Automating Design

Automation in technology is old news, and automation in design is limited to a few areas. In an earlier post, we talked about the automation you can achieve with Rules Based Engineering. This is impressive stuff, and if you talk to the management of companies that make extensive use of it, it can help you to huge profitability. But who designs the automated systems?  

Beyond RBE, which has some limitations from the point of view of what kinds of products can be automated, and then even limitations within the scope of the automation itself, there is another type of design automation that is more general, but also more one-off. We have a lot of people who are skeptical of this type of automation, but I believe that for a certain type of design, it is going to continue to grow in importance, and may even revive a certain style of design. 

Of course I’m talking about generative design. In concept, generative design is where you establish inputs and requirements, and the software fills in the blanks. For example, the bicycle frame has been around for a long time, and in the last 50 years or so designers and enthusiasts have tried to design a more efficient, stiffer, lighter frame. These designs have in a prescient way been trending toward the more and more organic. You may have noticed that when nature designs a structure, say your bones, the lines it produces are not straight lines and right angles. Natural structures are by definition “organic”, which has come to mean something special in design circles – curvy, in a word. Look at shellfish shells. Your skeleton. The shape of dolphins and birds. Nature doesn’t make shapes like these because that’s the style it learned in design school, or because they think it’s cool. Nature designs for efficiency. That’s the same goal that most of us engineers want to design for. The problem is that nature is far less limited in its manufacturing processes, where humans are limited in some cases to processes that produce mainly straight lines for mass production or right angles to accommodate machinery. 

Casting was an early process that was capable of producing organic shapes. Plastics molding really brought this to the mass markets. 3D printing will continue to democratize shape production. Except that these are just methods that make the shape possible – make it something you can finally bring into reality. We still need systems that can produce the shapes electronically.  

One method to do this is generative design. If you think this is going to be used for aesthetic designs first, I think you’d be wrong. The biggest benefit will come from structural shapes. Brackets, frames, supports. Especially stuff that has to be mass manufactured, where optimized material use matters, where symmetry and intuition don’t serve us. 

This is not something that has to wait for the development of AI, it just requires us to work backwards from systems we already have. Generative design is, in concept if not in practice, working backwards through stress analysis. Usually we start with geometry, and we apply loads, and ask the software to solve for the stresses, or displacement, or thermal conditions, etc. But with generative design, we specify where we want certain loads, and then assume that the stress is equally distributed across the structure, and ask the computer to find the structure. Instead of stress we might optimize fluid flow, or electrical properties, or light focusing or bending or whatever, The structure it finds is usually a little imperfect and lumpy – organic – like bones. Using this information of the optimized shape, we can go back through, and use our perfect design tools to clean these things up and make space age looking things, rather than things that look like they have evolved over millennia. 

Where does this get used first? Well, sales demos, to be perfectly crass about it, but then also in racing. Bicycle frames have been going through optimization for a long time. F1 racing. We will see buildings that use this combination of design and manufacturing methods. Think of Gaudi. Even before computers, he used organic shapes to solve computational structural problems in design.