Research On Time-frequency Order Analysis In Fault Diagnosis Of Unstable Rotating Machinery

The detection for non-stationary fault is important for rotating machinery. Order track analysis is a frequency conversion sampling method for the non-stationary vibration signal which involved hardware order analysis and time-frequency order analysis. Time-frequency order analysis could be lower-cost and more conveniently installed because of the neglecting of complex hardware devices. But the curve fitting of rotating speed couldn't be as precise as that in hardware order analysis.Based on researches of typical failure characteristics of rotating machinery, time-frequency analysis and traditional order analysis algorithms, we focused on improving the accuracy of the current time-frequency order analysis. New methods were put forward on the key issues such as instantaneous frequency estimation, peak search and curve fitting of rotating speed. We also achieved time varying order filtering and reconstruction of time domain waveform with no speed information input.Wigner-Ville distribution was compared with traditional Short Time Fourier Transform and be chosen to improve the precision of instantaneous frequency estimation in time-frequency order analysis. Window functions were used both in time domain and frequency domain to eliminate the cross-interference components generated in the Wigner-Ville analysis.An improved cubic spline polynomial fitting algorithm and first-order difference method were used in the speed curve fitting. The resolution and accuracy of order spectrum improved significantly in our method.With discrete Gabor expansions and its inverse transformation, time various order filtering and reconstruction of vibration waveform in time domain were achieved with no speed information input in this dissertation. With the combination of order filtering and vibration waveform reconstruction, arbitrary order component in the complex multi-component signal could be abstracted and reconstructed in time domain.The correctness and effectiveness of our method was finally validated through simulation and test on ST-5000A multi-function flexible transfer systems test bench.