During the course of our transmission development, we’ve utilized generative design tools extensively. A variety of modern CAD softwares have generative capabilities, and are extremely efficient at designing high-quality parts for additive manufacturing.

That being said, there are many designs that cannot be generated utilizing existing tools. Many of the novel and intricate mechanisms that power our CVT designs are far too complex to generate with current software.

Trial and error was not a strategy that inspired confidence, so we turned to the alternative solution — develop our own internal generative software.

Once we started exploring this, we realized we’re developing tools that have value not just internally but to engineers worldwide.

The first step in this vision is BracketCAD, Transcend’s generative tool made specifically for static structural generative design for uniform material objects.  

In generative design for a bracket, you have an indefinite number of fixing points, and an indefinite number of loading points. This means that BracketCAD will work for any type of bracket, not just one with 3 mounting points and 2 loading points, for instance. It’s a simple generative tool, but very valuable in practice. 

BracketCAD will be released in beta Q2 2025.

But that’s just the first step. Alongside BracketCAD, Transcend is developing an AI tool to check and fix 3D objects to be manufacturable. It utilizes internal research in novel methods of using neural networks to understand and manipulate 3D space. The end goal is a software as simple as inputting a file, finding out if it’s manufacturable, and if not, outputting a fixed version that is manufacturable via CNC or other forms of machining. This is incredibly beneficial, since this can help create a pipeline from generative design to CNC manufacturing.

We are also exploring a CAD software for flexure mechanisms. With brackets, your metrics for optimization are well-defined. You are trying to minimize compliance, minimize mass, and make it manufacturable. You don’t want it to move when loaded, and want it to still withstand the force you are applying. Flexures operate differently. You need a tool that has specific loading conditions, specific flexure requirements, degrees of freedom limits, amongst several other factors. This is clearly a much harder task, but albeit a possible one. We will take what we learn from the AI research in our manufacturability fix tool to help solve this problem in the future.

Ultimately, Transcend believes in the democratization of generative design technology. 

Current tools are not the most accessible, nor affordable. They also barely touch the potential of what’s possible with generative. With technologically complex, but easy to use generative AI design tools, the engineering world will be in a better place.

One potential option that aligns with that vision would be utilizing structured output of open source LLM models like Ollama. This was released only this past Friday, but this has the potential to bridge the gap between natural language input and special geometry output. There’s a future where you can describe what you want to design, answer a few questions on specification, and voila, the generative tool will design it for you.

We’re extremely excited about the future of generative design technology. If you are curious about Transcend’s current generative software (BracketCAD or the others mentioned above) or just interested to speak with us, please reach out at [email protected] or via the Contact Us form above.