Condition Diagnosis Method Based on Statistic Features and Information Divergence

In order to extract the features from the fault signal highly contaminated by the noise, and accurately identify the fault types, a novel feature extraction method is proposed based on the statistic features and information divergence for the condition diagnosis of reciprocating machinery. A root mean square (RMS) wave, called as the ¿RW¿, is defined in the time domain using the vibration signal. A method to obtain the RMS information wave (RIW) is also proposed on the basis of Kullback-Leibler (KL) divergence using the RW. Practical example of diagnosis for the outer-race defect of a bearing is provided to verify the effectiveness of the proposed method. This paper also compares the proposed method with the conventional envelope analysis technique. The analyzed results show that the feature of a bearing defect is extracted clearly, and the bearing fault can be effectively identified by the proposed method.