The Gearbox Fault Diagnosis Based On Generalized Morphological Filtering And Ensemble Empirical Mode Decomposition

With modern technology and automation level of science continues to improve, thegearbox as an important transmission components, is widely used and playing an importantrole in transportation, energy, power, metallurgy, chemical, aerospace and other fields, so thefault diagnosis of gearbox is also very important.This paper studies the dynamic model of gearbox, and all kinds of gear, bearing and shaftfailure mechanism are analyzed in detail, and summaries gearbox vibration signalcharacteristics of various types of failures. Then introduces the three main methods ofgearbox vibration signals include traditional time-domain, requency domain andtime-frequency method. Aiming at the shortage of traditional filtering method, the paperconstruct a generalized morphological filter and eliminat the phenomenon of statistical biasoutput. Then introduce the types of structural elements and selection principle and throughsimulation signal to the traditional form and generalized morphological filter and filter effectswere compared to verify that the method of. In the process of obtaining the fault characteristicfrequency, for the EMD decomposition has end effects and modal aliasing, using theextension of extreme points and ensemble empirical mode decomposition (that is EEMDdecomposition) to overcome the above shortcomings, in addition the traditional packageenvelope spectrum analysis does not reflect the details of the signal, the resolution is not highand the energy leaks and other problems, the use of Hilbert marginal spectrum can effectivelysolve the above problems, and through simulation signal verify the effectiveness of combiningthe two methods.Hilbert marginal spectrum overcome the traditional envelope method needs to determinethe center frequency and lack of bandwidth of the bandpass filter. But because of the EEMD decomposition is susceptible to noise interference, so denoising becomes even more important.To solve these problems, this paper combines three ways of generalized morphologicalfiltering, EEMD decomposition and Hilbert marginal spectrum. Firstly, using generalizedmorphological to filter fault signal and reduce noise interference, then using EEMD todecompose de-noised signal, for the decomposition of each IMF component, using thecorrelation coefficient method to select out the right IMF component, which constitutes partialHilbert marginal spectrum. Through the judge of fault characteristic frequency, distinguishingthe failure of inner and outer ring, cage of bearing and gear wear and broken tooth amongthem successfully, so as to achieve to distinguish fault type of different parts of the gearboxeffectively. The results can be seen from the combination of the above three methods can begood at filtering out different types of noise in the environment, decomposing denoised signaland extracting signal feature of gearbox, so the results of this paper can be applied not only inthe gearbox fault diagnosis, but can also be other similar mechanical fault diagnosis, with amore extensive application prospects.