The Application Of Clustering Analysis Based On Amplitude Entropy And Power Spectral Centroid To The Fault Diagnosis Of Rotating Machinery

With the rapid economic development and continuously upgrade ofindustrialization level, the demand for rotating machinery is more and more larger, suchas electric power, metallurgy, aviation, petroleum, chemical industry, etc.Usually thefunction of rotating machinery is irreplaceable in the whole process of production. onceit breaks down,it will cause great economic losses, so its stability and reliability isbecoming more and more important. By the method of fault diagnosis, to prevent therotating machinery rotor fault and to ensure normal work and steady of rotatingmachinery is of great significance.In the process of rotating machinery fault diagnosis,the characteristics of faultanalysis will be different for different fault object,so it is important for the effectivenessof the final diagnosis result to choose a suitable characteristic.In consideration of thefeather of rotor of the rotating machinery，a two-dimensional characteristic H A,C is presented based on amplitude entropyH A and power Spectral Centroid C,thenwavelet denoising and cluster analysis based on the two-dimensionalcharacteristic (H(A),C) is used to carry out theoretical research and experimentalverification for demonstrating the suitability of choosing the two-dimensionalcharacteristic.In the first,the reasons and the feather of the common fault of the rotor of rotatingmachinery is introduced in detail.Secondly, clustering method and its applicableconditions often used in the present situation is introduced,and the concept and basicprinciple of grid clustering involved in this article is elaborated.Thirdly, the concretemethod of wavelet denoising is presented on the basis of being familiar with theprinciple of wavelet analysis technique.Fourthly,the theoretical foundation and themethod for working out the two-dimensional characteristic(H(A),C)is presenteddetailedly.Finally, the rotor fault simulation experiment is conducted then thetwo-dimensional characteristic(H(A),C)of data sample is analysed by using waveletdenoising and clustering analysis.Results show that several common faults of rotatingmachinery rotor can be distinguished by choosing the two-dimensional characteristic.