Generative Design: I Don’t Think It Means What You Think It Means

Here are some tips to help you shop around for your generative design solution.

It’s inconceivable how often I hear the term generative design pop up in conversations with product design experts and engineering software marketers. Based on these conversations, I’ve come to one conclusion: Though many keep using the term “generative design,” I don’t think it means what some others think it means.

Let me explain.

If a generative design tool lacks physics or relies on templates, it might be more art than engineering. (Image courtesy of Bigstock.)

If a generative design tool lacks physics or relies on templates, it might be more art than engineering. (Image courtesy of Bigstock.)

When most think of generative design, they imagine beautiful lattice or organic bone-like structures, produced using procedural computer algorithms and 3D printing. But regardless of the promise behind these parts, and the robot overlords that designed them, many have failed to make it into the mainstream. The good news for the design engineers reading this: your career appears safe from this AI takeover, for now.

So, what happened? Well, I think fellow engineering.com colleague Roopinder Tara said it best when he wrote, “let’s call it art” while discussing a generated design of a prototype wheel displayed by Volkswagen in 2019’s Autodesk University. Basically, if the geometry produced from generative design isn’t informed by physics-based simulations, mass manufacturing criteria and design tradeoffs, then it’s likely producing art instead of a well-engineered product.

This begs some questions. What should engineers be looking for when it comes to generative design technology, to ensure they get engineering over art? And if engineers take the plunge, is there a risk of the software replacing them in the workplace? Let’s dig into this.

Generative Design Tools as an Engineering Co-pilot

According to ABI Research, the leaders in the generative design space are Altair, PTC and Autodesk. What sets the leaders apart from the rest? One key difference is having a larger perspective when defining and developing generative design tools.

The leaders in generative design software, according to ABI Research. (Image courtesy of ABI Research.)

The leaders in generative design software, according to ABI Research. (Image courtesy of ABI Research.)

For example, Ujwal Patnaik, Global Business Development and Strategy manager at Altair, says “Generative design is a set of technologies that provide computational intelligence to augment the design process, helping users engage the power of simulation, AI and HPC to develop validated and manufacturing-ready designs.”

He goes on to explain that generative design should incorporate aspects of optimization, design exploration and artificial intelligence. The idea is that optimization breeds simulation-driven generated designs that are pre-validated based on the parameters set by the engineer. This means the design isn’t only going to survive the stresses it will experience in the field, but it will also be manufacturable based on a given process and material.

As for the design exploration aspect, Patnaik explains it needs to be guided by multiphysics simulations that evaluate and predict the performance of the parts. It isn’t enough that the parts meet the bare-minimum criteria to be deemed pre-validated. Those numerous pre-validated designs need to be compared based on their performance and ability to be mass-produced.

On that point, Uwe Schramm, chief vision officer at Altair says, “A lot of generative design is driven by 3D printing. It’s not right, because 3D printing is not the dominating manufacturing technology. It’s fringe technology and is always going to be because others can produce in bigger numbers much cheaper.”

As for the AI aspect of these tools, they can be used to train predictive models that will quickly generate physics-based data, interpret geometry and automate engineering judgements so that part generation takes minutes instead of hours or days.

Does this AI risk replacing the engineer in the design loop entirely? Since it’s the engineer plugging in the criteria the part is based on, and they also have final say deciding which design is ultimately produced, that seems unlikely.

“A co-pilot is a beautiful description of this technology,” says Schramm. “We’re not suggesting a single design drops into a bucket at the end of the process. Generative design is a tool for designers and decision makers. And it’s an exploratory tool. You’re not just getting one design and you live with the results.”

What Engineers Should Look for When Selecting Generative Design Tools

How can you tell if a generative design tool is going to give you art or engineering? It’s a good question, and the best way to answer it is to do a little digging into the tools available to your organization.

Ask developers questions such as:

  • Have parts designed using this tool been used in the field?
  • Have any of those parts been mass produced?
  • Are the geometry and/or lattice structures defined by loads, templates or both?

For these questions, developers should be able to tell you upfront answers. But for the final question, how loads and lattices are defined, there is a way to make an educated guess. Produce a part with a lattice structure using the tool. Schramm hints that if the lattice appears uniform, there is a chance that it was procedurally generated using templated structures. These templated structures might be optimized for 3D printing tools, but they may not produce parts that are structurally optimized or easily mass produced.

In other words, if each beam is informed by physics, then it’s unlikely they will all be evenly sized and distributed—their size, shape and placement will be based on where the loads are experienced the most.

Instead, look for software that requests data on the loads the part will experience while it’s in the field. If this happens, then there is a better chance that multiphysics simulations (like CFD, stress, fatigue, electromagnetics, thermal analysis and more) are in the optimization loops that generate the geometry.

Also see if the tool can incorporate the part’s manufacturing process when developing the geometry. Chances are you’re not using 3D printing to mass produce the part. So, ensure the tool can optimize the geometry based on the manufacturing methods that will be used, such as extrusion, casting, milling, molding and more. Any geometry the tool produces that could cause downstream manufacturing issues should be automatically eliminated from contention by the software before an engineer even sees it.

“It needs to be physics informed,” says Schramm. “We use physics simulation and optimization to make design decisions and to pick or develop designs. Manufacturing feasibility is also included in the generated design process.”

Finally, look for tools that support the creation of a digital thread, are intuitive, offer familiar workflows and interoperate with legacy cloud CAD, CAE and PLM systems. These final caveats will ensure the tool integrates easier into the organization.

Written by

Shawn Wasserman

For over 10 years, Shawn Wasserman has informed, inspired and engaged the engineering community through online content. As a senior writer at WTWH media, he produces branded content to help engineers streamline their operations via new tools, technologies and software. While a senior editor at Engineering.com, Shawn wrote stories about CAE, simulation, PLM, CAD, IoT, AI and more. During his time as the blog manager at Ansys, Shawn produced content featuring stories, tips, tricks and interesting use cases for CAE technologies. Shawn holds a master’s degree in Bioengineering from the University of Guelph and an undergraduate degree in Chemical Engineering from the University of Waterloo.