Researchers from Carnegie Mellon University's Department of Mechanical Engineering have developed a system that uses multiple large language models (LLMs) to monitor and correct the performance of 3D printers in real time.
Researchers from Carnegie Mellon University's Department of Mechanical Engineering have developed a system that uses multiple large language models (LLMs) to monitor and correct the performance of 3D printers in real time.
The researchers used four specialized AI agents based on large language models, led by a controller agent to provide optimization, writes Tom's Hardware.
The first agent is a visual language model that takes snapshots after each printed layer and analyzes them for quality and defects. The second agent checks the current printer settings to determine what needs to be changed or improved to fix the problems it finds.
The information received is passed to the planning agent, which develops a specific action plan. This plan is then sent to the execution agent, which directly interacts with the 3D printer via an API to achieve the desired result.
The work of all four agents is coordinated by a controller, who ensures that all information is up-to-date and relevant.
A key feature of this system is that it does not require specialized AI models trained on narrowly focused data to work effectively. Instead, the team used a basic version of ChatGPT-4o and their own structured prompts tailored to a specific industry. This makes the system easy to implement and allows for easy improvements in 3D printing quality and efficiency.
“This work now provides the foundation for truly intelligent and autonomous manufacturing systems capable of achieving unprecedented levels of precision and reliability,” said Amir Barati Farimani, associate professor of mechanical engineering.
