Generative design is a very broad design methodology. It includes all goal-driven and computational approaches to engineering where software is used to generate geometry based on a set of logical operations and user-defined rules. It offers so much more than just topology optimization.
In other words, generative design is a set of digital tools that augment your capabilities as an engineer or designer. Ideally, these tools should address the main engineering need: to efficiently and effectively explore the available design space during all phases of the product development process. By following a generative process you can quickly determine the best possible solution that meets all design requirements — both technical and non-technical.
Even though the basic concepts behind generative design have been around for a few decades, most “generative” tools available today fall short of addressing this basic need — or, more precisely, they only address a small aspect of this need.
In this post, we will describe why the ideas behind generative design are so appealing. In the next part we will present nTop’s alternative approach to generative design.
The Promise of Generative Design
The intrinsic value of generative design is that it flips the traditional model of first creating and then evaluating a geometry. Instead of asking, “does this shape meet my requirements?” an engineer following a generative approach asks, “what is the shape that best meets my requirements?”
Generative design is a tool that unlocks rapid engineering innovation. If used correctly, it can help your business unlock new opportunities for rapid engineering innovation. More specifically, generative design can:
- Accelerate your product development cycles by automating manual engineering design and verification tasks
- Lead to the development of better products as it produces unbiased solutions that may not immediately be obvious to a human designer
- Enable you to take full advantage of advanced manufacturing technologies, such as additive manufacturing or any other robotic or digital manufacturing process
Going a step further, generative design aligns well with the “digital thread” and Industry 4.0 initiatives. An effective generative engineering process promotes collaboration between different disciplines: mechanical design, manufacturing, and computer science. It also enables curious engineers to apply, document, and share with their team new and radical approaches to engineering that take full advantage of the now accessible computational power.
The promise of generative design is definitely alluring. But how do the currently prevalent approaches to generative design help you achieve this promise?
The Current “Black Box” Paradigm
If you are not up to speed with the latest developments in the world of engineering design, it may surprise you that most “generative” tools today are based on the same methodology: topology optimization; a simulation-driven approach first developed back in the ’80s.
Topology optimization is a powerful tool that can generate new concepts during early stages of the design process. It has been used with great success to identify shapes that satisfy engineering requirements — mostly, structural or thermomechanical — and manufacturing constraints. However, it does have certain limitations that hinder its wider use in real-world applications.
Even if you manage to nail down all technical aspects of topology optimization — such as accurately simulating the underlying physics, expressing the design requirements that matter in a measurable way, and streamlining the reconstruction of the resulting geometry into a usable for manufacturing format — there is still an underlying issue that is often overlooked: topology optimization is very sensitive to the inputs that you give it. Even a small change in a parameter or input can make a huge difference to the produced “optimum” result. You can see this in action in a recent nTop Live by our application engineers or in this presentation by one of the pioneers of topology optimization.
To improve the usability of topology optimization, many engineering software vendors went down the route of enhancing it with a layer of machine learning and AI. On a high level, this approach makes sense. After all, the key strength of AI is that it lowers the cost of prediction. Let the algorithm predict the values of the topology optimization parameters that will probably produce good results so the user doesn’t have to. The system is typically combined with a “design of experiments” interface that gives multiple feasible solutions and the illusion of control to the user. And since this requires multiple computational heavy simulations, cloud computing is utilized as desktop computers cannot run in a reasonable amount of time. Plus, a pay-as-you-go monetization model adds additional (and hefty) costs to your design process.
The question arises: is this “black box” approach to generative design actually effective? In our experience, engineering design is characterized by multiple iterations and frequent change. It is much more complex than what can be captured by this simplistic three-step approach: 1) specifying objectives and constraints, 2) wait for the algorithm to generate solutions, and 3) select the best option.
Plus, the burden of evaluating a design still falls mainly on the engineer and design validation is a complex and cumbersome process — remember: the results of heuristic approaches, like topology optimization, always require detailed validation. As one of the most popular online articles that critiques prevalent approaches to generative design puts it, engineers do not work this way.
What is missing from the current paradigm of generative design tools is an element of predictability and repeatability; an element of control over both the produced geometry and the whole optimization workflow. The best tools are the ones that you can modify to the specific needs of your application.
The real challenge isn’t the technology, it’s how the algorithms fit in and improve the engineering process. This philosophy is at the core of nTopology’s approach to generative design.
Note: This article was first published Here.
ABOUT THE AUTHOR
Alkaios Bournias-Varotsis, PhD, is the Product Marketing Manager with nTopology. He is an engineer and a digital marketer with deep expertise in additive manufacturing, Industry 4.0 and advanced engineering products.