recent project with nontraditional organic shapes, and we’ve looked at generative designs used for their form as well as their function. This category of CAD tools hold a lot of promise, and we’re optimistic they’ll soon become widely accessible so we can all put them to good use in our earthbound projects.

Possibly even before they fly to another planet.

[via Engadget]

14 thoughts on “Generative Design Algorithms Prepare For Space

  1. “NASA is famously risk-averse, taking cautious approaches because billions of taxpayer dollars are at stake and each failure receives far more political attention than their many successes. ”

    And people dying.

    ” Mimicking mother nature’s evolutionary process, these algorithms output very organic looking shapes. ”

    Incorporate into the next X-com video game.

    1. So you take a set of constraints, apply a generic genetic EA algorithm, and say cull 50% of each generation that score the lowest and have a standard 1% mutation rate.and run for 100 generations, sounds like no innovation at all, you are totally right :)

    2. Designing space frames, even plain vanilla, straight trusses, is pretty involved if there are strict constraints on things like weight and factor of safety. There is a reason most software only deals with standard designs or very simple cases of nonstandard designs.

      This goes into the free form realm, structurally, and allows fairly arbitrary constraints to be used, without requiring a lot of babysitting from the user.

      A new application for a moderately mature technology? yes. But the devil is in the details, and the analysis of these structures is hard. Even the designers that work with these structures every day tend to lose sight of exactly how hard they are to analyze, because it is hidden by the software. This was first driven home by the Hartford civic center collapse, and every few years there is another reminder.

  2. Waiting for OpenSource software to ride that horse, I´m not going to play with those CAD-cloud solutions, I want full control over the software I use, this SaS thing is not my taste.

    What promising, open-source, software or add-ons for existing software exist yet for generative + optimization CAD design ?

    1. One of the problems with evolutionary design is that is takes many generations and a large breeding stock. For the computation to be done in a reasonable amount of time it needs to be farmed out to many processors in many machines. It is one of the few areas where I actually see using large numbers of other peoples computers for a short periods of time as being extremely useful (currently known as cloud computing). Instead of an single evaluation taking months to years on a single machine, the processing can be reduced to several hours on a large enough server farm (the ranking/evaluations of each generation can happen in parallel, but you still need the results before you can decide on subsequent generations).

      What could be interesting would be to design and implement a generic public cloud, where your resources (specified by you – memory limit, bandwidth limit, monthly bandwidth cap, core count limit, ..) join for say a year and in return you gain access to use the clouds resources for several hours. Maybe add some cloudcoins – *laugh out loud*

  3. The software is free. It comes from Bentley (Generative Components) and Autodesk (Dynamo). It is fully parametric and has a steep learning curve. It is used in Architecture and in construction and most probably can be hooked to other CAD packages as well. The documentation is scarce but there are some examples and guides.

  4. There are a huge number of structural solutions to these problems. The usual limitation is on manufacturing time which usually results in uniform sections for simple production. Obviously that’s less of a concern for a one-off space craft.

    I would probably use hydroformed tubing expanded to provide the correct outside form and then, post forming, machined to reduce bulk where it isn’t required. End fittings would be furnace brazed or EB Welded, then the tubing pressurized with argon to a few hundred psi to prevent buckling from external loads. Pretty sure that the software won’t develop that, but then it’s limited to FDM or milling out large blocks of material.

    In one design I did the key to a successful solution was to use multiple structural materials, which also doesn’t appear to be part of this solution space. Part of the item needed the extra stiffness of steel, but only for a small volume. The rest was low-density aluminum. Pro-tip: The most troublesome mechanical problems don’t concern strength and weight but deflection and vibration. Eventually material deposition techniques will allow a smooth variation in strength and stiffness, but I think that’s a long way off.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.