Abstract
The conventional way for bearing fault diagnosis under variable rotational speed generally includes prefiltering, resampling based on shaft rotating frequency and order spectrum analysis. However, its application is confined by three major obstacles: a) knowledge-demanding parameter determination required by prefiltering, b) unavailable shaft rotating frequency for resampling as it is coupled with instantaneous fault characteristic frequency (IFCF) by a fault characteristic coefficient (FCC) which cannot be decided without knowing what fault actually exists, and c) complicated and error-prone resampling process. As such, we propose a new method to address these problems. The proposed method free from prefiltering and resampling mainly contains the following steps: a) extracting envelope by windowed fractal dimension (FD) transform, requiring no prefiltering, b) with the envelope signal, performing short time Fourier transform (STFT) to get a clear time frequency representation (TFR), from which the IFCF and the basic demodulator for generalized demodulation (GD) can be obtained, c) applying the generalized demodulation to the envelope signal with the current demodulator, converting the trajectory of the current time-frequency component into a linear path parallel to the time axis, d) frequency analyzing the demodulated signal, followed by searching the amplitude of the constant frequency where the linear path is situated. Updating demodulator via multiplying the basic demodulator by different real numbers (i.e., coefficient λ) and repeating the steps (c)–(d), the resampling-free order spectrum is then obtained. Based on the resulting spectrum, the final diagnosis decision can be made. The proposed method for its implementation on the example of simulated data is presented. Finally, experimental data are employed to validate the effectiveness of the proposed technique.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call