A study on automatic machine condition monitoring and fault diagnosis for bearing and unbalanced rotor faults

This paper demonstrates a simple and effective data-based scheme for the continuous automatic condition monitoring and diagnosis of bearing and unbalanced rotor faults. The key idea is to use a normalized cross-correlation sum operator as similarity measure for the automatic classification of machine faults using the k-nearest neighbor (k-NN) algorithm. This technique is both noise tolerance and shift-invariance. The experiments showed an error rate of 0.74% is achieved over a wide range of machine operating speed from 15Hz to 32Hz.