Beltus Nkwawir Wiysobunri, a PhD student in our group, has published an article titled “CNN-based server state monitoring and fault diagnosis using infrared thermal images” in the journal Soft Computing.
The study, co-authored with his advisors Assoc. Prof. Dr. Hamza Salih Erden and Prof. Dr. Behçcet Uğur Töreyin, presents a deep learning-based method for diagnosing server operation states in high-performance computing data centers. Using infrared thermal images, the proposed method can automatically detect issues such as fan failures, CPU load levels, and server entrance blockages with an impressive accuracy of 99.60%.
This work is based on Beltus’s master’s research and demonstrates the potential of intelligent, non-contact methods to improve data center reliability and efficiency.
Full article is available at Springer Link and at ResearchGate
Congratulations to Beltus for this achievement!