Fault identification and diagnosis of steam turbine generator unit is very important for safely and economically operation of a power plant. Currently, on-line monitoring is already widely utilized for alarming and recording in steam turbine. However, large amount of on-line monitoring data is not fully utilized to realize on-line fault diagnosis and identification, especially multi-concurrent fault. In the present study, model-based method was used for on-line vibrational fault identification and diagnosis based on rotordynamics. A 660MW supercritical steam turbine rotor system was modeled using FEM. Single faults, such as mass unbalance, local shaft bow and transverse crack, and multi-concurrent faults were simulated. Fault-induced changes of equivalent loads were analyzed to figure out fault type, location and severity. For model-based method, fault diagnosis accuracy is influenced by the model accuracy and signal noise. Model sensitivity was studied in this research by comparing the influence of different model error and SNR (Signal to Noise Ratio). It was used to evaluate the degree of confidence of the diagnosis result. The smaller was the model sensitivity, the higher was the degree of confidence. Based on this research, model-based method was utilized to analyze real vibration signals extracted from the historical data of this unit. Fault identification result was an effective basis for conditioned based maintenance.

This content is only available via PDF.
You do not currently have access to this content.