Using Simplified Models to Assist Fault Detection and Diagnosis in Large Hydrogenerators

Abstract

Based on experimental evidence collected in a set of twenty 700 MW hydrogenerators, this article shows that the operating conditions of large hydrogenerators journal bearings may have unpredictable and significant changes, without apparent reasons. These changes prevent the accurate determination of bearing dynamic coefficients and make the prediction of these machines dynamic behavior unfeasible, even using refined models. This makes it difficult to differentiate the normal changes in hydrogenerators dynamics from the changes created by a fault event. To overcome such difficulty, this article proposes a back-to-basics step, the using of simplified mathematical models to assist hydrogenerators vibration monitoring and exemplifies this proposal by modeling a 700 MW hydrogenerator. A first model estimates the influence of changes in bearing operating conditions in the bearing stiffnesses, considering only the hydrodynamic effects of an isoviscous oil film with linear thickness distribution. A second model simulates hydrogenerators dynamics using only 10 degrees of freedom, giving the monitored vibrations as outputs, under normal operating conditions or in the presence of a fault. This article shows that simplified models may give satisfactory results when bearing operating conditions are properly determined, results comparable to those obtained by more refined models or by measurements in the modeled hydrogenerator.