Application And Research Of Hilbert-Huang Transform Method In Fault Diagnosis Of Rotor System

Rotating machinery is the most widely used and prone equipment in modern industry.It is of great significance for condition monitoring and fault diagnosis of rotating machinery to ensure the normal operation of equipment and early warning and elimination of faults.Rotor is an important part of rotating machinery.Most of the vibration signals are non-linear and non-stationary.Therefore,how to extract effective fault features from rotor vibration signals and carry out pattern recognition and classification is particularly important.In this paper,Hilbert-Huang transform method(HHT)is introduced into fault diagnosis of rotating machinery.The empirical mode decomposition method(EMD)and the improved set empirical mode decomposition method(EEMD)are very suitable for processing mechanical vibration signals.They are widely used in the field of mechanical fault diagnosis and have achieved good results.However,there are still some problems such as modal aliasing,high decomposition iteration times and low efficiency.In order to solve the above problems,a rotor dynamics test-bed is built,and three kinds of faults are simulated and analyzed.The CEEMDAN based on adaptive white noise complete empirical mode decomposition(CEEMDAN)is adopted.The CEEMDAN method is compared with the EEMD method by simulation and rotor vibration signals.The reconstruction error of CEEMDAN is small and the number of iterations is greatly reduced.Low,confirming the superiority of CEEMDAN method.Secondly,aiming at the problem of false component,false component discrimination method based on correlation coefficient and index energy method are adopted.The results of simulation and fault signal analysis show that the method can effectively screen out the correct component.Finally,in the application of mechanical fault diagnosis,CEEMDAN method is combined with other mathematical methods to form a mechanical fault diagnosis method based on CEEMDAN.On the basis of singular value entropy,a method based on CEEMDAN singular value entropy and support vector machine is proposed,and on the basis of fuzzy entropy,a method based on CEEMDAN fuzzy entropy and Mahalanobis distance is proposed.The experimental data show that both methods can effectively extract fault feature signals and classify fault models.