The Research And Application Of Early Classification Method For Rotating Machinery Fault

With the development of the machinery equipment large scale, complication and automation, fault diagnosis technology of mechanical equipment is becoming increasingly important. Since the rotating machinery is an important part of the mechanical equipment, fault classification for the rotating machinery becomes the priority among priorities. It’s also very necessary to classify the fault and take appropriate action in the early failure that not only can improve economic efficiency but also can reduce accidents. This paper mainly researched the early fault classification method for rotating machinery, and researched three early fault classification methods for rotating machinery based on wavelet analysis and blind source separation, and developed the rotating machinery monitoring and fault diagnosis system in Lab VIEW platform.This paper researched an early fault classification method based on continuous wavelet transform.It reconstructs the optimal scale selected based on the wavelet entropy and classifies the fault through the envelope spectrum of the reconstructed signal. The simulated signal and weak fault bearings signal are used to verify effectiveness of the fault classification method.When we conducted the scale optimization based on continuous wavelet transform, we found that the wavelet scales only decompose low-frequency part. So this paper researched the band optimization method to classify early fault of rotating machinery based on wavelet packet decomposition, and this method decompose not only the low-frequency part but also the high-frequency part. It conducts the wavelet packet decomposition on fault signal and the reference signal first of all, and it computes the kurtosis difference of the corresponding sub-bands afterwards. According to the validation analysis of simulated signal and the measured signal the maximum kurtosis difference band is the optimal band. The envelopment analysis of the optimal band can well classify the early fault of rotating machinery.In this paper a single-channel blind source separation is researched based on time-frequency analysis in order to analyze vibration signal which have more vibration source and a limited number of data acquisition channels. It decomposes the single-channel signal with empirical mode decomposition (EMD) first of all, it estimates the number of vibration sources and selects the optimal observed signals according to the EMD results afterwards, and the single-channel signal and the optimal observed signals constitute the multi-channel observed signals; At last it conducts the blind source separation based on time-frequency analysis. This paper specifically describes the algorithm of blind source separation based on time-frequency analysis and verifies the method with simulated signal and gearbox fault signal. Compared with the traditional ICA-based blind source separation algorithm we found that the method based on time-frequency analysis has a good separation effect for the mixed signal.This paper develops a rotating machinery monitoring and fault diagnosis system based on LabVIEW. It achieves the8-channel data acquisition through secondary development of AVANT MI-7016data acquisition device based on LabVIEW, and it online shows the histogram, time-domain waveform and frequency-domain spectrum of vibration signals, and it conducts the offline analysis through data storage such as the trend analysis, time-domain analysis, frequency-domain analysis and wavelet analysis.