Notre Dame, IN — Artificial intelligence systems that can process multiple types of input and reasoning can enhance worker safety in manufacturing and the services industry, results of a recent study suggest.
An interdisciplinary research team from the University of Notre Dame’s Keough School of Global Affairs collaborated with experts to gather welding images from several industries: recreational vehicle and marine, aeronautical, and farming. Two data sets of weld images were collected – one directly from welding experts and the other from images downloaded online.
The images were used to help a multimodal large language model assess whether the welds would work for different products.
The researchers determined that the AI tool showed promise in assessing weld quality but performed much better when analyzing curated online images than actual welds. They note a need to “incorporate real-world welding data when training these AI models, and to use more advanced knowledge distillation strategies when interacting with AI.” In addition, “context-specific prompts may enhance the performance of AI models in some cases.”
Using larger or more complex AI models, however, didn’t necessarily lead to better performance.
“As AI adoption in industrial contexts grows, practitioners will need to balance the trade-offs between using complex, expensive general-purpose models and opting for fine-tuned models that better meet industry needs,” study co-author Yong Suk Lee, a professor of technology, economy and global affairs at Notre Dame, said in a press release.
The study was published in the journal Information Fusion.
McCraren Compliance offers many opportunities in safety training to help circumvent accidents. Please take a moment to visit our calendar of classes to see what we can do to help your safety measures from training to consulting.
Original article published by Safety+Health an NSC publication