The latest generation of generative design applications can be found integrated with your design software.

“Everything should be made as simple as possible… but no simpler.” —Albert Einstein
There are a lot of reasons to make parts lighter—and few reasons not to. Lighter parts save energy. Anything that flies or floats in air or space requires energy to lift or to keep it off the Earth’s surface. With powered flight, lighter parts save fuel costs. Each pound of weight saved on a commercial airplane will saves 14,000 gallons of aviation fuel a year. On the ground, accelerating the mass of a vehicle takes a proportional amount of force (F=ma) and fuel is also burned forcing the vehicle through air. Governments regulate average fleet gas mileage (such as the NHTSA’s CAFE standards in the U.S.).
Lightweight: It’s the Law
While regulations are going forward to decrease fuel consumption, consumers are going in reverse as the average consumer-driven vehicle gets larger, and therefore heavier. The advent of the minivan and the SUV has increased the average weight of personally driven vehicles by 25% in the last 45 years.
That leaves automakers still held to increasing fuel efficiency mandates, scrutinizing each part of their vehicles and trying to make them lighter, almost as frantically as aircraft manufacturers.

For automakers worldwide, the wakeup call came in 1973 when members of OPEC imposed an oil embargo. Everything changed; gasoline became expensive and small, foreign cars became popular. U.S. automakers, once enamored of huge V8 engines, chrome-plated steel and stylish detail with no functional purpose (remember tailfins?), now went small. But it wasn’t long before American affluence recovered and started thinking big again. They took to minivans, SUVs and most recently, big pickup trucks. It was up to the government to remind automakers that the sources of fuel were finite, and the burning of hydrocarbons was ruining our health and changing the climate.

The reasons to choose not to make parts lighter may be fewer, but they are nevertheless substantial and reasonable. A part that is too light will fail, so engineers will overdesign a part for fear of failure. Margins of safety of 10 and above on auto parts were not uncommon. Why skate on thin ice, like aerospace engineers who had to account for every ounce, had a space for “weight” to be filled in on the title block of their drawings, and who made their parts with margins of safety dangerously close to one? Fear and safety were the primary reasons for parts to be as thick and heavy as they were.
The reasons to make auto parts lighter was that a smaller part would use less material, but in the heyday of the American automobile (before the oil embargo), that was a secondary consideration.
But here we are—consumer demand meets government dictate. The consumer vehicle must lose weight, with the government like a doctor demanding it of an overweight patient.
Lightweight: It’s Also a Good Idea
A leaner part will also demand less of the environment in terms of resources used. The United Nations is increasingly concerned with our decreasing supply of fresh clean water available to a rapidly growing population. With over 7 billion thirsty inhabitants and industrial and agricultural pollution running rampant, governments all over are uniting to preserve clean water resources.
The United Nations has created Sustainable Development Goals, one of which is to ensure availability and sustainability management of water and sanitation for everyone. Dassault Systèmes has added its support for the UN’s initiative with its Progress is Human project and 3DEXPERIENCE will be showing the water that is used, changing it in real time to reflect changes in weight, material and manufacturing process.
What is Lightweighting?
Lightweighting is a word coined by its practitioners. While not recognized formally as an engineering discipline, it nevertheless has its specialists who work on lightweighting full time in big manufacturing companies, as well as many more non-specialists, product engineers, designers and analysts who perform lightweighting part time. Historically, lightweighting has been initiated as a revision to an already designed part or assembly—a refinement of the original design. Engineers will take a working design and try to reduce the weight of the part, either by switching to a less dense material or by changing the geometry.
For the purpose this article, we will detail an emerging technology that promises to do more than try to reduce weight.
Enter Generative Design
Generative design is design generated by algorithm rather than human intuition. It is a generic term and not specific to any particular algorithm. It may be considered computer aided design in the truest sense, as the computer system is actually doing the designing. Proponents of generative design will say computer aided design (CAD) was only making your pencil sketches and drawings neater and more precise—and relatively recently, adding a third dimension to your part model. But to the engineer, the part was already designed in their head in 3D and living color. CAD software was only documenting the design. That’s really not very creative, is it?

Let’s say you are in a race to create a part, you in one lane with your CAD software, and a generative design program in the other lane. You both have the same design parameters, including a volume envelope not to exceed, interfaces you must meet and loads you must withstand. You, an experienced design engineer, will come up with a good design and a 3D model of it on your trusty workstation. In the same amount of time, the generative design program working with near-supercomputer resources in the cloud, can come up with thousands of creative variations of a design, all satisfying the same design parameters.
Generative design makes use of what may be the first time AI comes to part design: topology optimization, also referred to as shape optimization. Topology optimization is a specific type of generative design, one that is conceptually easy to understand yet still genius, and which nature has been using forever.
Borrowing from nature, topology optimization will “grow” a part along paths where stresses will develop and where material is necessary. And like nature, it will remove material where no stress develops. The resulting shape will theoretically be as light as possible, but no lighter. A car generated this way would have no tailfins, for example.

