AI technologies can help eliminate engineers’ biases and reduce complex design questions to basic principles. Generative design is quickly becoming a vital tool for the profession.
Words by Elle Hardy
This article was originally published in the November 2024 issue of create with the headline “Intelligent design”.
Hype around artificial intelligence (AI) remains sky-high, with some predicting it could add $200 billion per year to the Australian economy. But we still don’t fully understand how it will affect engineering. But as engineers have thrown themselves at the task, solutions have started to appear. Some are now being implemented, others are in testing and more are in design.
Forget large language models (LLMs) such as ChatGPT. AI-powered generative design is showing strong potential as the leading technology to help engineers design faster and more sustainably, and achieve better outcomes.
It’s an iterative process that allows engineers to collaborate with AI algorithms to create designs within a certain set of predefined constraints, it is changing the way engineers look at design. And while it’s best known for making significant advances in additive manufacturing – where weight, strength and minimising waste are the name of the game – experts believe that’s only the tip of the iceberg.
Long time coming
Andy Harris, Senior Research Manager and Head of Manufacturing Industry Futures at software firm Autodesk, said generative design is drastically accelerating and enriching traditional ways of design and manufacturing.
“It boosts innovation, speeds up the development cycle and promotes sustainable material use,” he said. “If we look at it through a manufacturing lens, it facilitates the design of complex and resource-efficient parts, often suggesting dozens or hundreds of ways to get to better outcomes, and it does so in less time than a designer could have explored a single concept.”
By automating options for generation of design, engineers have an increased capacity to focus on optimisation and customisation, as well as improving product performance – without added complexity or cost.
“While topology optimisation focuses on material distribution within a space for structural efficiency, [AI in] structural design takes a leap to explore a wider array of possibilities,” Harris said. “It generates multiple solutions that consider functionality, manufacturability and cost, fuelling a more versatile approach to design challenges.”
Sky’s the limit
Alongside traditional manufacturing methods, generative design can suggest larger and more complex objects than were previously thought possible. With it still in its relative infancy, engineers are continuing to explore new applications for generative design technology across a wide range of disciplines.
In additive manufacturing, for example, engineers can prototype products by quickly using test materials and perfecting designs before moving them into a live production environment. For injection moulding or casting, engineers can find a balance of mass and strength – a crucial step for selecting the metals and alloys for final production parts and structures.
Harris pointed to Toyota Japan, which used generative design to make a seat frame thinner to create more space in its vehicle cabins.
“The outcome was so positive it proved to be a roadmap for the company to consider other components to be optimised.”
Advancing on the back of machine learning research and training methods, “early adopters are testing the results in real-world applications, and that’s producing invaluable data from systems such as the industrial internet of things (IoT),” Harris said. “It’s an example of how automation prompts new disciplines and skills to create a new set of jobs rather than replacing manufacturing staff.”
Generative design is also helping industry manufacture more sustainable products. A study by McKinsey, which looked at the technology’s impact on automotive, aerospace and sporting goods, found that generative design reduced material mass by up to half.
Sporting retailer Decathlon has already employed generative design to halve the amount of plastic used to produce its performance swim fins.
Human vs machine
Engineering consultant Jacqueline Rohrmann said that a less heralded aspect of generative design is that it asks engineers to drill down into what they are trying to achieve, the limits they can operate within and which constraints need to be satisfied.
Design parameters are often not linear in their effect on each other, so engineers work by trial and error. That’s exactly what generative design algorithms automate, and they navigate design options at a faster speed than humans.
“The technology can rule out the bias of the engineer and reduce design questions to basic principles,” she said. “Simple optimisation problems – say, finding the biggest square to fit inside a circle – can easily be solved mathematically by an engineer.
However, in real-life applications, our engineering challenges are usually defined by many, often conflicting properties. Reducing one undesirable factor might increase another and eliminate a third.”
More questions
Of course, no technology is without its pitfalls, and Rohrmann believes that the biggest obstacle in the implementation of generative design isn’t related to regulation, compliance or manufacturing constraints. Instead, it lies in the fact that humans think and work differently to generative design processes.
“To use generative design, we need to describe our problem as a set of goals that we want the algorithm to optimise towards,” she said. “In order for the software to do so, each goal needs to be matched with a quantifiable indicator.
“Factors such as revenue, material usage or energy efficiency can easily be measured. But what about the more elusive factors that turn an engineering project into a masterpiece? How can we quantify appeal, comfort or usability in a way that is possible for an algorithm to score? With too much focus on design automation, could we rationalise away what makes human design special, amazing, jaw-dropping, thought-provoking, challenging, newsworthy – and human?”
Harris adds that technological advancement cannot come at the expense of precision, accuracy and trust, and generative design technology will be in a constant battle to maintain them. Nowhere is this conflict more prevalent than in eliminating hallucinations in LLMs that power so much AI technology, alongside ethical considerations.
“The industry needs to keep a close eye on ensuring the protection and privacy of information sitting within corporate firewalls, while still aggregating adequate databases to inform generative design and generative AI,” he said.
Engineers also need to solve for technological challenges when the complexity of design increases beyond the inputs to the model. Harris said that this means that the engineer and software must work together to solve problems.
“One of the reasons generative design is so impactful in the conceptual design stage,” he said, “is that you have maximum flexibility in your approach to the problem, and you are looking to find a set of solutions that will allow you to move more towards a detailed design.”
Diving in
For engineers wanting to dip their toes into the world of generative design, curiosity is paramount.
Rohrmann said that, beyond knowing how to handle the software, engineers need to be willing to learn about genetic algorithms and multi-objective optimisation, and be able to think both abstractly and structurally.
“Generative design systems can only create a design within the space that we define,” she said. “No computer has had an original thought yet – and if you ever worry about AI taking your job, go ask a large language model like ChatGPT to create and describe a floor plan of a simple one-family house to you. The response you get might sound professional at first glance, but if you were to draw the proposal, you’ll soon realise it’s a bunch of gibberish.”
As with any technology, Rohrmann said engineers should use AI in a way that serves them, their workplaces and communities – and just maybe makes the world a better place.
At this free Engineers Australia event in Melbourne, explore how AI-driven 3D printing is changing engineering.
Further reading
