nTopology Pulls Ahead in Generative Design with v3.0

The recent update includes variable mesh that adapts to pressure solved with GPUs.

Already with the intellectual lead in generative design technology for the variable geometry it uses to fill a shape, nTopology goes a long way to extending its lead with its recent major update, version 3.0.

Step on it. A pressure map of a foot is input to the sole of a lattice-filled shoe sole. nTopology’s lattices can vary in shape and, therefore, in stiffness. (Picture courtesy of nTopology.)

Step on it. A pressure map of a foot is input to the sole of a lattice-filled shoe sole. nTopology’s lattices can vary in shape and, therefore, in stiffness. (Picture courtesy of nTopology.)

The latest version of nTopology takes advantage of the graphics processing unit (GPU) acceleration—if your workstation has them—to greatly hasten lattice generation. The 10,000+ beam lattice in the shoe sole shown above was created in “real time” according to the company’s material.

We see a beam-based lattice in a shoe sole, a bicycle seat and a football helmet. Why would you need beams in your shoe soles, bike seats and helmets, we wonder? Isn’t foam, the usual material insulating material, enough?

Good question, nTopology would say. Glad you asked.

The “sit” bones of the pelvis. (Picture courtesy of SQlabs.)

The “sit” bones of the pelvis. (Picture courtesy of SQlabs.)

The sit bones create this pressure map on a bicycle seat, which will be used to create a varying lattice, with higher density lattice located where pressure is expected to be the greatest. (Picture courtesy of nTopology.)

Foam provides cushioning uniformly and evenly. nTopology’s variable length and diameter beams in the lattice varies stiffness according to the anticipated pressure on an area. As a result, there’s a tight mesh of beams in high pressure areas and sparse beams where there is little pressure. For example, a bicycle seat supports two pressure points caused by the lower points of the pelvis, commonly referred to as sit bones (not spread across the buttocks, as most people believe). Those two pressure points, shown in red in the figure, can be supported by a dense mesh in the saddle, with the result being a more ergonomic fit. The rest of the saddle can be filled with just enough material to keep the saddle full. The same principle applies to running shoes, which place a denser mesh in the heel of the shoe and the area under the ball of the foot, with the rest of the sole being light as a feather.

Varying the lattice may be more economical, too, since less material is used for a sparse mesh.

Football has come under scrutiny for its inability to protect players against concussion, which can lead to chronic traumatic encephalopathy (CTE), a progressive and fatal brain disease. nTopology and BASF Forward, a design and engineering service in the Detroit area, are working on a solution that includes a lattice layer for shock absorption.

GPUs and Real-Time Solution

GPUs are now wildly popular for applications that require massive amounts of computation, especially if the computation can be done in parallel processes, such as rendering and simulation. Ansys has used GPU power to create the first real-time finite element analysis (FEA) solution (Discovery), which instantly solves for stresses. The joy of seeing stresses change as parts of an assembly are in motion is indescribable for those accustomed to simulations being done overnight—or, even as was the case a few years ago—over lunch breaks. Shape optimization, which does a simulation each time it takes an optimization step (thousands of times during a single optimization), takes computing demands to a new level and, before the existence of GPUs, was practical only for cloud servers and high performance computing (HPC) networks.

Since an optimized solution satisfies structural parameters (such as maximum stresses not to exceed X), a generative design is what Bradley Rothenberg, CEO and founder of nTopology, calls an “unbreakable” solution. But arriving at unbreakable solutions was agonizingly slow on local computers—until GPUs could be pressed into play. With enough GPUs in local workstations, engineers can avoid having to send optimizations to the cloud—the method relied on by most generative design programs. Not only does this stop running up the bill (cloud credits can be quite expensive and a normal generative design can require many) but it also enables engineers to create designs more interactively with their computers, rather than having to stop a design, upload it to the cloud, wait for a result, and download the solution—all to be repeated multiple times.

Support for Supports

nTopology v3.0 considers supports in its optimization and will remove supports that are not needed. (Picture courtesy of nTopology.)

nTopology v3.0 considers supports in its optimization and will remove supports that are not needed. (Picture courtesy of nTopology.)

nTopology 3.0 adds support for 3D printing—literally. You can now use nTopology to optimize the number of support structures required in a 3D print. 3D printing software can be very liberal in adding support structures, guessing it is better to be safe than sorry (i.e., having a 3D print collapse of its own weight in the tank). But there are many advantages to having fewer supports: less time snapping off hundreds of supports, less time sanding the nubs that remain, less material used and wasted, and so on. nTopology will consider support structures in its optimization the same way material in the part itself is considered: if it is a zero-force element, it is removed.

nTopology’s command-line interface can call Python and MATLAB scripts, enabling automated on-the-fly optimizations of products, such as the heat exchanger. (Picture courtesy of nTopology.)

nTopology’s command-line interface can call Python and MATLAB scripts, enabling automated on-the-fly optimizations of products, such as the heat exchanger. (Picture courtesy of nTopology.)

Rather than run optimization manually, users can automate optimization with nTopology as the application uses a command-line interface. A command sequence can be captured and modified so that optimizations can be run in batches. And for those with programming skills, nTopology offers a programming language and environment that can call Python and MATLAB scripts for the ultimate, hands-off optimization that can run given a starting geometry and a few parameters—perfect for custom product configurations.

Unique to nTopology is the ability to use output from simulation programs as input for optimization. For example, flow field results from an Ansys CFD simulation can be loaded into nTopology, which will then get to work optimizing the flow by varying the volume by trial and error (as optimization algorithms do) until the pressure gradients are leveled. Along with fluid flow (CFD), additional simulations and parameters in nTopology are modal analysis (natural frequency), linear static (stress, deflection) and even nonlinear simulations (buckling, transient, and nonlinear thermal simulations).

Unbreakable

A design engineer in today’s world creating a new product will consider many designs before going with one of them into production. Designs that are considered may be subjected to simulation and testing. Those that break will go no further.

nTop, as Rothenberg refers to his product, gives you only unbreakable designs, citing the underlying principle of generative designs to solve for stress, flow, etc., and adjust geometry so that failure criteria are met.  Presenting engineers with only designs that will work should give them a head start, right?

nTopology Days

Bradley Rothenberg is the boy genius who you want to dismiss because he is on a higher plane, but you can’t because he is solving the problems you struggle to articulate.

This was apparent as Rothenberg addressed his company’s virtual conference (nTop Week), which recently concluded. nTopology is arguably the most sophisticated and capable generative design technology that has not been swallowed by a design software giant. Rothenberg attempted to explain the concept of generative design with fields.

“Everything is a field. Stress. Flow,” he said

Rothenberg favors “engineering design” to “design engineering,” demanding that computers and software assist with design in a significantly helpful way—not just simply document a design that has already taken shape in a design engineer’s head.

Rothenberg sees us attempting to create objects and shapes using 1980’s technology—and failing miserably. B-rep, or boundary representation, the basis for our most popular geometry kernels, cannot represent but a few of the shapes we need to model. Modeling organic shapes with any CAD program is maddening. Modeling detailed minutiae, such as lattices or cell structures, are either overwhelming or impossible. Point clouds and meshes are the same.

Rothenberg has his bachelor’s degree in architecture (Pratt Institute), but here he is rearchitecting shapes for engineers … and trying to explain it with fields—which, unlike mathematicians and physicists—engineers understand only in their discretization, like the finite elements that break a continuum into easily solvable problems. We are reminded of another individual who tried to find a unified field that encompassed gravity and electromagnetism. But whereas Einstein was unsuccessful, we give Rothenberg, who is taking on the smaller universe of generative design, a fighting chance.