Dr. Rabinarayan Sethi

My research focuses on making rotating machines and their critical components more reliable, intelligent, and predictable. In many engineering systems, components such as bearings operate continuously under changing loads and speeds. Their degradation is often gradual and difficult to detect at an early stage. A small fault, if missed, can eventually lead to unexpected machine failure, costly downtime, safety risks, and loss of productivity. I work on developing intelligent condition-monitoring and health-prediction methods that can detect these changes from the signals generated by machines during operation. My research combines mechanical-system understanding with signal processing, similarity estimation, machine learning, and deep-learning approaches. Rather than simply identifying whether a component is faulty, I am particularly interested in understanding how its condition is evolving and how much useful operating life may remain. A central challenge is that real machines rarely operate under ideal or identical conditions. Changes in speed, load, operating environment, and the nature of the machine itself can make conventional diagnostic methods unreliable. My work therefore explores methods that can learn meaningful patterns from complex vibration and operational data and distinguish genuine degradation from changes caused by normal operating variability. I have also investigated hybrid learning approaches, including CNN–LSTM–GRU architectures, for predicting the Remaining Useful Life (RUL) of components. Another important part of my research is the development of health-monitoring techniques for composite and other advanced bearing systems, where conventional assumptions may not adequately describe their behaviour. Through similarity-based estimation and intelligent data-driven approaches, I aim to make health assessment more interpretable and useful for practical engineering decisions. The broader purpose of this work is not simply to develop another machine-learning model. It is to bridge the gap between laboratory-based intelligent diagnostics and dependable engineering practice. A useful monitoring system should provide information early enough for maintenance to be planned, reduce unnecessary replacement of healthy components, and help prevent catastrophic failures. In this way, my research contributes to the larger goal of moving industrial maintenance from a largely reactive approach toward predictive, data-informed, and condition-based maintenance. I see this research as an interdisciplinary effort at the intersection of mechanical engineering, dynamics, materials, signal analysis, artificial intelligence, and reliability engineering. Ultimately, I want machines to provide better information about their own health, allowing engineers to make earlier, safer, and more economical decisions.