When the term ‘generative design’ came up in a press release about Mazda, Al Dean got excited – but it turns out the Japanese automaker is applying the technology to the design of its electrical harnesses
If you’ve been reading DEVELOP3D for a while, you’ll know I’ve got a serious interest in the subject of optimisation and generative design. If you’ve been through the process of education typical of a designer or engineer, you’ll have a natural curiosity about ways to solve challenges and overcome barriers.
Put the two together, and it’s easy to see how these new technology approaches can offer brand-new ways to boost our already-considerable talent for problem-solving.
So you can imagine how my interest was piqued when I received a press release from Siemens regarding Mazda’s use of generative design.
‘Finally’, I thought, ‘a large company venturing beyond an academic led research project or a vague promise of future deployment, and instead going public with its use of generative design on live projects’.
That would be something, eh?
According to the release: “automated generative design flow helps Mazda automotive design teams manage design complexity and changes across the entire vehicle platform, minimising errors and reducing costs.”
That all sounded brilliant, but my spirits were soon dampened when I read that Mazda is talking about maximising innovation in a very specific context – the design of next-generation automotive electrical systems.
Now, electrical design is a subject that I know so little about, it’s dangerous.
We’re talking here about the inability to rewire a plug without Googling the correct connections. I couldn’t even be 100 per cent sure whether live is brown or blue these days.
That said, the idea that you could use a generative approach to design the wire harness inside an automobile is actually pretty mind-blowing. And digging into the release, it seems that Mazda is looking to take this work further.
At Mazda, from system design, harness design and verification, down to manufacturing and service documentation, the data generated and output by each process or task is generated in its natural language, requiring designers to translate between processes and fill in the missing pieces using their talents and skills.
In other words, just as there is a translation process between mechanical, electronic and electrical design, so there is between the different disciplines of electrical design.
This introduces the potential for error, additional work and a lack of efficiency. This is the key challenge that Mazda is looking to overcome.
Kazuichi Fujisaka, technical leader at Mazda, explains it like this: “In order to remove ambiguity while maintaining the diversity of expressions that are characteristics of these natural languages, we applied formal methods to eliminate the loss in information transfer and set a goal to build a development environment that is consistent and connected all the way through the manufacturing phase.”
He continues: “Furthermore, we also aim to shift to a development methodology that allows us to optimise the vehicle as a whole, with all possible variations being considered in the early development stage.
“To make this happen, we needed a development environment to visualise the entire vehicle circuitry and standardise our language, tools, and processes without compromise, creating standard models across the company.”
This is what Mazda is looking to achieve, using the combination of a systems-based approach (which abstracts the data and tasks to a common language so everything can be considered in a lingua franca), and a generative design approach, to not only explore design options for its electrical systems, but also to automate much of the heavy lifting involved in taking those automatically generated concepts into production.
When you consider the challenges facing the automotive industry – customisation, electrification, autonomy (at whatever level) – it may be the case that this approach is going to become more commonplace.
There’s simply too much work to do in increasingly short timeframes. The only answer is to find ways to work smarter and more efficiently, from the very formative stages of design.
