Hyper-spherical distance discrimination: A novel data description method for aero-engine rolling bearing fault detection

A novel method called hyper-spherical distance discrimination (HDD) is proposed in order to meet the requirement of aero-engine rolling bearing on-line monitoring. In proposed method, original multi-dimensional features extracted from vibration acceleration signal are transformed to the same dimensional reconstructed features by de-correlation and normalization while the distribution of feature vectors is transformed from hyper-ellipsoid to hyper-sphere. Then, a simple model built up by distance discriminant analysis is used for rolling bearing fault detection and degradation assessment. HDD is compared with the support vector data description (SVDD) and the self-organizing map (SOM) in rolling bearing fault simulation experiments. The results show that the HDD method is superior to the SVDD and SOM in terms of recognition rate. Besides, HDD is applied to a run-to-failure test of aero-engine rolling bearing. It proves that the evaluating indicator obtained by HDD method is able to reflect the degradation tendency of rolling bearing, and it is also more sensitive to initial fault than the root mean square (RMS) of vibration acceleration signal. With the advantages of low computational complexity and no need to tuning parameters, HDD method can be applied to practical engineering effectively.