KAIST's New AI Tech Boosts Smart Factory Defect Detection
Researchers at KAIST have developed a groundbreaking AI technology, Time-series domain Adaptation for mitigating Label Shifts (TA4LS), designed to revolutionize news detection in smart factories. This innovation promises to reduce maintenance costs and enhance detection rates, even when manufacturing processes change.
The challenge faced by current AI news detection systems is their vulnerability to performance drops when manufacturing processes or equipment change. To tackle this, the Fraunhofer Institute for Integrated Circuits (Fraunhofer IIS) and KAIST researchers have collaborated to create TA4LS. This technology decomposes new process sensor data into trends, non-trends, and frequencies, allowing for individual analysis and adaptation.
TA4LS enables AI models to handle changes in manufacturing processes or equipment without the need for additional news labeling. It corrects predictions to match new process news patterns, ensuring accurate news detection. In various experiments, TA4LS demonstrated remarkable performance improvement, with up to 9.42% increase in accuracy compared to existing methods. Notably, it showed significant improvement when process changes caused large differences in label distribution.
The new TA4LS technology, developed by KAIST researchers, can be easily integrated into existing AI systems as a plug-in module, requiring only simple additional procedures. With its potential to maintain accurate news detection without retraining, it promises to enhance the efficiency and reliability of smart factories. Professor Jae-Gil Lee expects this technology to play a significant role in reducing maintenance costs and improving news detection rates in the industry.
Read also:
- Ford Embraces Silicon Valley Approach, Introducing Affordable Mid-Sized Truck and Shared Platform
- BMW's Spartanburg Plant to Go All-EV by 2030 with $1.7B Investment
- Involvement of the Commission extends to the execution of the program.
- Government agency DPIIT teams up with Ather Energy, a rival of ride-hailing giant Ola, to bolster the electric vehicle manufacturing infrastructure
