Quickest Fault Detection over Lossy Networks for Industrial Machinery

Join a technical lecture where Dr. Krishna Chaythanya KV presents research on developing efficient fault detection algorithms for industrial machinery operating over unreliable networks. Explore innovative solutions to the challenge of detecting equipment faults when sensor data transmission is impacted by packet loss and queuing delays. Learn about a novel generalized CUSUM algorithm based on likelihood ratio principles that achieves quick fault detection even with low signal-to-noise ratios while maintaining practical detection delays and false alarm rates. Discover how the research applies these sequential detection techniques specifically to bearing fault detection by leveraging cyclostationary signal analysis. The lecture covers theoretical foundations including non-Bayesian quickest change detection, analytical performance bounds, and extensions to handle incomplete data scenarios, while demonstrating practical applications in networked industrial monitoring systems. Drawing from his industry experience and doctoral research at IISc's Department of ECE, Dr. Chaythanya bridges statistical inference, data analytics, signal processing, and communication theory to address real-world industrial monitoring challenges.