Method for Extracting Fault Features from Shaft Vibration Signals of Large Turbine Units Based on Gabor Transform Time-frequency Filtering

Abstract

A method based on Gabor transform time-frequency filtering is proposed to extract special abnormal frequency components from measured shaft vibration signals for stability analysis and fault identification of shaft systems. The abnormal vibration phenomenon appeared on large steam turbine-generator units is analyzed using vibration signals continuously sampled during the process of speed up or speed down. The signals are transformed into time-frequency domain by means of Gabor transform, and abnormal time-frequency components are filtered out using tracking threshold and then the filtered signal is reconstructed so that the fault feature is clearly enhanced. Abnormal vibration response characteristics are investigated and signal features are extracted. The effectiveness of the method is proved by a practical abnormal vibration case in large steam turbine-generator unit