Most topology optimizations are based on “bone growth” algorithms. A broken bone will send “shoots” across the break. Eventually enough shoots make it solid and the fracture disappears. However, overshooting the gap will make a bump, but eventually—almost magically—the bump disappears, too.
Engineers, not believing in magic, will explain this naturally occurring healing process with principal stresses. Where there is little principal stress, nature removes the material. Where there is too much stress, such as when there are too few shoots across the gap, nature adds material.
In application, a part that is given a gross shape, often a rectangular volume, will be composed of thousands of elements. Restraints and forces are applied in the same way as a finite element analysis. The topology optimization application will do a stress analysis to determine which elements have no stress and eliminate them. Similarly, the part shape can be more refined, as would be the case of an existing design. The part may be smoothed over and elements made smaller in subsequent passes, with each refinement triggering another stress analysis. This would result in another shedding of nonstressed elements and shoring up by adding elements where additional stresses may have travelled to with the previous shape change. Finally, after multiple iterations, the material left is all the material that is needed—nothing more, nothing less.
The lightweighting problem is solved, but one question remains.
What About Manufacturing?
Topology optimization, hatched from academic research decades ago, was spectacular for creating light parts—but it was forced to stay in the labs because of the impracticality of manufacturing the stringy, organic shapes that resulted from its use. A theoretical solution of a particular load case can, from an engineer’s point of view, result in some funky-looking parts. The emergence of 3D printing allowed these funky parts to be produced. “Complexity is free,” said the 3D printing community, welcoming generative design. But the slow adoption of 3D printing by industry was a deterrent to the adoption of generative design.
However, modern generative design software, by automatically applying constraints (straight, smooth, round and reach of cutting tools, for example) can come up with a design that can be cast, forged and machined, whether milled or turned.
The most advanced generative design software will provide toolkits to help detail the part, taking out bumps here, thickening a too skinny section there or making a not-quite-round gap into a circular through hole where needed, all to get the part to where it can be produced by traditional manufacturing techniques and with material engineers are confident in.
For 3D printing, process-specific simulation can predict the distortion occurring during the printing through cooling of metal parts, even morphing the part’s 3D solid model to compensate for the expected distortion.
Next Generation Generative Design
The evolution of generative design from an academic exercise into a useful and practical design function is what CATIA users know as Cognitive Augmented Design.
“Science (cognition) augments the ability of the human being to create innovative design,” says Daniel Pyzak, worldwide mechanical industry process consultant and senior director at Dassault Systèmes, who sees generative design as the evolution of computer aided design (CAD) to finally live up to its name.
“The creator simply provides the design space and specifications, and new concepts are generated,” says Pyzak. Use of a common design platform, one that shares a user interface and data at the onset of design, is critical.
A design engineer is not trained or at ease with complicated, multi-variate optimization software. Therefore, it is important for modern generative design software to do quite a bit of its business under the hood, sight unseen with its complexity hidden. The ability of the generative design software to gather its defining volume from an existing part you need to lightweight, recognize mating parts as boundary conditions or restraints from the context of an assembly of solid model parts, and possibly loads, possible when both design and optimization software are integrated in a common platform, not only saves a lot of time, it also makes generative design much more useful.
The alternative is manually defining the volume in a foreign generative design program with an interface that you have not mastered, which requires looking up unfamiliar terminology, and in which you are unable to import loads from FEA or CFD programs and have to resort to manual entry—it’s enough to make you give up.

Better Products Faster
Performance-driven generative design, as Dassault Systèmes calls the next generation of generative design, allows designers to explore the design space, validating each design variation with simulation. This allows the potential for the engineer to be able to make informed decisions on weight reduction with no sacrifice in part strength, flow or other outcomes. Since generative design operates freely without bias, traditional restraints or habit, a generated design may be an original one never before imagined, with significant, rather than incremental, benefits.
With design, optimization and manufacturing under one roof, so to speak, as it is with an integrated platform, data is not lost in translation and the overall process can be executed in less time. Or, if granted the same amount of time, more concepts can be evaluated and more design iterations done.
Design, in the truest sense of the word, is a function finally being fulfilled by a computer, thanks to generative design.
For more on Performance Driven Generative Design on the Cloud, visit Dassault Systèmes.
Engineers, Designers and Managers:
Is lightweighting important to you? Do you use generative design to make your parts lighter?
Tell us why – or why not – by taking a short survey.