No abstract

This paper describes a robust diagnosis method for a rotor system with a journal bearing. To enhance the robustness of a journal bearing diagnosis system, it is of great importance to define an optimum datum unit for featuring anomaly states of the rotor system. To support the research goal, this study makes use of three measures for class separation, including Kullback-Leibler divergence (KLD), Fisher discriminant ratio (FDR), and a newly proposed measure: probability of separation (PoS). From the viewpoint of class separability, this work found that PoS is more attractive than other methods for quantification of class separation. PoS offers favorable properties like normalization, boundedness, and high sensitivity. A generic algorithm integrated with one of three measures consistently suggested the optimum datum units among the feasible datum units. Optimum datum units were found to be one-cycle for time-domain features and sixty-cycles for frequency-domain features. The support vector machine (SVM) classifier with the optimum datum units was used for diagnosing a normal and three anomaly states. The health classification results showed that the proposed optimum datum units can outperform other datum units.

The power generator is typically maintained with a time-or usage-based strategy, which could result in a substantial waste of remaining useful life, high maintenance cost, and low plant availability. Recently, the field of prognostics and health management offers diagnostic and prognostic techniques to precisely assess the health condition and robustly predict the remaining useful life (RUL) of an engineered system, with an aim to address the aforementioned deficiencies. This paper explores a smart health reasoning system to assess the health condition of power generator stator bars against moisture absorption based on the statistical analysis of the capacitance measurements on bar insulators. In particular, a relative health measure, namely the directional Mahalanobis distance, is proposed to quantify the health condition of a stator bar. The smart health reasoning system is validated using five years' field data from seven generators, each of which contains 42 turns. . His current research interests include system risk-based design, prognostics and health management, and energy harvester design. Dr. Youn's research and educational dedication has led him to many notable awards, including the winner of the international PHM competitions in 2014 hosted by the IEEE PHM and PHM Society, and the ISSMO/Springer Prize for the Best Young Scientist in 2005.

Boiler waterwall tube leakage is the most probable cause of failure in steam power plants (SPPs). The development of an intelligent tube leak detection system can increase the efficiency and reliability of modern power plants. The idea of e-maintenance based on multivariate algorithms was recently introduced for intelligent fault detection and diagnosis in SPPs. However, these multivariate algorithms are highly dependent on the number of input process variables (sensors). Therefore, this work proposes a machine learning-based model integrated with an optimal sensor selection scheme to analyze boiler waterwall tube leakage. Finally, a real SPP test case is employed to validate the proposed model’s effectiveness. The results indicate that the proposed model can successfully detect waterwall tube leakage with improved accuracy vs. other comparable models.

No abstract