How to ... use generative design to improve the outcomes of manufacturing
Find out how generative design tools helped optimise the manufacture of a Mars lander.
NASA’s search for life is expanding to the outer bodies of our solar system. But to reach these destinations, we need smarter design tools. Autodesk was tasked to produce a Mars lander for NASA’s Jet Propulsion Laboratory (JPL). This non-operational model would be a proof-of-concept to explore the capabilities of our software and demonstrate how it could inform future decisions. The project required Autodesk to create the structure of the lander to meet the requirements for space travel – from scratch, not improving upon an existing model. It also meant working with already successful engineers to understand their needs, to develop software tools to meet their demands, and introduce them to the very different way of working using generative design.
What is generative design?
Generative design is a multi-objective exploration approach. It allows engineers to explore specification complexities and to evolve and fine-tune a design according to changing constraints and costs, to produce more innovative solutions before entering into production.
There are always multiple viable outcomes for any multi-objective design optimisation problem, so generative design provides a framework to help identify and compare these different outcomes. It equips the user to handle requirement changes down the line, such as if your payload or budget changes, then you need to be able to quickly generate new solutions in as fast a time as possible.
Generative and optimal
This approach allows the user to evaluate trade-offs between many design conditions, which is particularly useful for ones that may not have been in the initial optimisation scope. For instance, Autodesk software for designing mechanical elements enables designers to isolate individual components to help define the problem definition. For instance, the user can colour code elements to define regions to preserve or those outside of the remit, and once the boundary conditions are defined, optimisation objectives and manufacturing methods can be addressed.
Solution options are generated in the Cloud, where users can visualise the trade-offs between the different solutions, comparing, for example, a minimum factor of safety to mass, or the impact of using various alloys. Then the user can export the best solution as a solid geometry. From here, geometric modifications can be made according to softer design elements such as aesthetics. In this way, it is possible to address design intent, which may not have been in the initial optimisation scope.
Applying it to manufacturing
When developing the lander, the team realised during the conceptual phases that typical topology optimisation methods would only go so far in terms of mass reduction. So it was necessary to start researching and prototyping different synthesis strategies, combining various optimisation methodologies and stiffening methods.
Ribs, lattices and hollow structures were introduced, before performing tests to evaluate the structural performance of these new synthesis methods to compare deflection versus mass. Autodesk also evaluated the best manufacturing methods for the new design strategies. Here, the choice of material was important.
Information on fastening geometry, load cases, manufacturing details like tool access requirements, as well as integration and assembly requirements, was gathered and from it the result was synthesised. It came back as a nice, clean, solid model, which we could then bring through the rest of the design process.
The final structure is hollow with variable wall thicknesses. After minor detail adjustments, it was possible to address elements emitted from the generation, such as detailing of the interfaces. Following this, the solid geometry could be directly simulated using tools JPL trusts.
Autodesk used three manufacturing processes – casting for the external structure, as it is suitable for very large, complex structures, metal additive manufacturing for the internal structures to achieve fine detail and specific properties, and machining on the legs. Generative design software and proven manufacturing methods produced a structure significantly lighter than what is possible with human design.
For the internal structure, Autodesk achieved a 10:1 payload to structure ratio, compared with JPL’s typical 5:1 ratio. A 30% mass reduction for the external structure was also achieved.
Daniele Grandi is an Autodesk Senior Researcher.