Fault identification of rolling bearing based on difference entropy improved by period scale

It is often hard to precisely identify a bearing fault due to noise interference and faintness and complexity of fault information. With a consideration to inherent periodicity of vibration signal from rotary machine, the paper brings forward a fault identification method of rolling bearing based on difference entropy (DE) improved by multiple scales. To make comprehensive assessment of fault information in vibration signal and solve the problem with an inaccurate calculation process of traditional DE (TDE), a calculation method for probability of absolute value of each point in the sum of absolute values of all points in TDE is converted to the method of calculating the probability of occurrence frequency of each point in all data points. Furthermore, multiscale algorithm is combined with improved DE (IDE) and scale factor interval is extended to period scale to embody adaptive description of period fault characteristic contained in vibration signals of rotational machines. The mean value of IDE of signal in time sequence of each scale is treated as characteristic parameter to build feature vector and depict the characteristic information of bearing in different fault types. The proposed method is compared with other typical methods to verify its effectiveness by vibration data of rotor-rolling bearing tester and public dataset. With the two types of datasets, the identification rate of proposed method for unknown samples exceeds 99%.