Feature Extraction Methods Based On Signal Processing For Rolling Element Bearing Fault Diagnosis

In chemical and petrochemical industries, equipments and machines are important tools which directly influence the stability and security of industrial processes. In industrial settings, large rotating machines demanding high performance criteria are being increasingly used, and the machine failures can lead to great economic loss and safety problems. The prediction and diagnosis of machine faults are important for improving operational safety and reliability. Rolling element bearings are the most important and frequently encountered components in a wide variety of rotating machinery, and their failure is one of the most frequent reasons for machine breakdown. According to statistics,30% of rotating machinery faults is caused by rolling element bearings. In recent years, fault identification of rolling element bearings has attracted extensive attention and research efforts, and different methods have been developed. This thesis aims to develop effective feature extraction methods for rolling element bearing fault diagnosis, which will be tested by vibration data obtained from Bearing Data Center in Case Western Reserve University (CWRU). The main works and results include as follows:(1) The envelope spectrum of rolling element bearing vibration signals are obtained from discrete wavelet transform and Hilbert transform. Then, correlation analysis is used to identify various bearing faults. The diagnosis results show that the proposed approach can perform successfully in identifying various incipient bearing faults, the level of fault severity and also the defect location of outer race faults. Compared to the traditional envelope spectrum analysis, the proposed method performs better in bearing fault identification, which can make full use of the frequency information over the signal's envelope spectrum.(2) Fault diagnosis method based on morphological analysis is studied. A double-dot line structuring element is proposed for morphological operation, and the multi-scale opening operation combined with correlation analysis is utilized to diagnose the rolling element bearing faults. Pattern spectrum, obtained from multi-scale morphological operation results, is used as a feature extraction index. A correlation analysis gives the final identification result by utilizing information over full pattern spectrum. Moreover, the definition of scale spectrum is also proposed on the basis of multi-scale morphological operation. Different from the approach mentioned in the above part, this method attempts to extract signal features from multiple scales. According to the test results, it shows that this method is more valid in incipient fault diagnosis and fault level identification, certainly which is also available for detecting defect location of outer race faults. That indicates the double-dot line structuring element performs well in signal feature extraction. In addition, the results also show the validation of scale spectrum in bearing fault diagnosis, which even works better than pattern spectrum in indentifying defect location of outer race faults.