Rutgers-led AI research boosts Additive Manufacturing in extreme environments
Two studies, led by Rajiv Malhotra, Associate Professor, Department of Mechanical and Aerospace Engineering, researchers from Rutgers School of Engineering, Piscataway, New Jersey, USA, have illustrated how artificial intelligence can support the broader use and faster evolution of Additive Manufacturing technology.
Expeditionary Additive Manufacturing
‘Scalable control of extraneously induced defects in in-field Additive Manufacturing’, published in The Journal of Manufacturing Processes, focused on making parts in spaces outside stable, controlled factory conditions such as battlefields or in disaster zones. These environments are unpredictable and impose disturbances such as vibrations or temperature shifts that can ruin a build job. Another issue may arise from the fact that the operators in these settings may lack specialised Additive Manufacturing experience.
To address these issues, Malhotra and collaborators developed a new AI technique called conditional reinforcement learning. The method uses a camera to monitor the build process and instantly adjusts the Additive Manufacturing machine’s settings when it detects a defect. The method also does not need retraining to adjust to new conditions as it does not need to understand what went wrong beforehand when fixing issues.
Article: Inside Nikon’s metal AM strategy
Part 2: Scaling industrial production in Long Beach
| Read now |
“We don’t have to anticipate anymore,” Malhotra stated. “Whatever disturbances come, we can deal with it without throwing away the part or stopping failure, both of which are bad for mission assurance.”
He explained, “We trained the AI to expect the unexpected, rather than expect the expected. We have created a new AI technique that ‘robustifies’ expeditionary manufacturing beyond the reach of literature. It reduces defects by 10x or more, increasing quality by similar amounts even when the disturbances are not known in advance.”
On this paper, Malhotra worked with Jeremy Cleeman, a doctoral student, and Adrian Jackson, an undergraduate from the Rutgers School of Engineering. He also partnered with Shane Esola from the US Army Armaments Graduate School, Chenhui Shao from the University of Michigan, and Hongyi Xu from the University of Connecticut.
Expediting development
In ‘Large language models for extrapolative modeling of manufacturing processes’, published in The Journal of Intelligent Manufacturing, the Rutgers research team created an AI system which leverages Large Language Models (LLMs) and refines its knowledge with real-world data.
The AI system is capable of reading scientific papers for useful information and, combining it with experimental data, can build predictive models. Rather than running hundreds of experiments, the team reportedly achieved accurate results within thirty samples. The team anticipates the AI system could accelerate development in a range of industries including aerospace, automotive, electronics and defence.
“The AI acts like that PhD expert. It tries a few times, and then it gets it right,” Malhotra said. “We cut short the samples that you have to make. That means you’re doing things much faster.”
“This method reduces the need for human interpretation and large experiments, speeding up innovation for new or complex manufacturing processes,” he concluded.
This study saw Malhotra collaborate with Kiarash Naghavi Khanghah and Hongyi Xu from the University of Connecticut, and Anand Kumar Patel, a doctoral student from the Rutgers School of Engineering.
