Condition Evaluation of Steam Turbine Generator Using Minute-Interval Integrated Vibration Signals

Abstract

Vibration levels are a primary indicator for condition monitoring of Electrical Machine (EM) systems. This work employs Singular Spectrum Analysis (SSA) to analyze vibration time series of shaft and bearings of a real 250MW industrial power plant steam turbine Synchronous Generator (SG) in Greece. The SG operates under reduced power due to increased vibration conditions according to threshold guidelines. This work attempts to retrieve fault indications from the minute-interval integrated signal sample via its trend and periodicity components to predict vibrations under full load operation and establish correlations to discern whether the reduced output is justified. The successful application of this novel procedure is important due to ease of storing and handling low frequency signals, especially in remote environments. Rules are established using fuzzy logic due to the nature of the dataset and industrial application. The predictor employed is based on Sugeno-based Adaptive Neuro Fuzzy Inference System (ANFIS).