On the question of turbine fluid-flow system lifetime prognosis

摘要

In his previous studies, the author has investigated the problem of steam turbine fluid-flow system lifetime prognosis on the basis of vibration-based diagnostic symptoms. Preliminary results have been encouraging, but more detailed analysis has pointed out that two issues are crucial for a reliable prognosis, namely - selection of the most informative symptoms, and - proper processing of measurement data, in view of strong influence of interference and control parameters. Both these questions are addressed in the paper. Symptom selection employs SVD (Singular Value Decomposition) and ICM (Information Content Measure) methods, the latter significantly refined and suited to the particular class of objects. Data processing involves mainly removing 'upper' and 'lower' outliers. Influence of outlier removing thresholds on the prognosis is discussed. Examples are provided for a steam turbine operated by an utility power plant.