NAMIC’s Global Additive Manufacturing Summit Tackles AI & AM

On day two of the National Additive Manufacturing Innovation Cluster (NAMIC)’s Global Additive Manufacturing Summit, the focus is on AI, design, design for additive manufacturing (DfAM), and more. Many are excitedly exploring how to use software, code, and machine learning more generally. The combination of computing power, software, and code is expected to significantly accelerate manufacturing. CNC and other processes remain remarkably analog. Factories often consist of rows of disconnected machines. While traceability is common in many industries, it is still generally rare. As computing power increases and computers become more accessible, the integration of manufacturing and computing is expected to continue. More advanced sensors, process monitoring, the elusive digital twin, and especially AI are powerful technologies on their own. The combination of these technologies could revolutionize manufacturing.

I had the opportunity to open the conference, speaking briefly about the AI hype versus reality. I drew a comparison to another technology with many potential applications: the laser. Would we have been able to predict LASIK surgery or the Compact Disc? We likely would have overestimated the impact of laser cutting but missed opportunities in lithography and other processes driving the microchip revolution. After my segment, Mahendran Reddy of the Fehrmann Tech Group spoke. Fehrmann, which produces aluminum, has developed an AlMigty variant of aluminum. This alloy has been optimized for a decarbonized world, suitable for both 3D-printed and traditionally manufactured lightweight parts. The company also introduced matGPT, a large language model-powered product that processes tens of thousands of engineering documents and research papers, making their data accessible. The hope is that this tool will accelerate alloy development by simplifying the search for the correct data, properties, and process information.

Next up was Daghan Cam of Aibuild, which received $8.5 million in Series A that saw participation from Nikon. Daghan showed a graph comparing the disappointing share prices of 3D printing firms with the soaring ones of tech companies. He believes the reason for this disappointment is that we underestimated how difficult it is to make the right parts correctly. He argued that solving this challenge requires superhuman intelligence. He also introduced a natural language interface for their software, allowing users to type normal sentences that are then interpreted as commands. For example, you could ask the software to find or slice a file. Daghan’s goal is for the software to function like a conversation with an application engineer.

Following his presentation, Michael Robinson, CTO of Hyperganic, took us on a tour of the geometry of a 3D-printed part as represented by CAD. He demonstrated how parts are not currently represented with full fidelity and discussed an AI-driven future.

Omar Fergani, CEO of 1000 Kelvin, then spoke about how his company is using physics-based AI for faster simulations. The startup’s AMAIZE platform helps quickly resolve manufacturing issues in LPBF for complex components. Omar gave an example of a part that would normally take 40,000 years to simulate with finite element analysis (FEA), but his software could solve it in just two hours. The differences between traditional FEA and computational fluid dynamics tools, and the varying results they produce, are areas of growing interest. Is physics AI the future of FEA, or will traditional methods prevail? Omar is focusing on users who need to analyze parts 100 times faster today and envisions a future where 1000 Kelvin will augment or replace build prep and process simulation.

Then we had a panel on AI in Additive Manufacturing. I was pleased with how measured and specific everyone was about where AI could be effectively applied. Overall, the conclusion seemed to be that AI is something companies will need to explore. At the same time, the specific areas where AI and Additive Manufacturing intersect now appear to be in physics-based simulation, large language models for interfaces or research, interpreting time-series data, tracking and identifying defects, and optimizing settings. I came away from the discussion more interested in AI. Yes, there is a lot of hype, but there is also a great deal of substance. While I know very little about AI, I now realize that I need to learn more.