Condition monitoring and fault diagnosis of equipment and processes are of great concern in industries. So far, various techniques in time and frequency domains have been developed for this purpose. However, there still exist certain serious and persistent problems in the situation where the signals picked up for condition monitoring and fault diagnosis demonstrate a transient or nonstationary nature. Owing to their intrinsic drawbacks, traditional techniques such as spectral analysis are often ineffective in dealing with such signals.

Wavelet analysis provides a promising tool to overcome the above-mentioned problem. In recent years, CWAIP has developed several wavelet-based techniques for the feature enhancement and feature extraction of transient and nonstationary signals. These techniques are much more effective than traditional techniques and have been successfully used in the condition monitoring and fault diagnosis of some typical electrical and mechanical systems. These techniques are applicable to the condition monitoring and fault diagnosis of

• Manufacturing equipment and processes
  
• Electrical and mechanical equipment in power systems
  
• Equipment in petrochemical and oil industries
  
• Construction machinery
  
• Multimedia and internet-based applications;
  
  

An application example ---- Failure detection of rolling element bearings

Rolling element bearings are used in various machines. Failure detection of such bearings has received considerable attention. Localized defects such as surface spall and surface crack are a typical failure form of rolling element bearings. These defects reduce the service life of bearings severely and are among the foremost breakdown causes of rotating machinery. Vibration generated in a normal bearing is usually dominated by the components caused by shaft rotation, stiffness variation, load fluctuation, etc. When a localized defect is induced, repeated impacts will be generated due to the passing of the rolling elements over the defect. The wide-band energy of the impacts often sets off some modes of resonance of the bearing elements, the neighboring structure and the sensor. This adds additional impulsive components into the vibration and results in vibration signals of a non-stationary nature. The main difficulty in the detection is that these featured components are often submerged in the background vibration especially in the early stage of failure development. Enhancing and extracting these components are the main tasks in detecting the defect. Here is an example of detecting a localized defect induced on the inner race of a ball bearing. Comparative study shows that the wavelet-based technique performs better than others.


Fig. 1: (a) A vibration signal collected from a ball bearing with a localized inner race defect, and (b) its power spectrum.

Fig. 2: Envelope of the signal detected using Hilbert transform (a widely used traditional technique for bearing defect detection)

Fig. 3: Time-frequency distribution of the signal, where the vertical patches show the impulsive components of the signal.

Fig. 4: Signature extracted with a wavelet-based technique. It clearly shows the presence of the defect